This PDF file contains the front matter associated with SPIE Proceedings Volume 12451, including the Title Page, Copyright information, Table of Contents, and Conference Committee listings.

Taking the subject collected in the web of science database as the literature data of the marine Internet of things as the research object, using document metrology and citation analysis methods, with the help of histcity citation visualization software, this paper reveals the citation context of the global marine Internet of things research field, clarifies the development track of the marine Internet of things research field, and analyzes the core authors and documents supporting the marine Internet of things technology, Provide reference for the development of marine Internet of things system. It is found that the research of the marine Internet of things has experienced the stage of preliminary development, prosperity and deepening development, and a number of core documents and authors have emerged. The trend of Group in-depth cooperation needs to be strengthened, and the discipline needs the deep integration of computer science and marine science. The architecture of the marine Internet of things and the key technologies of hybrid satellite ground communication of the marine Internet of things have become important citation objects and research hotspots in the field. The marine IOT communication mode will be extended in the direction of diversification and intelligence, taking into account the security development of the marine IOT cloud.

Based on the survey of 43 Marine ranches and 260 consumers, this paper uses the diversified and orderly Logit model to study the significant factors affecting the development of recreational fishery in Marine ranches. The study found that six factors, including consumer gender, individual economic strength, consumer demand for Marine pasture products, brand construction level of Marine pasture recreational fishery, online channel promotion level of Marine pasture recreational fishery and the impact of COVID-19 epidemic, had a significant impact on the development of Marine pasture recreational fishery. This paper divides tower into product development and brand building, marketing and daily operation three dimensions, suggest operators improve the level of recreational fishery product development and brand building, develop differentiation price, develop online channels, strengthen the whole process communication, innovation under the outbreak of daily operation mode, expand the market share, so as to enhance the competitiveness of Marine pasture recreational fishery.

In order to overcome the limitations carried by single-objective optimization of freight (railway and highway) network, in this paper we propose a dual-objective optimization model with minimum cost and shortest mileage using the augmented epsilon constraint algorithm. Specifically, the dual-objective optimization model is based on the characteristics of the actual traffic flow, and the relevant constraints. This paper uses a random example to demonstrate the distribution of non dominated solutions. The results show that our model and its algorithm can find high-quality representative non-dominated solutions within a reasonable time, which is significant for solving the loaded and empty car flow path problem in the railway network.

Textile trade requires a uniform digital softness evaluation index in the context of well-developed online trading as opposed to the sensory evaluation provided by actual touch. The fabric touch tester is used to assess the softness scores of 150 fabrics and 18 mechanical indexes in response to the existing issue that the fabric softness evaluation model does not have a wide application value. A subjective assessment experiment and objective score clustering are used to compare the validity of the objective scores. The samples are divided into three thickness categories based on one-way ANOVA and cluster analysis: light, medium, and thick. Correlation analyses are done between mechanical indexes and fabric softness ratings in each category, and stepwise regression is used to establish four prediction models for fabric softness. The findings indicate that thickness and weaving method considerably impact the fabric's softness among the fundamental variables. The bending, thermal conductivity, compression, and surface properties—in descending order— among the mechanical parameters—have an effect on the fabric's softness. The R² of the established prediction models for the softness of light, medium, and thick fabric, respectively, reached 83.2%, 88.7%, and 84.1%. It can help customers and businesses come to an understanding about the softness of fabrics, increase the number of transactions, and improve the effectiveness of online purchasing.

This paper starts with circular packaging, combined with the characteristics of low-carbon supply chain, expounds that circular packaging includes circular outer packaging and circular inner packaging, analyzes its application bottleneck in low-carbon supply chain, and puts forward suggestions to promote the application from multiple angles, hoping to play a certain role in promoting the development of low-carbon supply chain for logistics enterprises, governments and social organizations.

The total nitrogen (TN) in the effluent of wastewater treatment plant is an important index to evaluate the effect of water treatment, and its accurate prediction can provide data support for the optimal operation of wastewater treatment process. The Quantum Weighted Minimal Gated Unit (QWMGU) neural network by introducing a quantum computing mechanism into MGU network is presented to predict the effluent total nitrogen. Firstly, a multi-dimensional single-step time series prediction method was used to construct the TN dataset of wastewater; Then, the above index set was input into QWMGU to complete the TN prediction; Lastly, some comparison experiments with the water outlet data of the actual sewage treatment plant were performed. The results show that the proposed TN prediction model based on QWMGU has higher prediction accuracy than QWGRU/MGU/GRU/LSTM, and faster convergence speed than QWGRU/GRU/LSTM.

In this paper, The construction disturbance effect of subway tunnel on the cable tunnel of various cross-sectional sizes under various spacing is studied using numerical simulation. The two common working conditions are orthogonally undercrossing and obliquely underpassing cable tunnels under subway undercover tunnel. The main conclusions are: (1) The settlement and tensile stress value of the existing cable tunnel generated by the orthogonally undercrossing is higher than that of the obliquely underpassing; (2) The larger the cross-sectional size of the existing cable tunnel, the higher its structural longitudinal stress and the lower the safety.

Based on the application status of VR technology in modern education and teaching practice, this paper determines that Unity3D software is used as the development engine, and 3ds Max software is used to complete the realization of 3D digital model and virtual scene. Under the guidance of learning and cognitive theory related to virtual simulation experiment teaching, the design and implementation of integrated circuit virtual simulation experiment system are completed. With the help of the practical characteristics of VR technology, 3D digital modeling technology and visual interaction technology, and combining the efficiency and convenience of Web application, the system puts forward a comprehensive application solution for the problems existing in the current training process of integrated circuit professionals in colleges and universities. Systematically and innovatively combining online teaching with virtual simulation experimental teaching can provide students with an immersive experimental learning environment, making teaching activities more vivid, intuitive and understandable. It is not only conducive to the mastery of students' professional knowledge, but also promotes the improvement of students' ability to integrate theory with practice, and further accelerates the development of information teaching construction in colleges and universities while promoting the reform of teaching mode.

In order to detect the conductivity of electrode materials for supercapacitors, a new method is used in this paper, which is called Magnetically Mediated Thermoacoustic Detecting Method (MMTDM). Firstly, the principle of Magnetically Mediated Thermoacoustic Detecting Method is studied. Secondly, the simulation model of Magnetically Mediated Thermoacoustic Detecting Method is built, and the ultrasonic signal of the model is obtained. Finally, the self-made supercapacitor electrode material and pure aluminum foil were used as experimental targets for detection and comparison. The results of numerical simulations and physical experiments show that the Magnetically Mediated Thermoacoustic Detecting Method can realize the non-contact detection of the conductivity of energy storage materials. This paper provides a good theoretical and experimental basis for the application of Magnetically Mediated Thermoacoustic Detecting Method in the non-contact detection of energy storage materials.

This paper paid attention to the issue that there is jitter in the folding action of sidelobes antenna in the existing radar frame withdrawal system. The working principle of the hydraulic system and the mechanism is started to focus on the analysis of the cause of the failure. It is considered that differential loads in the system lead to pressure shock and jitter. The electro-mechanical and hydraulic system co-simulation technology is used to simulate the load dynamic characteristics of the folding system of the primary and secondary antennas and to reproduce the faults. The simulation test and parameter optimization of the system are carried out by adopting a sequential valve to absorb the pressure pulsation. Finally, the system was built in the laboratory to complete the functional test. The folding process is smooth and the result is consistent with the simulation. This verifies the rationality of the optimization method. It provides a reference for the development of the hydraulic system of the radar frame.

To realize the prediction of carbon trading volume of power generation enterprises, this paper takes a power generation enterprise as an example, and introduces the causal inference theory to study the carbon trading behavior of enterprises. Data mining technology is introduced to obtain the operation data of power generation enterprises in the market. It is also used to collate the obtained multi-channel data, and cluster them according to the data category to extract the attribute characteristics of carbon trading behavior of the power generation enterprises. Based on the causal reasoning relationship and referring to the SEM model in the causal reasoning relationship, the carbon trading behavior model is constructed. According to the market access conditions of sellers and sellers, the balance of power supply and demand is taken as the starting point to predict the carbon trading volume of power generation enterprises and complete the analysis of carbon trading behavior of power generation enterprises. Taking a power generation enterprise as an example, related application experiments are designed. The experimental results show that the prediction results are reliable by using the designed method to predict the carbon trading volume of enterprises, which can be used as the key basis for the decision-making of carbon trading behavior of power generation enterprises.

Nowadays, tunnel managers are paying more and more attention to the real-time sensing of tunnel morphology. At the same time, many researchers have proposed methods for real-time monitoring of tunnel morphology based on fiber-optic grating sensors, but these research methods have rarely discussed the impact of the layout of the monitoring scheme on the reconstructed shield tunnel profile. This study explores the problem of optimizing the layout of sensors in shield tunnel sections by using the DQN algorithm based on reinforcement learning algorithms. A cross-river shield tunnel is used as a model, and its daily loading conditions are simulated to verify the effectiveness of the DQN-based shield tunnel section layout optimization. It is also compared to the commonly used uniform and the traditional optimized layout. The results show that the optimization method in this study improves the accuracy of reconstructed section curves by 78.29% over the uniform layout and 70.11% over the traditional optimized layout, proving that the method proposed in this study can better reconstruct the realistic shield tunnel section morphology.

To improve the response time of public opinion on early warning of power service and expand the coverage of early warning, this time, combined with the theme of web crawler technology, the research on public opinion on early warning technology of power supply service is carried out. According to the power service situation, collect the public opinion data of power supply service, analyze the public opinion situation according to the service results, and conduct topic discovery. Based on this, the overlapping web crawler public opinion early warning model is designed, and the public opinion threat estimation method is used to realize early warning processing. The final test results show that: Compared with the traditional clustering public opinion early warning test group and the traditional sentiment calculation public opinion early warning test group, the final early warning response time of the theme web crawler public opinion early warning test group designed in this paper is controlled within 1.5s. It shows that in the practical application process, the method in this paper has a fast early warning speed for public opinion, and the error of data and information processing is small, which has practical application value.

As the communication network technology develops, the switch is of special importance in the secure communication configuration. A switch can be connected to the computer directly through the "Console" port and the common port. Especially when the user is in the office or on a business trip, it is necessary to manage the device remotely. At this time, the switch configuration cannot be configured locally, but needs the Telnet or Web browser to achieve the switch configuration. Usually, if multiple VLANs are needed for division, there will be different gateways for each VLAN. In practice, there are many multi-VLAN multi-triple convergence scenarios. Even the intermediate device has a route, if the device is connected by remote management without a route back, the data packet will be in a single-pass state. Although it can be sent out, the path is presented as inaccessible without a route4. The design is based on the Cisco simulator and the combination of VLAN trunking protocol (VTP) technology and VLAN technology of port division with Layer 2 and Layer 3 switches as core devices to achieve unified configuration and management of VLAN and remote password login management mode. Users can access the device management through a highly secure encrypted channel to avoid illegal user intrusion to obtain management resources and harm the network. In turn, it can restrict the remote IP address that allows access to the switch through Telnet and Web and guarantee the security function of network management.

In the paper, the automobile auxiliary instrument panel is taken as an example to illustrate the production process, and Flexsim software is employed to establish a simulation model of the production line of automobile auxiliary instrument panel to analyze the bottleneck operation in the production process of the product. Then the production line balance principle and ECRS principle are employed to adjust the production process flow and to reasonably configure machines and operators of the production line so as to improve the production efficiency of the production line.

ZrO2 electrolyte is one of key parts of solid electrolyte oxygen sensors. The heat storage performances and heat storage stability of ZrO2 electrolyte influence the performance of solid electrolyte oxygen sensors directly. Sensors work in high-temperature changing environment and the prominent non-harmonic effect of atomic vibration of ZrO2 electrolyte materials can influence heat storage performances and heat storage stability of sensor significantly. To improve performances and thermostability of electrolyte, non-harmonic vibration of atoms were taken into account in this study. The variation laws of heat capacity and heat storage stability of ZrO2 electrolyte materials with temperature and time were investigated by solid physical theories and methods. Moreover, influences of non-harmonic vibration of atoms were discussed. Results demonstrate that the Debye temperature achieves a linear growth as temperature rises due to the nonharmonic effect. The constant-volume specific heat presents a nonlinear growth as temperature rises and it shows an approximate negative exponential attenuation as time goes on. The constant-volume specific heat decreases due to the non-harmonic effect. The reduction amplitude increases firstly and then decreases with time, but it increases gradually as time goes on. Heat storage stability weakens firstly and then enhances with the increase of temperature and it is poorest at about 160K. Heat storage stability is weakened to a small extent due to the non-harmonic effect. The results are basically consistent with other literature results. Research conclusions can provide theoretical references to study stability and service life of solid electrolyte oxygen sensors.

In commercial flight trajectory prediction, the models based on deep learning have high versatility and accurate forecast. However, these prediction models have the problem of poor effect when using small data sets. This paper proposes a commercial flight trajectory prediction model based on representative trajectory. This model consists of two parts: Representative trajectory generation and Trajectory prediction. Firstly, our model clusters the commercial flight trajectories on the same route, then generate representative trajectories to represent the route pattern. Secondly, our model uses the Savitzky-Golay filter to eliminate the noise characteristics in these representative trajectories, then combines those with a deviation correction algorithm to predict flight trajectories. By comparing with several other prediction models, the experiment results support that our model performs better in a small data set of the commercial flight trajectory.

For the high-resolution remote sensing building image, the object shape is irregular and the texture characteristics are similar to those in the foreground and background. Meanwhile, traditional manual methods spend considerable time extracting features. In this paper, a Deeplabv3+ algorithm is modified and improved to better meet the accuracy, massive, and real-time requirements in semantic segmentation of high-resolution remote-sensing building objects. First of all, the lightweight network MobileNetv2 is implemented to replace Xception as the backbone to extract the network, so as to reduce the number of network parameters and improve the training speed. Secondly, ASPP (Atrus Spatial Pyramid Pooling) obtains the depth effective feature layer and integrates three parallel channel attention mechanisms to improve the extraction effect of depth features. The newly proposed algorithm is tested on the Massachusetts dataset. The results show that the IOU and accuracy of evaluation indicators are increased by 8.31% and 2.11%, respectively, compared with the original Deeplabv3+ model. While reducing the number of model parameters and training time, the segmentation effect of the new algorithm has also been partially optimized.

It has become the development trend of garden plant industry to use 3D digital model for aided design, effect display and later management and protection. It is the key element of garden plants. By discussing the necessity of building and applying the 3D digital model of garden plants, this paper uses the 3D plant modeling software PlantFactory to explore the 3D digital model building and build a 3D digital model information database about garden plants. Through the application of 3D digital model of garden plants in plant morphology cognition, plant scene design and expression, and the construction of plant garden information model, this paper explores its applicability and effectiveness. It provides a brand-new way for information protection of garden plants. In this process, researchers gradually get rid of the limitations of hand-drawn drawings with the help of computer and digital technology, which greatly improves the efficiency of the whole research process. Drawings are stored in digital form, and fast retrieval is supported. At the same time, the expression of animation can be output through the system webpage, which meets the diversified needs of the display forms.

Aiming at the problems of low efficiency, high labor intensity and poor real time of manual detection of inner glue defects in high frequency transformers, and the difficulty of traditional detection algorithms to quickly and accurately detect the defects of high frequency transformer images with low contrast and strong reflection, a high frequency transformer inner glue defect detection system is designed to achieve the purpose of automatic and rapid detection of workpieces. The hardware side of the system uses CMOS camera and LED low-angle ring light source by analyzing the workpiece structure and the image characteristics of the defect area. The software first uses VisionPro vision software to design the image pre-processing program to locate and crop the original image to obtain the image of the inner glue area, then uses VisionPro Deep Learning software to build the defect detection model, and finally integrates the image preprocessing program and the detection model and designs the upper computer operation interface. The experimental results show that the detection accuracy of the system is 99%, and the average detection time per image is 46.7ms, which can quickly and accurately detect the defects of inner glue of high frequency transformers and meet the requirements of real-time inspection of factory products.

As an important agricultural product, the market price of chemical fertilizer is regulated by various factors such as raw material price, exchange rate policy, supply, etc., and the fluctuation of chemical fertilizer price has a significant impact on the national economy. It can be seen that if the price of chemical fertilizer can be accurately predicted, its impact on agriculture and even the national economy can be minimized. Therefore, in order to accurately predict the price of fertilizers, this paper proposes a bidirectional long-short term memory neural network model based on the attention mechanism(A-Bi-LSTM). The model makes full use of the contextual relationship between the forward and backward directions on the time series, which can effectively adopt the long distance information for the sequence data relying on. Combined with the trend data of fertilizer transaction prices in recent years, this paper performs regression fitting on the data to generate a training model which is used to predict the price of chemical fertilizer. Finally, the root mean square error on the test set is 0.011, and good prediction results are obtained.

With the development and progress of science and technology, the Internet of Things, as a new generation of information technology, has been widely used in oil and gas production and other industries. The Internet of Things in oil and gas production (A11) is the application of a new generation of information technology in oilfield production. In the planning and construction of PetroChina, it is further determined that the system construction of the Internet of Things (A11) for oil and gas production is the key direction of future development. This paper studies the application of key technologies of Internet of Things (A11) in oilfield production. It uses Internet of Things technology to realize remote monitoring of the production process, and uses smart equipment to intelligently control production dynamics, thereby effectively improving production and management efficiency. Therefore, this article will first briefly introduce the Internet of Things technology, then discuss the application of the Internet of Things technology in major domestic and foreign companies, and finally discuss the specific practice of the Internet of Things technology in the oilfield.

The article first summarizes the meaning and characteristics of the "last mile" logistics distribution model in the fresh food e-commerce industry, then builds a multi-layer evaluation index system based on the products and demand characteristics of the fresh e-commerce industry, and evaluates it. On the basis of the index system, the weight of each index is calculated using the analytic hierarchy process, and the different distribution modes under a single index are selected by analysis and comprehensive evaluation. Finally, it concluded that the self-operated mode is the optimal distribution mode. Corresponds to the mainstream distribution mode of fresh food e-commerce enterprises, it verifies the rationality of the method, and provides a good decision basis for related companies to carry out the selection and evaluation of the "last mile" logistics distribution mode.

The positive interaction between tourism and agriculture is conducive to the win-win outcome of the industry. Based on the statistical data of tourism and agriculture in Liaoning Province from 2010 to 2019, the coupling evaluation model and index system of tourism and agriculture were constructed on the basis of discussing the interaction between tourism and agriculture development. With the help of the coupling model, the coupling coordination relationship between tourism and agriculture and its spatio-temporal variation law were analyzed. The main factors affecting its coupling and coordinated development are explored by using grey relational model. The results showed that :(1)the evaluation function values of tourism and agriculture in Liaoning province showed an upward trend, but the tourism development showed obvious fluctuation; The coupling degree was in the running-in coupling stage from 2010 to 2012, and then entered the high level coupling stage. The coupling coordination degree showed a continuous and rapid upward trend, realizing the transition from the dissonance stage in 2014-2016 to the coordination stage in 2017-2018.(2)Through the grey correlation analysis, it can be seen that travel agencies, star hotels, domestic tourism income, per capita disposable income of rural residents, agricultural gross output value and other indicators are the main internal factors affecting the development of the two systems. On this basis, relevant suggestions are put forward for the coordinated development of tourism and agriculture in Liaoning province.

Automobile design review is an important link of automobile product development process, which need to balance the contradictions between various objectives, to promote the progress and quality of automobile design review. With the development of virtual reality technology, it has realized the evaluation of automobile product modeling in the virtual scene based on VR glasses. From the perspective of hardware improvement, the paper designed a VR helmet that can improve motion sickness and enhance user experience, to expand the application field of virtual reality technology. Then the hardware and software environment are integrated to build a collaborative vehicle virtual evaluation system, which provide a referential and feasible technical scheme for automobile enterprise design review.

Total amount accounting of pollutants into river is the key link of total amount control system,under the condition that non-point source pollution has not been monitored effectively,according to the principle of pollutant input-output balance, MATLAB software is used to optimize the solution function, the Qing He river basin was taken as the demonstration area,the water quality monitoring data、 pollution source census data and monitoring data from 2009 to 2013 were used to build a binary statistical model for the total amount of pollutants discharged into the river. Nash efficiency coefficient (NSE) was used to verify the model coefficients, the statistical model is established and four model coefficients are proposed. The simulation results can accurately calculate the monthly amount of ammonia nitrogen into the river,the established statistical model has strong data assimilation ability, it can accurately calculate the annual amount of ammonia nitrogen into the river, the test results can meet the needs of practical research.

With the gradual maturity of solar thermal power generation technology and the business model of solar thermal power plants, solar thermal power generation has become a good way to accelerate the penetration of new energy to replace traditional energy. A new energy ecosystem is taking shape. The uncertainty of wind energy and load can not be ignored, and effective measures must be taken to ensure the safety of the energy system. In this paper, based on the chance-constrained Gaussian mixture model, the uncertainty set of wind power forecasting error and the uncertainty set of load forecasting error are constructed. The robust economic multi-objective dispatching model is established by incorporating the robustness into the collaborative optimization objective to obtain the robust economic dispatching optimization scheme with higher comprehensive satisfaction. The research in this paper can provide scientific guidance for the robust economic dispatch of cogeneration microgrids with solar thermal power plants, and also provide new ideas for increasing the consumption of clean energy. At the same time, the proposed robust economic dispatch method can be applied to other microgrid dispatchs, which has certain academic and engineering practical significance.

Based on the increasing number of smartphones used by the elderly in China and the difficulties encountered by the elderly when they seek medical treatment, a new appointment registration method based on Android and a medical convenience assistant designed by navigation are adopted. The App has four functional modules: login registration, medical route planning, medical consultation, and registration order management. The function of the client and the compatibility test of different mobile terminals proved that the App could realize intelligent registration, simulated payment, route planning, and traffic planning as one of the intelligent appointment registration medical services.

To the relevant policies of the carbon trading market in China, this paper adds the optimization algorithm that longicorn must use based on the traditional BP neural network, which increases the convergence speed and global optimization ability of the original BP neural network. The empirical study takes the carbon trading price of Shenzhen city, Guangdong Province, China, from 2015-to 2022 as an example. The results show that the base-BP neural network can significantly reduce the probability of local convergence in the iterative process and reduce the mean square error of prediction by 3.08% compared with the traditional BP neural network.

Chemical mechanical planarization (CMP) can effectively take into account the global and local flatness of the matters surface. The accurate control of CMP process largely depends on the understanding and of its material removal mechanism. At present, people do not fully understand the removal mechanism of CMP, but still control the CMP process through empirical or semi-empirical means. Therefore, exploring the mechanism of CMP is an important basic research to improve the technical level of CMP. The model is a chemical-mechanical synergetication in the CMP in terms of change in the chemically modified surface layer and wafer surface topography. The model proposed takes process parameters including pressure and velocity and other important input parameters including the wafer hardness, pad roughness and abrasive size into account to predict the material removal rate. At the end of the article, the key technical and future research focus issues that need to be further resolved are put forward.

In this paper, based on previous literature, it summarizes the influencing factors of dry bulk shipping market in such time dimensions as instantaneous, seasonal, short-term, medium-term and long-term, demarcates the boundary of dry bulk shipping system and its endogenous and exogenous variables, and obtains the endogenous causality diagram of dry bulk freight fluctuation by system dynamics analysis. Furthermore, by looking for the regression relationship between the main variables in the same or early or late period of the series, the dynamic prediction model of the shipping market is established to deduce the change of supply and demand of the international dry bulk shipping sector and the development trend of the annual BDI index, in order to achieve the purpose of providing reference for shipowners, investors and freight derivatives dealers in the international dry bulk shipping.

In the transition process of tailrace system of large hydropower station, there may be the phenomenon of free-surfacepressurized flow. In order to obtain the accurate operating state of constant flow in the free-surface-pressurized flow section, four calculation methods including Runge Kutta method, one-dimensional numerical simulation software, physical model test and three-dimensional flow field analysis software CFD are used to obtain the water surface profile under constant flow conditions. The results show that the calculation results of each method are relatively consistent, indicating that the research method and route are correct and reasonable. The established mathematical model is reasonable and can be used to study the hydrodynamics of free-surface-pressurized flow. Considering the difference between the outlet of tailrace tunnel and one-dimensional simulation method obtained from physical model test and three-dimensional flow field analysis, the water surface lines of the free-surface-pressurized flow section of tailrace tunnel under constant flow conditions obtained by various analysis methods are basically coincident. The analysis of the local flow field at the outlet of the tailrace tunnel shows that it is feasible to use the three-dimensional flow field simulation instead of the physical model test to simulate the more complex working conditions. The research results can provide reference for the design and research of similar water conveyance systems.

With the continuous development of China's economy, the country's infrastructure and the rapid development of computer industry, concrete mixing station weighing control system for improving production efficiency plays a decisive role. As the core part of concrete production, the weighing control system of concrete mixing station plays an important role in the weighing of concrete materials and the performance of finished concrete. In this paper, the mixing station structure and construction technology analysis of the mixing station weighing system, using control algorithm to achieve the mixing station weighing function to replace the control of conventional instruments. The stability of system operation is improved, the intelligent diagnosis function is increased, and the hardware cost is reduced.

In order to improve the reliability of pipeline design, a large amount of rich spatial data is required in the early stage of the project. Usually, the project in this stage does not require high data accuracy, but requires a large spatial coverage. At the same time, the construction period is usually very tight, and it is inconvenient to carry out a large number of field data collection works. This paper proposes to convert the route to SHP format through the FME tool, cut out the corresponding part of the TIF data, convert the raster data into vector data and perform area calculation, and then classify and output the statistical area. Through this process of FME, the surface coverage within 2 kilometres along the pipeline can be quickly and effectively calculated, and the vector data of the corresponding area can be output at the same time. It provides an important data basis for optimizing the route selection scheme and calculating the restoration of soil and water conservation plants and the compensation of surface economic crops.

In view of the damage that cannot be detected during the long-term operation of the elevator wire rope, the system adopts a magnetic detection method by comparing several representative non-destructive testing methods for ferromagnetic material damage. By establishing the wire rope leakage magnetic field analysis model, the U-shaped sensor array and the corresponding signal acquisition and pre-processing circuits are designed. Through practice, the non-destructive testing scheme of the wire rope proposed in this paper can provide monitoring and guarantee for the safe operation of the elevator. At the same time, it can also provide a reference for the maintenance and management of other on-site wire ropes in China.

This paper introduces a paper counting device. The system consists of STM32 core board, FDC2214 module, OLED display module, buzzer module, key module, and two parallel plates. Two copper clad plates with a side length of 50mm square are used as capacitance plates, and the paper to be measured is placed in the middle, and the capacitance value between the two plates is collected by FDC2214 and the collected data is transmitted to the MCU controller through the IIC serial bus. The controller through the algorithm converted to the corresponding number of paper values displayed in the OLED display. The system sets self-calibration mode according to adjust the size of capacitance.

IDS (Intrusion Detection Systems) include IDS based on host (HIDS) and IDS based on network (NIDS). Based on Spark’s unified host-based intrusion detection framework, including centralized Machine Learning (ML) in the cloud. FEDFOREST is a new layout based on NIDS design, the local-global computation paradigm. This paper compares the frameworks in two different systems under IDS, analyzes the difference of algorithm logic and makes a summary comparison. It focuses on threat, accuracy and efficiency. After research, it is found that HIDS using spark in cloud is superior to NIDS using FEDFOREST framework in three aspects. This is helpful to study the threat, accuracy and efficiency of NIDS.

The path planning of high-speed aircraft at land-sea interface area determines the accuracy of the whole path planning. In this paper, based on the establishment of scene-ring, the scene matching method of SIFT-GPU algorithm is used to judge whether there is any deviation in the position and heading of high-speed aircraft at key waypoints, so as to adjust the cruising state of high-speed aircraft in time. It has certain practical value and reference significance for the study of path planning of high-speed aircraft.

This paper proposes a prediction approach based on MLP-Mixer with FFT (The fast Fourier transformation). The wind speed series dataset was transformed using the FFT. Extract high dimensional features initially, then a deep learning time series prediction based on MLP mixer is introduced to explore and exploit the implicit information of wind speed time series for wind speed forecasting. We compared the different wind speed forcasting results by setting the lookback premeter to 4, 8, 12, and 16 hours. On the basis of two years of test dataset, the performance of the proposed FFT-MLP-mixer is effectively validated for short-term wind speed forecasting one hour in advance. The best wind speed prediction results are obtained when the lookback is 4, where the wind speed has an inflection point and the prediction results have slightly later features than the observation data.

Vehicle target detection and road scene segmentation in real time are essential tasks in the automatic driving system. This paper proposed an end-to-end inference model by improving the YOLOP. The improved inception structure was added as the backbone. CBAM module was used to reduce the loss of space caused by pooling. To obtain the richer feature information, the fusion of dilated convolution and standard convolution module was applied, and mixed mode of group convolution and depth separable convolution was used in different structures to improve the speed of the model and reduce the parameters. The experimental results show the accuracy of detection is slightly improved. Furthermore, the speed of improved model can reach to 54.8FPS, up by 33.1% from the original YOLOP model.

In this paper, the wireless module designed by domestic LoRa module chip is scanned in the near field, and the near-field electromagnetic radiation data and temperature data are obtained when the module is running. At the same time, the color chromaticity diagram and three-dimensional diagram generated by the data parameters visually display the characteristics of the module, which provides electromagnetic compatibility guidance for designing the whole machine using this wireless module in instruments and meters. It increases the stability and anti-interference ability of the product circuit.

At present, the novel coronavirus variant strain has broken out in some cities in China. After prevention and control, the epidemic situation is basically in the final stage of control. To assess the resilience of cities in different regions, this paper established a COVID-19 resilience assessment model based on TOPSIS and fuzzy mathematics. Firstly, according to the concept of urban resilience, three first-level indicators are selected, namely elasticity, resilience and adaptation. Then 17 second-level indicators are selected to quantify the three first-level indicators. Finally, based on the results of the model, possible approaches to improve resilience in COVID-19 prevention and control are proposed.

Through the analysis of the operation of the power market, the dynamic monitoring system of the power market is studied, the basic requirements and basic principles of the on-line monitoring of the power market are put forward, and a dynamic monitoring system with a three-layer structure is established, and the basic architecture, function and technical characteristics of the system are introduced in detail.

Visible light communication (VLC) has become a research hotpot in the Industrial Internet of Things (IIoT). In VLCIIoT systems, the overlapping areas between adjacent transmitters cause severe signal interference. The random movement of mobile terminals in these areas seriously affect the robustness of system communication. In this paper, a cubic receiver with optimal rotation angle that adapts to user’s movement is proposed to solve the problem. Firstly, there are adaptive channel estimation and signal recovery method for rotating cubic receiver to achieve signal receiving and combining in both non-overlapping and overlapping areas. Secondly, we maximize the channel capacity to optimize the rotation angle for further improving communication performance and channel gain. Results show that rotating cubic receiver can better support user mobility in all directions and system performance is enhanced by optimizing rotation angles. Especially, the bit-error-rate (BER) performance gains more in non-overlapping areas than overlapping areas under optimal rotation angles.

Spectroscopy is an important method in the field of nondestructive testing of fruits, and spectral pretreatment and feature screening are the difficulties and keys, which have a greater impact on the accuracy of the model, so this paper aims to use the lightweight neural network (MoblieNetV1) combined with near-infrared spectroscopy to improve the accuracy and stability of the model, reduce its computational amount, and verify its feasibility and effect. First, the fruit dataset is classified and labeled, and apples, pears, citrus, and peaches are labeled as 1, 2, 3, and 4, respectively, and then the sample spectra are put into model training in batches. MoblieNetV1 extracts features through the bottom layer, and then further extracts features at a deeper level, and finally obtains the classification of the target. Experimental results show that the moblieNetV1 used has achieved a high recognition accuracy, which is more suitable for use in the field of spectral detection than other neural networks, and avoids the cumbersome process of manually extracting features.

Digital equipment image retrieval is a typical fine-grained problem, In order to solve the fine-grained problem, necessary data enhancement methods will be essential. Propose a data enhancement method of deep learning in digital device image retrieval. Using the data preprocessing method of artificial optimal data enhancement strategy, the enhanced data set is generated through image enhancement(ChannelShuffle, ColorJitter), geometric transformation, noise interference, and the feature distortion in the process of data augmentation is improved. It can not only effectively solve the problem of large amount of manual labeling samples in deep learning, complete data expansion, but also help to improve the generalization ability of the retrieval model. The experimental results show that the CNN model trained by data augmentation method improves the mAP accuracy of digital equipment image retrieval by an average of 10.4.

Paper cultural relics are one of the most important cultural carriers of ancient human civilization and precious material materials. Due to age and human factors, a large amount of paper cultural heritage is being corroded by foxing, which seriously affects the safety of paper cultural heritage. At present, the detection and control of insect and mould diseases on paper cultural relics is still a recognized problem in the world. In order to solve the practical problem of foxing detection on paper cultural relics, this paper proposes a paper cultural relic foxing based on one-dimensional convolutional neural network based on the characteristics of nondestructive detection of hyperspectral imaging, aiming at the spectral characteristics difference between the foxing part and the healthy part. The spot detection method realizes the non-destructive detection of foxing on paper cultural relics. The experimental results show that the classification accuracy of the model in this paper is 99.17% for the foxing area and healthy area of paper cultural relics. Compared with traditional methods such as K nearest neighbor method and support vector machine, the convolutional neural network model can fully extract the deep spectral features of the foxing region, and can effectively avoid overfitting. The method of combining deep learning theory with hyperspectral provides a methodological basis for the protection of paper cultural relics.

Speech recognition system is a pattern recognition system in essence, including feature extraction, pattern matching, reference pattern library and other three basic units.This paper implements a speech recognition system based on DNNHMM structure with Kaldi toolkit. For multiple pronunciators, it can have synchronous acquisition of the tongue body morphology, lipping, pronunciation and other multiple physiological signals of Tibetan Lhasa dialect and process the data for feature extraction, model training, voice recognition, etc. The result shows that recognition results may vary according to different pronunciator models or articulation parts and the pronunciators of different articulation parts have low dependent recognition error rate but high independent recognition error rate. It also shows that ultrasonic tongue imaging and lip video image sequence includes pronunciator’s acoustic feature information, which provide great significance for the study of voice recognition, particularly the speech synthesis and silent speech interface.

Industrial wastewater classification is the basic work of water pollution prevention and control and water resources management, water chemical oxygen demand is the core indicator to measure the quality of water, compared with domestic sewage detection, the research on industrial wastewater classification is relatively lagging behind. Aiming at the shortcomings of COD classification model of industrial wastewater using convolutional neural network or long shortterm memory network alone, this paper constructs a hybrid model based on CNN and LSTM. The COD data of industrial wastewater were measured by ultraviolet-visible spectroscopy, abstract features were extracted from the data using CNN, and finally entered into LSTM, and the results of classification of industrial wastewater according to COD concentration were obtained, and the classification accuracy of the model reached 98.66%, which was 3.36% and 4.7% higher than that of CNN and LSTM alone.

In order to effectively solve the problems of inefficiency and single form in the generation process of communication solutions, a knowledge graph-based intelligent recommendation method for communication solutions is proposed. By constructing a knowledge map of solutions and mapping the entities and relationships in the knowledge map to a lowdimensional and dense vector space based on the TransH model, the solution similarity is calculated; the user-rating matrix is generated based on the collaborative filtering recommendation algorithm, and the solution similarity is calculated based on the matrix; the solution similarity based on the knowledge map and the user behaviour is fused to calculate the comprehensive similarity, and this is used to generate a recommendation list to overcome the problem of inefficient and unidimensional communication solutions. The recommendation list is generated based on the combined similarity, completing the intelligent recommendation process for communication solutions with significantly higher efficiency compared to the manual calculations.

In this paper, we propose an automated method to colorize old black and white videos. The model consists of several trainable components, containing a coloring module for video frame coloring and a refinement module for spatiotemporal color refinement. Video coloring is a multimodal problem. We satisfy the need for diversity in the coloring results of video frames by using a video frame coloring network to generate diverse coloring results. Then we use the color refinement module to enhance the coloring effect of the initial coloring results. We also introduce the optical flow method to constrain the consistency of coloring between video frames. By predicting the creation of multiple different coloring results, the user has control over these coloring results. This is essential in practical engineering applications. Experimental results show that this method has better result performance than existing video coloring methods.

Bridge deflection is an important indicator to evaluate bridge safety, and accurate deflection prediction is an important way to ensure the healthy operation of bridges. At present, the methods applied to bridge deflection prediction are mainly long-term short-term memory networks (LSTM), but for bridge deflection data, the prediction effect of LSTM in the case of large mutation data often does not meet the ideal standard. In this paper, this paper proposes an LSTM+GAN hybrid network based on wavelet transform, which decomposes the deflection of the bridge into flat data and live load data by the wavelet transform, uses LSTM to predict the flat data, and the GAN network generates the predicted live load data, and finally sums the predicted values of LSTM and GAN to obtain the final predicted value. Experiments show that the proposed method has a better prediction accuracy improvement than LSTM.

At present, the image stylization method is commonly based on the artistic style transfer method, so the generated results usually have the following problems. Comparing the result of the generated image with the content image, there are obvious content distortion and deformation, accompanied by loss of details, distorted light-dark relationship, and unbalanced pixel distribution. In solve above problems, this paper suggests a photorealistic image style based on attention mechanism and histogram loss. The method based on the channel attention mechanism proposed in this paper can preserve the semantic information of content image fully. At the same time, the proposed histogram loss module can better measure the high-level perceptual and semantic differences among images and optimize the high-dimensional feature distribution of images. In addition, this paper also proposes an evaluation method based on the color gamut space histogram to analyze the image generation results more effectively. Experimental results show that, compared with other mainstream methods, the proposed network model can achieve more realistic style transfer while maintaining the details of image content.

The ICAO document clearly states the need for a fuel-saving effectiveness review and evaluation of the continuous and step-by-step flight modes of the climb and descent phases. In order to identify the flight mode efficiently, a flight mode recognition method based on DE-RFE-RF is proposed according to the movement characteristics and flight characteristics of the vertical section of the track. Firstly, the random forest recursive feature elimination (RFE-RF) algorithm was used to search for optimal feature combination packages and classifier training simultaneously. Then, differential evolution (DE) algorithm was used to automatically optimize RFE-RF parameter settings. The method guarantees both the accuracy of feature combination selection and recognition results. Experimental results showed that the method can realize the automatic recognition of climbing/descending mode.

Aiming at the disadvantages of traditional manual identification method of chip welding cavity X-ray inspection, such as high technical threshold, strong subjectivity of inspection results and long inspection time, an automatic identification method based on digital image processing is proposed in this paper. Through histogram equalization, denoising, image enhancement and threshold segmentation on X-ray images containing welding cavity defects, welding cavity defects can be identified quickly and accurately, and the welding cavity rate of products can be calculated, which improves the efficiency and accuracy of inspection test.

In recent years, with the development of optically pumped magnetometers (OPMs), researchers have increasingly begun to construct on-scalp magnetoencephalography (MEG) systems based on OPMs. In the context of the increasing use of OPM-MEG, general-purpose rigid helmets represent the first choice for sensor placement. However, such helmets often cannot meet the requirements for the head fit; thus, depth adjustments are often required in actual use. Accurately acquiring the sensor depths and complete co-registration to magnetic resonance imaging (MRI) are critical for source localization. To date, no study has focused on the co-registration of OPM-MEG and MRI involving depth. In this paper, an OPM-MEG and MRI co-registration method (including an acquisition of sensor depth) is proposed. The sensor depth can be quickly acquired. Moreover, an initial co-registration problem of the helmet in the traditional iterative closest point (ICP) co-registration process is solved by using reflective markers.

The present 2-D direction finding methods for coexisted independent and coherent signals are mainly adopting array containing large number of array sensors, which have great computational complexity and bad utilization rate of array aperture. In this article, a 2-D DOA estimation method is presented based on the L-shape array. The estimation of the independent and coherent signals is conducted separately. the DOAs of the uncorrelated signal are estimated firstly, and the information of the coherent signal can be eliminated by exploiting the property of the coherent signals. Besides, utilizing the Toeplitz property of the uncorrelated data covariance matrix, we can obtain the data matrix containing the information of the coherent signals only. Sufficient theoretical analysis and simulation results are conducted, which show that the presented method has smaller computational complexity, higher array aperture and better performance comparing with the present methods.

Lung cancer has become a major threat to human life and health due to its characteristics of difficulty in early diagnosis and high mortality. Therefore, to develop a reliable method to predict lung cancer is very important. This study provides a new prediction method based on artificial intelligence technology. Firstly, a data set of patients' physical health status was collected, and the correlation between various variables and lung cancer was analyzed using random forest model. Then, the data sets were brought into some common data sets for training and the accuracy of these models was obtained. Then, the neural network technology is used to preliminarily process the data into feature vectors, and then the feature vectors are input into the support vector machine model, and the ANN-SVM hybrid model is established and trained. The results of this model are optimized by adjusting various model parameters and compared with the results of other common prediction models. Finally, a lung cancer prediction APP is made by using this model. Random forest results show that gender, alcohol consuming, and peer pressure are the main factors leading to lung cancer. The prediction accuracy of Ann-SVM model is up to 95.1613% shown by the cross-validation, which is higher than 2.8780%-8.5446% compared with other common prediction models, indicating that the model and prediction APP have practical value.

With the goal of carbon reduction and elimination in various countries, how to manage and achieve carbon footprint quantification and greenhouse gas emission accounting has become one of the key research tasks in low carbon manufacturing of electromechanical products. In order to provide users with a deeper understanding of the carbon emissions of purchased and used electromechanical products, we propose an idea of constructing a carbon footprint label (CFL) for the life cycle of electromechanical products using a hierarchical structure tree of electromechanical products. After the bibliometric analysis of the industrial-level knowledge graph research contents in the past decade, the construction method of knowledge graph in the field of electromechanical products is clarified. Based on the constructed CFL ontology model of mechanical and electrical products, a CFL representation model of electromechanical products based on knowledge graph is established. The CFL records the carbon emission data in the life cycle of mechanical and electrical products, making the CFL a new attribute of products, and providing targeted CFL representation of electromechanical products for low carbon designers.

Image segmentation is an important branch in the field of artificial intelligence and an important part of image understanding in machine vision technology. Based on computer vision, this paper makes a classification study on image segmentation technology, including the basic method of image segmentation, image edge detection algorithm and the application of connected area analysis, area growth, watershed and other algorithmic ideas to achieve image segmentation. In the basic method of image segmentation, the adaptive segmentation method and otsu method processing effect and processing range are better, in image edge extraction, the canny operator processing effect is obviously higher than other operators.

Model consistency is an important guarantee to ensure good interoperability and correct simulation results of large-scale parallel and distributed simulation applications (DSA). There are many factors affecting the consistency of the model, including internal, external, subjective and objective ones; Model inconsistency will eventually lead to poor interoperability performances and then result in the deviation from reality in DSA; Taking the typical distributed simulation application named Unmanned Aerial Vehicle (UAV) penetration as an example, the paper expounds the problems of model consistency in detail, and puts forward some consistency control strategies of simulation models in DSA from two aspects of technology and management, so as to lay a foundation for improving the accuracy of DSA.

Aiming at the problems of strong nonlinearity, high dynamics and strong disturbance during the return process of reusable launch vehicles, this paper proposes a model prediction control method based on time-varying prediction model and disturbance observer. Firstly, the principles of predictive function control and disturbance observer are introduced. Then the time-varying prediction model is used to predict the state of the system to overcome the problem of high dynamics of the system; the disturbance observer is introduced to overcome the problem of strong disturbance of the system, the deviation between the real system and the prediction model is observed, and the compensation is made in real time at the input end. By solving the quadratic optimization problem, an explicit solution to the control input is obtained. Simulation results show that the control quality of the control method is very good and the method has strong robustness.

Observing the fundus vessels in a noninvasive manner can play an auxiliary role in the clinical diagnosis of a variety of diseases. It is inevitable that misdiagnosis will occur by means of manual analysis. Aiming at the problems that the accuracy of computer-aided segmentation of fundus vessels is not high and the use of convolution layer sampling will cause the loss of global context information, an improved fundus vessel segmentation network based on TransUNet is proposed. Firstly, the transformer layer of TransUNet can effectively obtain the connection of global context information. Secondly, the convolution unit sampled under TransUNet is replaced by residual space attention block (RSAB) to improve the attention to the vascular region and prevent the information from disappearing and over fitting in the process of passing through the network. Finally, mini-UNet and TransUNet are cascaded to make the model correct the segmentation results and improve the accuracy of vascular segmentation. The experimental results show that the improved algorithm improves the accuracy of drive dataset by 0.23% compared with UNet network. Each evaluation index is better than most of the compared segmentation models.

UV-Vis absorption spectroscopy is often used to measure COD values. At present, certain points or wavelengths in the spectrum are often obtained for modeling, resulting in deviations in predicting COD values. Aiming at this problem, this study adopts a data fusion method of multi-source spectroscopy (Raman spectroscopy and UV-Vis absorption spectroscopy) to improve the accuracy of predicted COD values. In the experiment, two groups of single spectrum and multi-source spectrum were used for comparison. Potassium hydrogen phthalate was used to simulate the COD value of water quality. The range of COD was 5-100mg/L, and the concentration gradient was 5mg/L. The data fusion of multisource spectra in the experiment uses the normalization method to remove the dimension and then cascades into a new data set matrix. The eigenvalues are screened out through the importance analysis of projection variables, and finally the processed matrix is combined with the partial least squares regression method. The experiment was modeled with 19 training groups, and 1 group was used to verify whether the model was good or bad. The experimental results use the two dimensions of root mean square error and coefficient of determination to evaluate the quality of the model. UV-Vis absorption spectrum R2=0.9993, RMSECV=1.8525; data-level fusion R2=0.9668, RMSECV=1.6113; feature-level fusion R2=0.9766, RMSECV=1.2957. The experimental results show that multi-source spectral fusion has a positive impact on improving the detection of water quality parameters, and the fusion model has stronger predictive ability.

A feature attention based multi-stage network for image deblurring is proposed. The feature attention module is introduced into the model. This module is composed of channel attention and pixel attention mechanism. More attention is focused on the blurred pixels and important channel information, solving the problem of uneven blurred distribution in images effectively. We also introduce the atrous residual block and the context module between the encoder and the decoder. The atrous convolution are combined with the residual, and the context module adopts the multi-layer atrous convolution, which effectively increase the receptive field of the network and better capture the multi-scale contextual information. Experiments were conducted on the public dataset GoPro to evaluate the performance of our method. The results show that the PSNR of the proposed model reaches 30.51, and the processing speed reaches 0.035s, which outperform that of the most current deblurring methods.

Vibration is a widespread phenomenon in nature, and vibration detection and analysis of objects can improve the safety of precision equipment. However, real-time performance in the existing laser Doppler interferometry data analysis and processing system remains an ongoing challenge. This paper implements a laser Doppler vibration software based on overlapping frames, according to the data processing method of real-time spectrum analyzer in RF signal measurement and parallel programming method. Under different overlap rates, this paper compares the difference in refresh rate caused by three data processing methods. Experimental results show that the higher the overlap rate, the better the real-time performance of the software. The system can meet the design index of refresh rate(13Hz), has good display fluency and real-time display, and has certain reference significance for related research.

In modern computer science, tasks such as optical character recognition are of great importance. Yet, for artificial neural networks it’s never easy to distinguish what a scrawled image indicates.In December 2016, Google post a drawing game which would randomly pick up a word and ask the player to draw an outline within 20 seconds. An AI was used to judge if your work looks similar to what you were asked to draw. With more and more people playing this game, Google has collected over 1 billion drawings.Freehand sketching is a dynamic process where points are sequentially sampled and grouped as strokes for sketch acquisition on electronic devices. While most existing models simply convert the sketch vectors into binary images and pay little attention to the time information, we organize training data and adapt traditional image recognition networks in a novel way. In this report, we will introduce our solution to this doodle recognition problem, including data pre-processing and image recognition. We would also explain experiment process and the main challenges we met in detail.

In recent years, the explosively growing scientific and technological research achievements of China have led to challenges for the review expert selection. A review expert recommendation model based on a knowledge graph (KG) was proposed. The model took the results of knowledge graph embedding (KGE) and expert-term interaction matrix as inputs. After that, the embedding vectors of TransD and graph attention networks (GAT) were used to recommend review experts based on vector similarity. The results show that, compared with other graph embedding models, the proposed model could better capture the characteristics of the research domain of experts and is more conducive to distinguishing the research directions of different experts, so as to recommend relevant experts for projects or papers.

With the development of fluorescence super-resolution microscopic imaging technology, super-resolution optical microscopic imaging technology, which breaks through optical diffraction, has rapidly developed into one of the powerful research technologies in the field of life sciences due to its nanometer-level spatial resolution and low damage to samples. Since 2016, deep learning-based methods have been proposed for reconstructing super-resolution microscopy images. In this paper, we make two improvements based on the Deep-STORM network, first replacing the original full convolution with the dilated convolution. Then, the skip connection was added on top of the dilated convolution to reconstruct the microtubule dataset of the EPFL·SMLM challenge site. The experimental results show that NMSEs are reduced by 36.84% and 37.99% respectively on the two improved methods compared to CEL0. Compared to FALCON, NMSEs are reduced by 24.69% and 25.84% respectively. Compared to Deep-STORM, the NMSEs are reduced by 2.2% and 3.35% respectively. Using this method, super-resolution microscopic images at the nanometer-level can be reconstructed quickly and accurately, thus meeting the requirement to observe subcellular microtubule structures beyond the diffraction limit.

At present, the deblurring algorithms based on deep learning have achieved good performance, but the current deblurring public data sets are obtained in the scene with good light conditions In the actual scene, when the photosensitive device collects images, it is often disturbed by the external environment, such as night, overcast and other environments, which will cause the quality of the collected images to become worse, and will bring greater challenges to the deblurring task This paper proposes an end-to-end depth network for video deblurring based on dark light enhancement Firstly, this paper proposes an end-to-end video deblurring network based on STFAN; Secondly, the dark light enhancement module is proposed, which can improve the dark light image quality and further enhance the deblurring effect; Finally, we introduce our own dark light scene data set, and train and verify the proposed method on the dark light data set Experimental results show that the proposed algorithm has better deblurring effect in dark scenes than other deblurring algorithms High performance dark light deblurring algorithm will play a vital role in the fields of photography, photography, medicine and criminal investigation.

A knowledge graph, consisting of concepts, entities, and their relationships, is the most effective way to handle vast amounts of information. This study presents a relationship extraction approach for designing and constructing a graph structure, given that current domain knowledge graphs are unable to identify links between various kinds of data structures utilizing particular building methods. Data from the Network Encyclopedia's categorization system and web page classification labels, both structured and semi-structured, provide the basis for hyponymy. Non-superior connections are covered as well. The convolution residual network that is based on cross-entropy loss function has also been enhanced as a consequence of the suggested solution. Using a comparable existing item as a reference, we evaluate the design to see whether it performs as predicted.

Knowledge graph is a kind of storage structure which conforms to human cognition form and can accurately describe the relationship information between things. In this paper, a visualization and retrieval system of knowledge graph is built starting with the original Textual corpus. This paper first analyzes the requirements of the visualization and retrieval system. Then in the system implementation stage, in order to realize the visualization and retrieval system, we divide the system into three modules (Data Module, Visualization Module and Query Module) and implement them, respectively. For data module, we based on the entity relation joint extraction model processing text data into relational data (triples). For visualization module, we visualized the triples into node-link diagram based on the force-directed algorithm. Apart from this, we realized some interactive functions based on the D3.js visual framework and HTML technology.

Multi-hop Knowledge Graph Question Answering (KGQA) requires reasoning about multi-hop inference relations between topic entities and answers on the knowledge graph(KG) and returning correct answers. The difficulty of obtaining the implied inference relations in multi-hop questions described in natural language and the sparse knowledge graph bring challenges to the multi-hop KGQA. In this paper, we propose an embedded knowledge graph multi-hop Q&A model based on relational paths, which exploits the relational chains in the knowledge graph and the semantic similarity of multiple questions to improve the accuracy of multi-hop KGQA task. The comparative experiments prove that our method significantly outperforms the state-of-the-art counterparts. Comprehensive ablation experiments also validate the effectiveness of our approach on multi-hop KGQA tasks.

In the traditional ultra-wideband (UWB) positioning technology, the influence of the ranging error on the positioning model and the influence of the motion parameters of the moving object on the positioning accuracy are not considered, especially for high-speed moving objects, it is difficult to achieve real-time, stable and exact positioning. Therefore this paper proposes a ranging calibration algorithm to minimize the distance error. The algorithm constructs a cost function based on the distance error, and combines the acceleration gradient descent (NAG) method to screen the value of the large distance error in the cost function. For the positioning of high-speed moving objects, the speed parameters of the moving objects are introduced, the vector equations about the anchor and the moving objects are established, and the positioning solution model of the moving objects is obtained, so as to realize the precise positioning of the high-speed moving objects.

In view of the complexity and long time required of the current Mongolian pattern handwritten sketch, an improved Canny operator is proposed to extract the vectorized edge of the Mongolian pattern, use the adaptive median filter to denoise the image, calculate the gradient size and direction combined with the Sobel template, and detect and connect the edge based on the adaptive threshold combined with iteration and Otsu (Otsu method). The experimental results show that the improved Canny operator has an average improvement of 25.7% and 13.6% over the traditional edge detection operator in SSIM (structural similarity) and PSNR (peak signal-to-noise ratio), respectively.

In this study, a novel mechanism using for the satellite three-axis attitude adjustment is proposed. The aim of this paper is to study the magnetic field of the spherical magnetohydrodynamic attitude controller. According to the results of numerical simulation, the optimal design parameter combination of the magnetic field design of the spherical magnetohydrodynamic attitude controller is obtained. This work can provide a reference for the subsequent flow field optimal design and experiment.

Efficient and accurate large-scene point cloud reconstruction and semantic segmentation are key issues in 3D scene understanding and environmental intelligence perception. this paper proposes a point cloud reconstruction and segmentation method based on UAV aerial image, which use the improved MVSNet to reconstruct the UAV aerial image, and then the generated point cloud is segmented by using the RandLA-Net model. The experimental results show that compared with COLMAP and MVSNet, the accuracy of the proposed method is improved by 12.75% and 11.86% respectively, and completeness is improved by 37.19% and 20.87% respectively on DTU data sets. In addition, the proposed method can also effectively segment the largescale point cloud. This method can meet the needs of the smart city and provides new technical references for 3D digital city modeling.

In recent years, the rapid development of deep learning has greatly promoted the progress of computer vision and image graphics. Many new image style transfer algorithms based on deep learning have made a huge leap in image generation quality and operation speed. Online animation is the expression of animation and comics, including animation books, films, television, audio-visual products, stage plays and the development of animation products based on modern information and communication technology. However, most of the cartoon faces are similar, and the facial features are single, but if a more detailed description is carried out, the production cost is high. In this paper, a cycle-consistent generative adversarial network is used to convert the real face image to the comic style face image. The network does not require paired matching data, so that the network can use a wide range of data and is less difficult to obtain. Through experiments, it is found that the improved CycleGAN model has a good effect on generating cartoon face images, which can save a lot of cost for cartoon makers, and the generated cartoon faces have the advantages of being more realistic and detailed.

A news recommendation system is crucial to help users locate interesting news and relieve information overload. How to accurately model news text is the key to a news recommendation system. Traditional news recommendation systems use news encoders that directly model the text of all historical reading news. However, pre-trained language model (PLM) trained on large-scale corpora often have more robust reading comprehension capabilities and can extract more profound implicit expressions of news texts. Users have their preferences when reading different categories of news. Using a large number of training samples to learn such preferences will consume many computing resources. Adding category awareness to the news recommendation system can efficiently and accurately model users and build a personalized News recommendation system. Therefore, this paper uses Electra as a news encoder to capture a more robust contextual implicit representation of news texts. At the same time, additional weights are used to model different news categories. On this basis, users' behavior logs are added to build a news recommendation model. The cosine similarity is used as a loss function to obtain the recommendation probability of candidate news. Experiments show that the AUC of this method is 2.37% higher than the current state-of-the-art methods on the Mind-small datasets.

For the existing 3D point cloud model classification method PointNet only extracts global feature information, but for the problem of insufficient local feature information extraction, a network that can take into account the combination of global features and local features is proposed, so as to realize 3D point Classification of cloud models. This paper proposes a classification network based on local location information. First, the point cloud model is sampled in multiple groups to obtain the center point of each group of samples, and then the radius is set to obtain its spherical neighborhood, and the points in the neighborhood and their center points are used as local location features. In information fusion, the acquired local information features are used as high-dimensional features of the neighborhood space through feature mapping. By analogy, multiple sets of sampled local features are integrated to realize the classification of the point cloud model. The experimental results show that the classification network can achieve a classification accuracy of 92.5% on the ModelNet40 data set, which is higher than the current mainstream algorithm.

This paper is about the study of computer graphic based on appearance modelling. and. subsurface scattering mainly discussing that the purpose of computer graphics is to make the picture more beautiful and nicer because of the application of these two technologies. It describes in detail the typical physical process of light, the tracking of object features, and the recognition and transformation of this feature by computer from the two directions of appearance modeling and subsurface scattering. It also describes the outstanding theoretical achievements of some predecessors on computer graphics and the successful application of these technologies in the film and animation industry. At present, the development of any science has its own outstanding advanced aspects, as well as its own limitations and problems. Computer graphics has experienced a long disciplinary bottleneck period. Now, the in-depth development of computer digitization, informatics, physical optics and other sciences has brought infinite vitality and broad prospects to this science, let it better reproduce nature and draw more infectious and realistic beautiful pictures to service the growing demand for this technology. The appearance modeling techniques are constantly improving and people are also learning more and more about the subsurface scattering knowledge to better serve the computer graphics.

Pseudo inverse method is widely used in dynamic positioning engineering. This method is simple, stable, efficient and reliable. Its essence is to use Lagrange function method to solve the optimal value. In the process of matrix operation, the problem of pseudo inverse solution is involved, so it is called pseudo inverse method. In this paper, under the simulation environment, the effectiveness of the pseudo inverse method in thrust distribution is theoretically verified by the different control commands given by the controller.

Establishing an accurate and effective quality evaluation system is very important in the practical application and development of image fusion technology in various fields. In the field of image fusion, there is no systematic evaluation system. In order to solve these problems, this paper uses a number of different objective evaluation indicators to design a set of fused image comprehensive evaluation system. The quality of fused results can not only comprehensively evaluate the advantages of the algorithm but also be used as an information feedback mechanism to optimize and improve the image fused algorithm. Through the evaluation and analysis of six image fusion methods, the rationality of the evaluation system is verified.

To improve the accuracy of infrared ship detection and more effective background suppression, we propose an infrared ship detection algorithm based on latent low-rank decomposition and salient feature fusion. First, to avoid the influence of the number of superpixel segmentation on the integrity of the GR saliency detection, multiple salient maps with different numbers of superpixels are taken large, and a single foreground coarse salient map of the ship target is obtained using a high-level central prior and graph regularization. Secondly, to suppress the noise influence and refine the foreground coarse saliency map, sparse salient features of the ship target image are extracted using a method based on latent low-rank decomposition. Then, the sparse and null salient features are fused with the feature values. Finally, the accurate ship target is segmented based on the OTSU algorithm. The subjective comparison experiments and quantitative analysis show that the infrared ship segmentation algorithm proposed in this paper has better performance in terms of accuracy and visual effect.

Snoring is a typical symptom of obstructive sleep apnea hypopnea syndrome (OSAHS), and it is a widespread sleep disorder. Accurate detection of snoring can help to screen and diagnose OSAHS. However, the current voice recognition methods based on deep learning can not achieve satisfactory results. To accurately identify snoring, this paper proposes an automatic snoring detection method based on a convolutional neural network (CNN) and constructs a snore dataset. For each sound segment in the snoring dataset, we calculated the time-domain waveform, spectrogram, and Melspectrogram. The proposed method classifies snoring and non-snoring sound segment images through a new convolutional neural network MBAM-ResNet to accurately identify snoring. Experimental results show that spectrogram can better reflect the difference between snoring and non-snoring images and the accuracy of the proposed network for snoring on the spectrogram is 91.11%.

Lung cancer is the leading cause of cancer-related death worldwide. Early detection and diagnosis of pulmonary nodules are crucial to improve the patients’ relative survival rate. Although existing deep learning-based methods have achieved good results in distinguishing benign and malignant pulmonary nodules, most works mainly focus on designing new models to obtain deeper features, rather than on effective feature representations of the discriminative pulmonary nodules. In the present work, a benign and malignant pulmonary nodule classification method based on Swin Transformer model is developed. The multi-head self-attention hierarchical architecture of Swin Transformer capable of flexible modeling and linear computational complexity, enables the deep models simultaneously extract the local and global features, paying more attention on the key areas with a large amount of information in CT images, while suppressing other irrelevant information. The experimental results on LIDC-IDRI dataset demonstrate that the presented model is an effective and competitive method for the classification task of benign and malignant pulmonary nodules, compared with recent state-of-the-art approaches.

Conventional public transportation is an important part of public transportation, and it has always been the focus of urban transportation research to excavate the characteristics of public transportation and analyze residents' travel patterns. In 2020, the new crown epidemic broke out. The outbreak and continuation of the epidemic have caused shocks and challenges to conventional public transportation, and the characteristics of conventional public transportation have developed significantly. Taking Guangzhou as an example, this paper conducts bus IC card mining based on multisource data fusion, and conducts research on the characteristics of changes in Guangzhou's regular bus travel rules under the influence of the new crown epidemic. Research shows that under the continuous influence of the epidemic, the scale of bus trips has dropped significantly, the attraction of conventional buses to commuter passengers has been weakened, special groups are important users of public transport, and the ride code has become the most important payment method.

This research uses 61,780 patents related to electric vehicles in the global patent database to analyze the full text of electric vehicle patents in two major technology areas (1. transport, 2. electrical machinery, apparatus, energy) and other technology areas using the STM model for thematic analysis, and combines them with patent values. We look for the thematic bias of high-value patents in each technology area, so as to determine the most valuable development direction in the field of electric vehicles. In our analysis, we found that innovations in core technologies bring the most value, including energy conversion efficiency and battery module technology performance, while sustainability is another focus of core technologies and the green effect is gaining importance. This study compares the full-text themes of all technology patents since the concept of electric vehicles was first created, and considers the weighting of words of different core degrees to provide a baseline for subsequent research.

How to improve the employment competitiveness, employment rate and employment quality of graduates is one of the essential concern of college educators. To achieve this goal, college educators must provide in-campus students with courses and training which are consistent with the job skills required by the recruitment market according to the recent industry trends, employers' recruitment demand and job competency requirements. As college educators, how to obtain thousands of enterprise recruitment information in the industry in an efficient and low-cost way, and then dig out the skills that students need to master most? One of the effective ways to achieve this goal is to collect, sort out, mine and analyze big data based on python[1]. Firstly, with a Python based crawler program, this research collects the companies, positions, academic requirements, work experience requirements and salary range of massive IT positions from June 2021 to June 2022 in Guangzhou published on zhaopin.com. Then matplotlib is used to analyze the total market demand, average salary, salary changing curve over experience, and the minimum academic requirements. Finally, some suggestions for college educators are supplied so as to provide better college IT training.

Sequence-to-sequence models provide a feasible new approach for generative text summarization, but these models are not able to accurately reproduce factual details and subject information. To address the problem of unconstrained and uncontrollable content generation of generative text summarization models, this paper proposes a generative summarization method KGIT that uses Transformer as a skeleton and incorporates both BERT pre-training model and keyword information. The model uses a comprehensive keyword extraction algorithm, uses two results extracted by LSTM and TextRank as vocabularies respectively, and uses pointers keywords are selected and the extracted keywords are used as the guiding information to generate the summary based on the guiding information. KGIT model can associate the source text and keywords to avoid generating a summary of irrelevant topics. The ROUGE value is used as the evaluation criterion for text summaries, and the summaries generated by the KGIT model can contain more key information and are more accurate and readable when compared with the mainstream summary generation models on the NLPCC2017 Chinese news summary dataset.

In practice, it is very important to determine the requirements items of the proposed software system from the requirements document. However, due to the problems of ambiguity, redundancy, ambiguity, and difficulty in traceability of changes in the requirements documents described in natural language, the estimation results are subject to a certain degree, and the process is labor-intensive and cost-intensive. In this paper, we propose a new method to automatically extract requirement entries from requirement text by leveraging a set of natural language processing techniques and machine learning models. Our method is inspired by imitating the process of expert extraction of itemized requirements, which usually consists of three main processes: locating requirement locations and boundaries, building models, and extracting fine-grained requirement semantics. We performed evaluations in the field of military arguments, and the results showed that our model was nearly 80 percent accurate. Our approach can provide sound advice to help industry practitioners extract requirements items faster and easier.

For the limitation of existing quantitative indexes of Poincaré plot, the included angle cosine measure is proposed in this study. The novel index has the ability to assess the degree of the change of three consecutive RR intervals and remedies the deficiency of the Poincaré plot analysis partly. Finally, PhysioNet/PhysioBank is applied in this study and the experiment results demonstrate the effectiveness of the included angle cosine measure

The millionaire’s problem is the origin of secure multi-party computation. This problem has always been the focus of scholars' research, and it is also the design basis and important part of other secure multi-party computation protocols. An in-depth study of the millionaires’ problem is of great significance to promote the further development of secure multi-party computation. Based on millionaires’ problem and secure vector dominance protocol, we propose an improved protocol and an extended protocol. The improved protocol solves the problem of coarse granularity when comparing elements in the original protocol, and can realize that two parties can compare the relationships of greater than, equal to or less than between their elements by executing the protocol once. The improved protocol and the extended protocol are further applied to solve the problem of finding the cardinality of the intersection of the two-party sets in two cases. In addition, based on the integer division problem, we give a method to privately evaluate the greatest common factor of two integers.

To improve the low estimation accuracy and poor robustness of invisible human joints in 2D human pose estimation in computer vision, a convolutional network based on intra-layer feature fusion is proposed to improve the estimation accuracy. Firstly, the convolutional block based on intra-layer feature fusion is proposed to obtain abundant local features and reduce the influence of local information loss. Secondly, intermediate supervision is introduced to reduce the influence of vanishing gradient in deep neural network. Finally, several proposed convolutional blocks form the neural network in cascade, through continuous Up-Sampling and Down-Sampling operations, convolutional blocks obtain local information from images with different resolutions and have the information fully fused, the local information after fusing can establish the connection between joints and accurately estimate the joints. The proposed neural network achieves first-class results on standard benchmark including the LSP dataset and its extended dataset.

The temporal convolutional network (TCN) model has the characteristics of strong parallelism and stable gradient in time series processing. The structural depth of the model is related to the input length, convolution kernel and dilation factor. In order to further improve the accuracy of prediction, this paper proposes a software aging prediction framework based on TCN model optimized by grey relational analysis. Collecting available memory data as the input of the framework, determine the length of the input nodes of the TCN model through gray correlation analysis, and then conduct training and prediction, and evaluate the efficiency of the model by checking the average error between the predicted output memory and the actual memory. Then change the length of the input chunk to carry out a comparative experiment, which verifies the effectiveness of the grey relational degree analysis.

To address the problems of low cache hit ratio and high routing hops to obtain content in Named Data Networking (NDN), this paper proposes a Cache placement Policy based on Dynamic Popularity (DPCP). When the requested content arrives at the routing node, the latest popularity is calculated by combining the number of requests in current cycle and the popularity of the previous cycle, and the caching threshold is dynamically adjusted according to the occupancy of caching space, only content with popularity exceeding the current threshold is cached, thereby retaining the content with the highest probability of being requested in future time periods. This cache placement policy is also implemented using Programming Protocol-independent Packet Processors (P4) deployed on programmable data plane. Experimental results show that the proposed algorithm can effectively improve the cache hit ratio and reduce the average routing hops compared with the traditional caching policy.

In order to meet the requirements of mass data transmission between various communication systems today, this paper proposes a high-speed LVDS data transmission system based on FPGA, which uses a high-speed serial LVDS bus for communication. The transmission system makes full use of the advantages of LVDS, such as strong anti-interference ability, low power consumption, and fewer external cores. Combined with the characteristics of FPGA field programmability, it can transmit data efficiently in real time. According to the LVDS-related transmission characteristics, the LVDS-related transmission layer is built through FPGA, including the conversion of serial data and parallel data, the synchronization of clock and data, the framing and de-framing of transmission data, etc. detailed design. Finally, the loopback simulation test shows that the system has high transmission rate, stable and reliable operation, simple design and broad application prospects.

Color perception, a kind of visual-motor awareness, is a critical information processing ability to precept the movements of objects. This ability is not only closely related to the daily life of autistic people, but also profoundly influences their social behavior and adaptive behavior. This paper illustrates a serious game to assist children with autism between four and six years old to recognize color and adapt to real life. In the game, human-computer interaction technology is applied to give instructions and feedback to autistic children. The study selects the ABA round teaching methodology and four dimensions to evaluate the behaviors of autistic children, and analysis the score given by experts after observation of autistic children’s immersion in video games. The result of the study showed an encouraging future trend of a serious game in the treatment field for autistic children.

With the accelerated urbanization and rapid socio-economic development, the construction process is also accelerating. Construction workers have more and more influence on the construction, and the effectiveness of construction worker management directly affects the quality and efficiency of the whole project construction. However, the traditional way of managing construction workers has problems such as incomplete information data of workers, difficulty in supervision, weak awareness of workers' safety and frequent safety accidents. Therefore, this paper designs and develops a construction worker management service information system based on IoT technology and effective integration of Building Information Modeling (BIM) and Radio Frequency Identification (RFID) technologies. The system realizes collaborative information sharing, achieves intelligent management of construction workers, improves construction site safety and quality management, and promotes the sustainable and healthy development of the construction industry, which in turn promotes the rapid development of China's economy.

The computer-based interlocking system is a control system to ensure driving safety, with the safety level of STL4, which is a typical safety-critical system. In order to ensure driving safety, before the computer-based interlocking system is applied on site, the manufacturer of the computer-based interlocking system will formulate test cases to test the interlocking logic according to the computer-based interlocking technical specifications and interlocking logic. Due to the huge amount of testing, manual testing cannot be completed in a short time, and there are many repetitive tasks in the testing process, which are suitable for computer completion. In order to improve the test efficiency, this paper intends to develop a test software, and propose a closed-loop control method based on state machine. The control method is used to control the test process, which is of great significance to the test of the computer-based interlocking system.

In this paper, the V-column nodes of the National Snowmobile and Bobsleigh Centre of the Winter Olympic Games are used to analyze the force performance under different parameters such as axis angle, and thickness. The nodal stress variation of different V-column nodes during finite element loading is analyzed, and then the parametric variable analysis is carried out for 2 different parameters of uniaxial tension nodes of crossbeam V-columns. The effect of different parameters on the force performance of the nodes is analyzed to study the force performance of the V-nodes.

In the task of text classification, Chinese short text has the inherent characteristics of short text, sparse text features and multi ambiguity, which makes it difficult to extract and express the features of short text. To solve this problem, a short text classification algorithm based on Topic BERT and BiGRU is proposed. The algorithm integrates BERT and BTM as Topic BERT model as word embedding layer, and extends the semantics of the short text by integrating BTM topic model, so that the integrated semantic vector includes topic features. Then BiGRU is used to extract the depth information feature of vectorized text. The test results on the THUCNews header data set show that compared with the existing model, The F1 value of the comprehensive evaluation index is better than other existing models, which proves the effectiveness of this model in dealing with Chinese short texts.

Training a dialogue generation model requires large-scale, high-quality dialogue data. Data augmentation is a common method to expand a dataset. Some common data augmentation methods such as back-translation and synonym replacement are used to augment dialogue datasets, but these methods are only aimed at word or sentence level changes and do not consider the relationship between dialogues to generate new dialogue data and the meaning of the generated sentences may conflict with the context of the dialogue. To solve this problem, a new data augmentation method is proposed for expanding the task-oriented dialogues. Solve the problem of insufficient training data by looking for a dialogue in the original dataset, then replacing the dialogue after comparing the belief state and dialogue acts. Experimental results on MultiWOZ dataset show that this method can improve the Combined Score by 0.99, 0.24 and 1.5 respectively compared with the data augmentation method at the character-level, word-level and the sentence-level.

Learning to rank is an effective recommendation approach since its introduction around 2010. Famous algorithms such as Bayesian Personalized Ranking and Collaborative Less is More Filtering have left deep impact in both academia and industry. However, most learning to rank approaches focus on improving technical accuracy metrics such as AUC, MRR and NDCG. Other evaluation metrics of recommender systems like fairness have been largely overlooked until in recent years. In this paper, we propose a new learning to rank algorithm named Pareto Pairwise Ranking. We are inspired by the idea of Bayesian Personalized Ranking and power law distribution. We show that our algorithm is competitive with other algorithms when evaluated on technical accuracy metrics. What is more important, in our experiment section we demonstrate that Pareto Pairwise Ranking is the most fair algorithm in comparison with 9 other contemporary algorithms.

Nowadays, the rapid development of big data is related to the development of various industries, and the field of education is also inevitable. It is necessary to research and discover hot issues in education and teaching through big data technology research.This paper uses the web crawler techniques to obtain all the Weibo data released by 11 private undergraduate universities in Shaanxi Province from May 2021 to May 2022, and analyzes the forwarding, commenting and liking of Weibo issued by 11 universities, and studies the peak time period, content preference, form preference and other characteristics of the forwarding, commenting and liking situation. The study found that Weibo users of private undergraduate universities pay more attention to social news, campus life, campus activities and other theme Weibos, and prefer Weibos in the form of hashtags and pictures, while the peak of user behavior in a day appears at 7:00PM-12:59PM. According to the above statistics, universities can update their ideas, innovate the form and content of blog posts, increase the activity of official Weibos of universities, strengthen interactive publicity, and enhance users' sense of belonging.

At present, emergencies occur frequently. How to better adapt to the changes brought by big data in the analysis of network public opinion for emergencies is an important issue of emergency management in the era of big data. In this paper, combining the actual needs and the shortcomings of the existing public opinion pre-warning system, data mining technology, emotion computing technology and artificial intelligence methods are introduced into network public opinion pre-warning system, build a network public opinion pre-warning system model based on Hadoop big data architecture, which includes four layers: public opinion data collection layer, data preprocessing layer, public opinion analysis layer, public opinion pre-warning and decision support layer. The model integrates the important functions of the whole process of network public opinion analysis and pre-warning, in order to provide accurate pre-warning information and scientific auxiliary decision-making scheme for emergency management.

We propose a two-stage label noise learning framework. Classification by deep learning with datasets with noisy labels. Filtering noise samples based on the loss of the warm-up stage is a common method, but it is impossible to judge the optimal warm-up length and thus suffers from memory effects. This paper is inspired by the use of contrastive learning self-supervision as pre-training to replace the warm-up part, and pseudo-labels for subsequent loss calculations, which can avoid the influence of noisy labels on pre-training. The second part is a semi-supervised algorithm that applies data augmentation. The adopted boosting strategy is to use weak boosting for any loss analysis task and strong boosting for gradient descent. Learn a more accurate representation for each image. Through these two steps, the performance of the network can be improved. Experiments show that our framework is more robust to data classification problems with noisy labels, and also works well in datasets with high noise rates. At the same time, since it is a two-step framework, it is also easier to split and apply in combination with other noisy label learning algorithms.

With the development of information age, enterprises' demand for software is increasing sharply, and the problem of software development platform has become the focus of the industry. MDA model driven architecture is an architectural solution to solve the problems of efficiency and consistency in the development process; Microservice is an advanced technology that uses services to build applications, maintain service independence, and reduce mutual interference and dependence. The core of microservice application development platform is a secondary development software framework, and software developers can quickly develop various application systems based on the application development platform. Therefore, a microservice application development platform based on MDA is proposed, which takes metadata as the core, transforms the traditional manual development into the realization of software business functions based on microservice architecture through visual modeling operations, and accumulates and hierarchically precipitates software assets through metadata.

UE4 (Unreal Engine 4), a new digital information technology software, involves film and television production, game production, chat platform, and live broadcast applications. Model engineering files with maps in MAYA, 3dMax, and ZBrush can be imported into UE4, where the materials and special effects of this software can be distinguished. In addition, plug-ins can be imported and designed by blueprints, level editing, node editing, and C++ programming language editing, leading to the increasing use of UE4 in various industries. It is not only animation design or real-time demonstration and experience of game operation, but also simple operation and improved efficiency play an important role in promoting the whole industry. In this paper, we use principal component analysis (PCA), analytic hierarchy process (AHP) and neural network model to study image display through machine learning and other algorithms. By preprocessing the data to make it meet the conditions of neural network operation in machine learning, we will use neural network architecture for analysis and make a horizontal comparison with the live broadcast feedback of other E-sports platforms. The neural network model focuses on the image effect, production cost and playback feedback of the more intuitive sequencer scene in the Unreal Engine4 virtual live broadcast and the traditional film and television live broadcast, and obtains the specific advantages and disadvantages through algorithm comparison.

Aiming at the difficulty of verification and test caused by the complex structure and timing logic of multi-source data fusion flight control software algorithm, this paper puts forward the digital simulation technology of flight control software, which simulates the reference clock of flight control software, the external link equipment of flight control software, and the functions of sensor data, actuator data, data communication interface, hardware driver, thread scheduling and so on. The simulation software technology enables the self-development stage of flight control software to get rid of the constraints of hardware simulator debugging conditions, quickly and conveniently verify the correctness of the timing logic and algorithm implementation of flight control software, greatly shorten the cycle of software development and unit test, save the development cost and improve the efficiency and quality of flight control software development.

Training with more data has always been an effective and convenient way of improving performance in the deep learning era. Therefore, in practical engineering projects, data is generally collected as much as possible. However, as data volume and class size increase, maintaining a balanced dataset across a number of classes is challenging because data are always long-tailed in practical engineering projects. As a consequence, the key of the practical deep learning application is to design solutions to solve the long tail distribution problem according to the characteristics of specific projects. At present, many existing methods are often not designed according to actual engineering problems, and there is a certain gap with the actual application scenarios. In this paper, we will design a strategy based on the actual application scenarios and solve the imbalance distribution problem in practical engineering projects by adjusting the classifier according to the data.

In today's society, consumers can no longer meet the traditional travel services, for a variety of travel needs are also more needed, the use of Spring Boot + SSM technology constitutes an integrated mobile travel platform. Through Spring's mature MVC architecture model, we continue the basic configuration of the Spring Boot framework, realize the system architecture of multi-layer modules,and on the basis of integrated transportation, it can be known that the service plays its own resource capabilities.will also propose a variety of means of transport for better development, city and city transportation, city and township between the county and township links, are very necessary. Serving the residents of the urban area to facilitate their daily passage, while providing passenger flow distribution services, further taking rail transit as the main core, and integrating with different means of transportation. Realize that users can seamlessly connect from travel to return, and provide integrated travel services for residential users.

Massive amount of vehicle trajectory data is an important data source and has been widely used in many research fields. However, due to the huge volume and variety of application scenarios, it is still not easy to achieve the application-oriented and efficient retrieval of vehicle trajectory. The paper proposes a multi-granularity vehicle trajectory index method based on the concept lattice model. The method identifies the multi-granularity feature regions for application requirements to segment vehicle trajectory and employs the concept lattice model to maintain the complex relationships between vehicle trip and the feature regions, to realize the vehicle travel trajectory data retrieval for the multi-granularity feature regions. In this experiment, shanghai city is divided into 21 feature regions according to the county-level area of Shanghai, and the concept lattice index structure of vehicle trips is constructed. The availability and efficiency of the proposed method are verified by the comparative experimental results.

Aiming at the problem of maintenance data quality of substation equipment in the power grid industry, an evaluation method of substation equipment health status based on unified semantic representation was proposed. Firstly, the current health data of substation equipment and historical data of integrated equipment archives are analyzed. Multi-source heterogeneous substation equipment data is processed by data governance and structured equipment health record database are constructed. Secondly, the selection and construction of data features are designed by observing the data scale of archives. Then, a unified semantic distributed representation model is constructed, and the word bag vector compressor is used to reconstruct the structured text data of the station into a standard state vector. Finally, a state evaluation method based on neural network is proposed to evaluate the state vector. Experiment result shows that the method proposed in this paper is feasible in the health status evaluation of substation equipment, and the accuracy rate is up to 80%.

Unit step signal or function u(t) appears in many fields, which plays an important role in the research and solution of many problems. But for a long time, this signal has a clear definition when t ≠ 0, but there is no definite definition at t=0 - this paper calls it u (0) uncertainty. Although u (0) uncertainty will not affect the application and related theoretical research of u (t) in many cases, we find that it will do in some cases. We also find that the definite definition of u (0) has been implied in the existing basic theory of signals and systems, that is, u (0) =1/2. A physical meaning of u(0) =1/2 is also given by the double switch circuit. This paper makes the understanding of u(t) and its related theory more profound, comprehensive and accurate, and lays a more solid foundation for establishing a more perfect theoretical system of information, signal, system and so on.

Place names are also an important part of Tibetan unregistered words. Place name recognition is the basic work of Tibetan natural language processing and a key issue in machine translation, information extraction, and text classification. Aiming at the characteristics and difficulties of Tibetan place name recognition, this paper analyzes the internal words and external boundaries of Tibetan place names, and designs four features, including vocabulary, label, combination and dictionary, and proposes a multi-feature fusion Tibetan place name recognition method. The accuracy of this method on the test set of this paper reaches 92.26%, and the F value reaches 88.96%, indicating that the recognition method proposed in this paper is effective.

Session-based recommendations have aroused widespread concern, and they can predict the next item of interest in the chronological order according to a given anonymous behaviour sequence. Current session recommendations benefit from the ability of graph neural network to obtain the complicated transformation patterns between items, but they still have shortcomings. Most recommendation models based on graph neural network only consider the information between items in a session but not the target item information. This paper thus proposes a session recommendation model that combines the target attention mechanism and directed graph, which considers information between items in a session as well as information about the target item. The session-level embedding is obtained using an attention mechanism and a graph neural network, followed by the target embedding obtained using a target attention mechanism, and finally the session-level and target embeddings are stitched together to get the potential transformation information of items in the session. The model has experimented on two publicly available datasets. Compared with the existing benchmark model, the proposed model has improved in all indicators, proving the effectiveness of the proposed method.

Runway occupancy time (ROT) has become a major factor limiting runway capacity. This paper presents a method based on airport operation data driving to analyze the factors that lead to the change of runway occupancy time. Based on the airport asde-x radar tracking data, different factors affecting the variability of ROT are analyzed for 36 major airports in China. The results show that the exit used for aircraft leaving the runway and aircraft type are the main influencing factors leading to the variability of ROT, and how these two dominant factors affect the change of ROT is analyzed in detail.

The Three-River Source region is located at a high altitude in the hinterland of the Tibetan Plateau, the complexity of the region and its biodiversity make its ecological data of the region very invaluable. In order to realize data sharing and further processing of the Three-River Source’s ecological data, it is essential to formulate a unified data collection and management standard for ecological data. In the paper, a method based on DOI mechanism for ecological data collection and management of the Three-River Source is proposed, and the corresponding software model is implemented and verified. Specifically, for the collection and transmission of ecological data, the metadata-based data collection model and a TCP/IP protocol-based data transmission structure are adopted to ensure the effectiveness, integrity and availability of ecological data collection and transmission between different sites. In respect of data management and storage, the data identification method based on the Digital Object Unique Identifier (DOI) is adopted in the software, meanwhile MySQL is used to realize the storage of descriptive information under DOI flag data, so as to realize the efficient management of multi-source ecological data, as well as improve the accuracy of data retrieval by users. Through testing results, the specific software model implemented based on the design scheme proposed in the paper is consistent with the expected results and can meet the requirements.

In the information processing technology, when confront the problem of multi-dimensional data processing capability limitation, a quality data information resource exchange prototype system is designed, and the key technologies involved in the system design are analyzed to realize the exchange and sharing of irrelevant information between different networks.

Based on 1200 journal papers retrieved from "big data" and "Ideological and political education" in the full-text database of CNKI from 2013 to October 2020, this paper uses the visual analysis software CiteSpace to mine and analyze the contents of journal documents in the field of Ideological and political education research, and uses the big data visual knowledge map to describe the internal relationship between various high-frequency keywords, so as to objectively reflect the basic situation of this research field, Hope to provide reference for researchers in this field.

Information visualization is an important part of bibliometrics to describe the structure of scientific system. It can gain insight into the trends of various contributing factors to the advancement of scientific publications on a certain field of research. As an example, the global scientific production related to water saving management was evaluated by the information visualization method. A total of 2302 articles published were obtained from Web of Science core collection database. From the high- frequency keywords, the literature paid high attention to water-saving management. Much of the research has focused on field of agriculture and land use. “Climate change”, “Model”, “Footprint”, “Virtual Water”, “Reuse”, “Trade” and “Water-Use Efficiency” were hot issues. This research could provide reference value and research guidance for future research on water saving management.

The spread of COVID-19 has caused irreparable and enormous damage to many families around the world, so using mathematical models to further study the changing pattern of the infection’s population caused by the spread of the coronavirus can help people to predict the trend of its changes. In this paper, on top of the logistic growth and classical SIR epidemiological models, the author develops a new SIRV model, including the effect of reinfection and breakthrough infection, to illustrate some properties of the spread of COVID-19. This study identified several fundamental properties and basic reproduction numbers of this SIRV COVID-19 model and further searched for the steady-state or equilibrium point of the model using dimensionless methods. This study demonstrated the following: first, the author proved that the solution of the model is positive under non-negative conditions. Second, the author applied the next generation matrix method to determine the basic reproduction number of the COVID-19 virus in the model and found that the calculation of the basic reproduction number in the model is the same as in the classical SIR model. Finally, the author used the dimensionless method to obtain expressions for the equilibrium points of the model in both disease-free and diseased cases.

In this paper, the node V column to analyze mechanical performance under different parameters, the selection of the ultimate bearing capacity of criterion at the same time, analyzes the beam column V node in finite element node stress changes in the process of loading, and V in beam column under uniaxial tension node parameters for the two different parameters of the variable analysis, analysis the influence of different parameters on the node force performance.

The transformer has improved slow training, the most criticized disadvantage, of RNNs and achieves fast parallelism by self-attention mechanism. Also, the transformer can be enhanced to a great depth to fully exploit the characteristics of the deep neural network model and improve the model’s accuracy, which is widely used in machine translation. Deep models, however, have the inevitable disadvantage of being prone to overfitting. We consider introducing the adversarial training method in the translation model to improve the robustness of the model through regularization to alleviate the overfitting problem. We conducted analytical experiments on the English-Chinese translation dataset which shows that the adversarial training approach can improve translation quality.

As the basic problem that introduces the idea of secure multi-party computation (SMC), millionaires’ problem (MP) is one of the research highlights in the field of SMC. However, most of current protocols for MP are only secure in semi-honest model, which using public key cryptography schemes, and there are few secure schemes in malicious model exist security defects. Therefore, this paper analyzes the Li et al.’s protocol for MP that secure in malicious model. Firstly, we point out several behaviours that can be successfully implemented in Li et al.’s protocol. Secondly, by using the zero-knowledge proof (ZKP) for the equality of discrete logarithm (DL), cut-and-choose method etc. We propose the improved protocol and show how to detect malicious behaviours. Finally, we analyze the correctness and security against malicious behaviours of the improved protocol.

Short oral texts have the characteristics of sparse features and vague expressions, which lead to poor performance when applying matching methods to them. Aiming at these problems, this paper proposes a matching method based on word vector text representation, which comprehensively considers part-of-speech, semantics and word order to map oral text to operation instructions. First, this paper uses the Skip-gram model to train the selected corpus to obtain the word vector representation of the corpus; then uses the cosine similarity to calculate the similarity of the feature words; then, combined with the text characteristics in this scene, use the WMR(Word Matching Rate) to calculate the similarity between short texts; finally, the method in this paper is evaluated on the test set. The results show that the method described in this paper has better precision and recall rate than other methods, and effectively improves the matching of oral text to instruction text.

Cross-lingual summarization (CLS) is the task of generating summaries in one specific language for source documents in different languages. Existing methods simply divide the cross-lingual task into two tasks: summarization task and translation task. Pipeline-based methods can cause problems with error propagation. In this paper, a round-trip translation strategy (RTT) is used to construct a Tibetan-Chinese cross-language abstract dataset from an existing monolingual Tibetan abstract dataset, and then an end-to-end framework is used to train the Tibetan-Chinese crosslingual abstract task. Experimental results show that our method achieves significant improvements in performance metrics compared to traditional pipeline methods.

To realize the effective application of interactive 3D technology in IETM, this paper proposes an interactive 3D method based on three.js. Using three.js, IETM can conveniently load and obtain the object information of the 3D model. According to the obtained information, IETM can further create and display the animations of the 3D model. Based on three.js, IETM can also read the information stored in the data module to establish cross-references between 3D model objects and other elements in IETM. By controlling the web pages generated by IETM, users can control 3D model objects and achieve full interaction. In order to avoid operation conflicts, view, animation and manipulation modes are set in the proposed method to realize the full interaction between users and the 3D scene.

The inability to extract and utilize features in semantic segmentation leads to the loss of detailed information and the discontinuity of semantic information, and this study proposed an encoding and decoding model CAF-Deeplabv3+ for semantic segmentation based on coordinate attention and feature fusion. First, the coordinate attention module (CA) is embedded in the backbone network ResNet for the feature extraction; Secondly, a strip pooling branch (SP) is added to the ASPP module in encoding structure to help capture contextual information; In the process of feature reuse, a feature fusion module (FFM) for fusing low-level features and high-level features. It has been verified that the model can enhance the feature extraction capabilities. Our method achieved excellent results on Cityscapes. The training and validation on the Cityscapes dataset showed that the experimental results achieve 75.01% of the mean intersection over union (mIOU), which is improved by nearly 2%.

Encryption algorithm is the basis for securing satellites, achieving reliable user access and secure data transmission. However, since the electromagnetic information generated by satellite encryption operation has certain regularities, the risk of leaking plaintext is high when the side channel is attacked, especially using a single algorithm. To solve this problem, this paper analyzes encryption algorithms with different types, summarizes four coarse-grained operators, and proposes a coarse-grained reconfigurable encryption module (CGREM) for satellites based on FPGA. Based on the coarse-grained reconfigurable architecture (CGRA), we implement a reconfigurable configuration subsystem and a computing array on a 0.15 μm CMOS FPGA. The result of the performance test show that our module can reconstruct encryption algorithms in 1.76 us. Compared with the previous method, our design is three orders of magnitude faster and more efficient. Our method can obtain the resource consumption rate of LUT and FF less than 60%, which means it can achieve a derating effect on satellite components.

Multi-behavior recommender systems model and learn user preferences by using users' multiple behavioral interbehavior. However, the existing multi-behavior recommendation models usually ignore the independent features of each independent behavior and the personalized interbehavior relationship of users, and cannot fully model the item features. To this end, this paper proposes a multi-behavior recommendation algorithm based on multi-view item semantic learning. Graph neural network is used to mine item semantics from the perspective of independent behavior, and multi-head attention mechanism is used to learn the correlation between different behavior under the graph learning paradigm. At the same time, iterative propagation captures the collaborative signals in the user's multi-behavior information, and learns item semantic information from multiple perspectives. Compared with existing multi-behavior recommendation algorithms on Yelp and BeiBei datasets, considering item multi-view semantic learning can effectively improve recommendation accuracy.

NEON technology is the implementation of the ARM Advanced Single Instruction Multiple Data(SIMD) extension. The purpose of NEON technology is to provide flexible and powerful acceleration capability for consumer multimedia applications and the trade off between the complexity of hardware implementation and so-obtained performance improvement. In this paper, we design a JPEG compression coding system that uses NEON instruction set. It uses Discrete Cosine transform (DCT) is AAN algorithm. This design is validated on a processor FT2000/4 made in China and a Debian operating system platform. The convincing results show that the performance of JPEG compression coding system with the NEON instruction has been greatly improved, compared with the not optimized compression coding system.

The depth maps obtained from current LiDAR scans are relatively sparse, while dense depth maps are needed. In this paper we propose a fusion-based RGB image and semantic image to jointly guide the deep completion. The RGB images, semantic images, and sparse depth images are fused after feature extraction, respectively. Specifically, we first perform depth estimation on the RGB image to generate the color depth. Then the color depth is fused with the semantic image to estimate the semantic depth. Finally, the color depth and semantic depth together lead to the complementation of the sparse depth map. To generate more accurate dense depth maps. In this paper, the fusion module (AFF) is added in some branches to fuse the color depth and semantic depth sparse depth maps. Under the test of public dataset KITTI, the experimental results show that the depth map predicted by the proposed depth completion model in this paper is more accurate compared with the backbone of the depth completion method guided by RGB only.

How to adaptively learn important signature features and use a lightweight model to achieve high-precision signature verification is still a challenge in the field of online signature verification. In response to this challenge, we proposed an attention mechanism depth-wise separable convolution residual network(A-DWSRNet) for online signature verification. First of all, the weight of signature features is adaptively learned through the convolutional attention module to improve the representation learning ability of the network. Next, the depth-wise separable convolution and depth convolution modules are introduced to improve the standard residual structure to construct a depth-wise separable residual unit, which reduces the overall parameter amount of the model and alleviates the loss of feature information of the multi-step residual structure. Finally, the proposed method has achieved 2.88% equal error rate (EER) and 4.17% EER in MCYT100 database and SVC2004-task2 database, respectively. The results show that the proposed method can effectively improve the accuracy of the Online signature verification system.

Aiming at the problems of low monitoring accuracy and short warning time of the existing fixed frequency monitoring and early warning method for the separation of machine and card, this paper puts forward the research on the fixed frequency monitoring and early warning technology for the separation of machine and card based on the improved least square method. Multiple monitoring points are set in the area of machine-card separation, and a single authentication negotiation protocol for machine-card separation monitoring with fixed frequency is designed. According to the protocol, the improved least square method is used to fit the early-warning function of machine-card separation, so as to shorten the early warning time and realize the efficient monitoring and early warning of machine-cards separation. The practicability of the proposed technology is proved by comparative experiments, and the experimental results show that this technology has higher monitoring accuracy and shorter warning time, which is of great significance for the rapid repair of the separation of machine and card.

As one of the extremely important components on the transmission tower, the insulator has two functions of electrical insulation and wire fixing, which directly affects the operation of the power system. Defects in insulators can impair the service life of transmission lines. UAV aerial photography of electric power towers has problems such as small number of defective insulator samples, small area, large aspect ratio of insulator strings, and variable inclination angle, coupled with the influence of environmental factors such as light, interference, distance, etc., which lead to low detection accuracy of insulator defects. Aiming at the above problems, an improved YOLOv5 insulator defect detection algorithm is proposed. First, screen the aerial images and use data augmentation to obtain a sufficient number of defective insulator images to enrich the dataset and avoid model overfitting. Secondly, the convolutional attention module CBAM is introduced to improve the expression ability of defect insulator features and strengthen the network's ability to identify targets. Finally, the Leaky ReLU activation function of the hidden layer of the original YOLOv5 algorithm is replaced by the Mish function to improve the generalization ability of the network. The experimental results show that compared with the original YOLOv5 algorithm, the average precision mAP (IOU=0.5) of the improved algorithm is increased by 7.8%, which effectively improves the problems of false detection and missed detection in the original algorithm. Compared with other mainstream object detection algorithms, the algorithm proposed in this paper has better detection effect on insulator defects.

In this work, a layout algorithm is proposed based on depth-first search algorithm. The concepts of graph theory, adjacency matrix, the characteristics of configuration diagrams and the problems of fixed shafts are systematically analyzed to give a general solution for the layout optimization of marine transmission configuration diagram. The algorithm is proved to be able to solve the layout problem with many-to-many connections, and can clearly express the node relationship of each layer while reducing the coincidence of nodes. At the same time, a visual configuration diagram modeling tool for the conceptual design stage is realized. Based on this tool, the above algorithm and other two general Sugiyama algorithms are compared in three aspects: total area, total edge length and the number of node cross. The results show that the algorithm has strong pertinence, good layout effect, clear hierarchy, less node intersection, small total area, and excellent applicability to different node sizes. And it is applicable to the layout optimization of the configuration diagram of marine transmission system diagrams of any scale.

In view of the deficiency that target threat assessment based on three-way decision(3WD) model is difficult to completely classified the targets, target threat assessment based on sequential three-way decision(S3WD) model was introduced in this paper. Firstly, the comprehensive information matrix of targets was established under intuitionistic fuzzy environment. Based on the comprehensive information matrix, VIKOR method was used to evaluate the conditional probability and calculate the threshold pair. Besides, two important parameters in VIKOR method with respect to classification effect in S3WD model were systematically analyzed. Finally, the numerical examples show that compared with traditional methods, this method can achieve the dynamic threat assessment by adjusting the threshold pair when attribute is updating, which is more suitable for modern air combat.

At present, the main problem of college course selection system is that when facing a large number of requests in a short time, there will be server downtime. The high concurrency solution adopted by large-scale distributed systems will use a large number of server resources, which is not suitable for small and medium-sized web applications deployed on a single server. Therefore, it is necessary to propose an available high concurrency processing scheme for small and medium-sized web applications. This paper mainly uses micro service architecture combined with flow limiting technology to design a mainstream high concurrency scheme. This scheme can be applied to the course selection system to improve the efficiency of students' course selection. Using micro service architecture, the complex business functions are divided into several small service modules, which solves the challenges faced by the traditional system, so as to build an efficient, A stable course selection system.

With the construction of urbanization, large-scale commercial buildings and large-scale hubs appear in people's lives. Indoor positioning technology plays an important role in many fields. However, in complex situations, in the indoor dynamic environment, it is easy to suffer from multipath interference, and the positioning accuracy will be significantly reduced, resulting in a large error in RSSI observation. This thesis presents an improved BP neural network ranging model algorithm. Compared with the traditional logarithmic distance ranging model, the algorithm does not need to estimate the parameters, which reduces the accumulation of errors. After the BP neural network model is trained, the RSSI data is input into it, and the corresponding distance will be obtained. The experimental results show that the model established by BP neural network can reflect the effect of RSSI value attenuation with distance, which is more in line with the real situation of signal propagation in complex indoor environment. Therefore, the BP neural network proposed in this thesis can improve the ranging accuracy and reduce the error.

In order to solve the problems of public safety and user parking irregularities caused by the current unmanned parking management, a design based on the combination of block chain and parking integral system is proposed. By integrating parking management into urban civilization integrity score and combining with blockchain technology, we can explore the establishment of a system to regulate personal parking behavior with integrity score. Based on the HyperLedger Fabric blockchain development platform, a set of common users, traffic management departments, personal credit system and other joint participation of the alliance chain, the use of blockchain technology can not be tampered with the characteristics, but also to solve the traditional integration system of poor liquidity, user rights and interests can not be guaranteed.

As a security standard, OAuth's access certificate is an anonymous token without any protective measures. It is easy to be malicious monitored or stolen in channel transmission. In order to enhance the security transmission of access token and ensure the confidentiality of protected resources. This paper discusses the authorization and delegation function of oauth2.0. Based on the operation mechanism of standard oauth2.0 protocol, it is proposed that the access credentials issued by the subject authorization server are transmitted in JWT structured token format, a handshake is added between the authorization server and the resource server, and the combined encryption CPK system is used in the transmission of access tokens to strengthen mutual authentication among servers, The formal security analysis of partially optimized oauth2.0 protocol is carried out with AVISPA tool as the carrier. The verification results show that this method improves the security of the protocol to a certain extent.

Based on Python language and Django framework, this paper completes the construction of cloud computing network security level protection evaluation system. The system will focus on many problems existing in the current cloud computing network security level protection evaluation work, focusing on the implementation steps, and through the comprehensive application of network information technology, try to provide convenience for the cloud computing network security level protection evaluation work with convenient and efficient Web applications. The system will highly meet the actual application requirements of users with different roles, and realize the full-cycle management and supervision of the evaluation of level protection with rich functional modules, which effectively reduces the workload and difficulty of evaluation institutions and personnel, and improves the controllability of the evaluation process of cloud computing network security level protection and the accuracy of the evaluation conclusion.

Popular island detection methods generally include active detection methods, passive detection methods and communication detection methods. The active detection method will cause instability in the power quality of the power grid due to the addition of other parameters, and the passive detection method will cause a large detection blind area because the threshold is difficult to determine, resulting in inaccurate island detection results, and the islanding phenomenon may cause serious harm to the power grid and related personnel when it occurs. Optimization problem has always been a hot problem in scientific research, because the traditional optimization scheme has many defects in solving problems such as large dimensionality and multimodality, this paper aims to study the artificial fish swarm algorithm, and combine the advantages of various algorithms, and proposes a grid-connected photovoltaic system island detection method based on the improved artificial fish swarm algorithm. Using the optimization principle of artificial fish population to extract the relevant electrical characteristics, the large number of data sets are placed in the algorithm environment, matlab/Simulink and Python are used to establish a photovoltaic system grid-connected model, simulation and verification are carried out and the experimental results are compared with common intelligent detection methods, and the results show that in the algorithm environment, the blind area of island detection is small.