This PDF file contains the front matter associated with SPIE Proceedings Volume 12328, including the Title Page, Copyright information, Table of Contents, and Conference Committee lists.

Image captioning tasks based on deep learning encompasses two major domains: computer vision and natural language processing. The Transformer architecture has achieved leading performance in the field of natural language processing, There have been studies using Transformer in image caption encoder and decoder, the results proving better performance compared to previous solutions. Positional encoding is an essential part in Transformer. Rotary Transformer proposed Rotary Position Embedding (RoPE), has achieved comparable or superior performance on various language modeling tasks. Limited work has been done to adapt the Roformer's architecture to image captioning tasks. The study conduct research based on the positional encoding of Transformer architecture, our proposed model consists of modified Roformer as an encoder and BERT as a decoder. With extracted feature as inputs as well as some training tricks, our model achieves similar or better performance on MSCOCO dataset compared to “CNN+RNN” models and regular transformer solutions.

Recently, metalens for circularly polarized incidence have received increasingly attentions due to their superior functionalities on the wavefront manipulation of light. However, the previous work on focusing metalens usually only deals with the single output component of incident light, that is, either the co-polarization component or the quadrature polarization component. In this paper, a reflective metalens working at 1550 nm is proposed and designed using a twodimensional nanoresonating structure of a metal elliptic-square ring. By combining the resonant phase with the geometric phase (i.e., Pancharatnam-Berry (PB) phase), the metalens is able to separate and focus two orthogonal polarized components of a circularly polarized incident light. When the orientation of the metal elliptic nanofin is opposite to the helicity of the incident light (i.e., left-handed circular polarization (LCP) or right-handed circular polarization (RCP)), two distinct focal points can be obtained can be obtained while only one focus can be obtained when the orientation of the metal elliptic nanofin matches the helicity of the incident light. The focusing spot of the metalens approaches the diffraction limit (1.22λf/D=1.2 μm). The different responses of the proposed metalens to LCP and RCP light has a broad application prospect in the polarimetric imaging and detection of chiral molecules.

The structural parameters of the quantum well have a very important influence on the performance of InGaN laser, including output power, optical field distribution, electron leakage, etc., so it needs to be considered in the design. The photoelectric performance of InGaN/(In)GaN quantum well lasers with varying thickness of quantum well/barrier layers are theoretically investigated with the simulation program Crosslight. For three In_0.15Ga_0.85N/GaN quantum wells violet laser diode (LD) with lasing wavelength around 410 nm, the performance of threshold current and optical output power of the laser degenerates with the uneven well thickness. This is attributed to the deterioration of the carrier distribution and the mode gain in quantum wells. When the quantum well structure adopts barrier layers with non-uniform thickness, the threshold current of InGaN quantum well laser degenerates a little while the lasers’ optical output power increase slightly. Based on the In_0.15Ga_0.85N/In_0.02Ga_0.98N quantum well, the larger refractive index difference between the barrier layer and the well layer significantly improves the distribution and concentrates of the optical field near the active region. Meanwhile, compared with In_0.15Ga_0.85N/GaN structure, In_0.15Ga_0.85N/In_0.02Ga_0.98N quantum well laser is more effective in reducing the electron leakage. Moreover, the output power of gradually thickening barrier laser achieves 2.6 times that of the In_0.15Ga_0.85N/GaN symmetric quantum well structure. Our results prove that the asymmetric quantum wells with higher refractive index and gradually thickening barrier layer are beneficial to realize low threshold current and high output power laser.

The emerging hot industries such as 5G communications require a higher demand for Si-based photo detectors, especially the SiGe heterojunction phototransistor with internal gain. But there is a big contradiction between the responsivity optimization and working speed promotion. These cause the difficulty of optimizing the common vertical-illustrated SiGe phototransistor. This paper proposed a novel SiGe ridge waveguide phototransistor and analyzed the influence of the waveguide structure, including base thickness, ridge waveguide width and length, on its optical characteristic frequency and optical gain. The ridge waveguide structure is optimized and obtained finally. When the thickness of the base region is 40nm, the doping concentration is 1.0×10¹⁹cm^-3, the waveguide length is 50μm, and the ridge width is 1μm, the designed SiGe ridge waveguide phototransistor obtained the optical gain of 6.24 and the maximum characteristic frequency of 12.8GHz under incident light of 850nm.

We propose and analyze an optically transparent microwave perfect absorber which is based on a metamaterial structure. The absorber consists of a layer of a high-square impedance ITO film and a resonant cavity that provides zero phase and total reflection. The resonant cavity is formed by a metal grid structure on the upper surface of a transparent polyethylene terephthalate (PET) and a metal grid on the bottom surface as the ground. Almost perfect absorption could be achieved because of the interference phase extinction within the high-square resistance ITO film. In order to achieve optical transparency in the visible range, PET film is used as the substrate with a grid-like metallic pattern. Absorption of higher than >90% from 13.88-16.92 GHz and average optical transmittance of ~83% in the visible-near-infrared (400-1400 nm) are achieved both theoretically and experimentally. The suggested approach and structure provides a way to realize an optically transparent microwave absorber which is of significance in real applications.

We propose a method for designing diffractive lenses, which is called binary variable slope search (BVSS) algorithm. The diffractive lens designed by this algorithm have a fixed-position focus at several prescribed wavelengths, which we refer to as achromatic diffractive lenses (ADLs). The BVSS algorithm greatly reduces discontinuity in depth of the diffractive microstructure, thus it allows one to consider the technological limitations associated with the fabrication of the diffractive microrelief. As example, we designed an ADL, which can focus five wavelengths at a point to reduce chromatic aberration. The simulation and experimental results show that the deviation between the designed and fabricated profiles amounts to 8%. The obtained results confirm the efficiency of the proposed method in practical applications.

With the increasing integration of optical devices, optical integration technology is becoming more and more important. In this paper, we propose double-stage end coupler based on SiO₂. The double stage taper structure greatly shortens the size of the coupler and ensures a high coupling efficiency. In this paper, we take an arrayed waveguide grating output waveguide and a ridge waveguide of photodetector as application scenarios, the cross-sectional areas of the two are 4.5× 4.5 um and 2×1.04 um respectively. Then, the coupler parameters are designed. By adding couplers, we can improve the coupling efficiency from 73.3% to 96.4%. At the same time, the length of the proposed coupler is only 900um. In addition, the model also has wider operation bandwidth, lower polarization dependence loss and larger alignment error tolerance. The coupling efficiency of the model in the 1270 - 1350 nm band is higher than 94.8%. At the same time, the polarization dependence loss is only 0.35dB, and the alignment error tolerance of 1 dB is more than 1500 nm, which ensures the model can be well applied to the field of photonic integration.

A microwave photon signal generation method based on electrical frequency sweeping is proposed. The frequency sweeping range of the microwave photon signal is 4 times that of the frequency-swept microwave source, which can reach 380 GHz. In addition, the spectral purity of the signal is relatively high, and its optical sideband suppression ratio is greater than 96 dB, which can serve as a swept-tuned optical local oscillator (LO) of the coherent spectrum analyzer to improve the resolution of spectral analysis.

Dynamic manipulation of the polarization state of light has essential applications in many fields, including optical communication, holographic display, and imaging. To date, graphene-loaded polarization devices have been reported mainly working in mid-infrared and THz wavelength bands. Here, we propose a tunable reflective polarization generator working at the wavelength of 1.55 μm based on graphene-loaded ultra-thin silver nanoslits array. The physical mechanism of the device is based on the perturbation effect of plasmonic resonance. The Fermi level (E_f) of graphene is adjusted to change its permittivity around the plasmon mode excited by the incident light whose polarization direction is perpendicular to the direction of nanoslits, and subsequently to change the resonance frequency to obtain a dynamic tunability of the polarization state of the reflected light. For a normal linear incident light with a polarization orientation angle of 35º with respect to the slit direction, the polarization states of the reflected light can be tuned to right-handed circular polarization (RCP), left-handed circular polarization (LCP), and linear polarization (LP), when graphene’s E_f is adjusted to 0.53 eV, 0.71 eV, and 1 eV by applying a gate voltage, respectively. Also, continuous ellipticity of circular polarization state of reflected light can be tuned by continuously changing graphene’s E_f. The proposed tunable device makes it possible to rapidly and dynamically encode polarization information in optical communication bands, which will have tremendous applications in the field of polarization division multiplexing.

In recent years, methods based on autoencoders (AE) in deep learning have received extensive attention for hyperspectral unmixing. The purpose of hyperspectral unmixing is to estimate terminal members and their respective abundances. This is similar to the learning process of an autoencoder, which is trained to find a set of low-dimensional hidden layers and combine them with their corresponding weights to reduce the reconstruction error. Therefore, AE is well-suited to solving the problem of unsupervised hyperspectral unmixing. Aiming at the problems of being unrobust to noise and the unmixing accuracy to be further improved, this paper proposes a convolutional autoencoder unmixing network (CAA-Net) based on attention mechanism. First, an attention mechanism is introduced to improve the unmixing performance. Then, a total variation regularization term is introduced to exploit spatial information and facilitate piecewise smoothness of abundance maps. The paper conducts experiments on the Samson dataset and Jasper dataset, and compares with other classical methods to obtain higher accuracy.

The organic material with numerous technological advantages and desirable properties promotes the development of highperformance organic photodetectors (OPDs), possessing wide applications in various fields. The photomultiplication (PM) type OPDs with sandwich structure of ITO/PEDOT: PSS/PBDB-T: PC71BM (mass ratio, 50:1)/Al are successfully demonstrated. The effect of methanol treatment (MT) on the performance of PM type OPDs was investigated in this work. The dark current of PM type OPDs with MT is higher than that without MT under working voltage, and the external quantum efficiency (EQE), responsivity and specific detectivity of PM type OPDs is less than that without MT. The champion EQE value of OPDs with or without MT at -40 V approaches 280% or 330% under 680 nm light illumination, respectively. The PM effect can be ascribed to the hole tunneling injection assisted by interfacial band bending. This work confirms the unavailability of MT on improving the performance of PM type OPDs, providing new insight into achieving excellent performance PM type OPDs.

In this work, we propose a new calibration technique for locating the projector in the structured light measurement system. Using Brox optical flow, the estimation of three coordinate components of the projector only requires two images captured before and after the motion of the calibration plate. The calibration principle presented on geometry depicts the relation between the position of the projector and that of the camera, to the optical flow caused by the movement of the calibration plate. The accuracy of the proposed method is verified by numerical simulation. The measurement system is calibrated by using the proposed calibration method and by Falcao’s method simultaneously for comparison. After calibration, eight-step phase shift technique of a recognized high precision is performed to reconstruct the 3-D shape. The results show that the height distribution of the specimens is in good agreement with its true values, and the maximum absolute error is less than 0.1 mm.

High speed and high efficiency photodetectors are the key components in optical transmission, optical sensing and optical processing systems. Except performance requirements, cost control is also an aspect that must be considered in the current technological development. Therefore, silicon-based high-speed and high-efficiency chip level photodetectors that can be monolithic integrated with very large-scale CMOS integrated circuits have become the focus of research. SOI structure provides performance improvement for silicon-based HPT. We demonstrate a SOI-based SiGe Heterojunction Phototransistor which has greater light absorption compared with Si-based HPT. The influence of SOI structure on absorption and collector current of SiGe heterojunction phototransistor (HPT) are simulated and analyzed in this paper. When the optical power is 10mw, the collector current in SOI-based HPT reported in this paper can reach 4.6mA with 5.83% of light absorption calculated, while the collector current in Si-based HPT is 3.35mA with 4.48% of light absorption calculated under 940nm wavelength. The peak responsivity with SOI structure exhibits 2.82A/W which is enhanced by 12.8% than that of Si-based HPT.

An image telecentric micro-projection system based on multilayer diffractive optical elements (MLDOE) has been designed for portable device. It consists of 0.26-inch Liquid Crystal on Silicon (LCOS), the polarization beam splitter (PBS) and hybrid refractive/diffractive eyepieces. The system has a focal length of 9.75 mm, a field of view of 40°, an entrance pupil diameter of 4 mm, and works in the visible light. The diffractive structure contained in it is a double-layer separated diffractive optical elements with an angle-band integral average diffraction efficiency (ABIADE) of 98.9% at an incident angle of 0-15° and working wavelength of 400-700 nm. The Modulation Transfer Function (MTF) of the system is better than 0.5 at 100 lp/mm, the distortion is less than 2%, the imaging quality is excellent, and the maximum diameter is 9.6 mm, the total length is 24.2 mm, and the weight is 2.37 g. It has the advantages of compact structure and light weight, and is satisfied with the miniaturization of the micro-projection system demand.

A reconfigurable optical FIR filter based on delay interference structure is proposed, which can realize band-pass and band-stop filtering. The interference structure is mainly composed of an adjustable beam splitter based on Mach-Zehnder interferometer(MZI) structure and a delay line with a phase shifter. Light can be guided to four paths that have different delays and different optical transmission spectrum can be achieved by tuning the amplitude and phase of the four paths by tuning the phase shift of the MZI and delay line. To optimize the transmission spectrum, the theoretical model of the delay interference filter is established and simulated by the transmission matrix method. The simulation results show that the extinction ratio of band-pass filter is greater than 15dB, the extinction ratio of band-stop filter is greater than 40dB, and the in-band fluctuation is lower than 1.5dB.

In this paper, a kind of absorber based on local surface plasmon resonances is designed by using the finite difference time domain (FDTD) method. The absorber can realize the nearly perfect absorption from visible light to the near-infrared wavelength. The absorber consists of the bottom metal tungsten (W) as substrate, the middle dielectric layer aluminum oxide (Al₂O₃), and the multilayered nanocircular disk and nanoelliptic disk. The four layered nanodisk structure materials are Ti- Al₂O₃ -Ti- Al₂O₃, and these nanodisks are periodically and symmetrically arranged. Numerical analysis shows that the average absorption rate of the absorber can reach 97.2% in the working band of 426~1947nm, and the absorption bandwidth is 1521nm. By analyzing the electromagnetic field distribution of the resonant wavelength, it can be seen that characteristics of the ultra-broadband and high absorption in the absorber are attributed to the local surface plasmon resonance (LSPR). This kind of absorber with ultra-broad band and high absorption absorber is expected to play an important role in photoelectric devices, solar energy collection and other fields.

In order to detect the hydrophobic characteristics of composite insulators, based on water drop edge feature recognition, the gradient amplitude of the image was suppressed by non-maximum to remove non-edge pixels. Double threshold detection algorithm and image closed algorithm were used to fill edges locally. Finally, the water drop surface regions were obtained, and the hydrophobic characteristics of composite insulators were calculated by using the maximum area ratio method. The results show that this method could effectively extract the surface area information of water drops which enhanced the image analysis effect of composite insulators. Based on this method, the hydrophobic characteristics level HC₁-HC₅ of composite insulators could be accurately identified.

This paper mainly studies the evaluation criteria of camouflage effect. Based on the visual characteristics of human eyes, this paper expounds the perceptual characteristics of brightness, color and boundary of human eyes, and establishes the evaluation system of camouflage effect, and puts forward a series of parameters to evaluate the camouflage effect. Camouflage effect evaluation standard design is the key supporting technology of variable camouflage system. Through this standard, the verification of camouflage scheme library is realized. In the process of camouflage decision, the evaluation of the existing camouflage effect can assist in the judgment and decision of changing camouflage scheme. After the completion of the work system design, the validity of the work is verified with the help of this standard.

The Driver Machine Interface(DMI) is a peripheral interface device of the onboard equipment of the high-speed railway train control system. It receives and processes the key control information such as train speed, alarm, level, and mode updated by the onboard equipment in real-time, meanwhile it outputs dynamic display and instructions to the driver, and thereby transmits the automatic protection status of the onboard equipment. To promote the automatic function test of onboard and DMI equipment, this research is proposed for the identification of the dynamic displays of DMI. Different schemes are mainly used for feature extraction and classification recognition of DMI so that reducing artificial recognition errors and assist in-vehicle equipment fault diagnosis.

Current high speed photon communication system has a big demand for high performance optical detector, especially the high speed silicon-based detector. In this paper, a SiGe heterojunction phototransistor with different Ge-profile in the SiGe base is designed for high optical characteristic frequency. Influences of three types of Ge-distributions (box, triangular, and trapezoidal distribution) and different Ge-content which is in the range of 15%~35% on the frequency performance of SiGe heterojunction phototransistors (HPTs) are analyzed in this paper. A characteristic frequency of 10.52GHz for an 850 nm incident light is achieved by using triangular distribution with 20% Ge-content in the SiGe base.

The photodetectors generally works as the significant unit in intelligent optoelectronic systems, converting optical signal into electrical signal. The photomultiplication (PM) type polymer photodetectors (PPDs) were successfully demonstrated based on sandwich structure of ITO/PEDOT:PSS/PBDB-T:polymer acceptor (mass ratio, 100:3)/Al. The performance of PM type PPDs were enhanced by solvent additive 1-chloronaphthalene (CN). The EQE value of PM type PPDs with CN under -15 volts bias approaches 860% and 600%, and is better than EQE values of 300% and 240% in PM type PPDs without CN under 350 nm and 670 nm light illumination, respectively, which confirms the effectiveness of solvent additives CN on improving the performance of PM type PPDs. The PM phenomenon can be ascribed to the hole tunneling injection assisted by interfacial band bending. This paper provides new insight into realizing excellent performance PM type PPDs.

This paper studies a dual-optoelectronic delay cross feedback dynamic control technology to how to obtain control-chaos of two coupled lasers. The dynamic control scheme is designed out: Step one, two photo-converters transform two lasing from two lasers into two photocurrents; Step two, the photocurrents are delayed to cross feedback to two lasers respectively to control two chaotic behaviors of two lasers so that two laser’s dynamic variations are dynamical guided. Step three, the dynamic control technology is performed on the two lasers to shift a controlled dynamic state to another controlled dynamic state, such as chaos in two coupled lasers having been dynamical guided to different stable-states, or laser one shifting from a controlled cycle-three state to a controlled stable state while another laser shifting from a controlled cycle-five state to a controlled stable state. We find that it takes different time for the control-chaos time of chaos being stabilized to a controlled state in different performing time. The physical mechanism is found that controlchaos is very sensitive not only to the two feedback level parameters, but also to the two delayed time parameters. Our results have certain reference value for the study of laser, optics, control science and chaos.

Vertical cavity surface emitting laser (VCSEL) is an excellent laser light source with small volume, low threshold, easy integration, and array arrangement, and has been widely used in sensing, communication, medical instruments, processing, and other fields since its birth. However, the application of VCSELs in some fields is limited, such as laser processing, projection and display, and medical equipment, because the laser light is a Gaussian distribution with concentrated energy. TopHat beams have received extensive attention due to their uniform distribution of energy over the spot area. In this paper, we tightly combine the metasurface with the VCSEL by means of on-chip integration and realize a metasurface-integrated VCSEL (MS-VCSEL) that can directly output the TopHat beam. The standard deviation is used to calculate the test results, and the uniformity of the spot was 70.5%. This work has important implications for the design of chip-scale optical systems, making it possible for laser chips with TopHat beams as light sources to be applied in future products.

Towards the improving apparent quality of optical film, it is of great significance to study the formation mechanism of apparent rainbow veins and analyze their variation laws and influencing factors. In this paper, the effects of coating thickness, refractive index, uniformity of coating thickness and the internal structure of the base film on the formation of rainbow veins were studied based on the single-coating test method on the surface of polyester film (PET) base film. On the basis of experiments, the formation mechanism of apparent rainbow veins in optical film was analysis.

Energy resolution, considered to be one of the most important parameters, reflects the performance of Silicon Drift Detectors (SDDs). A constant area spectral peak reconstruction algorithm is proposed to improve the energy resolution of SDDs in this paper. Without changing the peak position and peak area, the peak width is reduced by spectral peak reconstruction. The results show that the energy resolution of the SDDs is improved by 28.57% using the algorithm. Moreover, with the increase of energy resolution, the peak amplitude increases, which in turn increases the peak-to-back ratio by 18.71%. For sum peaks, it is effective to decompose sum peaks by the least square fitting. The method is applied to fit the curve to solve the relevant parameters and the peak widths are changed to reconstruct the Gaussian peak.

In view of the low segmentation accuracy of multiple overlapping images in traditional 3D point cloud segmentation, this paper proposes an overlapping citrus fruit point cloud segmentation method combining supervoxels clustering and European clustering. The color image and depth image of overlapping citrus fruits are obtained by Kinect V2 camera, and the three-dimensional color point cloud of citrus is obtained. Set the color difference threshold to obtain the fruit image, then carry out statistical outlier filtering, and then carry out European clustering segmentation. Set a certain number of point cloud thresholds to extract the point cloud clusters in adhesion state. The segmentation of citrus hypervoxel is realized by clustering the state of citrus hypervoxel. The results of field experiments show that the proposed method has a certain improvement in accuracy and time efficiency compared with the traditional segmentation method.

We present a two-type modulation method and study how to produce chaos and quasi-period in an erbium-doped fiber laser, as well as give a 2-level laser physical model including a digital photo attenuator controlling of lasing photon, and a modulator shifting the pumping. The laser appears chaos and some quasi-cycle when the pump of the laser is modulated and the digital modulation is operated to absorb lasing photon using the digital photon attenuator. We illustrate a route to chaos by altering the pump modulation levels and the absorption levels of the digital photon attenuator while a cycle, a three-cycle, a four-cycle, and other quasi-cycle and chaos show in the laser. We analyze the effect of the levels of the photon attenuation and the pump modulation setting in other values on dynamical behaviour of the laser, and give a route away from chaos while a quasi-cycle movement and chaos are found. And we find the threecycle distribution and chaos distribution via adjusting the absorption and modulation levels and frequencies. We find that the absorption modulation levels and frequencies lead to the dynamics behaviour of the laser. Our results have certain reference value to laser optics, and chaos.

Measuring the surface topography of the chip package can detect the quality of chip package. Aiming at the problem of the high aspect ratio of the chip package topography, this paper proposes a dual-wavelength digital holographic microscope measurement system based on temperature-controlled modulation of the laser wavelength. The system uses two identical lasers, and the temperature-controlled modulation of one of the lasers has a wavelength shift interval of 2.81 nm, and the difference between the two-wavelength lasers is 2.30 nm. Building a dual-wavelength digital microscope holography system to record the surface topography hologram of the 24μm standard specimen and the chip package specimen, experimental results verify the system’s performance.

In recent years, due to the rapid development of range-gated lidar imaging technology, electro-optical modulation devices of crystals that can realize high-speed time-slicing have become the focus of research. The modulator acts as a "shutter" in the gating system, and controls the on-off of the optical path by loading a voltage on the crystal. However, due to the small aperture of the crystal, the imaging field of the commonly used electro-optical Q-switched electro-optical shutter is too small, and it is more practical to use the conoscopic interference phenomenon of the crystal to study the imaging characteristics of the shutter. Traditionally, Matlab software is used for simulation, but this software has problems such as too complicated setting of biaxial crystal orientation, high simulation difficulty, complicated simulation system model steps, inconvenient parameter switching, and difficulty in rapid simulation. The Virtual Lab Fusion software used in the simulation in this paper improves the above problems to a great extent. Virtual Lab Fusion is a high-speed optical simulation platform developed by German Light Trans with the concept of field tracing.As the world's only high-speed physical optics simulation software developed based on the concept of field tracing, Virtual Lab Fusion has the following features: 1. It can quickly switch between ray tracing and light field tracing; 2. It has a rich light source library, medium The library, the detector library, can perform system modeling arbitrarily and quickly; 3. The parameter setting during simulation realizes data visualization, which can carry out rapid simulation, which is more conducive to speeding up research progress. In this paper, Virtual Lab Fusion is used to simulate the conoscopic interference of crystals.The system is modeled by selecting appropriate optical components. By adjusting the angle between the transmission axis of the polarizer and the analyzer, the crystal orientation, and the incident angle of the incident light, the influence of these factors on the conoscopic interference phenomenon is studied. theoretical support. In the latest Virtual Lab Fusion 2021.1 version, biaxial crystals can also be modeled in the optical system, and there are multiple ways to define the crystal coordinate system, which greatly reduces the simulation difficulty of biaxial crystals. The simulation results show that: 1. Changing the direction of the transmission axis of the two polarizers can provide theoretical support for the study of the imaging characteristics of the crystal-based electro-optic shutter to a certain extent. 2. The crystal orientation has a great influence on the simulation results. 3. The increase of the incident light angle increases the isochromatic lines of the interference pattern. The imaging distortion caused by the interference of the crystal cone can be eliminated by controlling the incident angle of the light. However, the problem of the small field of view of electro-optical switch imaging still exists in practical applications, and further research is required.

This paper aims to solve the problem of the fiber optic cable sensitivity measurement. A fiber optic cable sensitivity measurement method is proposed based on distributed acoustic sensor (DAS) system, and an experimental device is implemented. Five different optical cables are connected to the DAS system, and the acoustic pressure sensitivity of different optical cables at different frequencies is measured through the DAS system. The results show that the cable G654 has the maximum sensitivity of -148.84 dB(rad/μPa) at 120 Hz, GYTA53T has the maximum sensitivity of -160.68 dB(rad/μPa) at 140 Hz, and G657A2 has the maximum sensitivity of -166.07 dB(rad/μPa) at 160 Hz.

Ground glass pulmonary nodules have small grayscale differences and uneven distribution of grayscale values. Traditional random walks consider only grayscale features in the definition of weights, which are not accurate enough in the calculation of weights for regions with small grayscale differences, thus leading to segmentation discrepancies. In this paper, a segmentation algorithm based on region growth and improved random walk is proposed to improve the accuracy by introducing geometric distance and Gabor texture feature values in the weight calculation. Firstly, the coarse segmentation is done by using region growth; then the connected regions of the image are judged by grayscale and geometric features to locate the lung nodule region; finally, the improved random walk is used to complete the segmentation of lung nodules. Experiments show that the proposed algorithm can effectively improve the segmentation accuracy and Dice coefficient, and reduce the RVD to a large extent.

Aiming at the problems of low efficiency and large error of artificial puncture points for accurate correction of data in aerial photogrammetry hollow three-encryption, this paper proposes a right-angle image control point target detection method based on LSD (Line Segment Detection) algorithm. In this method, the color information of the image is first enhanced by the Retinex algorithm; then the image is filtered bilaterally, which can well preserve the edge information while removing noise; then the LSD algorithm is used to extract the line segments, and the line segments are combined with the least squares fitting; The outermost right-angle edge is filtered out by angle, distance and length information, and finally the right-angle image control point is obtained by intersecting. In the case of complex background and target distortion, it can still maintain a high accuracy, and the positioning accuracy is significantly better than that of artificial punctures. Compared with the Radon and PPHT algorithms, the detection accuracy of the image control point group of the algorithm in this paper is significantly improved, indicating that its detection accuracy is less affected by the shooting angle.

With the continuous growth of car ownership, intelligent transportation has penetrated into our daily life. As an important part of intelligent transportation, car detection has also been developed rapidly. It plays a vital role on the planning of urban public transportation and brings great convenience for citizens to commute. Due to the extremely complex urban conditions, car detection encounters many difficulties. By analyzing a large amount of vehicle color difference data, it draws the conclusion that illumination is the main factor of affecting detection and recognition. Based on the convolutional neural network framework, this paper focuses on low-light enhancement and car recognition, aiming to realize the task of car recognition in more complex low-light situations. Notably, in those scenarios, the YOLOv4 model with a basic training set can recognize automobiles well.

Pointer meters are widely used in industrial production, and it is very important to read the position of instrument pointer accurately for monitoring the running state of equipment. In view of the type of pointer meters and the environment of the factory, a method of detection of the angle of the pointer meter using the two-dimensional pan-tilt is proposed. First of all, the image pre-processing part meant to reduce the degree of interference, and then the use of Hough circle detection to find the position of dial, according to this position and the size of the image turn the pan-tilt and zoom. After the end, the image by bit and operation will be removed outside the dial, and then use Hough line detection to obtain all the lines in the image, through the distance between the straight line and the center of the circle to filter out the pointer, and finally according to the coordinates of the pointer to calculate the angle. The results show that compared with other methods, this method allows a certain distance and angle between the image acquisition device and the pointer meters, reduces the requirement of image acquisition, and successfully eliminate other interference information by bit and operation method, reduces the operating burden of the algorithm and improves the accuracy.

Biosensors have broad application prospects in biomedicine, pharmacy, chemical industry and environmental monitoring. Therefore, the research and development of biosensors has become a new hotspot in the development of science and technology in the world. A vertical cavity surface emitting laser (VCSEL) biosensor chip integrated with gold nanostructures is introduced in this paper. The chip mainly uses the packaging technology of semiconductor laser to realize the high integration of light source, hexagonal gold nanoparticle array, and detection system. We use anodic aluminum oxide film (AAO) as a mask to prepare hexagonal gold nanoparticle array, combined with a microfluidic chip to realize the sensing application that can be sensitive to the change of environmental solution. We modify the gold nanoparticles on the chip surface with specific antibodies, and then inject different concentrations of protein solution for detection. The output light power changes with the change of environmental solution, so as to detect the concentration and type of biological solution. The sensor has the advantages of low cost, high sensitivity, and high integration.

The composition spatial distribution and degradation properties of dissolved organic matter (DOM) in Micro Pressure Inner Loop bioreactor (MPR) at low temperature were analyzed by fluorescence excitation-emission matrix regional integration analysis (EEM-FRI) and ultraviolet spectroscopy (UV). The results showed that MPR had a good degradation effect on DOM in low temperature municipal sewage and the upper region can degrade high molecular weight DOM. At low temperature, the fluorescence components of DOM at different sites were not different, and there were tryptophan-like Peak T1 and Peak T2, but the fluorescence intensity of the same components was different. The content of aromatic protein-like substance in MPR was higher, the microbial metabolism and humic acid in sewage treatment process were lower and these fluorescent components were evenly distributed.

A diode-pumped solid-state laser with natural cooling is introduced. Since there is no limitation of air cooling and water cooling, the volume of the laser is greatly reduced and the application scenarios of the laser are expanded. The single-pulse energy of the solid-state laser is adjustable from 20 to 120 mJ, and the repetition frequency is adjustable from 0 to 20 Hz. Under the condition of single pulse energy of 80mJ and repetition frequency of 20Hz, it can work continuously for 2min, and the environmental experiments at high temperature of 60℃ and low temperature of -40℃ have been carried out. The trigger signal provided by the laser through the external interface can achieve various purposes such as irradiation, interference, and laser ranging.

The spatial optical solitons in nonlinear optics have been studied by many scholars in recent years, but the dynamic properties of spatial optical solitons under the boundary conditions of transmission media are rarely studied. This paper has done some research in this direction. In nonlinear thermal media with square cross section, the force and the corresponding torque exerted on the soliton by the boundaries is studied based on Ehrenfest's theory, and the propagation dynamics of the soliton under such force and torque are investigated. It is found that when released off center, the soliton's trajectory pattern will experience a transition from a horizontal one to a diagonal one when the releasing position is away from the 𝑥 axis. It is also found that there exists bouncing phenomenon when the soliton is released from four special positions.

Tunable diode laser absorption tomography (TDLAT) can reconstruct the non-uniform temperature field. Based on TDLAT, a two-dimensional temperature field reconstruction strategy for mesh optimization is proposed in this paper. Firstly, the temperature field is reconstructed once, the temperature distribution information of the field to be measured is extracted from the first reconstruction result, the meshing is optimized, and then the temperature field to be measured is reconstructed twice. The numerical simulation results show that the reconstruction error of the temperature field reconstruction results after meshing optimization is significantly improved compared with that before meshing optimization, and can better reflect the actual temperature distribution.

The far-field analysis of the finite difference time domain (FDTD) method has been presented to ascertain the valid range of the scalar diffraction theory (SDT) for periodic harmonic diffractive optical elements with continuous surface profile. And the near-field analysis was used to ascertain the source of error. The valid range of the scalar diffraction analysis is then determined as a function of the grating period-to-wavelength ratio, period-to-depth ratio, and angle of incidence. The evaluation of diffraction efficiency can help to choose the proper method to design and optimizing harmonic diffractive elements with the advantages of SDT’s simplicity and FDTD’s accuracy.

The sun acts as an obstacle to prevent the galactic cosmic rays from propagating from the direction of the sun to the earth. The detectors on the ground are used to observe the direction of the sun, and the absence of the number of cosmic ray particles is found at the center of the sun. This missing phenomenon is called solar shadowing. The magnetic fields of some planetary systems between the Sun and the Sun can be observed by near-Earth satellites. In this article, we utilize a simulation package developed to predict the single-day observational impact of the sun's solar shadow on galactic cosmic rays. And we compared with the magnetic field of the planetary system observed by the near-Earth satellite, and then established the theoretical prediction of cosmic ray solar physics.

At present, diesel generators have been widely used in many industrial fields around the world. Due to the characteristics of small loss, fast start-up speed and high power generation rate, diesel generator can effectively guarantee the emergency power supply of related fields and departments. In this paper, infrared thermography is mainly used to collect the infrared thermogram of diesel generator. Then, the infrared images are processed by bilateral filters, and then the infrared images are accurately located by classification. The experimental results prove the accuracy of this method.

Medical images are an important tool for doctors to diagnose conditions and treat diseases, and performing accurate medical image segmentation is the basis for disease diagnosis and treatment planning. Currently convolutional neural networks have achieved significant results in the field of medical image segmentation. However, considering the actual usage scenario, the model needs to run on resource-constrained devices, so the model needs to be lightweighted. ESPNet is a lightweight segmentation model structure. The ESP module effectively reduces the number of model parameters and computation, but in this paper, we note that directly reducing the number of model parameters by point-wise convolution will lead to the loss of model feature map information, which in turn leads to the degradation of model performance. In this paper, in order to further reduce the number of model parameters based on the ESPNet model, the number of channels of all model feature maps of the ESPNet model is halved, and in order to mitigate the resulting degradation of model performance, the feature maps of the model are grouped in the channel dimension using the modified Shuffle-ESP module. In order to avoid the loss of information interaction between different grouped convolutional feature maps, the channel information is artificially interacted using a channel shuffle mechanism before entering the atrous convolution of different dilation rate. It is experimentally demonstrated that the model in this paper decreases 54.99% and 54.99% compared to the original model parameters on two tumor data, and that the model performance metrics decrease by 1.59% and 1.76% respectively. The superiority of the proposed model in this paper is demonstrated through experiments.

The adaptive threshold based SIFT image registration algorithm (AT-SIFT) is proposed to improve the traditional SIFT algorithm which has the problems of large calculation amount, single matching constraint, and low registration accuracy. Feature points are detected and described by combining FAST operators with and SIFT descriptors. The FLANN algorithm based on adaptive threshold and the improved RANSAC algorithm are used to remove the mis-matching points and solve the transformational matrix. The experimental results show that the algorithm not only improves the matching speed and accuracy but also has relatively strong robustness.

Considering the growing security requirements, face security authentication (FSA) has attracted more and more researchers' attention, such as in financial payments, electronic devices unlock, secure access control, suspect matching and so on. At the same time, FSA is often seen as a kind of problem cost sensitive. This paper proposes a coarse-to-fine face security authentication method, which is named face authentication based on specific expression action mode (FSA_SEM) in cost sensitive learning. The highlights of this work are: (1) establish an identification dictionary to judge whether the visitor is member. (2) take a confirmation dictionary to confirm again whether the visitor is legitimate. (3) weight the dual dictionaries respectively by calculating the similarity relation matrix for adaptive weighting. The experimental results are encouraging, showing the effectiveness of FSA_SEM in different classification tasks. In practical application, the method can better deal with the cost sensitive FSA problem, and bring more convenient, safer and effective service for people's life.

In recent years, hand gesture has been widely used in human-computer interaction, virtual reality and other fields as an input channel. However, there are still some problems in gesture interaction, such as low resolution, occlusion and slow recognition rate, which lead to limited use scenarios. Therefore, based on the real scene of human-computer interaction, this paper builds a gesture recognition and object detection network according to the open source algorithms and frameworks, and constructs a human-computer interaction system based on gesture recognition. Experimental results show that the system can switch different modes according to hand movements and positions to realize interface interaction. The object detection and recognition in entertainment scenarios and the mathematical operations of virtual calculators in office scenarios are simulated respectively, which further improves accuracy and real-time performance of gesture recognition in the practical application scenarios.

Aiming at the difficulty of multi-target damage assessment in complex environment, a target damage assessment method based on Mask R-CNN is proposed. In the method, the pre-training weights are obtained by Mask R-CNN model, and the parameters of the Mask R-CNN backbone network are initialized according to the target size and damage scene. The initialized parameter of Mask R-CNN backbone network is improved based on the pre-training weight. The method solves the problem that a single target can’t provide accurate damage assessment effect. Through simulation experiments of multi-target damage, the reliability and accuracy of the method are improved obviously.

In this paper, the design and structure of a multi-functional weather environment simulation system for evaluating imaging detection is introduced. The system can be used to simulate common weather conditions such as rain, fog, and air flow and temperature variance, in a small spatial size, which covers a large spectral range from the visible to infrared band of 0.4μm~14μm. The key effects of various weather factors of affecting electronic imaging detection can be conveniently explored. Firstly, the overall system is designed, and then the main parameters are selected according to specific weather requirements. The main function and the accuracy requirement of each component in the system are determined. Through some processing methods such as turning and punching sheet metal, each functioned part of the system is designed and made and assembled. The function of the whole system is debugged. The liquid-crystal based plenoptic camera and the imaging system in the visible range are used to build the experimental platform. The experimental results show that the acquired simulation system already achieves the design aim, and thus can be easily used to explore the impact of complex weather factors on imaging detection.

With the rapid development of the information age, more and more image messy information appear in our lives. However, people prefer to focus on the information they are interested in. Based on this, we propose a method for extracting objects of interest from the interference information using U-net network. To achieve this goal, we specially design the dataset that the labeled images only retain objects of interest, so that the network model only needs to learn the feature information of the object of interest related to the task, which can extract and preserve the feature information of the most relevant objects in the different scene. The objects of interest can be reconstructed in different scenarios under small self-built datasets. The method avoids processing the global information of all objects in the scene, greatly reducing the storage and transmission of useless information, and will have far-reaching application prospects in object recognition, object classification, etc.

Aiming at the shortage of matching accuracy of ORB algorithm in image matching stage and the poor robustness of GMS algorithm under repeated texture conditions, this paper proposed an improved GMS image feature matching algorithm based on BEBLID descriptor. This algorithm firstly uses the ORB algorithm for image feature point extraction, secondly describes the feature points with BEBLID descriptor, after that discards false matching pair preliminarily by brute force matching. In order to improve the matching accuracy of the algorithm, two algorithms GMS and PROSAC are combined on this basis to obtain better matching pairs. The experimental results show that the algorithm has uniform extraction of feature points and high matching accuracy for different image feature matching, and its correct rate is improved by 10.97 percentage points to the GMS algorithm, which can meet the demand of large parallax image matching and improve the accuracy and efficiency of target acquisition in vision tasks.

Image super-resolution is widely used and research on its algorithm are also developed rapidly. In recent years, deep learning has been introduced into the process of image super-resolution, and the output image has been improved effectively. On this foundation, this paper proposes to reconstruct the mapping layer with a method that reduces the dimension of extracted features at first and then extends them at last. Also, a deconvolution layer is used at the end of the network to map an uninterpolated low-resolution image to a high-resolution image directly. Besides, smaller convolution kernels and more mapping layers are used in this algorithm. Comparative experiments demonstrate that the above methods can accelerate the speed and increase the effectiveness of the image reconstruction by optimizing the network structure and reducing the computational complexity.

In recent years, CRNN has been widely used in computer vision and has achieved remarkable results in the direction of text recognition. CRNN is a convolutional recurrent neural network structure, which is mainly used in image sequence recognition problems. The CRNN network model implements variable-length verification, combining CNN and RNN networks, using a bidirectional LSTM cyclic network for time series training, and then introducing a CTC loss function to recognize variable-length sequence texts. In the field of Tibetan text recognition, based on end-to-end recognition, it is usually to recognize a line of text. Due to the special structure of Tibetan syllables, the components of Tibetan characters can be split, and end-to-end recognition can be applied to the study of Tibetan single-character recognition. In this paper, a new split-based method is used for end-to-end recognition of single characters in Tibetan ancient books and single characters in Tibetan handwriting using the CRNN model, and a good recognition effect is achieved. It provides a new method for Tibetan character recognition research.

Based on the image chunking processing theory, this paper proposes an improved ORB-based UAV image stitching algorithm that takes the pyramidal optical flow characteristics into consideration to overcome the weaknesses of the traditional algorithm (proneness to feature point aggregation, uneven distribution, and low feature matching accuracy). Firstly, the ORB algorithm is used to detect the feature points using the image chunking method of constructing moving masks. Subsequently, the non-maximum suppression method is used to reject the feature points aggregated in each mask, and the Hamming distance is used for feature matching after traversal. Following the step, the pyramid optical flow method is used to correct the motion displacement vector of the feature points and reject the mis-matched pairs after matching. Finally, the RANSAC algorithm is used to filter the redundant pairs to further improve the accuracy, and then the image is stitched together using the optimal stitching seam strategy and the fade-in-and-fade-out fusion algorithm. Overall, the aforementioned procedure increased the alignment rate to approximately 97% in the alignment stage, achieving a better alignment accuracy and a more accurate stitching effect.

Aiming at the problem of segmentation and extraction of cropland parcels in remote sensing images with complex background, a method of segmentation and extraction of cropland in high-resolution remote sensing images is proposed. Firstly, the linear spectral clustering (LSC) algorithm is applied to the remote sensing image to obtain the segmentation results of superpixel blocks in the target area and the background area; then, the maximum similarity region merging algorithm (MSRM) algorithm is used to merge the superpixel blocks of two different areas separately, which effectively reduces the phenomenon of under-segmentation and over-segmentation of the image and obtains the binary image containing the cultivated land parcels and non-cultivated land parcels. Based on this, the total arable area is calculated using MATLAB. Finally, in order to verify the correctness and effectiveness of the proposed method, the remote sensing image data provided by Beijing Guosheng Xingmai Information Technology Co. The simulation results show that the proposed method can effectively segment and extract remote sensing cropland images.

Aiming at the precise measurement of numerical aperture which is a key optical parameter of gradient index (GRIN) lens. An incremental measurement method based on image recognition is proposed. Based on the definition of numerical aperture of GRIN lens by geometrical optics, the sinusoidal value of the maximum receiving angle is measured by imaging method. Incremental measurement means the GRIN lens images the light-emitting mesh twice, and calculates the sinusoidal value of the maximum receiving angle with the object heights of two images and the relative displacement. Two object heights are obtained by computer image recognition algorithm. According to the characteristics of mesh and the edge of field of view on the image, the complete object height in the longitudinal truncation is extracted which includes the central object height and the edge object height. The central height is determined by all the fringes on the truncation line, and the edge height that equals to the distance between the edge of field of view and the outermost fringe is determined by the extrapolation algorithm. The error analysis of the measurement system is carried out by establishing the error model, and the main source of the system error is determined. The experiment proves the stability of the sub-measurement method meets the practical needs. The measurement of the numerical aperture of a GRIN lens, which combines image recognition with incremental measurement, can reduce the measurement error caused by absolute measurement and discrimination of our eye, and it is simple, fast and accurate.

According to the actual requirements of real-time target tracking processing and display of short wave infrared camera with CameraLink interface, an embedded image processing system based on FPGA is designed. The LVDS and GTX ports of FPGA are used to realize CameraLink interface and 3G-SDI interface, it not only improves the integration of the system, but also saves the cost. The adaptive dual platform histogram equalization algorithm is adopted and implemented, which effectively improves the display effect and subsequent processing quality of short wave infrared image. The paper briefly analyzes the algorithm principle of adaptive dual platform histogram, puts forward the adaptive calculation method of dual platform value, studies the implementation method of the algorithm based on Xilinx's FPGA, and gives the solutions of key modules such as adaptive dual platform histogram equalization, image enlarge and 3G-SDI display in FPGA. The experimental results show that the designed embedded image processing system gives consideration to real-time and effectiveness, and realizes the functions of image amplification and high-definition display; The algorithm can calculate and adjust the upper and lower platform values of the histogram in real time according to the scene within 400 μs, effectively enhance the contrast of the image and suppress the background noise. The system has the advantages of strong algorithm adaptability, simple implementation, high integration and low power consumption. It can also be effectively applied to the image acquisition, processing and display of medium wave and long wave infrared cameras.

This design is an anti-fatigue driving system based on image recognition, using Raspberry Pi as the core processor. The image acquired by the video image acquisition unit is processed by the optimized Imutils library, and the model of the machine learning library Dlib is used to locate the key points, extract the driver's facial features, and calculate the eye aspect ratio through the algorithm. This parameter is used to judge whether the driver is fatigued, cooperate with the display screen and the light of the RGB small light to display the fatigue state, and issue a warning to the driver through the voice warning module, so that the driver can stay awake while driving and avoid traffic accidents. This design is implemented by an embedded system, which has the advantages of being mobile and deployable to the real driving environment and hardly disturbing the driver's driving. After the test of the simulated driving environment, the face detection and eye feature detection work normally, the anti-fatigue driving function has high accuracy in identifying the driver's fatigue state, and the system has high upgrade potential.

In high-density stereo correspondence, matching cost aggregation is one of the key links, and the non-local cost aggregation algorithm based on tree structure has attracted people's attention in recent years. On this basis, the clustering algorithm based on non-local tree is studied, and a new cyclic tree structure is given. Each pixel in the image is rooted in eight adjacent tree structures, which we call the first level. We made up the second floor with eight adjacent first floors. Since the algorithm has a natural location in the image pixel structure, it does not need to perform any operations on it. The performance of Middlebury's data set is evaluated, and the results show the application effect of the algorithm proposed in this paper in the current most advanced clustering algorithm.

With the revival of traditional Chinese culture, a wave of fine brushwork painting as a popular element has been set off, but there are few studies on fine brushwork painting coloring. Based on Generative Adversarial Networks, this paper presents an automatic coloring method for fine brushwork painting. The generator uses U-NET network to solve the underlying information sharing problem. PatchGAN was used to ensure the performance of details. Smooth L1 loss is added to the loss function to make the generated image color transition smooth and more realistic. In order to solve the problem of single color drawing, color guidance information is added to the line drawing, so that it has the function of interactive color painting, and the specified color is realized in the line drawing. The experimental result shows that the automatic coloring method can quickly color line drawings, and the generated simulated fine brushwork paintings have distinct colors. The unique interaction mode improves the diversity of coloring effects.

In this study, we propose a novel scheme for human action recognition that combines the advantages of both spatial and temporal representations. We use shape context (SC) as pose representation in the spatial domain, and explore the temporal feature by taking into account the correlation between sequential poses within an action. In terms of the pose matching with high-dimensional data, we provide a fast matching algorithm using pyramid match kernel (PMK) based on adaptive partitioning. Additionally, this work introduces a size-pruning based longest common sub-sequence (LCSS) alignment algorithm for action sequence matching, and obtains the final cost via the decision-level fusion. Experimental results prove the viability and superiority of the fusion of two descriptors and the proposed method outperforms the majority of state-of-the-art methods on Weizmann and KTH datasets.

Beethoven's Piano Sonata No.5op10. 1 is one of the representative works of Beethoven's early piano sonatas, creating a heroic work that belongs to Beethoven's uniqueness. This article uses music visualization analysis technology, combined with the relevant musical background information of Beethoven Piano Sonata No.5. The audio performance of the four pianists was selected as the research content, and the overall velocity of the work, the rhythm of the excerpted passages and the inner elastic rhythm were visualized through the Vums.net platform, and through a series of different pictures, the comparison of the speed velocity diagram and the IOI deviation analysis chart from the macroscopic to the microscopic, the trend of limiting the elastic expansion and contraction of the rhythm can be drawn, and the embodiment of the composition structure can be more and more obviously seen from the figure. The self-expression of emotions is becoming more and more restrained, and more and more intoxicated with the expression of the original spectrum.

In order to improve the dictionary sparsity of image compression perception, enhance the noise suppression ability of the dictionary and improve the quality of image reconstruction, a k-singular value decomposition dictionary learning algorithm based on decision-making gray wolf optimization is proposed. In this method, the decision-making grey wolf optimization algorithm is introduced into the atom update stage of k-singular value decomposition dictionary learning algorithm to further optimize the atoms, so as to effectively improve the sparse representation performance of the dictionary. At the same time, a priori experience is introduced into the grey wolf optimization model to guide the optimization direction-making of the wolf pack, and the motion dimension of the wolf pack is reasonably limited. The experimental results show that in the two image data sets, the dictionary trained by the dictionary learning algorithm optimized by the decision-making wolf has stronger sparse representation ability of the image, better image reconstruction effect, and has certain significance to suppress the noise of the image itself.

Multi-color resolution is an indispensable method in biological experiments. However, in the actual experiment, due to the wide fluorescence spectrum and the proximity of multiple fluorescence wavelengths, the crosstalk rate is high. As a result, researchers often cannot distinguish multiple different markers with the naked eye. In order to solve the problem of indistinguishable multi-target markers and limited resolution, we propose a multi-color super-resolution imaging method based on AOTF-assisted super-resolution optical fluctuation imaging. The emission intensity of each fluorophore is different under different excitation light. It has the characteristic of specific excitation spectrum. By using AOTF to scan a wide range of excitation wavelengths, different fluorophore groups are linearly superposed in the wide-field image. In this paper, the abundance distribution of each fluorophore is calculated by combining the separation algorithm for solving the linear equation with multicolor fluorescence microscope. The abundance distribution is used as the image of each color channel to achieve multicolor resolution. In addition, super resolution imaging can be achieved by combining with fluorescence microscope. Simple numerical simulation results show that the image resolution is improved by √2 times and low spectral crosstalk rate is achieved by combining the two methods.

When using satellite remote sensing images to identify vegetation areas, only using spectral information to extract information will cause the phenomenon of “same spectrum foreign matter”. Aiming at the limitations of the above methods, this paper proposes a vegetation area identification method based on vegetation index and texture features. Firstly, the normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) are used to classify the vegetation areas in satellite images. Then, the Gray-level Co-occurrence Matrix (GLCM) of the image is calculated and the image texture feature information parameters are superimposed. Based on the initial division result, the boundary of the vegetation area is accurately identified. Compared with the method of using only spectral information, the extraction result of vegetation area in this paper obtains the research result with higher accuracy.

The feasibility of the imaging reconstruction in the infrared domain according to the basic reconstruction method of Zernike moment in the visible range, is discussed. Due to the orthogonality of Zernike polynomials, any image within an unit circle can be represented by Zernike moments. Therefore, we propose an approach to reconstruct the target images inside the unit circle in the infrared wavelength range. Firstly, according to the light intensity distribution of the infrared object selected, a relatively clear image can be successfully reconstructed based on the Zernike moment reconstruction. And then according to the liquid-crystal Shaker-Hartmann wavefront measurement technology, the phase distribution corresponding to a tailored incident wavefront of the target radiation can be successfully recovered according to the refocusing spots of the liquid-crystal micro lenses over the CCD. Through combining wavefront measurement and recovery and further the infrared image reconstruction of the target, both the theoretical basis and the algorithm for recovering wavefront according to the intensity distribution of the infrared image are provided.