An improved image wrapping method that preserves visually prominent features while resizing image into arbitrary scaling ratios is proposed. In this method, a salient context aware model is adopted to construct salient map to sign prominent image features at the aspect of texture and semantics. The salient map contains not only prominent texture with strong gradient changes but salient region with significant semantics. This model is implemented with an end-to-end convolutional neural network which maps input image to corresponding feature salient features. The proposed method simultaneously learns detailed and salient context and fuse both features as the salience mask. Then the seam carving method is utilized to implement image resizing with arbitrary ratios combined with the salience mask. Experimental results of the proposed and classic seam carving method indicate that our formulation has more robust and effective performance.
Due to the wide application of welding in the modern industry, effective detection of weld surface defects is an important measure to ensure the quality of components, monitor the Service life of the structure, and ensure the safety of users. However, there are wrinkles and stains on the weld surface, which makes detection difficult. Based on the dynamic detection of pulsed eddy current thermography, a multi-feature fusion algorithm of infrared features and visible information is proposed in this paper. In dynamic detection, the relative position of cracks in the field of view is constantly changing, therefore, the thermal image sequences are spatially aligned to obtain the transient thermal response curve in static mode. Feature extraction and dimensionality reduction of thermal image sequences are carried out in time domain. The processed data is fused with the visible image features, and classified in pixel-level applying the pattern recognition network. The experimental results show that the proposed algorithm can effectively suppress the noise caused by weld texture and surface stains, and obtain more clear and accurate defect information. All 21 weld surface defects can be detected, and the detection ability is greatly improved.
The ultrasonic detection technology of railway locomotive wheels is of great significance to the safety of train operation. However, the current detection technology relies on experts ' manual operation, which has the disadvantages of low accuracy and high cost. In this paper, a detection method based on improved Exceeding YOLO Series in 2021 (YOLOX) is proposed. Firstly, a series of processing such as cutting and rotating the ultrasonic B-scan image obtained by LU system is carried out to obtain the B-scan data set of wheel ultrasonic defects after data enhancement. Secondly, we add an adaptive spatial feature fusion block (ASFF) to the tail of the Neck module of the YOLOX detection algorithm, and further improve the multi-scale feature map. Finally, the original BCEWithLogitsLoss in the loss function is replaced by FocalLoss to improve the ability to distinguish defects from similar backgrounds. The test results show that the detection rate of each type of the improved YOLOX model is more than 90 %, the false negative rate and false positive rate are less than 10 %, and the detection speed is 17 ms. Compared with the original YOLOX network and other mainstream detection models, the improved YOLOX model has the best detection performance. This study provides a new idea for the automation of ultrasonic testing of train wheels.
Three dimensional (3D) shape reconstruction based on structured light technique is one of the most crucial and attractive techniques in the field of optical metrology and measurements due to the nature of non-contact and high-precision. Acquiring high-quality 3D shape data of objects with complex surface is an issue that is difficult to solve by single-frequency method. However, 3D shape data of objects with complex surface can be obtained only at a limited accuracy by classical multi-frequency approach. In this paper, we propose a new robust deep learning shape reconstruction (DLSR) method based on the structured light technique, where we accurately extract shape information of objects with complex surface from three fringe patterns with different frequencies. In the proposed DLSR method, the input of the network is three deformed fringe patterns, and the output is the corresponding 3D shape data. Compared with traditional approach, the DLSR method is pretty simple without using any geometric information and complicated triangulation computation. The experimental results demonstrate that the proposed DLSR method can effectively achieve robust, high-precision 3D shape reconstruction for objects with complex surface.
Trains are an important means of transportation in China. With the popularity and speed increasement of trains, safety issues have received wide attention. The daily safety inspection of high-speed trains becomes crucial, the abnormal target detection for key component that is at the bottom of the train is an important part. Most of alarms which detected by machine vision based on global comparison method are false, thus, it cannot effectively monitor the key component. In this paper, the digital image processing technology is adopted to detect abnormal targets of the three key components, the steeve, the shaft cabinet and the core plate, and an algorithm is presented to detect these components of different types. The key component images are extracted from the train image by template matching. Traditional template matching method is often failed due to the strong reflection happened in the process of train bottom imaging. Therefore, the matching method based on structural similarity is proposed, which greatly improves matching accuracy. Finally, the abnormal target detection of three different key components of locomotive is realized by edge detection, shape detection and contour matching.
The detection of rail surface defects is of great significance for railway safety. To detect the rail surface defect, the laserinduced ultrasonic rail propagation model is established by the finite element method. The intrinsic relationship between the defect depth, of the defect on rail surface and the acoustic surface wave is investigated by discussing the variation of the reflected wave and the transmitted wave both in the time and frequency domain, respectively. Quantitative evaluation of defect depth is given based on the energy of the reflected and transmitted wave, which providing a promising theoretical way for the estimation of the rail surface defect feature.
Edge detection is a crucial task in image processing. Owing to the similarity in property between edges and noise, which demonstrates abrupt changes in image grayscale values, traditional edge detection methods are insufficient in detecting weak edges. Therefore, a local multi-threshold fuzzy inference method (LMFI) is introduced. Considering the binarization processing prior to conducting a fuzzy inference, to retain more edge information, a local threshold processing method and a triple threshold processing method are proposed. To reduce noise interference, an improved sigma filter and an improved fuzzy inference strategy are presented. The experimental results show that the effect of weak edge detection is improved by LMFI, when compared to conventional methods such as the original fuzzy inference algorithm and Canny edge detection algorithm.
X-ray testing is based on the attenuation of X-rays when passing through matter. Image detectors acquire the X-ray information which is defined by the local penetrated wall thickness of the tested sample. By X-ray absorption in the detector and following read-out and digitization steps a digital image is generated. As detectors a radiographic film and film digitization, a storage phosphor imaging plate and a special Laser scanner (Computer Radiography - CR) or a digital detector array (DDA) can be used. The digital image in the computer can then be further analyzed using many types of image processing. In the presented work the automated evaluation of wall thickness profiles are investigated using a test steel pipe with 9 different wall thicknesses and various X-ray voltages and different filter materials at the tube port and intermediate between object and detector. In this way the influence of different radiation qualities on the accuracy of the automated wall thickness evaluation depending on the penetrated wall thickness of the steel pipe was investigated.
In this paper, a detection system which combined with the line-structured light scanning technology that can visualize the pantograph surface and automatically locate and evaluate the wear condition of the pantograph is proposed. In this system, a three-dimensional camera and a line laser generator are used for acquiring surface data of pantograph. Then Laplacian filtering is used to smooth the data. Proceeded data and standard model are registered by using distance constrained ICP algorithm which combined with geometrical symmetry of pantograph. In this paper, a method to locate and quantify the wear area of the pantograph is proposed, which provides a feasible solution for inspection and visualization of pantograph wear.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.