Research On Track Damage Algorithm Based On Image Processing

Railway is the backbone of the transportation system carring a large number of passenger and freight transportation.With the rapid development of China’s economy,the characteristics of a strong railway country are gradually highlighted,but the railway tracks and other infrastructures are also under tremendous pressure.Moreover,track damage brings risks and hidden dangers to the transportation,so the detection of track damage is essential.But in our country,it is still in the stage of manual inspection,which consumes a lot of manpower and material resources.Based on the image processing,this article designs a series of algorithms for track damage detection by Open CV from the perspective of non-destructive inspection.The simulation is carried out based on i7-9750 H computer.The main work contents are as follows:Research on optimization of image preprocessing algorithm.Classic BM3 D algorithm is transformed into two-dimensional and one-dimensional transformation model,and the 3×3search window is enlarged to 6×6 to shorten the algorithm operation time,and ensure t the effectiveness and high quality of the algorithm processing results.The improved BM2D+1D algorithm preprocesses the track damage image,and obtains the processing results of different noise density images.The average operation time of the BM3 D algorithm is 48.638 s,and those of the improved BM2D+1D algorithm is 32.196 s.The operation efficiency is improved by33.8%,and the peak signal-to-noise ratio is improved by 0.388 d B.Research on optimization of edge detection algorithm.Using log-gamma operator by enhancing image edge,the Canny operator program is improved,which enhances the edge information and weakens the interference pixel information.It not only improves the efficiency of the operator,but also filters out most of the false edge information.The average operation time of the algorithm is reduced from 0.018 s to 0.003 s,and the operation efficiency is improved by 81.4%.The optimal threshold is obtained by comparing with the reference value of edge information ratio of the original image 0.0273 and combining with the visual effect.Research on optimization of shape detection algorithm.By using chain code to optimize Hough transform,the edge contour of the binary image of the previous processing result is described,which filters out some false edges and meaningless pixel information.Then the track scar shape is detected,and the scar location are marked.The operation time is reduced from1.086 s to 0.862 s,and the operation efficiency is improved by 20.6%.The effectiveness of the optimized Hough transform is verified by comparing with the reference value of edge information of the original image of 0.0273 and combining with the visual effect.According to the research results of visual effect and image quality evaluation system,including PSNR value,SSIM value and histogram evaluation,the effectiveness and universal applicability of this algorithm for track damage detection are verified,which provides the algorithm foundation for track non-destructive detection.In the future research work,it is necessary to continue to improve the operation efficiency of denoising algorithm to achieve the standard of short-term detection.