Research On Key Technologies Of Rail Defect Detection Based On Image

Rail surface defects cause potential safety hazards for railway transportation.It is of great significance to detect defects in time in order to ensure the safety of train operation.Rail defect detection technology based on image processing has the advantages of high real-time,strong anti-interference and high precision,which is the development trend in this field.This paper focuses on the related technologies of image-based rail defect detection.The research work is divided into the following points:(1)By introducing the basic methods and characteristics of collecting the original track image by rail inspection vehicle,analyzing the influence factors of the original image on the subsequent rail defect detection,the method of extracting the rail surface area is determined.Firstly,the original track image is preprocessed by histogram equalization.According to the gray distribution characteristics of the collected rail image,the line detection algorithm based on Hough transform is used to extract the rail surface area.The method can extract surface area of the rail quickly and accurately,and the simulation results verify the effectiveness of the method.(2)A weight factor seeker optimization algorithm(WFSOA)based on random weight strategy and asynchronous value factor is proposed.Combining with the 2D-Otsu algorithm,a method of 2D-Otsu rail defect segmentation based on WFSOA is proposed.For SOA,by introducing random weighting strategy,the algorithm can avoid falling into local optimum,which is beneficial to enhancing the diversity of the algorithm and improving the global search ability of the algorithm.Meanwhile,by introducing improved value factor,the self-interest ability of the algorithm is properly enhanced,and the altruistic behavior ability is weakened to achieve the effect of convergence to the global optimal.Combining WFSOA with 2D-Otsu,the optimal threshold solution in 2D-Otsu is transformed into WFSOA to solve the optimal solution problem.The distance measure function is used as the objective function to realize the image segmentation of rail surface defects.(3)An improved Canny edge detection algorithm is proposed.The image is filtered by5*5 Gaussian smoothing filter.Four directions are improved to four sectors for non-maximum suppression.The upper and lower thresholds of Canny algorithm are solved by VEM algorithm.This improved method not only reduces the noise of the image,but also enhances the applicability and practicability of Canny algorithm when calculating the threshold,and improved the edge detection effect of Canny algorithm.The improved algorithm is applied to rail defect edge detection,which improves the effect of rail defect detection.(4)The binary image of rail defect obtained by image segmentation and edge detectionis used to describe the features by geometric and shape features.The characteristics of the three defects of the rail scar,crack and wear are calculated,and the characteristic parameters of each defect are obtained.The features are extracted by describing the characteristic parameters.K-neighborhood classification is used to identify and classify rail defects.The simulation results show that the method has a high ability to identify defects,and there is no missing detection,which meets the requirements of actual detection.