Research And Application Of Automatic Detection Method For Train Parts Abnormal Based On Image Processing Technology

With the continuous development of China’s railway trains,railway trains bear the major responsibility of national transportation,and the safe operation of railway trains has received more and more attention.The train structure is complex and there are many parts.Manual inspection of the train requires a lot of time and manpower.However,the widely used image-based train component fault detection system can only detect faults on certain key parts of the train,and cannot complete the inspection of the train parts.Therefore,it is indispensable to research on the detection of train parts.In order to effectively complete the fault detection of small,this paper takes the train brake shoe bolts,brake shoe bolts and sand pot parts as the research object,and uses the ralated image processing to realize the automatic detection of the brake shoe bolts,brake shoe bolts and sand pot parts.The main content of this article is divided into the following:1.A method for identifying the lost bolts of train brake shoes based on GS-SIFT feature descriptor is established.Firstly,the brake shoe image is preprocessed.Because of the defects of SIFT,it is concluded that the GS-SIFT algorithm has stronger adaptability by contrastive analysis in different scenes,especially in feature matching of the brake shoe image and the template brake shoe image.Adaptability,then the exact matching points obtained by the RANSAC algorithm.The rotation and scaling correction of the image of the brake shoe to be tested,the target area is segmented.Finally,the normal bolt image is used as a template to traverse the image of the target area to determine whether the bolt is lost.The algorithm runs stably with an accuracy rate of 96%.2.Canny-based identification method for open angle fault of train brake shoe latch is estalished.Firstly,according to the regional structure of the fixed position of the brake shoe bolt and the bolt,the target area of the split pin is preliminarily positioned.the pre-latch area is pre-processed.the reason for the edge detection of the pin area by Canny edge detection algorithm is explained,the line segment detection algorithm by hough transform is detected Pin pin,calculate the pin angle and analyze the result.The algorithm is designed to meet the recognition rate requirements.3.The sand tube image anomaly detection and chain anomaly detection method based onimproved LBP characteristics are established.Firstly,the Canny-hough algorithm is used to correct the tilting of the sander image.The improved LBP template feature extraction is proposed,which can locate the target area of the sand tube well.Traverse the target area in order to find the pixel jump area and realize the crack check.Based on the fixed position structure of the middle sand pipe,the joint of the sand pipe and the chain are processed by threshold,and detect the abnormality of the chain region.The comprehensive evaluation index obtained by the algorithm is 97.7%,which meets the expected target.