Research On Close-Range Image Stitching And Application To Rail Fastener Detection For High-Speed Railway

Rail fastener is the middle link connecting parts of rail and sleeper,which paly an important role of connecting rail and sleeper and stopping a movement in the vertical and horizontal directions.Its structure stability is an important guarantee for train's running.The accurate detection and finding out which track fastener is missing have important significance for the safe operation of the train.Using the high resolution digital images acquired by the close-range photogrammetry device,we explore a set of method for high speed railway track close-range image stitching,realization technology and fastener missing recognition.The development of high speed railway track close-range image mosaic system and track fastener detection system provides a new support technique for detecting whether the rail fastener is missing or not and high-speed railway track close-range image stitching in a large and consequent area.Combining with the characteristics of high speed railway track close range image,the thesis uses the method of image matching based on feature matching to realize the close range image stitching.In this thesis,Harris algorithm and SIFT algorithm is used for matching and extracting feature points of close range image.Moreover,we use the RANSAC method to correct and tick the error of the matching results,finally getting appropriate numbers of homologue points and providing adequate and reliable data for image stitching.In order to obtain precise and stable model of image stitching and transforming among the adjacent high-speed railway track images,we select the good data first and then adjust it.Using the experimental images,the intermediate mosaic method is used to complete the perfect image mosaic for high speed railway image.After obtaining the stitching close-range image for high-speed railway track,the thesis improve the MeanSift method to locate the rail fastener instead of according to the relation between the rails,sleepers and fastenings of the three traditionally.At the same time,in order to avoid large quantity of repeated calculation and improve the detection effliciency,GrabCut algorithm is used to segment the high speed railway track close range images.HOG is used to describe the characteristics of the rail fastener and the nearest neighbor algorithm based on Hamming distance is used to identify and detect the fastenerAccording to the research of image stitching and the detection algorithm of rail fastener,the thesis develops a high-speed railway track close-range image stitching system and the track fastener recognition and detection system.The close range image collected by the close range image acquisition device is used as the test data for the system,and the results shows that the system can effectively complete the image stitching and the detection of the rail fastener missing detection.