Research On Track Fastener Detection Method Based On Morphological Processing And HOG Features

Chinese railway has grown from scratch,allowing the world to witness the rapid development of Chinese economy.Nowadays,railway travel has become one of the most important means of transportation for Chinese people,and ensuring the safety of railway travel has become a top priority.Rail fasteners are an important part of tightly fixing rails.With the development of computer technology,digital image processing and computer vision technology are gradually applied to the detection of rail fasteners,which makes rail detection gradually automated.The automatic detection of rails has many advantages,such as low labor costs,high detection efficiency,and intelligent data storage.In this paper,based on the detection of rail fastener defects,based on the research of rail inspection technology at home and abroad,a set of design methods for fastener defect detection methods based on morphological processing and HOG features are proposed.The design scheme of this method is divided into two parts: the research on the positioning algorithm of rail fasteners,and the research on the classification of fasteners based on the extracted features.(1)The track image collected by the track inspection vehicle is divided into left and right tracks.Each track image is distributed with several fasteners.In order to facilitate the detection and classification of fasteners,it is necessary to locate and extract the fasteners on the track images.In this paper,from coarse to fine,the cross positioning method and the morphological processing projection method are used.The positioning method based on the crosscorrosion expansion morphology method can effectively reduce the influence of lighting factors and achieve accurate positioning of the fastener area.(2)Based on the positive and negative samples,extract HOG feature images of the extracted fastener sample images.Before the classification,the feature vectors are subjected to PCA dimensionality reduction,and SVM performs classification training on the extracted sample feature information.Through the symmetry detection method of defective fasteners,the rate of missing detection of fasteners is further reduced.Finally,test samples are used to verify the classifier.