Research On Abrasion Defect Detection Technology Of Urban Rail Transit Pantograph And Catenary Based On Computer Vision

Pantograph-Catenary system is the key system of the power supply system of urban rail transit,which is responsible for the import task of transmitting the electric energy of traction network to electric locomotive.Pantograph and catenary are the key components of pantograph-Catenary system,and due to their bad working environment and strong electrical and mechanical effects,pantograph and catenary are prone to produce various diseases,the most common of which is abrasion.At present,the most commonly used detection method of subway related departments is still manual detection,which is timeconsuming,labor-consuming and inefficient.The main research content of this paper is to use computer vision related technology,through deep learning,image processing,binocular vision and line structured light and other methods,to realize the modeling and disease analysis of the wear characteristics of pantograph and catenary,so as to provide technical support for the automatic and intelligent detection of pantograph and catenary wear diseases of urban rail transit.The research content of this paper mainly includes the following aspects:(1)Pantograph region extraction and disease recognition based on deep learning.Aiming at the detection of the surface defects of pantograph slide plate,the network structure and parameters are adjusted based on Faster R-CNN framework by adopting deep learning theory method,and the Soft-NMS technology is applied to improve detection results.After the model training,the pantograph region is extracted from the original image and the disease category is identified.And after the model test stage,the method used is verified to have high accuracy;(2)Wear analysis of pantograph carbon slide based on image processing.On the basis of the pantograph area and general disease types extracted in the above chapter,the image processing technology is applied to realize the extraction and fitting of the wear curve of pantograph slide plate.According to the mathematical definition of different types of disease,the quantitative analysis of the actual collected pantograph slide plate disease is carried out.After comparing with the manual measured data,the detection accuracy is verified to be high;(3)Three-dimensional reconstruction and disease analysis of catenary surface based on binocular vision.In view of the limitations of single camera detection method,a stereo imaging detection system is built in the laboratory environment based on the binocular vision theory.After the system calibration,the contact network image pair is collected.After image correction,stereo matching and conversion calculation,the threedimensional reconstruction of the contact network surface is realized,and a disease definition method based on the sampling wear curve is proposed for different types of abrasion of catenary was analyzed;(4)Three-dimensional reconstruction of catenary surface based on line structured light.Aiming at the problems of the binocular vision scheme,such as the environment interference,the difficulty of feature extraction and matching,the line structured light scheme is adopted.After the system parameters are calibrated,the light plane calibration and the light strip center extraction are carried out,the three-dimensional coordinates of the light strip center point can be obtained with high accuracy.Finally,using the point cloud data generated by simulation,the three-dimensional reconstruction of the contact network surface is carried out.