Fault Detection Algorithm Of Support And Suspension Devices Of Catenary In High-speed Railway

Catenary is an important part of electric power supply system of high-speed railway, once the fault happened, it will cause a serious impact on the transport capacity of the railway. According to relevant statistics, catenary fault of traction power supply system accounted for about 70% of the total of all kinds of fault,and the manual inspection of OCS is time-consuming. In order to achieve efficient intelligent inspection of catenary, the railway company put forward the construction of high speed railway power supply safety detection and monitoring system(6C system). As an important part of 6C system, contact network security inspection device(C2 system) detects state of catenary by camera monitoring device installed in the locomotive cab. However, huge number of video image obtained by camera and complicated and variable background of image and image blurring caused by adverse weather and other factors will lead to recognition difficulties of the key components of the catenary, so a target detection technology that can intelligently distinguish Abnormal conditions of key components of catenary is urgently needed to reduce the human workload.In this paper, according to the huge number of catenary images obtained high speed railway contact network security inspection device(C2 system) and complex background of image and other issues,the exploratory research has been carried out about the relevant algorithm of machine vision intelligent identification of abnormal condition of catenary supporting device. Firstly, according to the structure and characteristics of the contact system, four kinds of sample database support suspension device of catenary that include normal, birds faults, pollutants faults and dropper defect has been established. At the same time, by using noise interference and effectively salving the illumination variation in the catenary fault detection by the method of image enhancement, and correctly identified the catenary pillar plate by the method of template matching. Secondly, the triangle area of cantilevers device of catenary is quickly and accurately automatic acquired by using a combined target tracking algorithm of Kalman filtering and Meanshift algorithm. And then, features are effectively abstracted from the image of key components of catenary by using wavelet moments. Finally, the abnormal state of key components of catenary is intellectually classified and identified by RBF neural network, and the hanging defect detection is achieved by Hough transform method.In practice, a lot of inspection images of catenary of the Beijing-Shijiazhuang high-speed railway are analyzed by the methods in this paper. By repeatedly comparing experiment, the methods can real-time accurate positioning the triangle area of cantilevers device of catenary and effectively extract the image features of catenary. The classification and recognition of the abnormal situation of the key components of the catenary are achieved, which include birds faults, pollutants faults and dropper defect and so on. Detecting platform of abnormal situation for the key component of catenary in high-speed railway is simply designed. The project practice shows that the method proposed in this paper is fast and accurate about fault detection of support suspension device image of catenary in high-speed railway, and effectively improves the work efficiency and reduces the manual workload. The method can pick out abnormal images of a large amount of images under complex background of key components of catenary and can provide a reference for intelligent and real-time catenary inspection in high-speed railway.