Research On Fault Detection Algorithm Of Contact Network Pipe Cap And U-shaped Holding Hoop Based On Machine Vision

The catenary power supply system,as an important part of the electrified railway,directly affects the normal operation of the railway system.Therefore,regular inspection of catenary power supply system and timely detection and removal of hidden troubles are of great significance to ensure the normal operation of electrified railway.In order to solve this problem,the Railway Department of China has formulated the “General Technical Specification of the Power Supply Safety Detection and Monitoring System(6C System)for High-speed Railway”,which collects video and image information of various components of the contact network power supply system through cameras,and then automatically analyzes these video and image information to realize the patrol inspection of the catenary power supply system.The contact network suspension state detection and monitoring device(4C)in the system is difficult to realize automatic recognition due to the complicated situation of the field device failure and the huge amount of image data.In view of the above problems,this paper uses image processing technology to carry out in-depth algorithmic research and improvement on the problems of the linear detection of the wrist of the supporting suspension device,the accurate positioning and the automatic fault identification of the cap,the accurate positioning and the automatic fault identification of the U-shaped hoop in 4C image automatic identification.The main contents are as follows:Firstly,the traditional methods of linear detection of supporting suspension devices is studied,and onthe basis of them,a wrist-line detection algorithm based on geometric features is proposed.The algorithm is based on the feature that the silver galvanized wrist arm of the supporting device is divided into a plurality of regions by the sleeve,and the straight line corresponding to each connecting region is found according to the nearest pixel method,then the straight lines of the regions of the same wrist are merged into one,finally obtains the corresponding straight lines of each wrist arm.The experiment shows that this straight line detection method is more suitable for this situation than the traditional straight line detection method.It can detect the wrist arm straight line completely and has better robustness,and the test results lay a foundation for the accurate positioning of the pipe cap and U-shaped hoop.Secondly,on the basis of the straight line detection results,the wrist arms with pipe cap are determined by using the different features of the wrist arm with pipe cap and the wrist arm without pipe cap.Then,according to the difference between the cap end and the non-cap end of the wrist arm with pipe cap,the precise location of cap area is located by pixel-by-pixel search method.According to the cap characteristics obtained by location,a limited angle rotation invariant HOG feature extraction algorithm suitable for the cap identification in this paper is proposed.,and the cap fault classification recognition is realized by combining the improved HOG algorithm and SVM support vector machine.Experiments show that this method has good accuracy,speed and robustness,and can complete the automatic diagnosis of cap faults in various situations.Thirdly,according to the characteristics that the other end of the U-shaped hoop is the end of the pipe cap,three wrist arms of the pipe cap are selected.Then,according to the different sleeve diameters of the non-pipe cap end of the three wrist arms,two wrist arms of the U-shaped hoop are obtained.On the U-shaped hoop wrist arm,the pixel-by-pixel search method is also used to determine the end of the hoop to achieve the coarse positioning of the hoop.Then,the binary projection map of the coarse positioning area of the hoop is used to accurately position and segment the hoop.The U-shaped hoops after location and segmentation are automatically classified and recognized by using depth convolution neural network.Experiments show that this method has good accuracy,speed and robustness,and can complete the automatic fault diagnosis of cap and U-type hoop.Finally,on the basis of Microsoft Visual Studio platform and OpenCV computer vision library,the algorithm of this paper is programmed,the man-machine interface is designed and developed,and the system is tested with several different actual line inspection data,which verifies the feasibility and reliability of the algorithm.