Research On Pantograph Fault Detection System Based On Computer Vision

With the development of electrified railways,people’s travel modes have also undergone great changes,but with the continuous increase of train speed,the requirements for train operation safety are also getting higher and higher.Pantographs,as the key electrical equipment on trains,transmits electric energy for the train,and with the increase of the pantograph’s use time during the operation of the train,the pantograph will produce faults such as skateboard wear,skateboard cracks and missing horns,which seriously threaten the safety of high-speed trains.On the basis of external research,an online detection system for pantograph faults based on computer vision is designed.The main contents of this paper are as follows:First,the improved distortion model to calibrate the camera internal parameters,the three-dimensional target to calibrate the camera external parameters and the threedimensional target to test the accuracy of the calibration results is used in this paper.In addition,for some low-illumination images taken on-site in the detection ooth,in order to improve the subsequent detection effect,the image is enhanced by the Retinex algorithm based on HSV space improvement.Secondly,in view of the difficulty of determining the high and low thresholds of the Canny algorithm due to the influence of light,an adaptive Canny edge detection algorithm based on the Otsu algorithm is used to detect the edge of the sliding plate,and Combined with the camera calibration parameters,the wear measurement of the pantograph sliding plate is realized.then,the feature extraction network of the YOLO V4 model is replaced with Mobile Netv2,which makes the backbone feature extraction network more lightweight,and the PANet network of the YOLO V4 model is improved.A dense block feature fusion network enhances the detection ability of small targets and realizes the detection of cracks in the pantograph slide plate.Next,the SURF algorithm is improved and used to analyze the partial image and standard of the pantograph horn.The feature point matching is performed on the pantograph horn image,and the missing horn is detected by comparing the number of matching point pairs and the set threshold.Finally,the hardware of the detection system is selected according to the requirements of the fault detection system,and the system interface is built through Python’s Tkinter to form a complete detection system,and then the test in the laboratory environment is carried out.The experimental results show that,the detection algorithm has high accuracy,feasibility and repeatability,and has a good application prospect.