Research On Pantograph Fault Detection Algorithm Based On Image Processing

The pantograph on the train roof takes power from the catenary to drag the train forward during urban rail train running.The pantograph constantly rubs against the catenary,so it is inevitably to cause the pantograph slide abrasion and other problems.Slide abrasion,slide crack and horns missing may lead to serious pantograph-catenary accidents.A method based on image processing to detect pantograph faults to prevent pantograph-catenary accident is proposed in the paper.The proposed method can detect slide abrasion,slide crack and judge horns missing.The Zhang Zhengyou calibration method is first introduced in the paper.An improved Zhang calibration method is proposed to overcome the shortcomings of the original method.The tangential distortion is taken into consideration when cameras are calibrated.The influence of the number of images on the stability of the camera intrinsic parameter is analyzed to get results.The detection of pantograph slide abrasion,slide crack and horns missing is the focus of the paper.In the slide abrasion detection,cascade filters is adopted to filter images and the intensity of cascade filters is dynamic.The adaptive threshold Canny edge detection algorithm is then performed on the pre-processed images to extract the edge of slides.The slide edges is located by Hough transform.Then the actual slide residual thickness can be calculated by the camera calibration method.In the slide crack detection algorithm,filtering noise is first performed on original images and then slide crack is detected by second generation curvelet transform.If the slide surface has crack,the crack length is calculated.In horns missing detection algorithm,an ASM model is built and the initial position of horns is determined.Combine the model with the original position to judge whether the pantograph has horns.The experiment has been done to validate the feasibility of the proposed algorithms.All pantograph images were captured when trains were in motion.Images captured in different situations were processed to detect pantograph faults in order to prove the robustness of proposed algorithms.The results showed that the slide abrasion detection accuracy is ±O.5mm.The recognition accurate rate of slide crack was 91.2%and the detection accuracy of crack length is ±0.3mm.The recognition accurate rate of horns missing was 79.45%.The proposed algorithm can be applied to detect pantograph faults in the flied.