Defect Detection of Pantograph Slider Based on Improved Faster R-CNN

Pantograph-catenary system is the key component in electrical railway system. In order to keep a steady power supply to the train, pantograph should keep a favorable contact with the overhead contact line, which makes the pantograph slider one of the most vulnerable parts in pantograph-catenary system. In this paper, an improved Faster R-CNN algorithm is used in the surface defects detection of the pantograph slider, which overcomes the limitations of the traditional detection methods, extracts the effective area of pantograph slider from raw image and achieve good classification results at the same time. After experimental analysis, the proposed method reaches a satisfactory mAP of 0.9763 in the collected pantograph image dataset. This can give a useful reference to the metro operation company and future related research.