Application of Target Detection Algorithms in Railway Intrusion

China’s railway construction level is relatively high, and the level of operation and maintenance needs to be developed. The development of railway video technology is of great significance for improving the safety of railway operations and reducing the occurrence of accidents. This paper mainly studies the application of target detection algorithm in railway scenarios. For the railway video detection, this paper adopts four kinds of target detection algorithms: unsupervised frame difference method, background difference method, supervised class based on deep learning YOLOv3, Faster-RCNN algorithm. By collecting the video image data of the railway scene, the data set training deep learning algorithm is created, and then the four kinds of target detection algorithms are used to process the collected railway video image data, respectively, and the test algorithm is used for the perimeter intrusion detection effect in the railway scene. By comparison, it points out its advantages and disadvantages in the railway perimeter invasion.