Automatic Learning Technology of Railway Based on Deep Learning for Railway Obstacle Avoidance

Abstract

Railway identification is the key to dividing the dangerous area and safety area of the rail surface. It is also the key technology for the detection of rail surface obstacles. The problem is that the traditional rail detection edge detection effect is not good and the illumination environment is seriously affected. This paper proposes a track recognition detection system based on deep learning. The system can not only identify the rails, but also plan the dangerous areas, apply the convolutional neural network, extract the feature information of the rails in different environments, and use the Mask R-CNN instance segmentation algorithm. The identification of the rails and the division of the dangerous areas are carried out to achieve accurate and rapid detection. The test results show that the system has high anti-interference and accuracy.