Detection of Railway Tunnel lining Based on Adaptive Background Learning

Abstract

Tunnel is the absolutely necessary part of railway line. Its state of service is often influenced by linings. Once the lining has fallen on the concrete bed, it will directly threaten the safety of trains in the tunnel. So, one timely lining detection system is a key assistant to keep the tunnel a good operating condition. In this paper, the approach of lining fall-blocks detection by intelligent video analysis based on adaptive background modeling is proposed. Without a large number of labeled samples, the unsupervised method based on gaussian mixture model (GMM) is applied to monitor the fallen block in real time. The basic idea is to set the learning rate automatically, which is mainly according to the critical attributes including image intensity and feature point of current frame. Because of the strategy the method has the advantage of modeling the background in variable scenes such as illumination-changing, camera-shaking, train-passing, etc. Finally, in our experiment, the results demonstrated the effectiveness of the proposed approach.