Subway Rail Fastener Locating Based on Visual Attention Model

摘要

Fastener defects detection based on computer vision is an important task in subway rail inspection systems, in which the first step is fastener location. In this paper, a fastener locating method based on visual attention model for subway track is presented, which can implement fastener location in the subway track mixed with ballast-less track in the main line and ballast track on turnout region. Firstly, rail location is obtained by the vertical projection algorithm. Secondly, the image is classified into ballast or ballast-less based on visual attention model with the fractal dimension. Thirdly, ballast image and ballast-less image are treated separately. On one hand, in the ballast image, the sleeper is located with the line segment detector algorithm, and then fastener can be located by the sleeper location combined with the rail location. On the other hand, in the ballast-less image, the fastener is located with horizontal projection algorithm combined with the rail location. Experimental results demonstrate that the proposed method can locate the fastener accurately.