Computer-Vision Based Visual Inspection and Crack Detection of Railroad Tracks

Surface analysis is a very important measurement for track maintenance for Railroad Tracks, because deviations in surface geometry indicate where potential defects may exist. A rail surface defects inspection method based on computer vision system is proposed in the paper. Various algorithms related denoising, filtering, thresholding; segmentation and feature extraction are applied for processing the images of Railroad surface defect and cracks. It has mostly been implemented on computers. For better speed and complexity, the algorithms need to be implemented on embedded platforms. These methods were designed for different software setups, namely MATLAB and C++ using the Intel OpenCV library. Then accurate region of Interest in respective to defect is extracted and recognized by adaptive thresholding and feature matching methods. Percentage of wear of rail head and length of cracks in surface are calculated next as an evaluation of flaw on inspected rail head section. Experimental results of the proposed algorithms are presented in the results sections along with bench marking with software algorithms. KeywordsOpenCV, Track Inspection, Dynamic Threshold, Computer Vision, Optimal Thresholding.