Application of computer vision algorithms in the problem of coupling of the locomotive with railcars

Problem of determining the distance between locomotive and railcar from a video image is considered. The camera is located on an unmanned locomotive. Several algorithms of computer vision have been analyzed and tested. The search for a track to the railcar or safe distance detection methods were developed using Canny edge detector and Hough transformation. The difficulty in applying the algorithm for track detection is found, determined by the presence of turns and railway switches. A scheme for computing the Inverse Perspective Mapping of an image is considered. Neural network algorithms and Haar cascades are used to search for images of railcars. A conclusion is made about the need to apply hybrid algorithms to this problem.