Effectiveness of Approaches for Rail Candidates Detection and Verification of the SVM

rails detection,computer vision,Histogram of Oriented Gradients,pixel-per-pixel method,Support Vector Machine

Rail candidates detection is the primary task in railway recognition systems based on recognition in images taken from the camera mounted on the board of the locomotive. In order to reduce the classifier complexity, effective and responsible rail candidates generation plays an important role without placing big decision responsibility on a further classifier stage. There are two basic options. Due to the rich complex environment along the track, pixel-per-pixel methods are often omitted. The second option involving a thorough investigation around a pixel is preferred. In this paper, we present comparison between two different approaches to rail candidates detection, each representing one of the basic groups, furthermore consequences in rail hypotheses generation. We introduce the finding that using the SVM is more efficient than the method based on pixel-per-pixel.