3D Image Stitching of Train Bottom based on SURF and Cross Gauss Laplace Descriptor

With the modern development of the city, highspeed train traffic plays a significant role in public transportation. However, high-speed trains run frequently through long routes. Once a train fault occurs, the personal safety of passengers will be in danger. Therefore, fault diagnosis of trains becomes particularly important. In order to speed up fault detection, the train bottom is scanned by 3D camera before the train enters the station so that train bottom fault can be detected before train enter the station. It will buy more time for staff to carry the corresponding tools to repair if faults are detected. But the obtained data needs to be stitching to obtain the train bottom. In this paper, the 3D data obtained from the scanning of the train bottom is stitching to prepare for fault detection. This paper presents a new method for image stitching. First, select potential feature points based on SURF, then calculate the Cross Gaussian Laplace Descriptor (CGLD) of these feature points, finally determine whether the two feature points are the same feature points by matching the approximation of the CGLD of the candidate feature points. Experimental tests confirm that the matching accuracy of the feature points after the secondary matching has been improved to a certain extent.