Track gauge measurement based on model matching using UAV image

Abstract

This study investigates the use of unmanned aerial vehicles (UAVs) for track gauge measurement, as an alternative to traditional methods that rely on specialized equipment mounted on rail inspection vehicles or trains. Such equipment not only consumes valuable rail resources but also poses safety risks to inspectors. The proposed method involves utilizing image data collected by a UAV to generate a point cloud of the railway scene. A hybrid segmentation algorithm based on augmented Region-Growing and improved alpha-shape techniques is then employed to extract rails from chaotic scene data. To determine the track gauge, a pre-built rail Building Information Modeling (BIM) model is discretized into a point cloud template, which is then matched with the extracted rail features using a template-oriented model matching approach. Validation experiments on four groups of data demonstrate the superiority of this method in terms of safety, efficiency, speed, and accuracy.