Image-based 3D reconstruction for rail profile measurement

Abstract

The routine inspection of railheads for defects such as wear and surface cracks is a tedious process, which, if not detected, can alter the wheel-rail contact interaction leading to catastrophic events. This study investigates and implements a railhead measurement method using image-based three-dimensional reconstruction, which enables rapid scanning of railheads and production of detail cross-sectional measurements for rail-wheel interface analysis. A complete workflow with a methodology for reconstructing railheads from images and extracting cross-sectional measurements from the reconstructed model is presented. In order to validate the proposed method in the field, a mobile automated system was equipped with an array of cameras specifically spaced to cover the areas of interest on the railhead. The system can automatically transverse along the railhead, acquiring images synchronously. Two case studies in the laboratory environment and the real railway site have been performed to evaluate the performance and accuracy against industry practices. The results show that the proposed method can accurately measure the railhead cross-sectional profile at an root mean square error (RMSE) less than 0.3 mm compared with MiniProf. Furthermore, continuous cross-sectional data and intuitive color information are provided by our method which can help inspectors to locate defects easily and more efficiently.