Automated extraction of structural elements in steel girder bridges from laser point clouds

Abstract

This study proposes a heuristic-based method to automate the semantic segmentation of laser point clouds collected from steel girder bridges and thus to facilitate the applications of laser scanning in bridge inspection and management. In the proposed method, domain knowledge on the geometric and topological constraints of steel girder bridges are utilized to recognize and extract individual instances of the main structural elements, which include the bridge deck, steel girders, cross-frames, piers and abutments. The effectiveness of the proposed method is validated using point clouds acquired by a terrestrial laser scanner and a UAV-based laser scanner. The results demonstrate that the proposed method on average can achieve a 98.3% element-level accuracy and a 96.1% point-level accuracy on terrestrial laser scanning data. For UAV-based laser scanning data, the average accuracy is reduced slightly to 96.0% in element-level and 93.3% in point-level.