Semi-automated geometric feature extraction for railway bridges

Abstract

In open-deck railway bridges, the timber ties constitute a major portion of the maintenance costs and must be replaced periodically. This procedure begins by sending surveyors to manually measure bridge and track geometry. The accuracy and efficiency of tie replacement procedures as part of bridge retrofitting projects can be significantly improved with the use of modern three-dimensional (3D) scanning technologies. This paper introduces a semi-automated geometric feature extraction framework specifically for the dapping process during tie replacement on railway bridges. First, a bridge must be 3D scanned to generate a point cloud. Next, the point cloud of the structure is pre-processed for alignment, sliced into 2D images for dimension reduction, and segmented into recognizable components. Finally, relevant features in every component are calculated and transformed into production tables or visualizable 3D models for manufacturing purposes. This framework is applied to an open-deck bridge in Lyndhurst, New Jersey. It is anticipated that with the introduction and further development of novel computer vision-based approaches, costly manual surveys of bridges can be avoided in the future.