A Framework for Automated Bridge Inspections and Assessments with Visual Sensing Technology

The use of visual sensing technology and autonomous robotic platforms provides significant capabilities to inspect, document and assess bridges for both routine inspection and after significant natural or manmade events. To advance these capabilities, this study presents an end-to-end framework for automated conversion of raw visual sensor data into meaningful information that is directly related to bridges. Three categories of information are considered: 1) object information that includes object identity, shapes, and spatial relationships; 2) surface damage information that includes both small deformations (e.g., cracks) and large deformations (e.g., bent members, alignment issues); 3) as-built bridge models that include solid geometry models and volumetric finite element meshes. With a focus on steel girder bridges, robust algorithms have been developed and used to validate the proposed framework based on real-world data collected in situ.