The aging of bridges coupled with increased vehicular traffic requires timely and accurate inspections for elevated highway structures. Recent studies have leveraged the advent of drones and computer vision to automatically conduct quick, safe, and effective inspections for elevated highway structures. However, such studies rarely offer insight or recommendations for an end-to-end integrated system that streamlines data collection, analytics, and reporting. Toward this goal, we present an end-to-end robotic bridge inspection system consisting of five tightly coupled methods to: (1) create automatic data collection missions; (2) assure visual quality of such missions; (3) reconstruct three-dimensional (3D) models of elevated structures; (4) detect and localize surface distresses in 3D; and (5) generate reports complying with highway agencies' requirements. We validate each developed method and the whole system on two representative inspection projects. Results show that our system can objectively satisfy requirements for data collection and provide up to 85.3% average precision over five defect types. We finally share lessons learned while deploying our system to 30 bridge inspection projects in the US and Japan, particularly for documenting, communicating, and following-up with bridge inspectors' recommendations.
|Original language||English (US)|
|Journal||Journal of Computing in Civil Engineering|
|State||Published - Mar 1 2021|
ASJC Scopus subject areas
- Civil and Structural Engineering
- Computer Science Applications