Use of Marker-less Motion Capture Approach for Construction Field Workers’ Biomechanical Analysis

We propose computer vision-based biomechanical analysis for construction field workers. Taking into account the fact that the number of fatalities in construction is the highest among all industries [1], and that the unsafe acts and behavior of workers cause 80% to 90% of accidents in construction [3,5,7], the systematic understanding of field construction workers’ unsafe behavior has great potential to contribute to the reduction of injuries and fatalities in construction. Though techniques do exist to study unsafe acts (e.g., surveys, focus groups, video analysis, and laboratory experiments), these techniques may not be suitable for measuring the physical demands exerted on workers under real conditions (i.e., biomechanical analysis). To address this issue, we aim to develop a computer visionbased marker-less biomechanical analysis system, which will extract workers’ skeletons (without any marker attached to the workers) captured from ordinary network surveillance video cameras; the skeletons will be used for biomechanical analysis with estimated force information. The paper first compares the accuracy of the proposed marker-less motion tracking system with that of a commercial system with ladder climbing in the lab. Second, it demonstrates how postures determined using marker-less motion tracking can be used to evaluate biomechanical loads for a quasi-static case.