Monocular Vision–Based Framework for Biomechanical Analysis or Ergonomic Posture Assessment in Modular Construction
Wenjing Chu, SangHyeok Han, Xiaowei Luo, Zhenhua Zhu
Abstract
Awkward and improper postures and motions reduce productivity and increase project costs in the modular construction industry. Ergonomic assessment is essential to identify, mitigate, and prevent these postures for safety and productivity improvement. Advanced computer vision technologies have made vision–based ergonomic assessment cost-effective in real workplaces. However, their accuracy and robustness still need to be improved. This paper proposes a monocular vision–based framework for conducting a biomechanical analysis or ergonomic posture assessment. The framework consists of four components: worker visual tracking, two-dimensional (2D) joint and body part detection, 2D joints refinement, and three-dimensional (3D) body model generation and joint angle calculation. The framework has been tested with videos recorded in real construction workshops. The results show that the framework could use the videos from a single camera to estimate a total of 14 joint angles with the average error of 11° and identify workers’ awkward postures and motions.