Work Posture Assessment in Confined Spaces Using Posture Estimation AI and Occlusion Correction

Abstract

This study applies posture estimation AI and occlusion correction technology to analyze work postures in a milking parlor of a dairy facility. In confined workspaces, occlusions and variations in work postures often cause missing joint information. However, the proposed method demonstrates that accurate posture estimation is possible in real-world environments using a single camera. Posture estimation and correction were performed on video data collected in the field, allowing for the analysis of posture changes and workload during milking tasks. This approach is expected to contribute to improving safety and optimizing work efficiency in dairy operations.