Automated Postural Ergonomic Assessment Using a Computer Vision-Based Posture Classification

In construction, workers are frequently exposed to manual handling tasks involving repetitive and forceful activities with awkward postures. As a result, construction workers are at about a fifty percent higher risk of work-related musculoskeletal disorders (WMSDs) than workers in other industries. Several ergonomic methods for evaluating workers' postural stresses through observations (e.g., OWAS, PATH, or RULA) have been applied to identify the risks of WMSDs during occupational tasks including construction. While inexpensive and practical to use, these methods are constrained by time-consuming procedures, observer bias, and the need for a well-trained analyst. These issues are exacerbated in the context of construction. In contrast to other factory-based (e.g., manufacturing) industries, construction is subject to irregular work patterns, relatively longer cycle time, and a wide variety of tasks. To address these issues, we propose computer vision-based posture classification that can automate existing observation-based postural evaluation methods such as OWAS. Specifically, the proposed method can classify workers' postures according to pre-defined ergonomic checklists using shape- and radial histogram-based features from video sequences. To test the feasibility of the proposed approach, we conducted a laboratory test with five subjects by simulating three representative postures according to upper limbs, back, and lower limbs. The results from the case study demonstrate that it can provide more reliable posture classification results with about 93% of accuracy. The proposed approach enables practitioners to identify the risk of WMSDs through automated ergonomic postural evaluation of on-going tasks only by taking videos at sites, which helps to prevent WMSDs for diverse occupational tasks including construction.