Automated Posture Assessment for Construction Workers

Abstract

Construction workers often suffer various kinds of musculoskeletal disorders (MSDs), which are injuries in the human musculoskeletal system. MSDs are often aroused due to sudden exertion such as lifting heavy equipment, or repeated and cumulative stressed motions. OWAS (Ovako Working posture Assessment System) may be used to evaluate the exposure of MSDs risk by sampling the snapshots of a worker's postures and categorize the postures. However, tracking and categorizing postures of different body parts by human eyes are tedious work with limited accuracy and easy to make mistakes even with facilitation of video recording. This research develops an automatic tracking and categorizing system, named Posture Assessment System for MSD (PAS-MSD) for OWAS using Microsoft Kinect. The PAS-MSD captures human postures during his/her movement, recognizes human skeleton, and assesses the risk of MSD. An experiment with typical construction activities such as handling and moving of materials, hammering, and tiling was conducted. Except for the hammering activity where the subjects' body parts were easily blocked by the target hammered box and could not be detected by Kinect, the posture identification accuracies for all other activities exceed 90% (i.e., 91.6%-93.9%). The OWAS categorization accuracies are also satisfactory, ranging from 85.4%-88.5%.