Evaluation of upper body postural assessment of forklift driving using a single depth camera

Abstract

Observational postural assessment methods which are commonly used in industry are time consuming and have issues of inter-and intra-rater reliability. Computer vision (CV) based methods have been proposed, but they have mainly been tested inside lab environments. This study aims to develop and evaluate an upper body postural assessment system in a real industry environment using a single depth camera and OpenPose for the task of forklift driving. The results were compared with XSens, an Inertial Measurement Unit (IMU) based system. Data from three forklift drivers performing seven indoor and outdoor tasks were recorded with a depth camera and XSens sensors. The data were then analyzed with OpenPose with additional custom processing. The angles calculated by the computer vision system showed small errors compared to the XSens system and generally followed the trend of the XSens system joint angle values. However, the results after applying ergonomic thresholds were vastly different and the two systems rarely agreed. These findings suggest that the CV system needs further study to improve the robustness on self-occlusion and angle calculations. Also, XSens needs further study to assess its consistency and reliability in industrial environments.