Authors
Nguyen KX; Zheng L; Hawke AL; Carey RE; Breloff SP; Li K; Peng X
Source
Comput Vis Image Underst 2023 Oct; 235:103780
Abstract
It is necessary to analyze the whole-body kinematics (including joint locations and joint angles) to assess risks of fatal and musculoskeletal injuries in occupational tasks. Human pose estimation has gotten more attention in recent years as a method to minimize the errors in determining joint locations. However, the joint angles are not often estimated, nor is the quality of joint angle estimation assessed. In this paper, we presented an end-to-end approach on direct joint angle estimation from multi-view images. Our method leveraged the volumetric pose representation and mapped the rotation representation to a continuous space where each rotation was uniquely represented. We also presented a new kinematic dataset in the domain of residential roofing with a data processing pipeline to generate necessary annotations for the supervised training procedure on direct joint angle estimation. We achieved a mean angle error of 7.19 degrees on the new Roofing dataset and 8.41 degrees on the Human3.6M dataset, paving the way for employment of on-site kinematic analysis using multi-view images.
Keywords
Musculoskeletal system disorders; MSD; Motion studies; Sensors; Biomechanics; Computers; Imaging techniques;
Author Keywords: Deep Learning; Computer Vision; Kinematic Estimation; Biomechanics
Contact
Kien X. Nguyen, Department of Computer & Information Science, University of Delaware, Newark, DE, USA
Document Type
Journal Article
Email Address
kxnguyen@udel.edu
Priority Area
Construction
Source Name
Computer Vision and Image Understanding
