Fast Three-Dimensional Posture Reconstruction of Motorcyclists Using OpenPose and a Custom MATLAB Script

Area of Science:

• Human-Computer Interaction
• Biomechanics
• Automotive Engineering

Background:

• Ergonomics investigates human interaction with their environment, crucial for understanding rider posture on motorcycles.
• Incorrect motorcycle riding posture can cause physical/psychological discomfort, impairing risk perception and vehicle control.
• Optimal rider posture is vital for motorcyclist health and road safety.

Purpose of the Study:

• To develop a rapid, cost-effective, and robust method for reconstructing motorcyclist posture in 3D.
• To enable detailed analysis of rider biomechanics during motorcycling.

Main Methods:

• Utilized stereo vision technology for depth perception.
• Applied OpenPose for real-time 3D keypoint detection and tracking.
• Calculated and evaluated distances between 3D keypoints to assess body segment lengths.

Main Results:

• Achieved accurate 3D reconstruction of motorcyclist keypoints and their temporal changes.
• Validated the stability and faithfulness of reconstructed body segment lengths against real-world measurements.
• Demonstrated the method's robustness for analyzing rider posture dynamics.

Conclusions:

• The proposed 3D posture reconstruction method is fast, cheap, and robust.
• This technique offers valuable applications in ergonomics, motorsport training, and dynamics analysis.
• Enhances understanding of rider biomechanics for improved safety and performance.