OptimalCap: Efficient and Robust LiDAR-Based Motion Capture in Free Environments.

Area of Science:

• Robotics
• Computer Vision
• Biomechanical Engineering

Background:

• LiDAR-based human motion capture is promising for large-scale, unconstrained settings.
• Existing methods struggle with noisy data and dynamic scenes, limiting practical use.
• There is a need for robust and efficient LiDAR motion capture solutions.

Purpose of the Study:

• To develop a robust and efficient LiDAR-based framework for accurate, real-time multi-human motion capture.
• To introduce a large-scale synthetic dataset for training and evaluating motion capture in noisy environments.
• To establish a new standard for scalable, real-world LiDAR-based motion capture.

Main Methods:

• Proposed OptimalCap framework integrating hierarchical skeletal modeling.
• Employed kinematic-aware temporal optimization for coherent motion capture.
• Introduced NoiseMotion, a synthetic dataset simulating human-object interactions in noisy conditions.

Main Results:

• OptimalCap achieves state-of-the-art accuracy, robustness, and temporal consistency.
• The framework supports over 20 individuals captured simultaneously.
• Demonstrated real-time performance at 60 FPS up to a range of 100 meters.

Conclusions:

• OptimalCap provides a significant advancement in LiDAR-based human motion capture.
• The framework is highly effective in challenging, noisy, and dynamic environments.
• OptimalCap sets a new benchmark for scalable and practical real-world motion capture applications.