Geodesic-Based Maximal Cliques Search for Non-Rigid Human Point Cloud Registration

Area of Science:

• Computer Vision
• Medical Imaging
• Biomechanics

Background:

• Non-rigid point cloud registration is crucial for human body pose analysis in diverse fields.
• Accurate registration is challenging due to non-rigid deformations.

Purpose of the Study:

• To propose an accurate non-rigid point cloud registration algorithm for human body pose analysis.
• To improve the accuracy of matching point pairs during non-rigid deformations.

Main Methods:

• Constructing graphs using geodesic distance measurement for point cloud sets.
• Employing maximal clique search to identify matching pairs.
• Driving human body model parameters to overlap sparse matching pairs.

Main Results:

• The proposed algorithm demonstrates improved accuracy in non-rigid point cloud registration.
• Geodesic distance measurement preserves matching information during deformations.
• Validation on FAUST and CAPE datasets confirms algorithm accuracy.

Conclusions:

• The geodesic distance-based non-rigid point cloud registration algorithm effectively enhances human body pose analysis.
• The method offers a robust solution for applications requiring precise deformation tracking.
• This approach advances the field of non-rigid registration for complex biological shapes.