PKSS-Align: Robust Point Cloud Registration on Pre-Kendall Shape Space.

Area of Science:

• 3D Computer Vision
• Geometric Computing
• Computer Graphics

Background:

• Point cloud registration is crucial for 3D data but is sensitive to transformations, noise, and incomplete structures.
• Non-uniform scales and defects in point clouds often lead to local optima during registration.

Purpose of the Study:

• To develop a robust point cloud registration method, PKSS-Align, capable of handling various challenges.
• To improve registration accuracy and efficiency for real-world 3D data.

Main Methods:

• Proposed PKSS-Align utilizes shape feature-based similarity measurement on the Pre-Kendall shape space (PKSS).
• Employs a manifold metric robust to Euclidean coordinate representations, avoiding point-to-point or point-to-plane metrics.
• Directly generates transformation matrices without requiring data training or complex feature encoding.

Main Results:

• PKSS-Align demonstrates robustness against similarity transformations, non-uniform densities, noisy points, and defective parts.
• The method achieves significant improvements in efficiency and feasibility through parallel acceleration.
• Experimental results show superior performance compared to state-of-the-art registration methods.

Conclusions:

• PKSS-Align offers a robust and efficient solution for point cloud registration challenges.
• The Pre-Kendall shape space measurement provides a novel approach for shape comparison in 3D vision.
• The method is practical for real-world applications due to its training-free nature and performance.