Advanced Planar Projection Contour (PPC): A Novel Algorithm for Local Feature Description in Point Clouds.

Area of Science:

• 3D Computer Vision
• Geometric Deep Learning
• Point Cloud Processing

Background:

• Local feature description is crucial for 3D computer vision tasks.
• Existing point cloud descriptors often lack robustness, have high dimensionality, and poor computational efficiency.

Purpose of the Study:

• To propose a novel local feature descriptor for point clouds that overcomes limitations of existing methods.
• To enhance robustness, reduce dimensionality, and improve computational efficiency in 3D feature description.

Main Methods:

• Constructed a Local Reference Frame (LRF) using covariance analysis.
• Projected neighboring points onto orthogonal planes defined by the LRF.
• Encoded projected points into convex hull contours to form the Planar Projection Contour (PPC) descriptor.

Main Results:

• The PPC descriptor achieved over 80% accuracy across all recall levels on B3R, UWAOR, and Kinect datasets.
• Demonstrated superior robustness against high noise and point density variations compared to classical descriptors.

Conclusions:

• The Planar Projection Contour (PPC) descriptor offers an effective and robust solution for local feature description in 3D point clouds.
• The proposed method addresses key challenges in 3D computer vision, paving the way for more reliable point cloud analysis.