A Coarse-to-Fine Keypoint Detection Method for 3D Model

Abstract

This paper proposes a coarse-to-fine 3D keypoint detection method based on Principal Component Analysis and Harris operator. At first, the local neighborhood of each vertex is determined according to the conception of "ring". Then the Principal Component Analysis method is performed on the local surface, and the ratio between the first two principal axes of the local neighboring surface is used for selecting candidate keypoints. Finally we compute the Hessian matrix of the local surface through paraboloid fitting, and the Harris operator is used to obtain final keypoint. Extensive experimental results have testified the effectiveness of the proposed method, and it is more robust to noise, especially to high level noise.

Keywords

• 3D model
• Keypoint detection
• Principal Component Analysis
• Hessian matrix
• Harris operator.