CI-SHOT: A Statistical Method for Binary Feature Descriptor on 3D Point Clouds

Abstract

3D feature descriptors play an important role in the field of computer vision because it is a pre-requisite for many 3D vision applications. Although many such descriptors exist so far, most of them represent in the form of floating vector, which leads to several problems such as computational complexity and costly memory footprint in the feature matching process. The algorithm proposed in this paper contains two variable parameters, simplified unit μ and the number of coding bits N. For different original descriptors, using the method proposed in this paper, different simplified descriptors can be generated by changing these two parameters. The Chebyshev Inequality mathematical model is then used to convert the floating vector into a binary code to obtain a simpler and more efficient feature descriptor CI-SHOT (Chebyshev Inequality Signature of Histogram of Orientations). Finally, we compare CI-SHOT and SHOT with another binary simplified descriptor B-SHOT. The experimental results on the dataset show that CI-SHOT has obvious advantages in keypoint detection and matching performance.