The Application of Regional Combined Feature Variance-Covariance Matrix in Point Cloud Similarity Measure

Abstract

Effective recognition of regions and regional topological features is an important issue in similarity measure of free form surfaces point cloud. The present study provides an effective solution to this problem via a combined feature variance-covariance matrix based on region segmentation coupled with matrix similarity estimation. Firstly, we defined point features, regional features, and topological features based on the characteristics of point, region and the topological relations between different segment regions. Secondly, the feature variance-covariance matrix of each segment point cloud was calculated and then the combined feature variance-covariance matric was calculated through a variance-covariance matrices synthesis algorithm. Finally, through the similarity measure of the combined feature variance-covariance matrix, the similarity of different point cloud could be estimated. Experiments show that our algorithm has high precision in recognizing regional and topological features of point clouds.