A method and device for local feature extraction of three-dimensional point cloud

Abstract

The local feature extraction method and device of the 3D point cloud provided by this application calculates the angle information and concave-convex information between the feature point to be extracted and the point of the adjacent voxel element based on the local reference system corresponding to the point of each voxel element, which can accurately Calculating the feature relationship between two points has the property of translation and rotation invariance, and because the extraction includes the concave-convex information of the local point cloud at the same time, it solves the inaccurate extraction caused by ignoring the ambiguity of concave-convex in the previous 3D local feature description The problem. In the normalization process, the index normalization process and the second normal form normalization process are used to solve the problem of inaccurate similarity calculation caused by a small number of elements in the vector being too large or too small during feature extraction, so that it can Improve the accuracy of the extracted 3D local features.

Description

technical field [0001] The present application relates to a method and device for extracting local features of a three-dimensional point cloud. Background technique [0002] With the rapid development of 3D laser scanning technology, 3D digital geometric model has become the fourth digital media form after digital audio, digital image, and digital video, and its related basic theory and key technology research has developed into a new discipline—— Digital geometry processing has gradually been widely used in computer-aided design, animation and game industry, biomedicine, digital cultural heritage protection and other fields. In addition, with the rise of hardware devices such as Microsoft Kinect and Primesense (a somatosensory technology device), the acquisition of 3D (3Dimensions, three-dimensional) point cloud information has become more convenient and quicker. In 3D vision, local feature extraction has always been the most critical part of point cloud processing, and lo...

AI Technical Summary

Problems solved by technology

[0004] However, the three 3D local feature descriptors in the prior art all ignore the concave-convex feature of the point cloud surface, which makes the extracted local features prone to ambiguity, so when applied to the processing of 3D point clouds, the estimation is often inaccurate situation occurs

The accuracy of local feature extraction in the prior art needs to be improved

Images