A partial overlapping point cloud registration method based on dynamic feature matching

Abstract

The point cloud registration method can effectively complete the registration of point clouds with different overlap rates and various sizes, and ensure the accuracy of the 3D reconstruction model.To address the above issues, a partial overlapping point cloud registration method based on dynamic feature matching named PPCR was proposed.Firstly, the point cloud was divided into sub-point clouds based on the Euclidean distance segmentation method.Secondly, the features of sub-point clouds were extracted.Since the sizes of different point clouds are varied, the scales of the extracted features are also different.Thus, a dynamic time warping (DTW) algorithm was proposed to map the sub-point clouds.Finally, the registration algorithm was iterated to obtain the translation and rotation matrix between point clouds.This matrix was used to complete the registration and stitching between point clouds.Experimental results show that the proposed method can effectively solve the registration problems of partially overlapping point clouds and point clouds with different scales.

Keywords

• lidar
• overlapping point cloud
• point cloud segmentation
• point cloud mapping
• point cloud registration