Research On Point’s Neighbors And Keypoint Descriptor Of Three-dimensional Point Cloud Model Data

With the Information Age coming, a variety of surface reconstruction software been improved as long as the rapidly developing of modern 3D scanning equipment. Querying point’s neighborhood on 3D point cloud data has become the main factor restricting the point cloud model reconstruction efficiency. After accomplish points’ neighbors, we extract the key-points & descriptors as point clouds matching reference. Therefore, proposing efficient and high robustness points’ neighbor searching and feature-points extracting becomes the key in the process of the three-dimensional reconstruction.This paper focuses on points’ neighborhood searching and feature-points extracting on three-dimensional point cloud data algorithm methods and related research, the main research work is as follows.(1) This article proposed an algorithm for searching neighborhood on search-window, aimed at the points’ neighbor searching algorithms on point cloud data with the problems of large fluctuations on edge points from realistic experiment with Kinect.(2) In order to accelerate further surface registration, reduce the size ratio set point, this article proposed an algorithm constructed a kind of key-points based on the surface normal fluctuations and improved the feature point histogram descriptor. This algorithm make statistical analysis on relations get the key-points, which the surface normal fluctuated wildly.Experiments show, compared with the existing searching neighborhood and feature descriptor extraction algorithms, searching neighborhood on search-window with adaptive parameter is more particularly suitable for point cloud model obtained by Kinect. The key-points & descriptors, which based on the surface normal fluctuations scaling simplify the original point. This algorithm will provide good initial values for the next step of registration.