Feature Extraction And Clustering Simplification Algorithm For 3D Laser Point Cloud

In recent years,three-dimensional laser scanning technology has been widely used in such disciplines as surveying,computer vision and graphics,and archeology.With the development of modern science and technology,the development of three-dimensional laser scanning hardware equipment is rapid,but its corresponding information processing technology lags to a certain degree,which restricts the development of three-dimensional laser scanning technology to a certain extent.Therefore,the relevant theory and algorithm of three-dimensional laser point cloud data The optimization study is particularly important.This article studies the issue of point cloud feature extraction,point cloud clustering and point cloud simplification.In the aspect of feature extraction of point cloud,this paper proposed an algorithm of feature points extraction of scattered point cloud based on multiple parameters hybridization method,which aimed to solve the problem that previous algorithms existed,including sharp feature points extraction were incomplete,and could not retain boundary points.The method calculate the data point curvature,average vector angle between the point and its k-nearest neighborhood points,distance from point to its neighborhood gravity center,average distance from point to its neighborhood points for each k-nearest neighborhood.According to the four parameters,the characteristic threshold and feature discriminant parameter are defined.A point recognized as the feature point when its value of discriminant parameter is bigger than the threshold.Compared with existing algorithms,the proposed algorithm can not only extract the steep feature points,but also have the ability to identify the boundary points.In the aspect of point cloud clustering,this paper proposed a point cloud K-means clustering algorithm based on adaptive octree,which aimed to solve the problem that traditional K-means clustering algorithm existed,including long running time,randomness and poor clustering results.The K value of K-means clustering and initializing cluster centers related to point cloud density distribution can be provided by adaptive octree.The experimental results show that the proposed method is better than the traditional K-means clustering in the evaluation function value and running time.And the randomness of clustering is avoided.Finally,in the aspect of point cloud simplification,feature points are extracted by the proposed feature detection method.Then the K-means clustering based on adaptive octree is constructed to the point cloud.At last,the entire cluster is replaced by the nearest data point from the center of gravity of the cluster in the clustering that do not have feature points,and delete other data points.In the clustering that have feature points,the clustering is subdivided to preserve the details of the model.The two data points in clustering with the largest curvature difference as the new initial clustering center.The subdivided will not be end until the maximum curvature difference of the data points in the cluster is less than the threshold or only one data point in the cluster.In the same way,the entire cluster is replaced by the nearest data point from the center of gravity.The experimental results show that the distributions of the reduced data points are even and without holes.The reduction error and the boundary shrinkage error are small.