Point Cloud Reduction Based On Multi Parameter K-Means Clustering

Compared with traditional measurement methods,3D laser technology has obvious advantages in data acquisition.Especially in recent years,3D laser technology has been widely used in various fields.With the development and update of 3D laser scanning equipment,the accuracy of data collected by the equipment is improving while the amount of data becoming very large,which brings great challenges to the subsequent work of data processing and data storage.In this paper,the feature model is constructed to extract the feature of point cloud,and the point cloud data is simplified by clustering.The main contents of this paper are as follows:(1)Research on point cloud feature extraction algorithm.Traditional point cloud feature extraction algorithms usually only focus on the single feature of data points,which can’t effectively extract multi class features,and it is easy to cause edge loss,missing details and other problems.Aiming at the problems of traditional algorithms,we can get the point curvature,neighborhood normal angle,neighborhood average distance,neighborhood pull,neighborhood point projection angle by calculating the neighborhood information of the current point A multi parameter feature extraction model is constructed to extract the point cloud features including corner points,edges and boundaries.Experimental results show that the algorithm can effectively improve the recognition efficiency of feature points compared with the traditional algorithm.(2)Research on point cloud clustering algorithm.In the case of large amount of data,the traditional K-means clustering method generally runs for a long time,and the clustering results are unstable.In order to solve these problems,triangular inequality is used to optimize K-means clustering to reduce unnecessary distance calculation,and octree space partition is used to provide initialization parameters for clustering to reduce the running time and stabilize the clustering results.The experimental results show that this method is superior to the traditional K-means algorithm in the stability of clustering results,clustering effect and running time.(3)Research on point cloud reduction method.At present,most of the commonly used reduction methods have some problems,such as data holes,lack of details and so on.To solve these problems,this paper proposes a point cloud reduction method based on multi parameter K-means clustering.The feature data with key information is retained by feature extraction,and the fuzzy set is constructed for the clustering results.The minimum fuzzy entropy is calculated,and the corresponding curvature is used as the threshold of clustering region division.The reasonable reduction strategy is adopted for different regions to complete the point cloud reduction.The experimental results show that this method is superior to the traditional point cloud reduction method,and has good point cloud reduction effect.