Research Of Segmentation Algorithm For Terrestrial LiDAR Point Cloud

Three dimensional laser scanning technology with fast,low-cost,high-precision gets massive advantage of point cloud data,is becoming “intelligent city” construction of three dimensional data in important ways.The point clouds acquired by the laser scanning system with three dimensional discrete,high density and the characteristics of large amount of data,brings the new challenges for data services.Relative to the rapid improvement of laser scanning data acquisition capacity,point cloud processing power lagged behind.How to extract object information from a massive point cloud data is current research focus.For this,this paper focuses on the ground point cloud remove and the segmentation of the adhesion problem.The main work is as follows:(1)Working principle of laser scanning system and the existing point cloud segmentation method are introduced.First introduced the principle of terrestrial laser scanning technology and analyzes the current ground segmentation method of the point cloud secondly.Finally,introduced several classic methods for segmentation of point cloud,and simulations to test data,pointing out the advantages and disadvantages of these algorithms.(2)In view of the point cloud in the ground and on the ground are connected,put forward a algorithm using supervoxel to realize the ground point cloud and the non ground point cloud separation.The algorithm generate local point cloud firstly.Then construct a 39 d descriptors using spatial distribution characteristics of the point cloud,color information,and so on.For each local point cloud,generate supervoxel using the Fuzzy C-means algorithm.On this basis,introducing Octree voxel space adjacency,using a up growing segmentation method for the segmentation of the ground point cloud and the non ground point cloud.The supervoxel method has good effect on retent local boundary characteristics,and the comparative experiments proved the effectiveness of the proposed method.(3)In view of the point cloud segmentation result that exist the adhesion phenomenon,combined with the spatial distribution and color information of the three-dimensional point cloud,introduce a over-segmentation method divides the points into supervoxels,and construct a weighted graph model.On this basis,realize the segmentation of point cloud adhesion area using Normalized Cut.In view of the trees,buildings,the experiment results show that the algorithm which used in the graph theory of Normalized Cut has a good segmentation effect about the adhension between the trees and buildings.Finally,summarizes the work being done in this paper,and looks forward to the follow-up research work of this article.