Segmentation Of 3D Point Clouds Of Outdoor Scenes

With the development of 3D point cloud acquisition technology,3D point cloud data has gradually entered people’s field of vision,which is concerned by more and more researchers.In 3D point cloud related processing technology,area segmentation technology is very important,and effective segmentation results are the basis of subsequent classification recognition,model reconstruction and other tasks.Outdoor scene is very complex and rich,and the volume of scene point cloud data is large,there maybe has many noise pollution.The 3D point cloud data segmentation of outdoor scene is a complex and meaningful work.This paper has done the main research in this area.For the segmentation of 3D point cloud data in outdoor scene,this paper proposes a segmentation method based on 3D Gaussian process regression.Firstly,a fan-shaped grid is established based on polar coordinates,and all the 3D point cloud data collected by laser scanner are divided and stored,and the candidate ground points in each grid are selected to form a candidate point set;then,the candidate ground points are fitted by using 3D Gaussian process regression,and through online learning and Mahalanobis distance judgment conditions,starting from the most center,the candidate ground point set is traversed from the inside to outside,select the seed points that meet the conditions,update the ground model online,use the maximum likelihood estimation method to optimize the super parameters,and use the sparse processing method to reduce the time complexity of the algorithm;then,judge whether the relative distance between the real height of each point and the height predicted by the final ground model meets the given threshold conditions;finally,the DBSCAN clustering algorithm is used to segment the non-ground point cloud.In order to test the segmentation effectiveness of this method,four outdoor scenes are selected for ground extraction and non-ground point cloud segmentation experiments.Compared with the segmentation algorithm based on line extraction,the experimental results show that our method has a high accuracy and good security;for non-ground point cloud segmentation,the experimental results show that our method is reliable and effective.