Research And Application Of DBSCAN Algorithm In Tree Point Cloud Segmentation

Trees(Individual Trees)are of great significance to the assessment and prediction of forest resources.As trees themselves are characterized by numerous details of branches and leaves and complex geometric structure,the obtained point cloud data is complicated,and it is difficult to distinguish individual trees,making it challenging to obtain information of individual trees.In order to improve the efficiency of obtaining single tree information and efficiently carry out forest management and forest resource monitoring,this paper obtains point cloud data through terrestrial laser scanning,and proposes a single tree segmentation algorithm and a single tree skeleton extraction algorithm based on DBSCAN(Density—Based Spatial Clustering of Application with Noise)algorithm.For the DBSCAN-based single-tree segmentation algorithm,it takes the point cloud acquired by Faro X330 terrestrial 3D laser scanner as the input,firstly preprocesses the point cloud,that is,the point cloud is de-noised and normalized after removing the ground points,then the pre-processed point cloud is vertically segmented,and the vertically segmented point cloud is clustered by DBSCAN,according to the distance between the centers of adjacent vertical segmented clusters,the point clouds of single tree trunk segment are matched until all clusters are matched,search for the nearest point from the line to complete the single wood segmentation.In this paper,the results of single tree segmentation based on DBSCAN and distance discriminant clustering based on normalized point cloud are compared.The experimental results show that compared with the distance discriminant clustering algorithm based on the normalized point cloud,the single tree segmentation algorithm based on DBSCAN performs better in the forest areas with different canopy densities,it can distinguish the phenomena of tree crown overlap,occlusion and migration in forest area,and can achieve the accurate segmentation of single tree point cloud in forest plot.For the DBSCAN based single wood skeleton extraction algorithm,firstly,the single wood point cloud obtained by the DBSCAN based single wood segmentation algorithm was imported and the single wood point cloud was vertically segmented.Secondly,the point cloud after each vertical segment is performed DBSCAN clustering and the cluster is obtained.Then calculate the center of the cluster and save the location information of the center of the cluster.Finally,by calculating the distance between the center points of the adjacent piecewise cluster,the center points of the cluster are connected to form a single wooden skeleton.In this paper,the extraction results of single wood skeleton based on DBSCAN were compared with VS(Voxel Switch)algorithm and single wood segmentation algorithm based on L1 median.The experimental results show that both algorithms can achieve skeleton extraction of single wood point cloud.Among them,the skeleton extracted by DBSCAN skeleton extraction algorithm is relatively complete and accurate,and can better reflect the real form of single wood.