Voxel-based Marked Neighborhood Searching method for identifying street trees using Vehicle-borne Laser Scanning data

Abstract

As an important method for acquiring close-range three-dimensional (3D) data of cities, Vehicle-borne Laser Scanning (VLS) system provide an efficient technique for acquiring 3D information of different objects on the two sides of urban streets. Segmenting and identifying different types of urban street objects from laser scanning point cloud data are urgent issues for the VLS applications. This paper presents a new Voxel-based Marked Neighborhood Searching (VMNS) method for identifying street trees and measuring their morphological parameters from VLS data. The VMNS method includes following five steps, including voxelization, setting voxel value, neighborhood search and mark, computing the biophysical parameters, and eliminating other pole-like objects. The feasibility of the method is proved through a case study. The results show most of the laser points that constitute an individual tree are identified and extracted correctly. The errors of derived biophysical parameters are analyzed by compared with the in situ measurement data. The paper shows that our VMNS method is effective for the street trees identification from VLS data.