Voxel-based 3D Road Extraction And Its Application In Road Network Construction

As the lifeline of the national economy,road plays a very important role in production and life.Airborne Light Detection and Ranging(LiDAR)technology can quickly obtain high precision three dimension(3D)point cloud data,which makes it possible to quickly extract the road 3D information.The data structures used in the classical method of road extraction based on airborne LiDAR point cloud are raster grid,triangular irregular network and point cloud.Among them,the first two is 2.5D data structure,the same plane coordinate corresponds to only one elevation,so the expression of the 3D point cloud will result in the loss of effective information,affecting the integrity of the road extraction results.Point cloud data structure contains all the information of the original data,but it is difficult to use the spatial relationship of data points,which leads to low accuracy of road extraction.So it is necessary to choose a simpler 3D data structure—Voxel,In this paper,we focus on the construction of a gray voxel data structure model,on this basis to extracting 3D road voxel,and then applied to the construction of road network,the innovation of this paper mainly includes:(1)Proposing a model of gray voxel data structure.The model divides the bounding box of the data set into 3D voxel grids,and according to the intensity value of the laser point as the voxel assignment.The model regularization the original LiDAR point cloud into a 3D voxel data using voxel as structural unit,lay the foundation for subsequent road extraction.(2)Putting forward a 3D road extraction method for extracting road voxel from the constructed gray voxel data structure model.Firstly,search for the seeds voxel,then labeling the seeds voxel and its 3D connected set,and finally,using mathematical morphology to optimize it,the method combines the elevation and intensity information of road,and extracting 3D road based on the idea of 3D connected construction.(3)Raising a method of constructing road network from 3D road voxel.Firstly,detecting and filling the road void,and extracting skeleton line,and skeleton line denoising.Constructing road network mainly based on the spatial index of road voxel in the gray voxel data structure model.Through two sets of experimental data with different road conditions to prove the proposed method,and quantitative analysis of road voxel and road network extraction results,experiments proved that completeness and correctness of road voxel both over 80%,and completeness and correctness of road network both over 85%,and qualitative analysis of road network extraction results,the two results also verify the feasibility of the gray voxel data structure model.