Research On Automatic Mapping Technology Using Vehicle-Borne LiDAR Point Cloud

The vehicle-borne LiDAR system can quickly obtain high precision surface point clouds for all kinds of objects,such as road boundaries,buildings,pole-like objects and so on.Compared with the point clouds acquired from the airborne and the terrestrial counterparts,it is more suitable for large-scale topographic mapping.The traditional topographic mappings are performed through a large variety of digital mapping techniques,such as total station mapping,GPS-RTK mapping and digital photogrammetry and so on.Total station mapping and GPS-RTK mapping are mainly used to collect the three-dimensional coordinate information of the objects directly by using the instrument for single point collection while digital photogrammetry is used to obtain 3D coordinate information of objects by using stereo image pairs,such as aerial triangulation,and so on.Compared with these digital mapping techniques,the vehicle-borne LiDAR has advantages of not only obtaining the massive three-dimensional coordinate information of the surrounding objects,but also obtaining the other related information such as the reflection intensity and color of the ground objects.With regards to point clouds classification,the existing researches mainly focus on the extraction of common objects in vehicle-borne LiDAR point clouds.Currently,the existing methods about the digital mapping from the LiDAR point cloud primarily pick up the positions interactively with low level of automation.To address the above issues,we analyze the characteristics of vehicle-borne LiDAR point clouds data and conduct the topographic mapping based on the automatic classification.The main contents are as follows:(1)For the vehicle-borne LiDAR point clouds data in the road environment,we analyze the characteristics of the point clouds and super-voxelize the raw point cloud for generating a set of super-voxels as primitives.Then,the point density and height difference inside each super-voxel and context among super-voxels are extracted to classify road boundary,building and pole-like object.(2)According to the diversity among different objects,the extracted objects are re-classified.On this basis,the different feature points are extracted from different objects and then these extracted feature points are arranged in a certain order for constructing the topology relationship.Finally,the DXF files are automatically derived and rendered by the simple code recognition function using the South CASS software.In this paper,the vehicle-borne LiDAR point clouds data in the city of Delft and a street of Panyu District(Guangzhou)are tested.The result of the experiment shows that the automatic mapping results can meet the requirements of the large scale topographic map,The shortcoming is that only the common simple objects can be automatically plotting.