Research On Airborne LiDAR Point Cloud Data Filtering And Building Extraction Method

Airborne LiDAR measurement technology is a measurement technology capable of direct observation of the ground surface,which has developed rapidly in recent years and is widely used in digital city construction and other fields.Airborne LiDAR measurement system can quickly obtain a large amount of highly accurate and dense 3D point cloud data,which is the original data for building extraction and modeling.However,the acquired point cloud data contains other categories such as ground,vegetation,and vehicles in addition to buildings,and it is difficult to distinguish the categories directly.To this end,this paper takes the original point cloud data acquired by airborne LiDAR as the target research data,and for the problem of difficult extraction of buildings from airborne LiDAR point clouds,this paper adopts a super voxel-based point cloud filtering and building extraction method,which can effectively filter out unnecessary ground,vegetation and other feature categories while extracting building and contour information,and the main work and research contents are as follows.(1)The filtering performance of five filtering algorithms is compared and analyzed.The basic principles of five filtering algorithms including slope filtering,mathematical morphology filtering,Irregular Triangulation filtering,moving surface fitting filtering and cloth simulation filtering are studied.The filtering performance of different filtering algorithms is compared and evaluated through experiments.The results show that in three standard data sets with different geomorphic features,The filtering performance of cloth simulation filtering algorithm and slope filtering algorithm is better.(2)A super voxel-based point cloud filtering and building extraction method is proposed.The principle of super voxel construction for point cloud data is described,proposed region growth algorithm taking into account neighboring super voxel properties and clustering method based on neighboring concavity,the similarity of each super voxel is judged to achieve the region growing merging in the next layer,and the concavity of super voxel clusters and neighboring super voxel clusters is judged to achieve the merging in the next layer,and finally the ground,vegetation and noise are filtered by calculating the geometric features of the clustered objects themselves to obtain the building point cloud.The experimental results show that the filtering performance of this algorithm is better than other common filtering algorithms,and the building extraction results are better than those of professional point cloud processing software.(3)Adaptive Alpha Shape algorithm is used to extract the contour boundary points of each roof plane of the building.Firstly,we use the super voxel-based region growth algorithm to segment the roof planes for the single building point cloud data,and then use the adaptive Alpha Shape algorithm to extract the contour lines of each roof plane after the building segmentation,and select four types of buildings with different shape structures for the experiment.The results show that the clustering and contour lines of each roof plane point cloud can be well extracted,and the topological relationship between the planes can be obtained,which has good application prospects.