Research On Airborne LiDAR Point Cloud Filtering And Building Extraction And Reconstruction

With the development of the knowledge economy and information society,the "digital city" represents the new trend of the world and the direction of urban development,and has gradually become an inevitable trend of the future development of society.As the most important part of the city,buildings have become the key to information extraction and analysis.In recent years,many scholars have conducted research on building extraction and modeling based on airborne LiDAR point cloud data.Airborne LiDAR is a new type of active remote sensing technology that can acquire a large amount of high-density and high-precision 3D laser data.Therefore,airborne LiDAR data has become an irreplaceable data source for building extraction and model reconstruction.In this paper,the research and analysis of filtering,point cloud extraction and 3D reconstruction of buildings in airborne LiDAR point cloud data are carried out.The specific contents are as follows:(1)Point cloud filtering method based on cloth simulation.This method is a new airborne LiDAR point cloud filtering method proposed in recent years.Its idea is completely independent of the traditional ground filtering idea,and the effect is very good.In this paper,this method is compared with mathematical morphological filtering and slope-based filtering algorithms.The cloth simulation filtering algorithm has the best adaptability and highest accuracy,but it cannot distinguish the bridges connected to the ground,and it will mistake the bridge into ground points.In this paper,the geometric information of LiDAR point clouds is combined with multispectral images to clearly distinguish bridges and roads.(2)A method for extracting point cloud from buildings based on histogram is proposed.Firstly,the filtered LiDAR point cloud is divided into regions based on the open source point cloud library(PCL).The local normal vector and the normal vector of the point cloud in each cluster obtained after the region growth and segmentation are calculated.Direction cosine and generate a histogram.According to the distribution characteristics of the histogram of the vegetation cluster and the building cluster,the buildings are separated from the non-buildings,and the point clouds of the buildings are extracted.The results show that the accuracy of the building extraction results in this paper is higher,with smaller class I extraction errors and class Ⅱ extraction errors.(3)Improved the method of building roof model reconstruction based on topology map.First,the roof point cloud is segmented by a RANSAC-based clustering algorithm;Second,the building boundary points are obtained using the Alpha-Shape algorithm;the roof line and roof topological maps are obtained based on the intersection of the roof patches;Combined to complete the ridge line expansion,and then obtain other key line segments of the relevant roof surface;Finally,based on the extracted roof key line segments,the closed polygons of each roof surface are constructed,and the wall surface is constructed by the height from the outer boundary of the roof surface to the ground,thereby Construct a complete 3D model of the building.Experiments show that the roof point cloud and the roof model have a high degree of fit,and can correctly reconstruct the ground auxiliary buildings next to the main building,ensuring the geometric accuracy of the building model.