Research On The Key Technology Of Airborne LiDAR Point Cloud Building Profile Extraction

Building profile extraction is closely related to urban modeling,road planning and other issues,is an important link in urban information construction.How to obtain the building profile quickly and efficiently is an urgent problem to be solved.Advanced airborne LiDAR system provides a new way to acquire building information because of its advantages such as high automation,high precision and no influence from external conditions.In recent years,building profile extraction from airborne LiDAR point clouds has become a research hotspot.Therefore,this paper focuses on the three key technical links: point cloud filtering,point cloud classification and building profile extraction.The main research contents and results are as follows:1.A point cloud filtering method based on multi-constraint connected graph segmentation is proposed.Point cloud filtering is the process of separating ground object points from ground points,and it is an important part of point cloud post-processing.However,the filtering effect of existing point cloud filtering methods is poor in terrain relief.Aiming at the problem,this paper divides the point cloud by constructing connected graph,and achieves the acquisition and screening of ground seed point sets based on ground coverage rate and grid elevation.Finally,the ground point set was optimized based on the distance between the points and the adjacent ground seed point set.The experimental results show that the filtering method can achieve good filtering effect for different terrain environment,and the average total error of 15 groups of sample data is 5.44%.2.A point cloud classification method based on multi-element eigenvector fusion is proposed.Due to the existing point cloud classification algorithm by the problem of how to extract effective feature vector,this paper based on the primitive using the characteristic value,and surface elevation information to extract the feature vector,based on object primitives extraction height and object bounding rectangle largest eigenvector,and combined with color information of the point cloud data using random forest classification,Random forest feature selection algorithm is used to remove redundant vectors.The experimental results show that the classification accuracy of multi-primitive classification method is higher than that of single primitive classification method.3.An improved Alpha-shapes building profile extraction method is proposed.The Alpha-shapes algorithm has the advantages of simple principle and easy implementation,but the profile obtained by the algorithm is jagged,which cannot reflect the regularity of the building.Aiming at this shortcoming,this paper firstly selects the initial profile points extracted by Alpha-shapes algorithm using RANSAC algorithm,then determines the key profile points using Douglas-Peucker algorithm,and finally realizes the accurate profile extraction through forced orthogonal optimization.Three groups of point clouds with different shapes and sizes of buildings are used for experimental analysis.The experimental results show that the improved method proposed in this paper can obtain more accurate edges of buildings and effectively overcome the jagged edges of the traditional Alpha-shapes method.