Research On Building Feature Extraction And Modeling Based On LiDAR Point Cloud Data

In recent years,laser LiDAR,as a new way to acquire spatial data,satisfies the high precision requirement of distance and spatial resolution,which provides a new data support for building reconstruction research.However,due to the high complexity of real buildings,there are still many technical difficulties and challenges in the reconstruction of the real and accurate building 3D models,such as the low efficiency of data storage and management,the missing of feature data in data compression,the incomplete extraction of feature points,the low degree of modeling automation,and the incomplete modeling of single data.Therefore,the building modeling technology based on LiDAR point cloud data is still in the rapid development stage,and there are many technical problems need to be solved urgently.This paper focuses on building LiDAR point cloud data processing and modeling technology.And the key issues such as point cloud data organization,point cloud data compression,feature point extraction and multi-source data fusion modeling are deeply studied.The main tasks are as follows:(1)The efficiency of point cloud data organization and nearest neighbor search by KD-tree method and Octree method are compared and analyzed.It is found that the KD-tree method is more efficient in the organization and management of building point cloud data.(2)The point cloud data compression algorithm based on uniform slice is improved.Firstly,the point cloud data is sliced in a certain direction,and the slice point cloud is projected onto the middle plane of two slice planes.When the chord height difference of a point is less than the set threshold value,it is deleted as a redundant point,and the whole point cloud is compressed layer by layer using the improved string height difference method.And comparing the existing two methods of point cloud compression experiment,this paper improved the methods on reservation of feature points at the same time gentle part of the larger curvature change will not appear holes by a compression,when high compression rates still can obtain better compression effect.(3)A method for building feature extraction based on normal vector estimation of moving least squares is proposed.First of all,the normal vector of point cloud is estimated by moving least square method,and the normal vector with ambiguity is redirected so that the normal vector of all could points to the exterior of the model.Then the average value of the normal vector angle change of k-nearest neighbor point is used as the significant index to judge the reserved feature point.Moreover,the redundant data is further reduced by sampling the extracted feature point set.By experiments,the clear,concise and complete characteristic lines are extracted for two different building models.(4)To solve the problems of low automation and incomplete models when modeling single point cloud data,the multi-source data fusion modeling technology is deeply studied.Also the coordinate system of multi-source measurement data,the coordinate unification ground laser point cloud + space-surface images fusion modeling are expounded in details.In addition,the modeling method of multi-source data fusion using Smart3 D software is discussed.The experimental results show that using ground laser point cloud + space-surface images fusion modeling can get a more accurate,complete and beautiful 3D model.To sum up,this paper studies the point cloud data processing and modeling technology of building LiDAR.The experiments analysis the point cloud spatial index technology,and improve the method of building point cloud data compression,meanwhile,construct the feature extraction of building point cloud and discuss the method of modeling the ground laser point cloud + space-surface images fusion.The experimental results show that all the methods proposed in this paper have good experimental results.