3D Buildings Fast Extraction By Fusion Of VGI Data And Vehicle LiDAR Point Clouds

Buildings are the main places for human activities.Accurate three-dimensional models of buildings can provide support for urban planning,emergency response,and BIM applications,etc.Therefore,the demand for three-dimensional reconstruction of buildings is urgent.Volunteered geographic information(VGI),risen with Web2.0 technology,is rich of semantic content,including a large number of urban buildings information,thus makes low-cost 3D modeling possible.However,the geometric position accuracy of VGI data is difficult to guarantee.Moreover there is a lack of detailed information on building facades.Only the floor level(LOD1)and roof structure level(LOD2)models can be reconstructed.In recent years,mobile measurement technology and unmanned vehicle technology have developed rapidly.The vehicle-mounted laser scanning system can quickly acquire 3D point clouds on the surface of roads and on both sides of the ground.The point clouds have high geometric accuracy and contains rich details of facades,so as an ideal data source of facade structure(LOD3)level models reconstruction.However,due to the complexity of real-world scenes,large amounts of point cloud data,and non-uniform density,existing methods for automatically extracting buildings from point clouds have problems such as high computational complexity,long time-consuming,and poor robustness.It is difficult to achieve good extraction results.In order to solve above problems,this paper proposes a set of three-dimensional rapid extraction method for buildings with VGI and vehicle LiDAR point clouds,which achieves complementary advantages of heterogeneous data and provides effective data support for true three-dimensional reconstruction of buildings.First of all,in order to quickly establish the matching relationship between two kinds of data and then realize the information fusion of the data,this method uses MLS track points and VGI road sampling points for fast registration,thereby simplifying the matching between two-dimensional vector data and three-dimensional point clouds.Then,this paper uses VGI data to assist in building extraction of vehicle LiDAR point clouds.The structured information of VGI data provides prior knowledge of building candidate areas,which can reduce the search scope of point clouds,thereby improving the processing efficiency;Also,The unstructured information of VGI data can not only assist in formulating building extraction rules,to achieve higher accuracy of building target extraction,but also enrich the semantic content of LiDAR point cloud data and facilitate the further realization of 3D model reconstruction.In this paper,the Paris-Lille-3D Benchmark vehicle LiDAR dataset and some point cloud data of Wuhan University areas are used to verification.The OpenStreetmap(OSM)is used as a typical example to illustrate the proposed method and process.The experiment verifies the validity and feasibility of the proposed process.In addition,this paper compares the building extraction results and accuracy evaluation of VGI-assisted and non-VGI-assisted methods.The experimental results show that the VGI-assisted vehicle LiDAR point cloud building extraction method has higher efficiency,higher accuracy,thus effectively improved the time-consuming problem,and the problem of high misclassification and leakage rate.In addition,the work in this paper also shows that VGI data represented by OSM and professional mapping data represented by LIDAR point cloud can achieve higher data value by effective data fusion,and further deepen data fusion between VGI and LiDAR research has great significance.