Construction And Visualization Of Large Scale 3D Scene Tiles Based On Cesium

In recent years,with the development of UAV,3D laser scanning and other data acquisition technology,the acquisition cost and time cycle of 3D spatial data are effectively reduced,and the data accuracy is improved.However,with the continuous accumulation of large-scale 3D spatial data,how to publish,transmit,visualize and analyze 3D spatial data efficiently under B/ s has become an urgent problem.With the development of webgl technology and 3D web development technology,there are many 3D visualization frameworks.Cesium is one of the excellent 3D visualization frameworks,which can add rich and diverse geographic elements to3 D Virtual Earth scene.Cesium team has designed and developed 3D tiles data structure to facilitate the network transmission and rendering of massive 3D spatial data.Although it is an open source set of specifications,but how to generate 3D tiles,the official did not give a reasonable solution.Of course,Ag I has opened some 3D tiles processing tools,which can process tilt photography data well,but there is no good solution for artificial large models and point clouds.In this paper,through a detailed study of cesium framework and 3D tiles data specification,a large-scale 3D scene tile construction scheme is proposed,and the visualization of massive 3D spatial data is studied.Major work of this thesis is as follows:1.This paper summarizes several popular 3D tile data specifications,and introduces the 3D tile construction technology and visualization research based on cesium in recent years.2.This paper describes the architecture of cesium,3D tiles scheduling and rendering related classes.The data format of gltf model is introduced in detail,including gltf node organization,data access rules and binary GLB file structure.This paper introduces the composition of 3D tiles,including several common formats of tile data,and the relationship and meaning between the attributes of tile data.3.Three dimensional tile construction technology research.A 3D scene segmentation algorithm based on octree is proposed,and the segmented scene is processed by mesh simplification or point cloud thinning,so as to generate scene hash file with multi level of detail.According to the 3D tiles data specification,the hash file is transformed to generate tile data of 3D tiles.Several spatial index structures that can be used in the tile data of 3D tiles and their applicable conditions are introduced.The tile data of 3D tiles is generated by using octree spatial index structure,and finally the 3D tiles data format with multi level of detail is generated.4.Research on massive 3D spatial data visualization technology.This paper studies the scene clipping technology of cesium,and first removes the invisible tile data through back clipping and visual genre clipping.The screen space error SSE is calculated,and the tiles that meet the condition are selected according to cesium's LOD scheduling strategy.According to the principle of cesium quadtree scheduling rendering,a massive 3D spatial data scheduling algorithm based on octree is proposed.By using jump loading,dynamic screen space error,and controlling the number of request file queues,the rendering speed is improved,and the rendering pressure of cesium for large-scale 3D spatial data is reduced.5.The construction of visualization platform.In order to verify the related algorithms,a WebGIS visualization platform based on cesium is designed and developed.The front end is integrated with Vue and cesium to form the client part.The server uses nodejs to publish the system,release the data,process the request,operate the database and so on.MySQL relational database is used to store tile attribute information,which is convenient for attribute information management.Combined with the cesium related spatial analysis function,the map display module,data visualization module and data analysis module are designed,and the related functions are tested,which meet the needs of the project and verify the effectiveness of the related methods in this paper.