| As the main source of 3D spatial data for cities,the oblique photogrammetry model can provide a unified spatial data reference in the construction of digital twin city.There have been many studies on web-based 3D GIS platforms for rapid visualization of oblique photogrammetry models.However,with the continuous increase in the fineness and data volume of oblique photography models,the existing 3D platform is gradually unable to meet the needs of large-scale and high-fidelity visualization of oblique photogrammetry model.The rendering of 3D scenes in the game industry is not only fluent,but also realistic,its characteristics match the requirements of digital twin cities.Therefore,in order to improve the rendering effect of the 3D GIS platform,the 3D data management capabilities of GIS technology and the 3D rendering capabilities of game engines can be combined,and the real 3D scene can be constructed based on the game engine.Geographic information system researchers have researched and realized large-scale oblique photogrammetry model visualization from aspects of data organization,scene scheduling,loading methods,and memory management,but there are still some problems for game engine platforms,for example,1)Based on the pyramid model of regional block,when organizing and visualizing multiple tile quadtrees in the model,there is a problem of low efficiency of node indexing in large-scale scenes;2)The selection and scheduling parameters of tiles depend on the geometric range of tile nodes during visualization,but the parameters are not accurate enough when tile quadtrees divided by different regions are loaded at the same time;3)In the process of real-time rendering,when the rendering range changes rapidly,there may be cross-layer parent-child nodes in tasks that are not loaded in multiple frames,that is,tile nodes with different levels and different degrees of fineness in the same tile tree and having "ancestor-descendant" relationship,which takes up too much GPU rendering time.After loading is completed,there will be cross-layer parent-child nodes in memory,resulting in memory waste.In view of the above problems,this paper makes the research on the fast visualization technology of oblique photogrammetry model based on unreal engine,and the main work is as follows:1.To solve the problem of data organization,this paper proposes a data organization method based on virtual reconstruction quadtree,which combines all quadtrees in the visual scene and reconstructs virtual nodes according to the quadtree structure,thus realizing the unified organization and index of large-scale oblique photogrammetry models.This method speeds up tile indexing.2.To solve the problem of inaccurate scheduling method,this paper introduces a parameter to measure the area of triangle mesh in unit pixel,and uses it as the threshold for real-time scheduling of tile nodes,so as to realize a unified data scheduling method for oblique photogrammetry models that does not depend on the geometric range of tile nodes.This method solves the problem that tiles cannot be loaded correctly when the geographical divisions are different.3.To solve the problems of loading and memory management,this paper proposes a quadtree coding method based on binary,and a binary bit operation method for quickly calculating the parent-child relationship between tile nodes.On this basis,we propose an asynchronous loading and memory management strategy based on parent-child relationship,which is helpful to reduce the data loading time and memory consumption in the visualization process.In order to demonstrate the effectiveness of the proposed method,we implemented a visualization platform for large-scale tilt photogrammetry model based on Unreal Engine(UE),and conducted three sets of comparative experiments under the premise of fixed roaming path.The results show that our methods can speed up the indexing of tile nodes and significantly improve visualization fluency and reduce memory usage. |
PDF Full Text Request