A Lightweight Visualization Method Of 3D Real-scene Model Considering The Levels Of Detail

With the development of geographic spatiotemporal data informatization and automated collection methods,the data scheduling and analysis carried by the real-scene 3D scene visualization platform continues to expand,and higher requirements are put forward for the efficient and dynamic visualization of 3D model data.In the limited computing resources and complex network environment,multi-level detail models with different richness of details according to the perspective change have been widely used in the 3D visualization platform.Through on-demand scheduling and rendering of multi-level detail models,the number of scene models and detailed features are jointly retained,and the visualization efficiency is greatly improved.However,the current multi-level detail model visualization methods have problems such as inefficient scheduling and rendering of high-detail level models,and the model rendering time fluctuates greatly with network and device performance differences,which affects the integrity and visual consistency of the 3D visualization scene.In response to the above problems,this paper proposes a lightweight visualization method of real scene 3D models that considering the characteristics of multi-level geometric details.The multi-level detail point cloud model is constructed through compressed sampling based on the real scene3 D model,and lightweight point clouds are used to replace the real scene at some scales and perspectives.The visualization of the 3D model has improved the visualization efficiency and visual effects of the 3D model in some scenes.The main content and research results of this article are as follows:(1)Propose a lightweight point cloud sampling method that considering geometric details.Aiming at the problem that the visualization efficiency of the real 3D model under high level of detail is not high,and the main structure is easily lost when the degree of simplification is too high,this paper uses the Poisson Disk sampling method to construct a point cloud model that considering the geometric details of the real-scene 3D model.Model to improve and optimize the visualization of a single real-scene 3D model.In order to obtain a uniformly distributed point cloud with low redundancy,this paper designs a comparative experiment to compare the sampling results of the random sampling method,the longest edge sampling method and the Poisson Disk method under different data sets and sampling points.The results show that the Poisson Disk method It has significant advantages in sampling uniformity and reducing data redundancy.(2)Based on the sampling results constructed the point cloud Levels of Detail model with texture color.In order to obtain a model closer to the characteristics of real ground objects,the point cloud is texture and color mapped using the triangular center of gravity method.Subsequently,by using the voxel down-sampling method to dilute the model layer by layer,a multi-level detail model of the point cloud is constructed.(3)Propose a multi-level detail model organization and management method base on mixture of quadtree and KD tree.Aiming at the problem of a single spatial data division and organization method that is difficult to considering the spatial distribution characteristics of the model,resulting in poor consistency between data scheduling and visualization efficiency,it is proposed to use a quad tree at the low-detail level and a KD tree that considering the number of model vertices at the high-detail level.The method of division makes the data distribution between model nodes more even,and the completeness of the scene is better when the model is visualized.(4)Based on the above research results,carried out a scale-adaptive lightweight visualization comparative test and prototype system.First of all,on different data scales,it is verified that the point cloud model has higher scheduling and rendering advantages than the3 D real-scene model.Secondly,developed a prototype system based on the Web client.Combined with actual data tests,it was shown that the system can run stably at 60 frames per second,visualization effects was good.