Research On Key Technologies Of 3D Geological Model Rendering In Cloud Environment

3D model is a 3D digital representation of objective things,which has been widely applied in fields like urban construction,disaster prevention and mitigation,medical research,biological science,industrial manufacturing,agricultural production,etc.As a special 3D model,3D geological model possesses the characteristics of 3D model and plays a fundamental role in geological survey,mineral exploitation,underground engineering and smart city construction.With the development of intelligent sensing technology and 3D geological modeling technology,the scale of 3D geological model data increases exponentially.Meanwhile,with the pace of large-scale underground engineering and smart city continuing to increase,3D geological model with fine large scenes is being eagerly required.The rapid growth of data and the refinement of large application scenes bring new challenges to the real-time dynamic visualization of 3D geological models.These challenges are mainly reflected in the new technical problems related to 3D geological model rendering :(1)File-based data management takes geological model as a single entity,which makes it difficult for current mainstream 3D geological modeling software to calculate and present large-scale 3D geological model.(2)The rendering process of a single computer with common configuration cost long time because of the increasing huge computation of large scene visualization,which is difficult to meet the requirements of dynamic visualization;However,the high performance computer is too expensive to widely use,and it is also difficult to meet the needs of dynamic data growth.(3)Although a large number of methods have been tried to do research on parallel visualization for massive geographical scene data,there is no proper framework for massive data management,scheduling and parallel visualization of 3D geological model in cloud environment.This thesis studies the above issues and puts forward the corresponding solutions.The validity of the technology has been proved by the simulation test of cluster cloud environment consisting of 5 computers.The technique has been applied in the construction of 3D geological information and visualization system in transparent Xiong’an.The main research work and innovations of this thesis are as follows:(1)The distributed storage mode of 3D geological model is established to improve the efficiency of data access.Firstly,the data organization mode of two common structures of 3D geological model(3D geological structure model and 3D geological high-precision grid model)is analyzed,and a distributed storage strategy of 3D geological model based on Mongo DB is proposed.Aiming at the characteristics of multi-layer data in z-direction of 3D geological structure model,an octree index mechanism is proposed to improve the efficiency of data scheduling according to the z-direction spatial information and layer information.In view of the regularization characteristics of 3D geological high-precision grid model,this thesis proposes that the grid is identified by the grid serial number rather than the geometric coordinate information stored in the database,based on which a strategy of fast and accurate query of 3D geological grid model by IJK serial number is proposed.The efficient query and rendering optimization algorithm of 3D geological high-precision grid model based on IJK index is supported by spatial query operator of 3D geological model constructed by regular grid.It can effectively integrate the requirements of model data organization and fast query.(2)The efficient rendering optimization strategy of single node 3D geological model is designed to improve the rendering efficiency of 3D geological model.The communication between memory and GPU is the main factor restricting the rendering efficiency of single node 3D geological model.Through the strategies of visibility elimination,LOD establishment,data merging and instance rendering optimization,this thesis effectively reduces the number of drawing calls and communication times.For the 3D geological structure model rendering optimization,the regular block rendering method is adopted,and the LOD model is generated by the geological triangulation network simplification;For the 3D geological grid model rendering optimization,the geological grid data is organized in the form of octree to complete the model coordination and integration.On the basis of improving the rendering efficiency of single node 3D geological model,the overall rendering efficiency of 3D geological model in cloud environment can be greatly improved with the help of effective node scheduling strategy.(3)A texture mapping method of 3D geological model based on machine learning is proposed to improve the texture mapping effect of 3D geological model.Aiming at the lack of geological body texture sample library and unsatisfactory texture mapping effect,a 3D geological texture mapping method using tangent space normal mapping technology and boundary seamless processing technology is proposed.(4)A multi-level distributed SCMP framework is proposed,which significantly improves the overall performance of 3D geological model rendering in cloud environment.SCMP framework integrates the advantages of cluster,GPU,distributed storage,etc.,to maximize the distributed computing ability of existing machines and improve the rendering efficiency in cloud environment.Spark implements multilevel cluster expansion,load balancing among nodes,and high availability.CUDA provides a large number of core-level fine-grained image fusion parallel computing;Mongo DB database engine can support efficient storage and query of massive 3D geological model data.During the rendering process,the rendering task is divided into coarse-grained subtasks,which are completed by corresponding sub-nodes that are uniformly scheduled by Spark.Fine-grained tasks are decomposed and completed within sub-nodes.CUDA multi-core computing is used to improve the computing power of a single node,realize the parallel running of multiple threads within the node,and finally improve the overall computing performance and speedup ratio.(5)A cloud-based 3D geological model visualization platform based on Spark big data technology was built to provide support for large-scale 3D geological model visualization.The platform includs data storage layer,data scheduling layer,cloud rendering layer and terminal visualization layer.The big data of 3D geological model is automatically divided by the strategy of "vertical stratification and horizontal block".In the process of 3D model calculation and 3D visualization,the distributed rendering,centralized fusion and visual display of 3D model can be automatically completed by using the fast parallel computing and task scheduling of “Spark Streaming + GPU”.From the experimental data,the node invocation optimization strategy with“GPU+CPU” can ensure that the frame rate of the four rendering nodes and the end-user scene in the cloud environment is stable at about 35 frames per second,and can achieve satisfactory cluster load balancing effect.