Task-aware Adaptive Visiualization Of Multi-modal Spatio-temporal Data

Compared with the traditional spatial information system,which mainly deals with the vector data and related attribute data of the physical space,the data processed by the newgeneration spatial information system has typical multi-modal characteristics.The multimodal spatio-temporal data fully characterizes the holographic feature information of the multigranularity spatio-temporal objects in the physical,social and cyberspace,such as the position,geometry,behavior and semantic relationship in the whole life cycle from birth to extinction.Multi-level visual analysis,such as description,diagnosis and prediction,has become an important way to perceive,recognize and control the ternary world of physical-social-cyber.The key to intelligent governance lies in the comprehensive analysis of multi-modal spatiotemporal data and cooperative visualization for decision-making: Firstly,multi-modal spatio-temporal data needs to be fully aggregated,correlated and deeply utilized,then intelligent early warning of urban anomalies,intelligent decision-making of key issues and coordinated handling of major events should be completed through visual analysis.The massive,high-dimensional,dynamic and other characteristics of multi-modal spatiotemporal data determine the diversification and high concurrency of diverse visualization tasks in applications.The scene content and visual representation are highly dynamic,and the storage,computing and rendering resources of the collaborative visualization system need to be efficiently coordinated.Although visualization of large spatial and temporal data has been studied a lot,there are still some points that have not been effectively solved.Firstly.In the face of high concurrency,multi-level and multi-modal spatial-temporal data visualization tasks,the visual representation of scene is too immobilized,which could not match flexible presentation requirements for unknown analysis results in time-space exploration analysis.Secondly.High concurrent I/O and high-performance rendering visualization mechanism is difficult to meet the requirements of high concurrent and diverse visualization tasks.Thirdly.The lack of effective collaboration between high-performance computing environment and diversified client environment makes it difficult to effectively support the new generation of spatial information systems and the application requirements of spatial-temporal data visualization.To address these issues,supported by virtualization and cloud computing,this paper intends to study task-aware adaptive visualization of multimodal spatiotemporal data.First,research on classification and construction of multi-level multi-modal spatiotemporal data visualization task model.Second,focus on task-driven multi-granular storage,computing and rendering resource collaborative scheduling method.Finally,design and construction taskaware multi-modal spatiotemporal data adaptive visualization engine.Desiring to achieve high-concurrency,multi-level visualization on demand.The main research work of this paper is as follows:(1)Traditional spatio-temporal data visualization methods only designed for the highperformance display task of a single spatio-temporal scenario,which can not meet the needs of diverse visualization applications.To address these issues,this paper classified the multilevel visualization tasks of multimodal spatiotemporal data as view-only task,analytical task and exploration task.The classification is based on the basic requirements of human spatialtemporal cognition,the relationship between multi-modal visualization application purpose(display,analysis and exploration),visualization driving force(data-driven,model-driven and interactive driving)and visualization task content(real-time rendering,parallel computing and interaction).The classification is described in four dimensions: multimodal spatiotemporal data,analytical model,human-computer interaction and rendering,and could establish a hierarchical semantic mapping relationship between multi-level visualization task requirements and visualization resources allocation and scheduling,it could provide a theoretical basis for collaborative scheduling storage,computing and rendering of resources,and dynamic construction of visualization scenes.(2)Aiming at the difficulty of handing high concurrency of multi-level and diversified visualization tasks in traditional data-centric spatial-temporal data visualization scheduling mechanism,a multi-level visualization task-driven multi-granularity storage-computingrendering resource collaboration workflow and service chain optimization scheduling method is proposed.Serving storage,computing,and rendering resources into multi-granular services,and design workflow for multi-level visualization tasks.Then research on multi-level visual service chain construction method based on workflow and multi-granular services,dynamically optimize the service chain by establishing an evaluation model for the quality of service,finally,forming task-centered adaptive visual data scheduling mechanism,and achieve rapid response to different levels of visual analysis applications.(3)Based on the above research,a task-aware multi-modal spatiotemporal data adaptive visualization engine prototype system was developed.In the engine multi-modal spatiotemporal data organization storage,data analysis and spatial information visualization are loosely coupled through GIS microservice architecture.In order to meet the requirements of high concurrent visualization applications at different levels,the engine adaptively schedules and combines multi-granularity data and analysis services according to different levels of visualization application tasks.At the same time,the engine provides rapid deployment and maintenance of multi-granularity services,automatically allocates the storage,computing and network resources needed for the normal operation of services on demand,and automatically scales horizontally according to the concurrency of tasks to ensure the high availability of multi-granularity storage,calculation,and mapping services.At the end of the paper,in order to meet the needs of smart city construction,three typical cases are validated and analyzed,namely,the visualization for description of the relationship between urban macro-situation and pattern,the visualization of dynamic simulation and analysis of dambreak flood spatiotemporal process,and the exploration of indoor fire escape scheme.The result shows that the proposed method can effectively provide the multi-level visual application of multi-domain in the whole life cycle of urban overview,planning,operation,maintenance,and emergency disaster response.