Adaptive Visiualization Of Landslide Disaster Emergency Scene Driven By Tasks

China is a landslide disaster-prone country,there are great economic and lives caused by landslide disasters each year.The construction of virtual landslide disaster scene not only can support the relevant users perceive disasters in virtual scene,but also represent the geographical environment and simulate the disaster evolution process,which can improve the information management ability of landslide disaster emergency.However,the current relevant researches don't consider the multitype users,diverse terminals and multilevel visualization tasks in disaster emergency,have failed to accurately represent the features of the disaster scene objects and their relationships.In addition,there is a lack of consistent semantic description between disaster scene data.These problems result in poor intelligence,adaptive abilities and low efficiency of disaster scene construction,which lead to hardly meet the requirements of multi-level visualization tasks.To address above problems,this paper proposes an adaptive visualization method of landslide disaster emergency scene driven by tasks.Key technologies and methods including the construction of a knowledge graph of landslide disaster emergency scene driven by tasks,the intelligent optimal selection of landslide disaster emergency scene data guided by knowledge graph,and the efficient visualization of landslide disaster emergency scene for multilevel tasks are discussed in detail,so as to achieve the adaptive visualization of landslide disaster emergency scene for multilevel tasks.The following is an overview of this dissertation.(1)A knowledge graph of landslide disaster emergency scene driven by tasks was constructed.First,we focused on the characteristics and relationships of the multitype users,multilevel visualization tasks and diverse visualization terminals involved in landslide disaster emergency,and established multilevel visualization task model.Then,the characteristics and relationships of disaster scene objects was analyzed and summarized to construct the schema layer of the disaster scene knowledge graph.Based on the above work,we established the layer and completed the semantic networks of the disaster scene knowledge graph through semantic association measurement and association reasoning.Finally,the efficient storage and expression of the knowledge graph was researched,and constructed the disaster scene knowledge graphs with complete association of all elements in landslide disaster emergency scene.(2)An intelligent optimal selection method of landslide disaster emergency scene data guided by knowledge graph was proposed.First,utilizing the semantic relationships of the knowledge graph,a hybrid collaborative filtering method combining semantic relevance and graph model was proposed for calculating the demand degrees.Then,the semantic similarity degrees were fused to enhance the user historical preference data,and the deep neural network named RippleNet was adopted to calculate the recommendation degrees based on knowledge graph and user historical preference data.Finally,the demand degrees and recommendation degrees are integrated to form a hybrid recommended data set to realize the intelligent optimal selection of landslide disaster emergency scene data.(3)An efficient visualization method of landslide disaster emergency scene for multilevel tasks was discussed.First,combining with the analysis of the visualization terminals performance and network environment involved in multilevel visualization task model,and the mechanisms of diverse organization and adaptive scheduling of scene data was discussed.Then,an optimization rendering method including data lightweight and catching strategies was proposed to support user interaction exploration in efficient and accurate disaster scene.Finally,we researched the method of knowledge self-renewal in the process of user interaction exploration,and ensure the accuracy and timeliness of the landslide disaster emergency scene knowledge graph during the entire of landslide disaster emergency.(4)By integrating the above researches,a 3D visualization prototype system for diverse requirements and tasks in landslide disaster emergency was proposed,and Jinshajiang landslide disaster was selected as the case for experiment.Experimental results show that the proposed method can ensure the recommendation accuracy is stable above 90% for multitype users and multilevel visualization tasks,the visual frame rate on high-performance workstation and intelligent large screen display device is stable at 50FPS-60 FPS,the visualized frame rate on VR device is stable at 80FPS-90 FPS,and the visualized frame rate on smart phones is stable at around 30 FPS.It is proved that the proposed method can meet the requirements of multitype users and multilevel visualization tasks in landslide disaster emergency to construct landslide disaster emergency scene with high efficiency and quality.