A Three-dimensional Visualization And Optimization Modeling Method Of Landslide Disaster Emergency Scenes Guided By Knowledge

Geological disasters occur frequently in China,in which landslide disasters account for more than 70% of the total number of geological disasters.The prevention and control work of landslide disasters is still facing severe challenges.It is of great significance to carry out scientific research on landslide disasters to improve the comprehensive prevention and control ability of landslide disasters.Virtual geographical environment,as a new generation of scientific experiment and analysis tools,not only can reproduce the spatio-temporal process of landslide disaster,but also promote disaster knowledge sharing,improve users’ cognitive efficiency of landslide disaster.However,the existing method of landslide disaster scene visualization has some problems,such as lack of modeling and visualization knowledge guidance,difficulty in expansion led by content solidification,inaccurate pre-disaster information acquisition,large difference between DEM pre-disaster and post-disaster,low scene rendering efficiency,poor user experience and so on.It is difficult to meet the requirements of dynamic visualization and efficient rendering of landslide disaster scene.To address the above problems,this paper proposes a three-dimensional visualization and optimization modeling method of landslide disaster emergency scenes guided by knowledge.It intends to focus on key technologies and methods such as knowledge-guided dynamic visualization of landslide disaster emergency scene,rapid acquisition the key information of deposit based on smartphones,and optimized modeling of landslide disaster scene considering visual saliency.Based on this,the dynamic visualization of landslide disaster emergency scenes,the rapid acquisition the key information of deposit,and scene optimization modeling are realized.And the modeling and visuzalization operations of disaster emergency scenes are simplified,the readability and rendering efficiency of scene information are improved.So as to promote the sharing and dissemination of landslide disaster knowledge,and povide support for disaster emergency response and management decisions.The major research works are summarized as follows:(1)A knowledge-guided dynamic visualization method of landslide disaster emergence scene is proposed.First,the objects and their characteristics of the landslide disaster emergency scene are comprehensively analyzed.Then,from the perspectives of user needs,scene object relationships and disaster evolution laws,a multi-level semantic constraint rule including information completeness,spatial semantics,information readability and logical relationship is constructed.Finally,on this basis,combined with expert experience,the knowledge template for landslide disaster emergency scene modeling and visualization is established.To realize the dynamic visualization of landslide disaster emergency scenes,and provide demand guidance for disaster data acquisition and scene optimization modeling.(2)A method of rapid acquisition the key information of deposit based on smart phone is designed.Considering the influence of the shooting distance of the stacked 3D scene image,an adaptive image pyramid construction algorithm is designed,the input parameters of the pyramid layers of the SURF algorithm are improved,and a fast-matching model for feature points of the jointly improved SURF and ORB algorithms is established.Then,based on the principle of binocular vision,the sparse reconstruction method of 3D scene constrained by the edge of landslide deposit is discussed.Finally,the interpolation method of the bottom surface of the landslide deposit considering the steepest gradient is proposed,to fits the terrain surface before the disaster,realizes the rapid acquisition the key information of landslide deposit,and provides the disaster data for the optimization modeling of landslide disaster emergency scene.(3)A landslide disaster scene optimization method considering visual saliency is constructed.Frist,the visual characteristics of the human eye and the principle of visual saliency are deeply analyzed.According to the range of binocular frustum,the scene is divided into mixed horizon and peripheral horizon.Then,taking the landslide disaster range as the center,a calculation method of the region of interest of the landslide disaster scene considering the visual saliency is designed,and the mixed horizon is further divided into the region of interest and the region of non-interest.Finally,a scene-level optimization modeling method that combines Gaussian blurring and high pass is proposed to improve the rendering efficiency of landslide disaster scenes and the user’s cognitive experience of landslide disaster scenes.(4)Integrating the above research results,based on computer vision open-source library Open CV,3D visualization engine OSG,Unity 3D,Steam VR and VR interactive device Oculus Quest2,etc.,developed a prototype system for rapid acquisition of landslide deposits disaster information and a prototype system for scene visualization and optimization modeling.Two landslide disasters were selected as cases to conduct an experimental analysis of dynamic visualization,the key information acquisition,and optimization modeling of disaster scenes.The experimental results show that realized the instantiated expression of the dynamic process of landslide disasters from disaster conception,occurrence,development to disaster.Obtained the key information of landslide deposit volume comparable to commercial software,and the efficiency is 1-2 times higher than other commercial software,which has the advantages of high efficiency and convenience.The scene rendering efficiency has been significantly improved,and the average frame rate has stabilized at more than 80 frames per second,which improves the efficiency of scene rendering and the user’s cognitive experience of landslide disaster scenes.The validity and feasibility of this method are fully demonstrated.