The Research On Distributed Virtual Geographical Information System

With the development of Web GIS technique and distributed computation and computer graphics, distributed virtual geographic information system (DVGIS) becomes one of the major areas of GIS research which extend distributed virtual geographic information system when integrated with virtual reality. Distributed virtual GIS, one of the.,most important research fields, will meet the demands of National Spatial Data Infrastructure (NSDI) and "digital earth". Distributed virtual GIS, which contains all kinds of spatial information and cultural information, is the basic framework of "digital earth".In this dissertation, the key technologies of distributed virtual geographical information system are discussed. This dissertation is to explore the integration of network GIS and Virtual Reality. Level-of-Detail (LOD) model is an efficient method, which can be used to improve the speed of 3D roaming and reduce the data volume of 3D rendering, so the dissertation focuses on the LOD methods how to be used in distributed network. The main contents of this dissertation are as follow:In Chapter 1, because the DVGIS is the integration of virtual GIS and network, the development trend of network GIS and research progress of virtual GIS are reviewed. So, we investigate some background and application domains of DVGIS. In terms of aspect of techniques, the system architecture of DVGIS, network-based the real-time rendering of massive 3D spatial data and shared space and operation for multi-users are presented.In Chapter 2, Comparing with the traditional application architecture of distributed GIS, the network model - CORBA-based Distributed Network Model (CDNM), which is a clearly structured and easily scalable network distributed model, is proposed. And CDNM-based system architecture is also introduced.In Chapter 3, we discuss the terrain modeling with continuous of leve!-of-detail, and present a new network terrain leve!-of-detail model. The network terrain LOD model includes two LOD model. The first LOD model is static terrain LOD model at server side. The second LOD model is dynamic terrain LOD model with is real-time constructed at client side. A quadtree based optimization algorithm for dynamically generating continuous levels of a given complex landscape in real-time is developed. In this algorithm, the conception of view-dependent dynamic nod error evaluation method is proposed. The way of recovering the cracks intrinsically caused by quadtree data structure is also improved. The network architecture of dynamical terrain is also presented. To increase the perceived realism of the terrain, the quadtree texture structure is used to handle the tremendous texture data in network environment. Experimental results show that these methods proposed can meet the demand massive terrain data real-time rendering at the multi-user's network environment.In Chapter 4, the researches about the 3D city model are review and the key problem of 3D city visualization is point out. After that, we discuss how to construct distributed the 3D city scene, andpresent a new network 3D city LOD model.Considering the network environment, comparing with other 3D city model, the point-indexed spatial data model, which is used less data volume to reconstruct the 3D city scene. Based on the point-indexed spatial data model, the network 3D city LOD model can fulfill the requirements of 3D reconstruction and the application of level-of-detail in the visualization of city scene.Image data is used to map texture on the surface of 3D building model, which is very important in 3D city model; it can enhance the reality of 3D city scene. According the user's observation habit, the building texture mapping is the farther part of 3D building model to the viewer needs lower resolution image on its surface, the nearer part of 3D building model to the viewer needs higher resolution image. Therefore, the multi-resolution LOD texture data structure is proposed in this dissertation.To further accelerate 3D city rendering, our approach integrates view frustu...