Large-scale Vector Data Displaying for Interactive Manipulation in 3d Landscape Map

Displaying large-scale 3D vector data in landscape map is very important, as 3D vector data can provide many important information, such as: precise geographic boundaries, areas, 3D text marks, different attributes identity, precise path and many important invisible information in real world (e.g.: underground things) and etc. In this paper, we present a technique that is capable of efficiently and precisely displaying 2D vector data in 3D landscape map. In order to release CPU from heavy computation, texture-based approach is selected to render 2D vector data on 3D terrain blocks in browsing mode. Geometry-based approach is selected in interactive manipulation mode or precise expression of vector data is required. We use static terrain mesh (STM) instead of dynamic terrain mesh on the base of SOAR algorithm, to improve efficiency of 3D geometry primitives creation from 2D vector data in real-time, and to support interactive manipulation of 3D vector object. In addition, several scales of 2D vector data map are used to build a pyramid of vector data. These strategies made large-scale vector data displaying easy in 3D landscape map. 1. INTORDUCTION The majority of the research works on 3D landscape map emphasize the reconstruction of the real view, which use luxuriant models to express buildings and vegetation in one virtual scene, however, geospatial data (include vector data and its attribute data) expression is generally ignored (Stephen M. 2004). The importance of expression of geospatial data is that invisible data in real view can be browsed, manipulated and analyzed in digital virtual scene. This is also the main purpose of utilization of geographic information system (GIS). In other words, the application of GIS relies on geospatial data. Therefore, if the expression and visualization of geospatial data in 3D landscape map is ignored, then the result utilization would only limit to landscape browsing. Some research works have been done on how to visualize vector data on digital terrain model or in 3D landscape map. They generally can be divided into two classes of method: class of raster method, which project vector data onto one image that is used as a texture map on DEM, and class of vector method which convert 2D geometry primitives into 3D geometry primitives, two classes have different drawback. Generally, raster method can’t avoid aliasing and memory cost problem, e.g. when the terrain mesh close enough to the viewpoint, texture has to be zoomed in, even 1 pixel precise can’t satisfy the expression of vector data. Vector method generally can’t avoid z buffer artifacts problem, as vector data in GIS are mainly expressed on 2D plane, when they were converted into 3D space, line segment and polygon plane cannot match 3D terrain mesh surface in details. In our approach, we use two classes of method respectively in different levels of detail. For the fast visualization of large terrain scene, a pyramid of 3D terrain model is built with levelof-detail (LOD) and indexed with quadtree, blocks of DEM are expressed with static terrain mesh. While one corresponding pyramid of large-scale vector data is built with several scales of map data, vector data blocks in the pyramid are converted into textures in pre-processing or rendered with P-buffer technique in real-time if hardware capability is permitted. Figure 1. Visualization of vector data in Pinggu area (Beijing) When textures of vector data have not stretched large than their original size in rendering process, vector data can be expressed precisely. In our approach, we switch to geometry-based method if a block of DEM move close to the viewpoint or one vector object needs interactive manipulation. 3D geometry primitives are created from static terrain mesh intersects with vertical planes that pass line segments of vector objects.