| With the application of Virtual Reality and Visualization, it is very necessary to build thousands of complicated and real 3D models in a virtual scene which contains unexpectedly large 3D scene. It is very difficult for a large-scale complex scene to build the 3D models, manage the data of the 3D models and real-time interact and walk-through in the scene. Therefore, two pivotal problems must be solved in the developing process of a large-scale scene's walk-through system: how to represent a large-scale scene, and how to realize real-time interacting and walk-through in a large-scale scene. And there is no good method to solving the two problems at present. This paper addresses the representations and fast rendering techniques for large-scale scene further.Based on the research of convecting method of 3D model, this paper proposes a new transplanting method of polygon model made by Maya: developing a Maya exporter plug-in, then extracting the data of some 3D models from Maya environment secondly, and organizing and saving these data according to real-time requirement.Consequently, this paper designs a M3D data model with standardized, optimized and possessed spatial data structures. Being applied to the large 3D virtual scene development, this data model conduces not only to fast building thousands of complicated and real 3D models in a 3D scene walk-through system, but also to real-time rendering fine 3D models.Based on the research of the horizontal and the visual characteristics, this paper proposes a self-adaptive contribution index β based on time, space and illuminance. As a result of taking advantage of β, this paper implements Occlusion Culling based Part-of-Detail Horizon (PODHOC) algorithm in order to speed up the rendering of large-scale 3D scene with the support of M3D model.Finally, based on M3D data model and PODHOC+LOD fast rendering algorithm, a large-scale scene walk-through system prototype has been realized using VC6.0 and OpenGL1.2. |
PDF Full Text Request