| Virtual environment, one of the most important research areas in virtual reality technology, is an interactive scene simulation environment generated by computer, which acts on users through vision, hearing, feeling etc. so as to make them feel personally on the scene. The early technology of virtual environment generation is mainly about geometrical modeling, which causes some problems such as low efficiency in scene modeling, large calculation, poor real-time, and incapability in rebuilding of complicated characteristic of various natural entities. In recent years, a method called IBR(image based rendering) based on real scene image is proposed by some international VR(virtual reality) experts, which generates virtual environment by real scene image and solves traditional problems effectively. The technology based on image doesn't depend on geometrical model but uses a group of actual scene images obtained beforehand, encode them, and after proper adjustment generate a new view by different viewpoints, which realizes full virtual environment wandering and reflects entities' authenticity. Therefore, research on such a subject is of both theoretical and practical importance.In this dissertation, with virtual battlefield environment as application background, several key technologies in virtual environment generation based on image are studied in order to instruct realization of virtual battlefield environment, which are detailed as follows:(1) The development process and realization methods are analyzed first, then the basic structure and key technologies of virtual battlefield environment based on image are presented, and finally a basic framework is designed.(2) The general principle of camera calibration is studied. After the discussion of an active vision algorithm of camera calibration, a non-linear optimizing algorithm is designed. Another algorithm is proposed to find the linear solution of inner parameter matrix according to polar coordinate, and its application in 3D scene rebuilding is also discussed.(3) Mosaic technology of panoramic image is studied based on overlapping image. The advantages and disadvantages of general algorithms are discussed first. Then we present an automatic mosaic method for color image, which combines feature block separation by color difference with feature line segments. A multi-view imagemosaic algorithm is realized and optimization rule is improved. An idea to implement mosaic according to MPEG video file is presented and image extraction and mosaic for MPEG video are also realized.(4) The basic theory and general algorithm of image and view morphing are studied. An image warping method is implemented, which generates control point grid by uneven distance interpolation. The theory of view morphing and view interpolation according to view point proposed by Seitz and Dyer is discussed and the experiment result is analyzed in detail.(5) The policy of image organization and dispatch is discussed in this paper by the combination of color relativity, vision psychology and LOD(level of detail) technology. Taking image sorting with central moment and color histogram as the research focus, we adopt color palette based on binary tree, which reduces the number of color effectively and keeps visual effect well. LOD model is established based on both the analysis of the function of vision characteristics in image organization from the point of view of psychology and physiology and the combination of distance, focus region and brightness features , which provides a good technological foundation for wandering realization.(6) The framework of a wandering system is established, which includes the design of architecture structure, function structure and dispatch policy, etc. Two types of wandering frameworks, Grid Plane Wandering and difference layer(according to depth) wandering, are established using LOD region division and layer-segmentation principle respectively. Buffer technology is applied to solve the real-time problem. The large scope scene wanderin...
|
PDF Full Text Request