Research And Implementation Of Visualization Techniques For Complicated Vector Field Data

Vector field is important conception in science and engineering, which describes directional information of physical procedure and developing trends of states. Vector field data visualization is of great importance. With rapid development of computer and computing technique, more and more large volume vector field data emerge, with increasing complexity, which keeps on imposing greater challenge on visualization. Combined with real application background, we conducted wide research and exploration on visualization techniques for vector filed data.Texture methods are effective global visualization techniques for two dimensional vector field data. In the thesis, based on frequency analysis to texture method, the inefficiency of texture methods is outlined and a new textured streamline method is proposed. The new method has more clear frequency meaning, and can be implemented more directly. Multi-frequency texture can be achieved by texture scaling, which directly represent magnitude of vector.For unsteady data, good visualization method should represent not only status of each instantaneous time steps, but developing process contained in sequential time step. Visualization of data resulting from implosion and explosion is more difficult due to its unsteady grid. In the thesis, this problem is solved by unsteady spot noise method, which utilize temporal and spatial coherence in data. Satisfying images are generated in experiments with implosion data.More complicated is three dimensional vector field data. Volume rendering gains interest because of its ability to reveal wholesome information in a single image. Its most weak points are time cost for irregular data and deficiency in revealing delicate structure in flow. In this thesis, ray casting method is simplified to speed up rendering, and traditional material classification is modified to be guided by curvature. In resulting images, flow structure is emphasized more clearly.The difficulty of unsteady three dimensional vector field data visualization stems from both data representation and management of large volume data. In the thesis, wavelet-difference method is proposed. Without increasing disk cost, time cost of overall time cost is reduced by reducing duplicated computation. With the introducing of wavelet template, data can be represented on varying accuracy level. Results from data reveals, this method is especially effective for data of strong time coherence.Parallel processing is a powerful way to reduce time cost in computation-intense task. In the thesis, effective visualization of steady and unsteady three dimensional vector field data is explored on network of workstation, and gained super-linear speedup onshared-memory MIMD architecture. Based on analysis on shared-memory and distributed memory architecture, a parallel ray casting policy is designed for heterogeneous environment.Computing flow imaging is a new focus in flow visualization. Comparable images are gained through computer emulation of optical flow visualization process. What encumbers it most is the integration of density function along rays in field. The method presents in the paper borrowed idea from volume method of ray casting, and greatly improves calculation efficiency.A visualization toolkit for implosion physics, LcpVis, is designed and introduced.