【作者】 周迪斌；

【导师】 郑耀；

【作者基本信息】 浙江大学 ， 计算机科学与技术， 2008， 博士

【摘要】 在实际的科学可视化操作中,通常有大量的多元数据需要处理,而且这些数据往往是不规则的,包含大量的采样点和时间采样步。如何有效地高质量可视化这类数据一直是可视化研究的难题之一。不断增长的硬件性能持续地推动了矢量可视化方面的研究工作,尤其是基于纹理的矢量可视化技术(纹理法)。纹理法能形成具有特定特征的纹理,具有很好空间时间连续性,能致密地表征流场,又不会忽略矢量场的任何细节,在矢量可视化研究中具有巨大潜力,应给予更多关注。高质量的可视化一直是可视化研究人员追求的重要目标之一。本文研究了基于纹理的高质量矢量可视化算法,并从时间相关性,空间相关性,粒子流动性和反走样控制四个方面详细分析了算法的主要影响因素及其改进策略。目标是得到一种比较通用的高质量矢量可视化算法,适合于平面,曲面和三维矢量场,包括静态和动态的矢量场。算法设计上应该充分利用可编程图形硬件以实现实时绘制。本文研究的重点包括:1.在空间相关性方面,传统纹理法往往存在流线不清晰、纹理质量差的缺点。为此,本文提出了有效的纹理增强策略,即在传统纹理法上引入垂直流线方向的1D高通滤波。实验表明该策略能显著地增强纹理间的对比度,又不影响流线本身的连续性。2.在时间相关性方面,本文采用纹理的循环卷积实现流场动画,并详细分析了帧间流线抖动与高通滤波之间的关系,确认了合理的滤波参数范围,提高了动画质量。3.在粒子流动方面,针对矢量场边界流入区域往往存在粒子运动趋势不明显的问题,提出了粒子卷积补充策略用以增强该区域的属性变化。其作用相当于外界随机粒子的流入,实验表明该方法显著地改善了流场属性变化程度,使得流场运动趋势易于观察。4.在反走样方面,研究了一些改善纹理质量的策略。考虑到卷积纹理部分区域存在属性强度突变的走样,引入了随机过程以减少此类走样,如采样抖动和采样距离随机变化。考虑到某些纹理走样与采样区域的矢量方向有关,引入滤波距离随矢量方向变化的策略,以减缓此类特殊区域的走样。考虑到在流场拓扑中心,往往存在卷积纹理模糊的情况,为此,提出了一种基于角度变化的拓扑中心检测方法,用于检测此类拓扑中心,从而控制卷积过程,避免纹理模糊。5.在三维矢量场可视化方面,本文提出了一种基于纹理的高质量三维矢量场可视化算法,采用了基于流面的一维高通滤波策略,有效地提高了流线的对比度,改善了纹理质量。同时结合权重显示用户感兴趣区域,避免了因卷积数据过多而导致的流线过于紊乱和相互遮挡的情况。6.在基于纹理的三维流场可视化算法基础上,提出结合Flow Volume矢量可视化算法。由于Flow Volume是粒子沿流场平流而成,其外表面是真实的流平面,所以绘制纹理能获得更加清晰平滑的流线。算法采用了一种加速策略,即仅对显示相关的纹理单元进行卷积,极大地减少了所需卷积的纹理数量,从而显著提升了算法绘制速度。综合以上技术研究,本文实现了一个基于纹理的高质量矢量可视化系统,支持动态和静态的二维、曲面和三维矢量场可视化。系统已被用于可视化计算流体力学数据,包括超燃冲压发动机的燃烧过程和仿生鱼的流场等仿真数据。更多还原

【Abstract】 Flow visualization has been a very attractive component of scientific visualization for a long time, where usually very large multivariate datasets require processing. These datasets often consist of a large number of sample locations and several time steps. It is always a challenging task in scientific visualization to effectively visualize these datasets.The steadily increasing performance of computers has recently become a driving factor for a reemergence in flow visualization research, especially in texture-based techniques. These techniques can provide a complete dense representation of the flow field with high spatio-temporal coherency, and ignore none of details in a flow field. They are very promising in flow visulization and should be paid more attentions to.The high quality of flow visualization is pursued by many researchers in scientific visualization. This thesis presents an algorithm of high-quality texture-based vector visualization, analyses the influential factors of its rendering quality and amendatory schemes from four aspects, namely spatial coherency, temporal coherency, particles fluidity and anti-alaising. The objective is to get a general-purpose high-quality algorithm for 2D, 2.5D and 3D vector fields, including the steady and unsteady flows. The programmable graphics hardware should be made best use of to realize real-time rendering performance.The main contents of this thesis include:1. Considering the shortcoming of traditional dense texture-based methods with unclear streamlines and low-quality textures, a scheme of streamline enhancement is presented in this thesis, by applying a 1D high-pass filter to flow textures along the orthogonal flow direction. Results show that it can remarkably enhance the intensity contrast among streamlines while have no influence on the continuity of a single streamline.2. Flow animation is achieved by iterative convolution on textures, the relationship between the streamline dittering and high-passing filtering is studied in detail, and proper filter parameters are acquired to improve animation quality.3. Considering the unobvious particles flowing in the inflow boundary regions, a scheme of boundary supplement is used to modulate the characteristic of particle flows, which can be regarded as the inflow of outside particles. Practice shows that the scheme can obviously improve the characteristic of the total flow field and make the particle flows easy to be observed.4. Some shemes are studies for anti-aliasing. To reduce the aliasing of property break in the convolved texture, some random processes are imported, such as sampling jitter and random sampling distance. To reduce the aliasing which often occurs in regions with special orientation, an orientation-based sampling scheme is designed. To reduce the texture blur in regions close to critical points of a flow field, a method based on the local orientation change is designed to detect these critical points, thus the convolution process can be controlled and the texture blur can be avoided.5. A texture-based high-quality 3D flow visualization algorithm has been presented, a streamsurface-based 1D high-pass filter is adopted to enhance the intensity contrast between 3D streamlines and improve the display quality of streamlines on streamsurfaces. In combination with the importance function, our algorithm is capable of displaying interesting or characteristic regions, thus avoid too much occlusion and clutter issues.6. A volume-based 3D flow visualization algorithm has been presented. Due to the surfaces of flow volumes aligned with the flow by definition, our algorithm is able to aquire high-quality convolved textures with clear and smooth streamlines. A scheme is adopted to accelerate the rendering speed by convolving the voxels related to the final display while neglecting the others.Based on these studies, a system of high-quality texture-based flow visualization has been developed, which supports the visualization of 2D, 2.5D and 3D vector fields, including the steady and unsteady flows. The system has been used for the visualization of CFD datasets, such as combustion in a scramjet and flows around a bionic fish.更多还原