GPU加速3D流场特征提取与多分辨率绘制

GPU-Accelerated Feature Extraction and Multi-resolution Visualization for Complex 3D Fluid Field

由于视线方向上的网格单元前后相互影响,导致3D流场可视化面临遮挡和混乱问题,为此提出一种基于流场特征的多分辨率绘制方法.首先利用基于GPU的BP网络流场特征提取方法对流场典型特征或用户关注的新特征进行选取、训练和识别;在此基础上,利用Voronoi图技术对特征数据构造特征树;最后基于鱼眼视图多分辨率技术进行绘制.对绘制和性能进行测试的实验结果表明,该方法能有效地提取流场特征,降低遮挡和混乱对可视化效果影响.

Due to the mutual influence of grid points on the view direction, visualization to a 3D fluid field often has problem of occlusion and cluttering. In this paper, we propose a novel approach for solving the occlusion and cluttering problem based on multi-resolution rendering. By the solution, first, a new fluid feature extraction method is presented by taking advantage of the strong non-linear ability of the neural network. Then based On the flow feature detected, a Voronoi graph method is used to organize field data. Finally, the field data is visualized by the ＂fisheye views＂ method. The test result of the visualization effects and performance to our method shows that our proposed approach is feasible and efficient in solving the occlusion and cluttering problem.