Streamline Selection around Critical Points of 3D Flow Fields by the Multi-View Clustering

Abstract

When visualizing streamlines around critical points, the complex and diverse characteristics of the flow field and the possible existence of common points or symmetries among streamlines may lead to the failure of conventional geometric or similarity-based selection methods. Therefore, a data-driven streamline selection method around critical points, MvCcp, is proposed. It is a method based on multi-view clustering algorithm. By voxelizing the flow field with different granularity, the location distribution view and the geometric feature view data based on the histogram of their distance fields are generated, and the streamlines are selected by the multi-view clustering algorithm. The qualitative visualization of the 3D flow field around six typical critical points such as HalfCylinder was compared with three other typical selection methods, and the comparison experiments based on quantitative metrics such as MSE, PSNR, SSIM, AAD showed that MvCcp had more excellent and more stable performance in all experimental 3D flow fields.