A vortex-tracking method based on multi-attribute fusion
Lin Zhu, Ningbo Guo, Yadong Wu, Liang Deng, Zhouqiao He, Cheng Chen, De Xing · Physics of Fluids · 2024
Tracking time-variant vortex features in unsteady flow fields represents a significant challenge in scientific visualization. In this paper, we propose an innovative new method based on attribute fusion that leverages unsupervised clustering techniques to track time-variant vortex features. This method has the potential for use in large-scale flow field analysis. We apply an attribute matrix integrating the spatial and physical variables of vortex feature points. We then fuse and reduce the dimensionality of this original matrix to create an attribute fusion matrix. Subsequently, we apply coarse clustering to the fusion matrices across time steps. By defining a distribution curve, we derive a similarity measurement matrix among different vortices. Vortex matching is performed based on this similarity measurement to enable the tracking of time-variant vortices. Multiple experimental datasets demonstrate the effectiveness and matching efficiency of the proposed method.