Flow Field Simplification Based On Clustering

Clustering-based flow visualization is one of hot research topics in scientific visualization. Traditional clustering-based methods for 2D flow visualization suffer from the problem that they are hard to display the original features of the flow field. For this reason, this paper firstly presents a feature-emphasized clustering method for 2D vector fields, where streamlines are placed to partition the original field. Due to the nature of streamlines, this method overcomes the disadvantage of other methods where the flow field can only be divided into convex polygon regions. On this basis, the similarity of streamlines is used as the similarity of field regions, and the regions are further clustered by k-means method. By this means, features of flow fields are well preserved in the results. Finally, a bottom-up hierarchical clustering method using the similarity of streamlines is proposed to overcome the disadvantage of k-means clustering method where the k value needs to be set in advance. The hierarchical clustering also achieves visual-pleasing results while preserving flow features. The main work accomplished in this paper is as the following:(1) A top-down clustering method based on similarity of streamlines is proposed, where the vector at the geometrical center of a field region is used as the region’s representative vector to construct the intermediate flow field that helps to determine the seeds of streamlines. The clusters obtained by this way can preserve flow features.(2) The k-means clustering method is used to study the effects of streamline similarity in flow field clustering. By analyzing the results obtained with different similarities, we find the similarity of streamlines that can preserve flow features rather well.(3) With the similarity of streamlines, a bottom-up clustering method is proposed to overcome the disadvantage of the k-means where the k value has to be set in advance.