Improved Particle Image Velocimetry Through Cell Segmentation and Competitive Survival

Abstract

A new model of cell segmentation and competitive survival (CSS) is integrated into the standard techniques of particle image velocimetry (PIV). First, a set of initial interrogation fields is identified in the images, and the cells are defined in the field by cross correlation. Each cell is then segmented into smaller groups of matching points with different degrees of correlation. These subcells compete with each other to define the properties of the cell; the winner, in turn, competes with the other cells. Finally, the velocity vector of the field is defined as the displacement of the winning cell's centroid between frames. The algorithm is applied to some real and synthetic particle images, and its results are compared to particle correlation velocimetry and recursive PIV approaches. These experiments demonstrate that the CSS approach is effective and practical.