Wushu Competition Field Segmentation Based on Expectation-Maximization Algorithm

Abstract

This paper proposes a valid Wushu competition field segmentation algorithm based on expectation-maximization algorithm. Firstly, Wushu competition image is modeled by Gaussian mixture model. Secondly, the parameters of the Gaussian mixture model are optimally estimated and several candidate thresholds are obtained by expectation-maximum algorithm. Finally, convex sets are segmented through the convex hull algorithm corresponding to candidate thresholds. Then Wushu competition field is inferred corresponding to the convex set with the maximum density.