Crowd motion segmentation via streak flow and collectiveness

Crowd motion segmentation problem is a hot issue in crowd behavior analysis and action recognition. A number of methods have been proposed to tackle this problem. Among the existing methods, most used flow dynamics to model the crowd motion, with only a few taking the collective property into consideration. In this paper, we proposed a crowd motion segmentation approach based on streak flow and crowd collectiveness, where the streak flow is adopted to reveal the dynamical property of crowd motion, and the collectiveness is incorporated to reveal the structure property. Experimental results show that our method can offer better segmentation accuracy than the state-of-the-art methods on the benchmark dataset.