Improvement of crowd flow segmentation using new streakflow similarity
Haiyan Yang, Ping Zhou, Wei Li, Hua-An Zhao · 2017
Recently, studies on monitoring system have been actively conducted. For the purpose of developing an automatic crowd monitoring system, we focus on automatic crowd flow segmentation. The segmentation classifies crowd flows into some different segments according to the flow of moving objects such as people or cars. In this paper, we propose a new similarity evaluation formula which is very important for improving segmentation accuracy. As experimental results, the proposed method on segmentation accuracy is better than previous methods.