Crowd monitoring based on Face Orientation Estimation

摘要

Aim to the different needs, crowd monitoring and control is regarded to the key of modern surveillance. But, existing methods are still not good enough to be used on the large-scale spot, for the reasons of effect and efficiency. In fact, witnesses, who stay on the spot, practically act as "smart" sensors to the incidents. Thus, to detect their responds to events are effective to surveillance. Among all kinds of human actions, face orientation is easy to be detected; meanwhile it also implies plenty of immediacy with respect to the event. This paper presents a method of crowd monitoring based on face orientation estimation. According to preliminary experiment, it is proved valid and feasible.