Detection of Abnormal Behaviors in Crowd Scene: a Review

Crowd analysis becomes the most active-oriented research and trendy topic in computer vision nowadays. Typically, crowd is a unique group of individual or something involves community or society where the phenomena of the crowd are very familiar in a variety of research discipline such as sociology, civil and physics [1]. Within the crowd, there exist many behavior anomalies or abnormality. There are many ways of detecting these abnormalities such as crowd density estimation, crowd motion detection, crowd tracking and crowd behavior recognition. All of these protocols normally involve three steps: pre-processing, object detection and event/behavior recognition. In this paper, we provide state-of-the-art of crowd analysis from 2000 until now. Based on our analysis from these substantial reviews, we propose a general framework and pattern taxonomy of detecting abnormal behavior in a crowded environment accordingly.