A review of deep learning-based crowd abnormal behavior detection
Xin Li, Liuguang Song, Luyu Dong, Yuqi Sun · 2022
Video surveillance plays a crucial role in improving people's well-being and reducing social crime rates. In order to assist observers to improve the efficiency of surveillance, a large number of scholars have used deep learning to construct models for abnormal behavior detection by learning positive and negative samples. In this paper, we summarize the common methods of abnormal behavior detection from both traditional and deep learning perspectives, introduce the basic principles and steps of detection, and the more commonly used public datasets. The current state of research at home and abroad in recent years is analyzed, the problems faced by anomalous behavior detection are summarized, and finally the future direction of anomalous behavior detection is prospected.