Video-based human abnormal behavior recognition and detection methods

Dec. 21,2021
Print Language:

As an important branch of computer vision, abnormal behavior recognition and detection technology has been widely applied in many fields, such as intelligent security, medical monitoring, and traffic control. The definition and discrimination methods of abnormal behaviors are closely related to scene factors. In practical applications, it is vital to ensure high warning accuracy by choosing appropriate methods of feature extraction and abnormal behavior recognition and detection according to the characteristics of different application scenarios.

In view of this, Wang Li et al. from the North China University of Technology reviewed the research status on human abnormal behavior recognition and detection based on computer vision. They presented the definition and characteristics of human abnormal behaviors and listed typical scenes to which the abnormal behavior recognition and detection technology was applied and the common abnormal behaviors therein. On this basis, discrimination methods of abnormal behaviors were analyzed. Moreover, feature extraction methods were classified and summarized, which directly influenced the method selection and accuracy of subsequent behavior discrimination. Then, the abnormal behavior discrimination methods were expounded on from two aspects of behavior recognition and abnormal detection. The public dataset common for abnormal behavior detection was provided, and the performance of some representative algorithms on the dataset was also given, with a correlation analysis made.

Depending on the analysis and discussion on the existing technology, they provided an outlook on the future research direction in the field. Video-based human abnormal behavior recognition and detection methods can be applied to the multi-feature fusion of human abnormal behavior recognition and detection, which is conducive to describing behaviors and discriminating abnormal ones more accurately. They can also be applied to online behavior recognition, such as action segmentation, abnormal behavior prediction, and online behavior detection. Meanwhile, they will play a role in the research on anomaly detection, environment relevance of abnormal behaviors, and abnormal behavior prediction.

The research is supported by National Natural Science Foundation of China, National Key Research and Development Program of China, and Natural Science Foundation of Beijing. The research findings are published in the journal Control and Decision entitled Overview of video based human abnormal behavior recognition and detection methods (http://kzyjc.alljournals.cn/kzyjc/article/abstract/20220102).

Corresponding Author: WANG Li
Email: wangli939@ncut.edu.cn
CNKI Press Officer: LI Jingjing YANG Na
Email: ljj6806@cnki.net yn6791@cnki.net

Disclaimer: Some of the images in this website are derived of the public network, whose copyrights still belong to the authors. If there exist any infringements, please contact us to delete them