Survey on anomaly detection in surveillance videos. (2022)

• Record Type: Journal Article
• Authors: Anoopa, S.; Salim, A.
• Is Part Of: Materials today. Volume 58:Part 1(2022)
• Page Start: 162
• Page End: 167
• Publication Date: 2022
• Subjects: Anomaly detection -- Video surveillance -- CCTV cameras -- Deep learning -- Representation learning; Materials science -- Congresses -- Periodicals; 620.1
• DOI: 10.1016/j.matpr.2022.01.171
• Languages: English
• ISSNs: 2214-7853

Abstract

Abstract: The demand for video surveillance has seen a staggering growth in the last few years due to the rapid growth of urbanization & industrialization. Video Surveillance is a system which observe scenes, activities, behaviour or other kind of information with the help of CCTV Cameras & associated software. The main objectives of video surveillance are detection and tracking of moving objects, loss prevention, traffic monitoring, and capturing variety of real world anomalies. The major challenges in Anomaly detection are extraction of appropriate features, addressing normal behaviour, environment divergence and occurrence of abnormal events in sparse manner. Many algorithms have been proposed in this area for object detection and tracking related applications, however, the problem is still challenging because of dynamic behaviour in anomalies. In this review paper a comparative study of different anomaly detection methods in deep learning and representation learning is performed and the limitations of each method are listed. Also, we outline open research challenges for proceeding further research. This paper will use as a quick reference to the researchers working in the same field.