Comparative Analysis of Suspicious Activity Detection Techniques in Surveillance Videos

The need to detect suspicious activities automatically in surveillance videos has become a significant challenge in recent time and became a hot topic for researchers to find and develop new and better techniques to solve this problem. This paper provides a detailed review of several papers and the methods used in those on anomaly detection in video surveillance using different machine learning and deep learning techniques which have been explored till now and gave promising results. CNN-based models, spatiotemporal autoencoders, object-centric autoencoders, and transfer learning-based algorithms are among the approaches examined in this paper. A comparative review of their techniques, accuracy, computing efficiency, and real-time application is presented.