Enhancing Museum Security with Advanced Scene-Based Action Recognition Techniques

The automated detection of anomalous behavior in computer vision has become increasingly critical, enabling security personnel to swiftly and accurately identify potentially malicious actions in real-time and respond promptly to potential threats. This paper presents a novel approach to detect anomalous behavior in surveillance videos captured inside museums. Our methodology involves the use of two highly effective deep convolutional neural networks for human action recognition, specifically designed to identify suspicious behaviors. Moreover, due to the scarcity of publicly available data, we created a supervised dataset of video samples. To the best of our knowledge, this study is the first to construct an anomaly detection system for a museum using an action recognition network. Our experimental results highlight the efficacy of our method, underscoring its potential as a valuable tool for safeguarding museum heritage.