Using deep learning to analyse behaviour in video surveillances

The use of video survemance has become crucial in, addressing the growing concerns of crime and terrorism in the world, leading to an increased need for automated data processing. This study focuses on the application of Deep Learning (DL) models to extract data from video sequences captured by police cameras. Specifically, it explores the use of DL models for crowd behavior analysis and anomaly detection. The study examines publicly available datasets and cutting-edge studies to classify abnormal behaviors accurately. The practical aspect of the study involves developing a 3D anomaly detection system based on a convolutional feature extractor that combines the strengths of 2D CNNs and RNNs. The proposed feature extractor’s complexity and ability to preserve temporal information improve the quality of data extraction, leading to better classification results than the original model, supporting our hypothesis. Overall, this study demonstrates the effectiveness of DL models in addressing the challenges of anomaly detection in crowds and highlights the potential for future research in this area.