A survey of video sequence analysis in CCTV systems based on deep learning in the context of public safety

The analysis of video sequences in a video surveillance context plays a crucial role in public safety. As the number of surveillance cameras increases, it becomes essential to automate the extraction of relevant information in order to detect and prevent incidents in real time. Deep learning, relying on convolutional neural networks (CNN) and advanced architectures such as YOLO, SSD, R-CNN, and Faster R-CNN, offers remarkable performance in detecting and tracking objects, including vehicles and people. This method improves video processing performance compared to traditional. This work explores deep learning techniques applied to video surveillance, focusing on object recognition, anomaly detection and the identification of suspicious behavior. The aim is to optimize the performance of video surveillance systems to enhance urban safety and facilitate law enforcement intervention in the event of an incident.