Reconhecimento e classificação de cenas de risco em vídeos de câmeras de segurança

Abstract

This work presents a system for recognizing and classifying risk scenes in security camera videos, using computer vision and deep learning techniques. The main objective is to develop a system capable of detecting objects and risk situations, such as firearms and knives, in real-time, and generating automatic alerts. The methodology employs the YOLO (You Only Look Once) model, which allows fast and accurate object identification in images, essential for dynamic security scenarios. The obtained results demonstrate high precision in risk detection and classification, with proven effectiveness in practical tests. The research also emphasizes the importance of choosing suitable databases for model training and the need for continuous algorithm improvement to handle scene complexity. It is concluded that the developed system offers an advanced tool for security, with potential for future enhancements.