A Framework for Real-Time Crowd Behaviour Monitoring System Based on Human Activity Recognition in Surveillance Videos

Data from India’s National Crime Records Bureau (NCRB) show that crime rates are rising, which poses a serious threat to social safety and economic stability. Although overall crime rates have somewhat decreased, alarming increases in certain crimes underline the urgency of improving surveillance and security measures. To address this, there is a growing need for proactive, technologically advanced solutions. Surveillance cameras, widely deployed in crowded areas, offer a means to monitor real-time behaviour and identify abnormal or suspicious activities. This paper introduces a comprehensive framework for real-time crowd behaviour monitoring based on human activity recognition in surveillance videos, encompassing data capture, enhancement, feature extraction, models for abnormality detection, semantic interpretation, and alert systems. The framework also includes integration plans with existing crime databases and predictive analytics for more efficient law enforcement. The suggested approach ultimately shows potential in developing monitoring technologies, enhancing neighbourhood safety, and reducing changing security concerns.