INTERNATIONAL JOURNAL FOR INNOVATIVE RESEARCH IN MULTIDISCIPLINARY FIELD

Uncontrolled crowd gatherings pose significant risks to public safety, especially during large events, festivals, transportation congestion, and emergency situations. Real-time monitoring of crowd density is therefore essential for effective crowd management and incident prevention. This paper proposes a drone-based crowd monitoring system capable of detecting and counting people in real time using computer vision and deep learning techniques. The system processes aerial video captured by unmanned aerial vehicles (UAVs) and applies a deep learning-based object detection model to identify individuals within each frame. Unlike traditional multi-stage approaches, the proposed method performs simultaneous detection and counting in a single phase, improving computational efficiency and enabling near real-time analysis. The detected individuals are used to estimate crowd density within the monitored region. When the calculated density exceeds a predefined safety threshold, the system automatically generates alerts and notifications to authorities for immediate action. Experimental observations demonstrate that the proposed approach provides accurate crowd estimation and can support public safety management in large gatherings. The system has potential applications in smart city surveillance, disaster response, event monitoring, and urban safety management.