Abstract :

Crowd density estimation and threat monitoring play a crucial role in ensuring public safety in crowded environments. Traditional surveillance systems rely heavily on manual monitoring, which is inefficient and error-prone. With the growth of smart cities, automated crowd analysis has become essential. This work proposes a deep learning–based framework using YOLO (You Only Look Once) for real-time crowd density estimation and threat detection. YOLO enables fast and accurate object detection from video streams. The system detects and counts individuals in a scene to estimate crowd density levels. Additionally, abnormal behaviors and potential threats are monitored using object interaction and motion patterns. Video frames are processed continuously to provide realtime insights. The proposed approach is capable of operating under varying lighting and environmental conditions. Crowd density is classified into different risk levels. The system supports early warning alerts to prevent ove