Deeptracknet: Advanced Crowd Dynamics Estimation with YOLO and Deep Sort Integration

Monitoring and analyzing crowd dynamics is essential for maintaining safety in high-density areas, enabling timely detection and prevention of potential hazards. This paper presents a deep learning-based system for real-time detection of crowd dynamics, ground clearance monitoring, and automated alert generation in environments such as malls and public spaces. The system comprises three primary components: video preprocessing and crowd behavior analysis, ground clearance detection, and anomaly detection with alert generation. Using object detection models like YOLOv8 and SSD for crowd identification, optical flow techniques for motion pattern analysis, and dynamic thresholding for anomaly detection, the system captures movement characteristics such as direction, speed, and head motion. Ground clearance issues, such as slips and falls, are detected by analyzing y-direction motion vectors and feature tracking. Finally, machine learning classifiers and transfer learning models like VGG 16 enable real-time alerts and monitoring. Performance metrics, including mAP, precision, recall, and F1 score, validate the system's efficacy in handling varying environmental conditions, demonstrating its robustness and scalability for surveillance applications.