Date of Publication : April 2026

Abstract: This document presents an AI-driven real-time surveillance framework aiming at early detection and prevention of stampede in crowded environments. It combines CSRNet, DBSCAN and LSTM models to perform crowd density estimation, anomaly detection and motion trend forecasting respectively. These modules work in sequence to deliver predictive situational awareness, enabling authorities to act before congestion escalates in an alarming situation. The system’s adaptive and data-driven design ensures high reliability across diverse event scenarios, reducing dependence on manual monitoring. By providing accurate, real-time insights into crowd dynamics, the system enables timely intervention, ultimately aiming to reduce stampede risks and enhance public safety at mass gatherings.

Reference :