AI-BASED ALGORITHMS FOR REAL-TIME DETECTION OF DENSE CROWDING AND ANOMALOUS BEHAVIORAL OBJECTS IN PUBLIC PLACES

Abstract

This paper presents AI-based algorithms for real-time detection of dense crowding and anomalous behavioral objects in public places. We propose a two-stage approach combining crowd density estimation and behavior-based anomaly detection using lightweight deep learning models optimized for edge deployment. The first stage employs a convolutional neural network with multi-scale feature aggregation to produce accurate crowd density maps and localize high-density regions.