Vigilance Nexus: An AI-Driven Theft Detection Module for Real-Time Video Surveillance

This work describes the theft-detection module of the AI-based video surveillance system called Vigilance Nexus, built for real-time shoplifting/bag-snatching prevention and unauthorized possession detection, among others. It combines spatio-temporal action recognition (CNN + LSTM) and resilient object detection and multi-object tracking (YOLO + DeepSORT), so suspicious human–object interactions can be spotted and ownership transitions traced over time. Preprocessing and data augmentation pipelines enhance resistance to crowded frames, occlusion, and illumination variation. Time-stamped alarms and incident logging for downstream analysis enable quick security response and long-term crime pattern extraction. It is a modular design compatible with pre-existing CCTV/IP camera installations, ensuring low-latency usage and scalable field deployment.