AI-Powered Lost Object and Entity Tracking Across Live and Recorded Camera Feeds

Abstract

The rapid growth of video surveillance systems in urban, industrial, and institutional areas has resulted in a huge increase in visual data streams. Manually monitoring these large video feeds has become inefficient, prone to errors, and labor-intensive. This situation calls for the development of smart, automated surveillance systems. This paper proposes an AI-driven framework that can detect, identify, and track lost objects and individuals in both live and recorded CCTV footage. The system uses advanced deep learning models, in-cluding YOLOv8 for real-time multi-class object detection and CNN-based face recognition for re-identification and tracking across different camera views. The framework includes modules for data prepro-cessing, feature extraction, and temporal tracking. These components work together to ensure continuous identity tracking and object localization. It also allows for multi-camera synchronization, which enables smooth cross-view tracking of individuals and misplaced items in complicated environments. We conducted experimental analysis on benchmark datasets and real-world surveillance footage. The results showed an av-erage precision of 94.2% and a recall of 91.6%, highlighting the model's strength under various lighting, oc-clusion, and motion conditions. In addition to high detection accuracy, the framework offers real-time per-formance, scalability, and flexibility for use in airports, railway stations, shopping malls, and law enforce-ment agencies. This research bridges the gap between traditional manual surveillance and modern autono-mous monitoring systems, contributing to the field of intelligent security solutions.