Person Re-identification and Tracking with Multiple Non-overlapping Cameras

In recent years, the demand for robust person re-identification and tracking systems has grown substantially, driven by the increasing deployment of surveillance networks with multiple non-overlapping cameras. Tracking individuals is a vital task in multi-camera surveillance environments, where challenges arise due to variations in poses and lighting conditions. The research dives into the methodologies employed, identifies unresolved challenges, and sheds light on promising future research directions. The study presents a comprehensive approach that leverages YOLOv7 for person detection, employs DeepSORT for initial tracking, and introduces an exact Root Mean Square Error (RMSE)-based re-identification algorithm. The proposed solution outperforms most current techniques, amplifying safety and delivering optimal multi-camera tracking solutions, effectively addressing real-world challenging conditions.