Person re-identification systems are essential in video surveillance applications. They use biometric characteristics, such as the face, and soft biometric characteristics, especially body silhouette. The process uses computer vision descriptors such as Oriented Gradient Histogram and machine learning models such as Support Vector Machines, which offer computational efficiency. The system is implemented in non-overlapping cameras to identify and re-identify people in real time, even without facial information. The experiments achieve 96.50% accuracy with facial features and 69.50% with silhouette, validated by the Precision-Recall curve in unbalanced scenarios. The CPU implementation demonstrates the system’s potential in environments with lighting variations, as it allows real-time tracking of individuals without relying on the face.