Robust Person Re-identification through fusion of discriminant and deep Features

Person re-identification is a crucial task within the realm of computer vision and surveillance, aiming to identify and track individuals across various camera views or frames. It is commonly applied in video surveillance and security applications. The primary challenge in person re-identification lies in coping with significant variations in appearance induced by factors such as changing lighting conditions, pose variations, occlusions, and diverse camera viewpoints. Various methods and techniques have been developed to address this critical task. In this study, we introduce a novel framework that combines deep-learned features with hand-crafted features. This integration is designed to extract richer information for matching people across cameras. Our approach's performance is evaluated on two standard benchmark datasets, including small-scale (VIPeR) and large-scale (PRID450S). Our approach gives good results compared to the state-of-the-art methods of person re-identification.