Overview on Human Reidentification: Machine Learning and Metaheuristic Optimization Approaches

Human Re-Identification (Re-ID) is a vital task in security and surveillance systems, enabling the recognition of individuals across multiple environments. This survey aims to develop to systematically evaluate and categorize the existing methods based on a taxonomy-driven framework including Machine Learning (ML), Deep Learning (DL) and metaheuristic optimization strategy methods. The primary aim of the survey is to evaluate these approaches address the Re-ID considerations of viewpoint variation, occlusion, and scalability. A comparative analysis is provided to show that while DL based methods like DCNN, BNN achieve high accuracy, the metaheuristic optimizers like ACO, BOA significantly assist in feature selection and in parameter tuning. The study also highlight clear gaps in the literature including the often unexplored concept of optimization and DL. Future directions include possible hybrid DL-metaheuristic approaches and near-real time deployments. The structured review has provided a path forward for researchers to improve the robustness and efficiency of Re-ID systems.