The person re-identification (PRID) problems with respect to color distributions. In PRID the color distribution is multimodal in nature, thus, is composed of unique pattern. However, the color distribution is affected with variation of illumination but some feature exhibit invariant pattern. This pattern is referred as intra-distribution patterns where considering diverse environment conditions it is invariant and at same time it discriminative enough for practical applicability. This work uses silhouette, local maximal occurrence (LOMO), and sparse based feature of person for constructing ranking list. Then, for improving accuracies of top-k recommendation of person re-identification presented a re-ranking model. Experiment is conducted considering dynamic illumination and viewpoint variation for evaluating performance of efficient feature extraction and re-ranking based PRID over existing PRID methods. The conducted experiment was based on the dataset of the standard like VIPeR and iLID. Experiment outcome shows superior accuracies is achieved using EFER based person’s re-identification method.

Color invariant signature, Feature extraction, Illumination & viewpoint, Person reidentification, Re-ranking, Surveillance applications