Person Re-identification Using Prototype Formation

Abstract

Person re-identification is to match the appearance based images of an individual who has already been captured by different camera perspectives. This paper presents an appearance based model for person re-identification. It consists of the formation of prototypes followed by a matching strategy. The prototypes are discovered from the visual appearances among the individuals with similar characteristics. When a probe comes along, it necessitates to be classified through prototype assignment. To determine the correct matching of a given probe, the similarity computation is performed between the probe and a subset of gallery images, that shares the same prototype with the probe. Thus the strategy of finding the correspondence within a subset of gallery instances reduces the computational overhead. This model is useful in scenarios where individuals appear with similar attire. The performance measure of the proposed method is evaluated on benchmark data sets and presented using cumulative matching characteristic (CMC) curves.