Series Patch Matching With Global Spatial Constraints For Person Re-identification

The target person recognition is a problem of person re-identification in multiple non-overlapping camera system. It has important applications in the field of intelligent video surveillance, which includes extraction of the goal person and non-overlapping target tracking. For the pedestrians who have most similar area and a small different part, these methods can not give accurate recognition results. In this paper, the problem of similar pedestrian unrecognized is studied and propose relevant solutions. The main achievements include:1) The human recognition technology is summarized, including the human recognition of supervised learning and unsupervised learning. These two kinds of method has carried on the detailed analysis and introduction. These summaries provide a certain degree of reference for subsequent researchers in the field of human recognition technology.2) In order to solve the problem of similar pedestrians are hard to identify. On the basis of densely patch-matching, we propose a matching method with spatial constraints(SCM). The method not only considers the process of local patch matching in two different images, but also considers the constraint of each patch in the vertical direction. In our experiment, the proposed approach has been proved to be the best effect so far.3) Human identification technique is applied to video tracking. We combine human recognition technology and TLD method. The experimental results show that it has significantly improved results.