Research On Key Technologies For Person Re-Identification

With the rapid development of video surveillance network,the massive data generated by the surveillance system has brought great challenges to the traditional video surveillance technology.As a new research direction,person re-identification aims at matching the specific person in the massive video resources automatically,which plays an important role in pedestrian retrieval,crime prevention,and accident prevention.However,due to the various changes in light,viewpoint,background interference,and occlusion,it is of great significance to achieve person re-identification accurately and efficiently.Based on the analysis of the key issues in person reidentification,this paper focuses on improving the performance of the person reidentification.In this paper,we propose a novel person re-identification method,which consists of a reliable representation called Semantic Region Representation(SRR),and an effective metric learning with Mapping Space Topology Constraint(MSTC).The SRR integrates semantic representations to achieve effective similarity comparison between the corresponding regions via parsing the body into multiple parts,which focuses on the foreground context against the background interference.To learn a discriminant metric,the MSTC is proposed to take into account the topological relationship among all samples in the feature space.It considers two-fold constraints: the distribution of positive pairs should be more compact than the average distribution of negative pairs with regard to the same probe,while the average distance between different classes should be larger than that between same classes.These two aspects cooperate to maintain the compactness of the intra-class as well as the sparsity of the inter-class.Extensive experiments conducted on challenging person re-identification datasets show that the proposed method achieves competitive performance with the state-of-the-art approachesThis paper also implements a person re-identification algorithm based on progressive fusion and residual attention model.The inception module is used as the basic module to construct the network.According to the size of the feature map,the convolutional network is divided into several stages.The output features of the adjacent layers are cascaded and merged.Then the merged feature is served as the input of the next layer.For different stages of the network,the features of the shallow layers are fused with those in the deep layers by using skipping connection.Thus,the outputs of the model contain multi-stage feature information at the same time.Besides,the residual attention model is used at different stages of the network to activate the discriminative areas in the feature map.It could increase the discriminative power of the output features.Experiments results demonstrate the effectiveness of the algorithm.