基于时空共现模式的视觉行人再识别EI北大核心CSCD

基于视频图像的视觉行人再识别是指利用计算机视觉技术关联非重叠域摄像头网络下的相同行人,在视频安防和商业客流分析中具有重要应用.目前视觉行人再识别技术已经取得了相当不错的进展,但依旧面临很多挑战,比如摄像机的拍摄视角不同、遮挡现象和光照变化等所导致的行人表观变化和匹配不准确问题.为了克服单纯视觉匹配困难问题,本文提出一种结合行人表观特征跟行人时空共现模式的行人再识别方法.所提方法利用目标行人的邻域行人分布信息来辅助行人相似度计算,有效地利用时空上下文信息来加强视觉行人再识别.在行人再识别两个权威公开数据集Market-1501和DukeMTMC-ReID上的实验验证了所提方法的有效性.

Visual Person Re-identification Based on Spatial and Temporal Co-occurrence Patterns

Person re-identification technology plays important roles in video surveillance of security and customer analysis of business,which is to associate the same person under the non-overlapping camera network.At present,the technology of person re-identification has made great progress,however,it still faces many challenges,such as the appearance changes and inaccurate matching of pedestrians caused by different camera viewpoints,occlusion,and illumination changes.In this paper,a method combining the appearance features with the spatial and temporal co-occurrence pattern is proposed.The proposed method strengthens the computation of the similarity between pedestrian images by using the association of surrounding pedestrians of the target pedestrian.The proposed method effectively utilizes spatiotemporal context information to enhance visual person re-identification.Experiments on two public data sets,namely Market-1501 and DukeMTMC-ReID,verify the effectiveness of the proposed method.