【摘要】 提出了一种基于人体部位特征重要性的行人再识别算法,该算法首先提取人体各部位的颜色、纹理以及形状等特征,然后对多个行人样本的每个部位分别进行聚类分析,使用误差积累的方法为每个分类计算一种更适合该分类的部位特征重要性权值向量,使得不同类型特征能更有效地应用在其适合的外观上。在公共数据集VIPeR上进行了实验,通过积累匹配特性(cumulative matching characteristic,CMC)曲线对实验结果进行评价,结果表明,该算法具有较高的再识别率,且对行人视角转换、光照变化、环境嘈杂和物体遮挡有较好的鲁棒性。更多还原

【Abstract】 In a video surveillance system,the same person may look different across different cameras,while different people may look the same in one camera,thus making re-identification ofindividualsa challenging problem.We carried out an algorithm based on the importance of partial features,firstly extracting features such as color,texture,and shape.Each partis clustered by classifying different appearances of body parts,using an error accumulation method to figure out weight vectorsindicating the significance of the featurethat fits the type of appearance.Similarity iscalculated using this vector to weight the features of each part,making the feature more suited to match with appearance.This algorithm indicatesthat some features are more important than others for parts with different appearances.We completedexperiments on the public VIPeR datasets,and evaluated the results using the CMC curve.These tests indicated this algorithm achieved higher re-identification rate and was more robust to viewing condition changes,illumination variations,background clutter,and occlusion.更多还原