中国计量大学学报2017，Vol.28Issue(2)：208-213,6.DOI:10.3969/j.issn.2096-2835.2017.02.012
深度哈希算法行人再识别技术研究
Pedestrian re-identification on deep hash algorithm
摘要
Abstract
As one of the key tasks of intelligent video surveillance, pedestrian re-identification is very challenging due to large variation in visual appearance across different camera views.Assuming that a person`s ID could be indirectly represented by a combination of semantic attributes in the image, a deep hash pedestrian re-identification algorithm was introduced.The hash function was obtained by CNN and the muti-objective loss function ensured the validity of the classification and the efficiency of the hash codes, which could help obtain similar pedestrian images corresponding to similar hash codes.In the end, the hamming distance between hash features was compared with the re-identification.The experimental results show that the deep hash feature can improve the efficiency of pedestrian re-identification.关键词
哈希算法/深度学习/汉明距离Key words
hash algorithm/deep learning/hamming distance分类
信息技术与安全科学引用本文复制引用
章东平,尹奕博..深度哈希算法行人再识别技术研究[J].中国计量大学学报,2017,28(2):208-213,6.基金项目
浙江省自然科学基金资助项目(No.LY15F020021),浙江省公益性项目(No.2016C31079). （No.LY15F020021）
