《计算机应用研究》|Application Research of Computers

基于全局特征拼接的行人重识别算法研究

Person re-identification algorithm based on global feature stitching

免费全文下载 (已被下载 次)  
获取PDF全文
作者 熊炜,杨荻椿,熊子婕,童磊,李利荣,王娟
机构 1.湖北工业大学 电气与电子工程学院,武汉 430068;2.美国南卡罗来纳大学 计算机科学与工程系,美国 哥伦比亚 29201
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2021)01-064-0316-05
DOI 10.19734/j.issn.1001-3695.2019.09.0578
摘要 针对目前行人重识别出现网络模型复杂化、识别率低的问题,提出一种基于全局特征拼接的行人重识别算法。首先利用卷积神经网络(CNN)提取全局特征;然后把不同卷积层提取的特征进行拼接,使特征信息互补;最后将拼接后的特征再次进行卷积处理,获得高表征能力的特征。网络训练时,采用聚类损失函数和标签平滑损失函数联合训练,同时引入了随机擦除和减小池化步长的训练技巧。在Market1501、DukeMTMC-reID、CUHK03和MSMT17数据集上进行了实验验证,实验表明所提算法具有良好性能,其中在Market1501上,Rank-1、mAP分别达到了95.9%和94.6%。
关键词 行人重识别; 全局特征拼接; 聚类损失; 标签平滑损失
基金项目 国家留学基金资助项目(201808420418)
国家自然科学基金资助项目(61571182,61601177)
湖北省自然科学基金资助项目(2019CFB530)
本文URL http://www.arocmag.com/article/01-2021-01-064.html
英文标题 Person re-identification algorithm based on global feature stitching
作者英文名 Xiong Wei, Yang Dichun, Xiong Zijie, Tong Lei, Li Lirong, Wang Juan
机构英文名 1.School of Electrical & Electronic Engineering,Hubei University of Technology,Wuhan 430068,China;2.Dept. of Computer Science & Engineering,University of South Carolina,Columbia SC 29201,USA
英文摘要 In order to solve the problems of network model complexity and low identification rate, this paper proposed a person re-identification(ReID) method based on global feature stitching. Firstly, it extracted the global features using convolutional neural network(CNN). Secondly, it stitched the features from different convolution layers together to complement the feature information. Finally, it convoluted again to obtain the features with high representation ability. In network training stage, it combined the cluster loss with label smoothing loss, and adopted random erasing augmentation(REA) as well as pooling step reduction techniques. It conducted extensive experiments to evaluate the performance of this proposed method on Market1501, DukeMTMC-reID, CUHK03 and MSMT17 benchmark datasets. Results show that the proposed method outperforms other state-of-the-art techniques, for instance, the Rank-1 and mAP on Market1501 are 95.9% and 94.6%, respectively.
英文关键词 person re-identification(ReID); global feature stitching; cluster loss; label smoothing loss
参考文献 查看稿件参考文献
 
收稿日期 2019/9/15
修回日期 2019/11/6
页码 316-320
中图分类号 TP183
文献标志码 A