行人重识别中度量学习方法研究进展
CSTR:
作者:
作者单位:

1. 山东理工大学 电气与电子工程学院,山东 淄博 255049;2. 不列颠哥伦比亚大学 工程学院,不列颠哥伦比亚 基洛纳 V1V1V7

作者简介:

通讯作者:

E-mail: zgf841122@163.com.

中图分类号:

TP391

基金项目:

国家自然科学基金项目(61601266,61801272);淄博市张店区校城融合项目(118228);山东省自然科学基金项目(ZR2015FL029,ZR2016FL14).


A survey on metric learning in person re-identification
Author:
Affiliation:

1. School of Electrical and Electronic Engineering,Shandong University of Technology,Zibo 255049,China;2. School of Engineering,University of British Columbia-Okanagan,Kelowna V1V1V7,Canada

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    行人重识别是计算机视觉领域极具挑战的研究课题.近年来,伴随大规模行人数据集推出和深度学习发展,针对行人特征提取与描述、距离度量学习两大关键技术的研究取得众多成果.已有综述文献主要对特征提取与描述方法开展了归纳总结,尚缺乏对度量学习方法的全面分析.同时,鉴于度量学习在提升重识别性能中的关键作用,有必要对行人重识别中度量学习研究现状进行系统梳理.基于此,从距离度量方式、度量学习算法和重排序3方面系统总结了行人重识别度量学习方法,比较了部分典型方法的实验效果,并对未来可能的研究方向作了展望.

    Abstract:

    Person re-identification is a very challenging research topic in the field of computer vision. In recent years, with the emergence of large-scale datasets and development of deep learning, many results have been obtained in the study of person feature extraction and description, distance metric learning. In the existing literature, feature extraction and description have been well summarized, but there is no comprehensive analysis of metric learning. Considering the key role of metric learning in improving person re-identification, it is necessary to systematically review the research status of metric learning in person re-identification. This survey gives a systematic summary of metric learning methods from three perspectives: distance metric methods, metric learning algorithms and re-ranking algorithms. Then, the performance of some representative methods are compared and analyzed. Finally, we make a prospect for the future research direction of metric learning.

    参考文献
    相似文献
    引证文献
引用本文

邹国锋,傅桂霞,高明亮,等.行人重识别中度量学习方法研究进展[J].控制与决策,2021,36(7):1547-1557

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2021-06-16
  • 出版日期: 2021-07-20
文章二维码