计算机应用 ›› 2020, Vol. 40 ›› Issue (3): 672-676.DOI: 10.11772/j.issn.1001-9081.2019081356

• 人工智能 • 上一篇    下一篇

基于注意力机制的行人重识别特征提取方法

刘紫燕, 万培佩   

  1. 贵州大学 大数据与信息工程学院, 贵阳 550025
  • 收稿日期:2019-08-05 修回日期:2019-10-13 出版日期:2020-03-10 发布日期:2019-10-31
  • 通讯作者: 刘紫燕
  • 作者简介:刘紫燕(1974-),女,贵州都匀人,副教授,硕士,CCF会员,主要研究方向:移动机器人、深度学习、大数据分析、无线通信系统;万培佩(1994-),男,湖北安陆人,硕士研究生,主要研究方向:深度学习、行人重识别。
  • 基金资助:
    国家自然科学基金资助项目(61863006);贵州省联合资金资助项目(黔科合LH字[2017]7226);贵州省科学技术基金资助项目(黔科合基础[2016]1054);贵州省科技计划重点项目(20191416);贵州大学2017年度学术新苗培养及创新探索专项(黔科合平台人才[2017]5788)。

Pedestrian re-identification feature extraction method based on attention mechanism

LIU Ziyan, WAN Peipei   

  1. College of Big Data and Information Engineering, Guizhou University, Guiyang Guizhou 550025, China
  • Received:2019-08-05 Revised:2019-10-13 Online:2020-03-10 Published:2019-10-31
  • Supported by:
    This work is partially supported by the National Natural Science Foundation of China (61863006), the Joint Foundation of Guizhou Province (LH-[2017]7226), the Science and Technology Foundation of Guizhou Province ([2016]1054), the Guizhou Provincial Science and Technology Plan Key Project (20191416), the Academic Talent Training and Innovation Exploration Special Project of Guizhou University in 2017 ([2017]5788).

摘要: 针对真实环境中非重叠多摄像头的行人重识别受到不同摄像机场景、视角、光照等因素的影响导致行人重识别精度低的问题,提出一种基于注意力机制的行人重识别特征提取方法。首先,使用随机擦除法对输入的行人图像进行数据增强,提高网络的鲁棒性;然后,通过构建自上而下的注意力机制网络增强空间像素特征的显著性,并将注意力机制网络嵌入ResNet50网络提取整个行人的显著特征;最后,将整个行人的显著特征进行相似性度量并排序得到行人重识别的结果。该注意力机制的行人重识别特征提取方法在Market1501数据集上Rank1达到88.53%,平均精度均值(mAP)为70.70%;在DukeMTMC-reID数据集上Rank1达到77.33%,mAP为59.47%。所提方法在两大行人重识别数据集上性能都有明显提升,具有一定的应用价值。

关键词: 行人重识别, 特征学习, 注意力机制, 数据增强, 显著特征

Abstract: Aiming at the problem of the low pedestrian re-identification accuracy with disjoint multiple cameras in real environment caused by different camera scenes, perspectives, illuminations and other factors, a pedestrian re-identification feature extraction method based on attention mechanism was proposed. Firstly, the random erasure method was used to enhance the data of the input pedestrian image in order to improve the robustness of the network. Then, by constructing a from-top-to-bottom attention mechanism network, the saliency of the spatial pixel feature was enhanced, and the attention mechanism network was embedded in the ResNet50 network to extract the entire pedestrian salient features. Finally, the similarity measurement and ranking were performed on the entire salient features of pedestrians in order to obtain the accuracy of pedestrian re-identification. The pedestrian re-identification feature extraction method based on attention mechanism has Rank1 of 88.53% and mAP (mean Average Precision) of 70.70% on the Market1501 dataset, and has Rank1 of 77.33% and mAP of 59.47% on the DukeMTMC-reID dataset. The proposed method has significantly improved performance on the two major pedestrian re-identification datasets, and has certain application value.

Key words: pedestrian re-identification, feature learning, attention mechanism, data enhancement, salient feature

中图分类号: