高级搜索

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于多尺度和注意力融合学习的行人重识别

王粉花 赵波 黄超 严由齐

王粉花, 赵波, 黄超, 严由齐. 基于多尺度和注意力融合学习的行人重识别[J]. 电子与信息学报. doi: 10.11999/JEIT190998
引用本文: 王粉花, 赵波, 黄超, 严由齐. 基于多尺度和注意力融合学习的行人重识别[J]. 电子与信息学报. doi: 10.11999/JEIT190998
Fenhua WANG, Bo ZHAO, Chao HUANG, Youqi YAN. Person Re-identification Based on Multi-scale and Attention Fusion[J]. Journal of Electronics and Information Technology. doi: 10.11999/JEIT190998
Citation: Fenhua WANG, Bo ZHAO, Chao HUANG, Youqi YAN. Person Re-identification Based on Multi-scale and Attention Fusion[J]. Journal of Electronics and Information Technology. doi: 10.11999/JEIT190998

基于多尺度和注意力融合学习的行人重识别

doi: 10.11999/JEIT190998
基金项目: 国家重点研发计划重点专项(2017YFB1400101-01),北京科技大学中央高校基本科研业务费专项 (FRF-BD-19-002A)
详细信息
    作者简介:

    王粉花:女,1971年生,副教授,硕士生导师,研究方向为模式识别和智能信息处理

    赵波:男,1994年生,硕士生,研究方向为计算机视觉

    黄超:男,1993年生,硕士生,研究方向为计算机视觉

    严由齐:男,1997年生,硕士生,研究方向为计算机视觉

    通讯作者:

    王粉花 wangfenhua@ustb.edu.cn

  • 中图分类号: TN911.73; TP391

Person Re-identification Based on Multi-scale and Attention Fusion

Funds: The Key Projects of National Key R & D Plan (2017YFB1400101-01), Beijing University of Science and Technology Central University Basic Research Business Expenses (FRF-BD-19-002A)
  • 摘要: 行人重识别的关键依赖于行人特征的提取,卷积神经网络具有强大的特征提取以及表达能力。针对不同尺度下可以观察到不同的特征,该文提出一种基于多尺度和注意力网络融合的行人重识别方法(MSAN)。该方法通过对网络不同深度的特征进行采样,将采样的特征融合后对行人进行预测。不同深度的特征图具有不同的表达能力,使网络可以学习到行人身上更加细粒度的特征。同时将注意力模块嵌入到残差网络中,使得网络能更加关注于一些关键信息,增强网络特征学习能力。所提方法在Market1501, DukeMTMC-reID和MSMT17_V1数据集上首位准确率分别到了95.3%, 89.8%和82.2%。实验表明,该方法充分利用了网络不同深度的信息和关注的关键信息,使模型具有很强的判别能力,而且所提模型的平均准确率优于大多数先进算法。
  • 图  1  多尺度和注意力融合模型框架图

    图  2  ResNet50网络架构图

    图  3  Conv2_x模块架构图

    图  4  多尺度结构图

    图  5  CBAM模块图

    图  6  3元组损失

    表  1  多尺度融合模型准确率验证实验结果(%)

    方法Market1501DukeMTMC-reIDMSMT17_V1
    Rank-1mAPRank-1mAPRank-1mAP
    SSAN94.987.986.167.781.466.3
    SSAN(+RK)95.393.786.075.684.673.8
    MSAN95.387.989.878.882.260.6
    MSAN (+RK)95.993.992.389.785.074.6
    下载: 导出CSV

    表  2  CBAM模块准确率验证实验结果(%)

    方法Market1501DukeMTMC-reIDMSMT17_V1
    Rank-1mAPRank-1mAPRank-1mAP
    MSN94.486.287.577.279.656.0
    MSN (+CBAM)95.387.989.878.882.260.6
    MSN(+RK)95.393.190.989.283.272.0
    MSN(+CBAM+RK)95.993.992.389.785.074.6
    下载: 导出CSV

    表  3  所提MSAN算法与其他先进算法的准确率对比(%)

    方法Market1501DukeMTMC-reIDMSMT17_V1
    Rank-1(%)mAP(%)Rank-1(%)mAP(%)Rank-1(%)mAP(%)
    SVDNet[21]82.362.176.756.8
    DPFL[22]88.672.679.260.0
    SVDNet+Era[23]87.171.379.362.4
    TriNET+Era[23]83.968.773.056.6
    DaRe[24]89.076.080.264.5
    GP-reid[25]92.281.285.272.8
    PCB[4]92.377.481.965.368.240.4
    Aligned-ReID[5]92.682.3
    PCB+RPP[4]93.881.683.369.2
    MGN[6]95.786.988.778.4
    BFENET[8]94.284.386.872.1
    IANet[18]94.483.187.173.475.546.8
    DGNet[19]94.886.086.674.877.252.3
    OSNet[20]94.884.988.673.578.752.9
    MSAN95.387.989.878.882.260.6
    MSAN(+RK)95.993.992.389.785.074.6
    下载: 导出CSV
  • [1] FARENZENA M, BAZZANI L, PERINA A, et al. Person re-identification by symmetry-driven accumulation of local features[C]. 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Francisco, USA, 2010: 2360–2367.
    [2] 周智恒, 刘楷怡, 黄俊楚, 等. 一种基于等距度量学习策略的行人重识别改进算法[J]. 电子与信息学报, 2019, 41(2): 477–483. doi:  10.11999/JEIT180336

    ZHOU Zhiheng, LIU Kaiyi, HUANG Junchu, et al. Improved metric learning algorithm for person re-identification based on equidistance[J]. Journal of Electronics &Information Technology, 2019, 41(2): 477–483. doi:  10.11999/JEIT180336
    [3] HIRZER M, ROTH P M, KÖSTINGER M, et al. Relaxed pairwise learned metric for person re-identification[C]. The 12th European Conference on Computer Vision, Florence, Italy, 2012: 780–793.
    [4] SUN Yifan, ZHENG Liang, YANG Yi, et al. Beyond part models: Person retrieval with refined part pooling (and a strong convolutional baseline)[C]. The 15th European Conference on Computer Vision (ECCV), Munich, Germany, 2018: 480–496.
    [5] LUO Hao, JIANG Wei, ZHANG Xuan, et al. AlignedReID++: Dynamically matching local information for person re-identification[J]. Pattern Recognition, 2019, 94: 53–61. doi:  10.1016/j.patcog.2019.05.028
    [6] WANG Guanshuo, YUAN Yufeng, CHEN Xiong, et al. Learning discriminative features with multiple granularities for person re-identification[C]. 2018 ACM Multimedia Conference on Multimedia Conference, Seoul, Korea, 2018: 274–282.
    [7] 陈鸿昶, 吴彦丞, 李邵梅, 等. 基于行人属性分级识别的行人再识别[J]. 电子与信息学报, 2019, 41(9): 2239–2246. doi:  10.11999/JEIT180740

    CHEN Hongchang, WU Yancheng, LI Shaomei, et al. Person re-identification based on attribute hierarchy recognition[J]. Journal of Electronics &Information Technology, 2019, 41(9): 2239–2246. doi:  10.11999/JEIT180740
    [8] DAI Zuozhuo, CHEN Mingqiang, GU Xiaodong, et al. Batch DropBlock network for person re-identification and beyond[C]. 2019 IEEE/CVF International Conference on Computer Vision, Seoul, Korea, 2019: 3691–3701.
    [9] WOO S, PARK J, LEE J Y, et al. Cbam: Convolutional block attention module[C]. The 15th European Conference on Computer Vision (ECCV), Munich, Germany, 2018: 3–19.
    [10] LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]. 2017 IEEE conference on Computer Vision and Pattern Recognition, Honolulu, USA, 2017: 2117–2125.
    [11] HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, 2016: 770–778.
    [12] HERMANS A, BEYER L, and LEIBE B. In defense of the triplet loss for person re-identification[EB/OL]. https://arxiv.org/abs/1703.07737, 2017.
    [13] ZHENG Liang, SHEN Liyue, TIAN Lu, et al. Scalable person re-identification: A benchmark[C]. 2015 IEEE International Conference on Computer Vision, Santiago, Chile, 2015: 1116–1124.
    [14] RISTANI E, SOLERA F, ZOU R, et al. Performance measures and a data set for multi-target, multi-camera tracking[C]. 2016 European Conference on Computer Vision, Amsterdam, The Netherlands, 2016: 17–35.
    [15] WEI Longhui, ZHANG Shiliang, GAO Wen, et al. Person transfer GAN to bridge domain gap for person re-identification[C]. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, USA, 2018: 79–88.
    [16] ZHONG Zhun, ZHENG Liang, CAO Donglin, et al. Re-ranking person re-identification with k-reciprocal encoding[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, USA, 2017: 1318–1327.
    [17] SALLEH S S, AZIZ N A A, MOHAMAD D, et al. Combining mahalanobis and jaccard distance to overcome similarity measurement constriction on geometrical shapes[J]. International Journal of Computer Science Issues, 2012, 9(4): 124–132.
    [18] ZHENG Zhedong, YANG Xiaodong, YU Zhiding, et al. Joint discriminative and generative learning for person re-identification[C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, USA, 2019: 2138–2147.
    [19] HOU Ruibing, MA Bingpeng, CHANG Hong, et al. Interaction-and-aggregation network for person re-identification[C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, USA, 2019: 9317–9326.
    [20] ZHOU Kaiyang, YANG Yongxin, CAVALLARO A, et al. Omni-Scale feature learning for person re-identification[C]. 2019 IEEE/CVF International Conference on Computer Vision, Seoul, Korea, 2019: 3702–3712.
    [21] SUN Yifan, ZHENG Liang, DENG Weijian, et al. SVDNet for pedestrian retrieval[C]. 2017 IEEE International Conference on Computer Vision, Venice, Italy, 2017: 3800–3808.
    [22] CHEN Yanbei, ZHU Xiatian, and GONG Shaogang. Person re-identification by deep learning multi-scale representations[C]. 2017 IEEE International Conference on Computer Vision Workshops, Venice, Italy, 2017: 2590–2600.
    [23] ZHONG Zhun, ZHENG Liang, KANG Guoliang, et al. Random erasing data augmentation[EB/OL]. https://arxiv.org/abs/1708.04896, 2017.
    [24] WANG Yan, WANG Lequn, YOU Yurong, et al. Resource aware person re-identification across multiple resolutions[C]. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, USA, 2018: 8042–8051.
    [25] ALMAZAN J, GAJIC B, MURRAY N, et al. Re-ID done right: towards good practices for person re-identification[EB/OL]. https://arxiv.org/abs/1801.05339, 2018.
  • [1] 周非, 郭浩田, 杨伊.  一种改进的虚拟力重定位覆盖增强算法, 电子与信息学报. 2020, 42(0): 1-7. doi: 10.11999/JEIT190662
    [2] 陈勇, 刘曦, 刘焕淋.  基于特征通道和空间联合注意机制的遮挡行人检测方法, 电子与信息学报. 2020, 42(6): 1486-1493. doi: 10.11999/JEIT190606
    [3] 易诗, 吴志娟, 朱竞铭, 李欣荣, 袁学松.  基于多尺度生成对抗网络的运动散焦红外图像复原, 电子与信息学报. 2020, 42(7): 1766-1773. doi: 10.11999/JEIT190495
    [4] 柳长源, 王琪, 毕晓君.  基于多通道多尺度卷积神经网络的单幅图像去雨方法, 电子与信息学报. 2020, 42(0): 1-8. doi: 10.11999/JEIT190755
    [5] 付哲泉, 李尚生, 李相平, 但波, 王旭坤.  基于高效可扩展改进残差结构神经网络的舰船目标识别技术, 电子与信息学报. 2020, 42(0): 1-8. doi: 10.11999/JEIT190913
    [6] 赵斌, 王春平, 付强.  显著性背景感知的多尺度红外行人检测方法, 电子与信息学报. 2020, 42(0): 1-9. doi: 10.11999/JEIT190761
    [7] 谢金宝, 侯永进, 康守强, 李佰蔚, 张霄.  基于语义理解注意力神经网络的多元特征融合中文文本分类, 电子与信息学报. 2018, 40(5): 1258-1265. doi: 10.11999/JEIT170815
    [8] 徐彬, 陈渤, 刘宏伟, 金林.  基于注意循环神经网络模型的雷达高分辨率距离像目标识别, 电子与信息学报. 2016, 38(12): 2988-2995. doi: 10.11999/JEIT161034
    [9] 胡站伟, 焦立国, 徐胜金, 黄勇.  基于多尺度重采样思想的类指数核函数构造, 电子与信息学报. 2016, 38(7): 1689-1695. doi: 10.11999/JEIT151101
    [10] 郭静, 曹亚男, 周川, 张鹏, 郭莉.  基于线性阈值模型的影响力传播权重学习, 电子与信息学报. 2014, 36(8): 1804-1809. doi: 10.3724/SP.J.1146.2014.00090
    [11] 刘龙, 孙强, 宋琦军.  面向目标检测的多尺度运动注意力融合算法研究, 电子与信息学报. 2014, 36(5): 1133-1138. doi: 10.3724/SP.J.1146.2013.00477
    [12] 邓苗, 张基宏, 柳伟, 梁永生.  基于全变分的权值优化的多尺度变换图像融合, 电子与信息学报. 2013, 35(7): 1657-1663. doi: 10.3724/SP.J.1146.2012.01183
    [13] 杨旗, 薛定宇.  基于双尺度动态贝叶斯网络及多信息融合的步态识别, 电子与信息学报. 2012, 34(5): 1148-1153. doi: 10.3724/SP.J.1146.2011.01012
    [14] 陈琳, 卢湖川.  基于ML-pLSA模型的目标识别算法, 电子与信息学报. 2011, 33(12): 2909-2915. doi: 10.3724/SP.J.1146.2011.00455
    [15] 邵飞, 伍春, 汪李峰.  基于多Agent强化学习的Ad hoc网络跨层拥塞控制策略, 电子与信息学报. 2010, 32(6): 1520-1524. doi: 10.3724/SP.J.1146.2009.01092
    [16] 王晓华, 杨新艳, 焦李成.  基于多尺度几何分析的复杂网络压缩策略, 电子与信息学报. 2009, 31(4): 968-972. doi: 10.3724/SP.J.1146.2007.01860
    [17] 刘怡光, 游志胜, 曹丽萍, 蒋欣荣.  基于Fisher变换的多尺度图像识别方法及其车形识别应用, 电子与信息学报. 2003, 25(12): 1603-1611.
    [18] 杨群生, 余英林.  多模式对连接权矩阵的一种神经网络学习算法, 电子与信息学报. 2001, 23(3): 280-285.
    [19] 杨绍国, 尹忠科, 罗炳伟.  基于分形和神经网络理论的多尺度图象分割方法, 电子与信息学报. 1998, 20(6): 727-732.
    [20] 水鹏朗, 保铮, 焦李成.  一种基于子波神经网络的多尺度差分方程求解新方法, 电子与信息学报. 1997, 19(6): 733-738.
  • 加载中
  • 图(6) / 表ll (3)
    计量
    • 文章访问数:  275
    • HTML全文浏览量:  176
    • PDF下载量:  31
    • 被引次数: 0
    出版历程
    • 收稿日期:  2019-12-13
    • 修回日期:  2020-06-17
    • 网络出版日期:  2020-07-20

    目录

      /

      返回文章
      返回

      官方微信,欢迎关注