《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (4): 1285-1292.DOI: 10.11772/j.issn.1001-9081.2024050566

• 多媒体计算与计算机仿真 • 上一篇    下一篇

基于扩张重参数化和空洞卷积架构的步态识别方法

霍丽娜1, 薛乐仁2, 戴钰俊2, 赵新宇2, 王世行2, 王威1()   

  1. 1.河北师范大学 计算机与网络空间安全学院,石家庄 050024
    2.河北师范大学 软件学院,石家庄 050024
  • 收稿日期:2024-05-09 修回日期:2024-08-29 接受日期:2024-09-13 发布日期:2024-09-27 出版日期:2025-04-10
  • 通讯作者: 王威
  • 作者简介:霍丽娜(1982—),女,河北平山人,副教授,博士,主要研究方向:计算机视觉、多模态学习、深度学习
    薛乐仁(2002—),男,广东清远人,CCF会员,主要研究方向:步态识别、计算机视觉、深度学习
    戴钰俊(2003—),男,江西九江人,主要研究方向:步态识别、计算机视觉、深度学习
    赵新宇(2003—),女,河北沧州人,主要研究方向:步态识别、深度学习、计算机视觉
    王世行(2004—),男,河南焦作人,主要研究方向:计算机视觉、步态识别、深度学习
  • 基金资助:
    国家自然科学基金资助项目(61702158);河北省教育厅重点科学基金资助项目(ZD2020317);中央引导地方科技发展资金资助项目(236Z0102G);河北师范大学科技类科研基金资助项目(L2024ZD15)

Gait recognition method based on dilated reparameterization and atrous convolution architecture

Lina HUO1, Leren XUE2, Yujun DAI2, Xinyu ZHAO2, Shihang WANG2, Wei WANG1()   

  1. 1.College of Computer and Cyber Security,Hebei Normal University,Shijiazhuang Hebei 050024,China
    2.Software College,Hebei Normal University,Shijiazhuang Hebei 050024,China
  • Received:2024-05-09 Revised:2024-08-29 Accepted:2024-09-13 Online:2024-09-27 Published:2025-04-10
  • Contact: Wei WANG
  • About author:HUO Lina, born in 1982, Ph. D., associate professor. Her research interests include computer vision, multi-modal learning, deep learning.
    XUE Leren, born in 2002. His research interests include gait recognition, computer vision, deep learning.
    DAI Yujun, born in 2003. His research interests include gait recognition, computer vision, deep learning.
    ZHAO Xinyu, born in 2003. Her research interests include gait recognition, deep learning, computer vision.
    WANG Shihang, born in 2004. His research interests include computer vision, gait recognition, deep learning.
  • Supported by:
    National Natural Science Foundation of China(61702158);Key Scientific Fund of Hebei Education Department(ZD2020317);Central Guidance on Local Science and Technology Development Fund(236Z0102G);Science and Technology Research Fund of Hebei Normal University(L2024ZD15)

摘要:

步态识别旨在通过人们的步行姿态进行身份识别。针对步态识别中有效感受野(ERF)与人体轮廓区域匹配不佳的问题,提出一种基于空洞卷积的步态识别方法DilatedGait。首先,采用空洞卷积扩大神经元感受野,缓解下采样和模型深度增加导致的分辨率下降,以提高轮廓结构的辨识度;其次,提出扩张重参数化模块(DRM),通过重参数化方法融合多尺度卷积核参数,优化ERF聚焦范围,使模型捕获更多的全局上下文信息;最后,通过特征映射提取判别性步态特征。在户外数据集Gait3D和GREW上的实验结果表明,对比目前的先进方法GaitBase,DilatedGait在Gait3D的Rank-1和平均逆负惩罚(mINP)上分别提升了9.0和14.2个百分点,在GREW的Rank-1和Rank-5上分别提升了11.6和8.8个百分点。可见,DilatedGait消除了复杂协变量带来的不利影响,能进一步提升户外场景下步态识别的准确率。

关键词: 步态识别, 有效感受野, 重参数化, 空洞卷积, 步态轮廓序列

Abstract:

Gait recognition aims at identifying people by their walking postures. To solve the problem of poor matching between the Effective Receptive Field (ERF) and the human silhouette region, a gait recognition method based on atrous convolution, named DilatedGait, was proposed. Firstly, atrous convolution was employed to expand the neurons’ receptive fields, thereby alleviating the resolution degradation by downsampling and model deepening. Therefore, the recognizability of the silhouette structure was enhanced. Secondly, Dilated Reparameterization Module (DRM) was proposed to optimize the ERF focus range by fusing the multi-scale convolution kernel parameters through reparameterization method, thus enabling the model to capture more global contextual information. Finally, the discriminative gait features were extracted via feature mapping. Experiments were conducted on the outdoor datasets Gait3D and GREW, and the results show that compared with the existing state-of-the-art method GaitBase, DilatedGait improves 9.0 and 14.2 percentage points respectively in Rank-1 and mean Inverse Negative Penalty (mINP) on Gait3D and increases 11.6 and 8.8 percentage points respectively in Rank-1 and Rank-5 on GREW. It can be seen that DilatedGait overcomes the adverse effects of complex covariates and further enhances the accuracy of gait recognition in outdoor scenes.

Key words: gait recognition, Effective Receptive Field (ERF), reparameterization, atrous convolution, gait silhouette sequence

中图分类号: