[1]刘志刚,王 淼,刘苗苗.基于姿态引导特征增强的遮挡行人重识别[J].计算机技术与发展,2024,34(04):89-94.[doi:10. 3969 / j. issn. 1673-629X. 2024. 04. 014]
　LIU Zhi-gang,WANG Miao,LIU Miao-miao.Occluded Person Re-identification Based on Pose-guided Feature Enhancement[J].,2024,34(04):89-94.[doi:10. 3969 / j. issn. 1673-629X. 2024. 04. 014]
点击复制

基于姿态引导特征增强的遮挡行人重识别()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
34
期数:
2024年04期
页码:
89-94
栏目:
媒体计算
出版日期:
2024-04-10

文章信息/Info

Title:
Occluded Person Re-identification Based on Pose-guided Feature Enhancement
文章编号:
1673-629X(2024)04-0089-06
作者:
刘志刚1; 2 ; 王 淼1 ; 刘苗苗1
1. 东北石油大学 计算机与信息技术学院,黑龙江 大庆 163318;
2. 黑龙江省石油大数据与智能分析重点实验室,黑龙江 大庆 163318
Author(s):
LIU Zhi-gang1; 2 ; WANG Miao1 ; LIU Miao-miao1
1. School of Computer and Information Technology,Northeast Petroleum University,Daqing 163318,China;
2. Heilongjiang Provincial Key Laboratory of Oil Big Data & Intelligent Analysis,Daqing 163318,China
关键词:
行人重识别; 遮挡; 特征修复; 特征补偿; 广义均值池化
Keywords:
person re-identification; occlusion; feature repair; feature compensation; generalized mean pooling
分类号:
TP391
DOI:
10. 3969 / j. issn. 1673-629X. 2024. 04. 014
摘要:
为解决遮挡行人重识别在特征提取过程中的特征丢失、特征匹配过程中的噪声干扰问题,提出了一种姿态引导的特征增强模型。 首先,在关键点信息的辅助下,设计一种对称区域特征修复模块,将被遮挡区域丢失的局部特征替换为未遮挡区域的局部特征;其次,为挖掘局部特征的内在语义联系,设计一种相邻区域特征补偿模块,通过相邻区域特征修正局部特征表示;最后,通过引入广义均值池化对特征图的中心区域再次进行特征提取,提升行人特征向量的表达能力,以获得更加准确的全局特征。 仿真实验表明,该模型在常见的全身数据集、半身数据集和遮挡数据集的 Rank-1 和 mAP 均优于绝大部分算法,其中在遮挡数据集 Occluded-Duke、Occluded-REID 上的 Rank-1 分别达到了 56. 7% 和 72. 4% 。
Abstract:
To address the problems of feature loss during feature extraction and noise interference during feature matching in occludedperson re-identification,a pose-guided feature enhancement model is proposed. Firstly,with the assistance of key point information,asymmetrical region feature repair module is proposed to replace the locally lost features in the occluded region with the features in the non-occluded region. Secondly,to explore the semantic relationship between local features,an adjacent region feature compensation module isproposed to update the feature representation of each local feature by combining the features of adjacent regions. Finally,by leveragingthe generalized mean pooling to extract features from the central region of the feature map,the expression ability of person feature vectorsis improved to obtain more accurate global features. Simulation experiments show that the proposed model outperforms most algorithmsin common holistic datasets,partial datasets,and occluded datasets in terms of Rank-1 and mAP,achieving 56. 7% and 72. 4% Rank-1on Occluded-Duke and Occluded-REID datasets,respectively.

相似文献/References:

[1]潘铭星,孙涵. 基于先验信息的交通标志检测[J].计算机技术与发展,2017,27(02):96.
　PAN Ming-xing,SUN Han. Traffic Sign Detection Based on Prior Information[J].,2017,27(04):96.
[2]韩 肖,马 祥.基于二进制标签松弛模型的遮挡人脸识别[J].计算机技术与发展,2022,32(01):1.[doi:10. 3969 / j. issn. 1673-629X. 2022. 01. 001]
　HAN Xiao,MA Xiang.Occlusion Face Recognition Based on Binary Label Relaxation[J].,2022,32(04):1.[doi:10. 3969 / j. issn. 1673-629X. 2022. 01. 001]
[3]张誉馨,张索非,王文龙,等.面向行人重识别的多域批归一化问题研究[J].计算机技术与发展,2022,32(01):91.[doi:10. 3969 / j. issn. 1673-629X. 2022. 01. 016]
　ZHANG Yu-xin,ZHANG Suo-fei,WANG Wen-long,et al.Research on Batch Normalization for Multi-domain PersonRe-identification[J].,2022,32(04):91.[doi:10. 3969 / j. issn. 1673-629X. 2022. 01. 016]
[4]陈江萍,张索非,宋 越,等.位置感知注意力及其在行人重识别中的应用[J].计算机技术与发展,2023,33(01):150.[doi:10. 3969 / j. issn. 1673-629X. 2023. 01. 023]
　CHEN Jiang-ping,ZHANG Suo-fei,SONG Yue,et al.A Novel Position-aware Attention Module and Its Use inPerson Re-identification[J].,2023,33(04):150.[doi:10. 3969 / j. issn. 1673-629X. 2023. 01. 023]

更新日期/Last Update: 2024-04-10