Please wait a minute...
浙江大学学报(工学版)
Journal of ZheJiang University (Engineering Science)  2025, Vol. 59 Issue (5): 929-937    DOI: 10.3785/j.issn.1008-973X.2025.05.006
    
Person re-identification method based on multi-part guided foreground enhancement
Junjing LIU1(),Wanlu ZHENG1,Ziqiang GUO1,Shaorong WANG1,2,3,*()
1. School of Information Science and Technology, Beijing Forest University, Beijing 100083, China
2. Engineering Research Center for Forestry-oriented Intelligent Information Processing of National Forestry and Grassland Administration, Beijing 100083, China
3. Beijing Virtual Simulation and Visualization Engineering Center, Beijing 100871, China
Download: HTML     PDF(1236KB) HTML
Export: BibTeX | EndNote (RIS)      

Abstract  

A multi-part guided foreground enhancement method for person re-identification was proposed in order to solve the problem that the performance of person re-identification model was overly depend on background environmental factors. The model’s attention to the person’s foreground was enhanced by employing mask-guided enhancement and self-enhancement strategies, while retaining some background information. This effectively reduced the model’s dependence on background information and improved its generalization ability. A bottleneck optimization module was integrated into the backbone network, utilizing dilated convolutions to enlarge the model’s receptive field while maintaining the original parameter scale, thereby improving the overall performance of the model. The experimental results demonstrated that the proposed model achieved Rank-1 accuracies of 95% and 88.3% on the Market1501 and DukeMTMC_reID datasets, respectively. The effectiveness of the multi-part guided foreground enhancement method was verified, which strengthened the foreground while incorporating appropriate background information, and significantly enhanced the performance of the baseline model.

Key words： person re-identification      human parsing      background change      foreground enhancement      mask-guided enhancement     
Received: 06 July 2024      Published: 25 April 2025
CLC:  TP 391  
Corresponding Authors: Shaorong WANG     E-mail: L_JunJing@bjfu.edu.cn;shaorongwang@hotmail.com
Service
E-mail this article
Add to my bookshelf
Add to citation manager
E-mail Alert
RSS
Articles by authors
Junjing LIU
Wanlu ZHENG
Ziqiang GUO
Shaorong WANG
Cite this article:

Junjing LIU,Wanlu ZHENG,Ziqiang GUO,Shaorong WANG. Person re-identification method based on multi-part guided foreground enhancement. Journal of ZheJiang University (Engineering Science), 2025, 59(5): 929-937.

URL:

https://www.zjujournals.com/eng/10.3785/j.issn.1008-973X.2025.05.006     OR     https://www.zjujournals.com/eng/Y2025/V59/I5/929


多方引导前景增强的行人重识别方法

为了解决行人重识别模型性能对背景环境因素过于依赖的问题, 提出多方引导前景增强的行人重识别方法. 该方法通过掩码引导增强和自增强策略，提升了模型对行人前景的关注，同时保留一定的背景信息，有效减轻了对背景信息的依赖，增强了模型的泛化能力. 在骨干网络中引入瓶颈优化模块，利用空洞卷积，在保持原有参数规模的前提下，有效增大了模型的感受野，提升了模型的整体性能. 实验结果表明, 提出的模型在Market1501和DukeMTMC_reID数据集上分别取得了95%和88.3%的Rank-1准确率. 验证了多方引导前景增强的行人重识别方法的有效性，通过前景增强并结合一定的背景信息，有效提升了基线模型的性能.


关键词： 行人重识别,  人体解析,  背景变化,  前景增强,  掩码引导增强 
Fig.1 Example of human parsing prediction
Fig.2 Pedestrian image processing
Fig.3 Strong baseline network
测试类型${\mathrm{Rank}} {\text{-}} 1 $${\mathrm{Rank}} {\text{-}} 5 $${\mathrm{Rank}} {\text{-}} 10 $${\mathrm{mAP}} $
Original93.297.798.784.1
Foreground74.887.090.851.8
Background4.38.812.51.4
Tab.1 Experimental result of training with original image from Market1501 dataset %
测试类型${\mathrm{Rank}} {\text{-}} 1 $${\mathrm{Rank}} {\text{-}} 5 $${\mathrm{Rank}} {\text{-}} 10 $${\mathrm{mAP}} $
Original85.393.496.074.6
Foreground52.267.172.431.5
Background6.313.518.02.5
Tab.2 Experimental result of training with original image from DukeMTMC dataset %
测试类型${\mathrm{Rank}} {\text{-}} 1 $${\mathrm{Rank}} {\text{-}}5 $${\mathrm{Rank}} {\text{-}} 10 $${\mathrm{mAP}} $
Original35.153.061.617.3
Foreground31.348.456.113.1
Background18.437.647.78.8
Tab.3 Experimental result of training with background image from Market1501 dataset %
测试类型${\mathrm{Rank}}{\text{-}} 1 $${\mathrm{Rank}} {\text{-}} 5 $${\mathrm{Rank}} {\text{-}} 10 $${\mathrm{mAP}} $
Original52.569.475.135.7
Foreground31.145.751.615.9
Background39.859.667.722.6
Tab.4 Experimental result of training with background image from DukeMTMC dataset %
测试类型${\mathrm{Rank}} {\text{-}}1 $${\mathrm{Rank}} {\text{-}} 5 $${\mathrm{Rank}} {\text{-}} 10 $${\mathrm{mAP}} $
Original68.486.391.345.4
Foreground90.696.197.775.5
Background2.04.35.80.7
Tab.5 Experimental result of training with foreground image from Market1501 dataset %
测试类型${\mathrm{Rank}} {\text{-}} 1 $${\mathrm{Rank}} {\text{-}}5 $${\mathrm{Rank}} {\text{-}} 10 $${\mathrm{mAP}} $
Original57.974.579.839.3
Foreground77.187.090.160.1
Background2.35.98.41.0
Tab.6 Experimental result of training with foreground image from DukeMTMC dataset %
Fig.4 Person re-identification model based on multi-part guided foreground enhancement
Fig.5 Bottleneck optimization module
硬件/软件型号/版本号
CPUIntel Xeon Silver 4210R@2.40 GHz
存储14 TB SATA
内存256 GB DDR4-2933
GPU4*NVIDIA RTX 3090 24 GB
操作系统Ubuntu 20.04.3
GPU驱动510.47.03
CUDA11.7
CUDNN8.5.0
Python3.8.0
Pytorch1.13.1+cu117
Tab.7 Experimental hardware and software environment configuration of person re-identification model
参数数值
训练次数(epoch)120
批大小(batch_size)64
初始学习率(lr)3.5×10?4
学习衰减率(lr_decay)0.1
优化函数(optimizer)Adam
动量大小(momentum)0.9
权重衰减系数(weight_decay)0.000 5
随机水平翻转概率(random horizontal flip probability)0.5
随机擦除概率(erase_probability)0.5
Tab.8 Configuration of training parameter of person re-identification model
方法Market1501DukeMTMC
Rank-1mAPRank-1${\mathrm{mAP}} $
TriNet[21]84.969.172.453.5
ARP[22]87.066.973.955.5
JLML[23]85.165.5——
AlignedReID[24]92.682.181.267.4
HA-CNN[25]91.275.780.563.8
PCB[26]93.881.683.369.2
Mancs[27]93.182.384.971.8
OSNet[28]94.884.987.673.5
Auto-ReID[29]94.585.188.575.1
HOReid[30]94.284.986.975.6
CBDB-Net[31]94.485.087.774.3
DRL-Net[32]94.786.988.176.6
MGFE(本文方法)95.087.588.377.9
Tab.9 Performance comparison of proposed method with existing person re-identification method %
训练集->测试集方法${\mathrm{Rank}} \text{-} 1 $/%${\mathrm{mAP}} $/%
Market1501(Original)->
Market1501(Foreground)		Baseline74.851.8
+MPG77.956.2
+BOM78.256.6
MGFE78.856.9
DukeMTMC(Original)->
DukeMTMC(Foreground)		Baseline52.231.5
+MPG54.534.6
+BOM54.834.9
MGFE55.335.5
Tab.10 Impact of module combination on model’s foreground recognition performance
训练集->测试集方法${\mathrm{Rank}} \text{-} 1 $/%${\mathrm{mAP}} $/%
Market1501(Original)->
Market1501(Original)		Baseline93.284.1
+MPG94.385.7
+BOM94.585.9
MGFE95.087.5
DukeMTMC(Original)->
DukeMTMC(Original)		Baseline85.374.6
+MPG86.075.0
+BOM87.476.9
MGFE88.377.9
Tab.11 Experimental result of effectiveness of each module in improving model recognition performance
训练集->测试集方法${\mathrm{Rank}} \text{-} 1 $/%${\mathrm{mAP}} $/%
Market1501(Original)->
DukeMTMC(Original)		Baseline28.115.1
+MPG30.715.9
+BOM30.016.4
MGFE31.416.9
DukeMTMC(Original)->
Market1501(Original)		Baseline41.217.2
+MPG42.918.5
+BOM43.819.7
MGFE44.720.2
Tab.12 Experimental result of effectiveness of each module in improving model generalization
[1]   YE M, SHEN J, LIN G, et al Deep learning for person re-identification: a survey and outlook[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44 (6): 2872- 2893
doi: 10.1109/TPAMI.2021.3054775
[2]   ZHENG L, ZHANG H, SUN S, et al. Person re-identification in the wild [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . Honolulu: IEEE, 2017: 1367-1376.
[3]   YAO H, ZHANG S, HONG R, et al Deep representation learning with part loss for person re-identification[J]. IEEE Transactions on Image Processing, 2019, 28 (6): 2860- 2871
doi: 10.1109/TIP.2019.2891888
[4]   ZHAO L, LI X, ZHUANG Y, et al. Deeply-learned part-aligned representations for person re-identification [C]// Proceedings of the IEEE/CVF International Conference on Computer Vision . Venice: IEEE, 2017: 3219-3228.
[5]   SU C, ZHANG S, XING J, et al. Deep attributes driven multi-camera person re-identification [C]// Proceedings of the European Conference on Computer Vision . Amsterdam: Springer, 2016: 475-491.
[6]   LIN Y, ZHENG L, ZHENG Z, et al Improving person re-identification by attribute and identity learning[J]. Pattern Recognition, 2019, 95: 151- 161
doi: 10.1016/j.patcog.2019.06.006
[7]   MATSUKAWA T, SUZUKI E. Person re-identification using CNN features learned from combination of attributes [C]// 23rd International Conference on Pattern Recognition . Cancun: IEEE, 2016: 2428-2433.
[8]   HUANG H, LI D, ZHANG Z, et al. Adversarially occluded samples for person re-identification [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . Salt Lake City: IEEE, 2018: 5098-5107.
[9]   VARIOR R R, SHUAI B, LU J, et al. A siamese long short-term memory architecture for human Re-identification [C]// Proceedings of the European Conference on Computer Vision . Amsterdam: Springer, 2016: 135-153.
[10]   SONG C, HUANG Y, OUYANG W, et al. Mask-guided contrastive attention model for person re-identification [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 1179-1188.
[11]   YE M, LIANG C, WANG Z, et al. Ranking optimization for person re-identification via similarity and dissimilarity [C]// Proceedings of the ACM International Conference on Multimedia. Brisbane: ACM, 2015: 1239-1242.
[12]   TIAN M, YI S, LI H, et al. Eliminating background-bias for robust person re-identification [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . Salt Lake City: IEEE, 2018: 5794-5803.
[13]   ZHANG X, ZHU X, TANG M, et al. Deep learning for human parsing: a survey [EB/OL]. (2023-01-29)[2024-07-06]. https://arxiv.org/pdf/2301.12416.
[14]   刘俊婧, 郑宛露, 王少荣. 基于双重注意力及边缘约束的人体解析方法[EB/OL]. (2024-04-26)[2024-07-06]. https://github.com/shaorongwang/HumanParsing/blob/main/Human%20Parsing.pdf.
[15]   FU J, LIU J, TIAN H, et al. Dual attention network for scene segmentation [C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Long Beach: IEEE, 2019: 3146-3154.
[16]   LUO H, JIANG W, GU Y, et al A strong baseline and batch normalization neck for deep person re-identification[J]. IEEE Transactions on Multimedia, 2019, 22 (10): 2597- 2609
[17]   PARK J, WOO S, LEE J Y, et al. BAM: bottleneck attention module [C]// Proceedings of the British Machine Vision Conference . London: BMVA Press, 2018: 147-160.
[18]   ZHENG L, SHEN L, TIAN L, et al. Scalable person re-identification: a benchmark [C]// Proceedings of the IEEE/CVF International Conference on Computer Vision . Santiago: IEEE, 2015: 1116-1124.
[19]   ZHENG Z, ZHENG L, YANG Y. Unlabeled samples generated by GAN improve the person re-identification baseline in vitro [C]// Proceedings of the IEEE/CVF International Conference on Computer Vision . Venice: IEEE, 2017: 3754-3762.
[20]   WANG X, DORETTO G, SEBASTIAN T, et al. Shape and appearance context modeling [C]// IEEE 11th International Conference on Computer Vision . Rio de Janeiro: IEEE, 2007: 1-8.
[21]   HERMANS A, BEYER L, LEIBE B. In defense of the triplet loss for person re-identification [C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Long Beach: IEEE, 2019: 1526-1535.
[22]   LI W, ZHAO R, XIAO T, et al. Deepreid: deep filter pairing neural network for person re-identification [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . Columbus: IEEE, 2014: 152-159.
[23]   LI W, ZHU X, GONG S. Person re-identification by deep joint learning of multi-loss classification [C]// Proceedings of the 26th International Joint Conference on Artificial Intelligence. Melbourne: IJCAI Press, 2017: 2194-2200.
LI W, ZHU X, GONG S. Person re-identification by deep joint learning of multi-loss classification [C]// Proceedings of the 26th International Joint Conference on Artificial Intelligence . Melbourne: IJCAI Press, 2017: 2194-2200.
[24]   ZHANG X, LUO H, FAN X, et al. AlignedReID: surpassing human-level performance in person re-identification [EB/OL]. (2018-01-31)[2024-07-06]. https://arxiv.org/pdf/1711.08184.
[25]   LI W, ZHU X, GONG S. Harmonious attention network for person re-identification [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . Salt Lake City: IEEE, 2018: 2285-2294.
[26]   SUN Y, ZHENG L, YANG Y, et al. Beyond part models: person retrieval with refined part pooling (and a strong convolutional baseline [C]// Proceedings of the European Conference on Computer Vision . Munich: Springer, 2018: 480-496.
[27]   WANG C, ZHANG Q, HUANG C, et al. Mancs: a multi-task attentional network with curriculum sampling for person re-identification [C]// Proceedings of the European Conference on Computer Vision . Munich: Springer, 2018: 365-381.
[28]   ZHOU K, YANG Y, CAVALLARO A, et al. Omni-scale feature learning for person re-identification [C]// Proceedings of the IEEE/CVF International Conference on Computer Vision . Seoul: IEEE, 2019: 3702-3712.
[29]   QUAN R, DONG X, WU Y, et al. Auto-reid: searching for a part-aware convnet for person re-identification [C]// Proceedings of the IEEE/CVF International Conference on Computer Vision . Seoul: IEEE, 2019: 3750-3759.
[30]   WANG G, YANG S, LIU H, et al. High-order information matters: learning relation and topology for occluded person re-identification [C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Seattle: IEEE, 2020: 6449-6458.
[31]   TAN H, LIU X, BIAN Y, et al Incomplete descriptor mining with elastic loss for person re-identification[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2021, 32 (1): 160- 171
[1] Shi-lin ZHANG,Hong-nan GUO,Xuan LIU. Person and vehicle re-identification based on energy model[J]. Journal of ZheJiang University (Engineering Science), 2022, 56(7): 1416-1424.
[2] Hao-yuan WANG,Yu LIANG,Wei ZHANG. Real-time smoke segmentation algorithm fused with multi-resolution representation[J]. Journal of ZheJiang University (Engineering Science), 2021, 55(12): 2334-2341.
[3] Yu-ming YAN,Jian-feng HE,Shu-zhao LI,Hui-min YU,Feng WANG,Zhou-feng REN. Identity recognition under real scenes[J]. Journal of ZheJiang University (Engineering Science), 2021, 55(11): 2022-2032.