[1]赵亚凤,于继超,孙骞,等.基于Cluster Contrast和ViT的东北虎个体重识别无监督学习框架研究[J].智能系统学报,2026,21(2):435-443.[doi:10.11992/tis.202507012]
ZHAO Yafeng,YU Jichao,SUN Qian,et al.Unsupervised framework for Amur tiger re-identification with Cluster Contrast and ViT[J].CAAI Transactions on Intelligent Systems,2026,21(2):435-443.[doi:10.11992/tis.202507012]
点击复制
ZHAO Yafeng,YU Jichao,SUN Qian,et al.Unsupervised framework for Amur tiger re-identification with Cluster Contrast and ViT[J].CAAI Transactions on Intelligent Systems,2026,21(2):435-443.[doi:10.11992/tis.202507012]
基于Cluster Contrast和ViT的东北虎个体重识别无监督学习框架研究
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
21
期数:
2026年第2期
页码:
435-443
栏目:
学术论文—机器感知与模式识别
出版日期:
2026-03-05
- Title:
- Unsupervised framework for Amur tiger re-identification with Cluster Contrast and ViT
- 关键词:
- 东北虎; 重识别; 深度学习; 无监督学习; ATRW 数据集; Cluster Contrast机制; vision Transformer; 坐标注意力
- Keywords:
- Amur tiger; re-identification; deep learning; unsupervised learning; ATRW dataset; Cluster Contrast mechanism; vision Transformer; coordinate attention
- 分类号:
- TP391.4
- 摘要:
- 针对东北虎个体重识别中野外数据标注困难、样本失衡等挑战，以野外东北虎 （Amur tiger re-identification in the wild，ATRW）数据集为基础提出一种“全局特征提取-空间位置强化-无监督均衡训练”的协同框架，完成无监督重识别。使用vision Transformer（ViT）自注意力机制捕捉东北虎条纹的全局长距离的依赖特征，通过坐标注意力机制加强对条纹空间位置的精确解析，解决传统卷积神经网络局部性导致的特征关联缺失问题。引入Cluster Contrast机制构建簇级内存字典，通过动量更新平衡不同样本量东北虎的特征优化速率，避免无监督学习中样本失衡导致的特征偏差。实验表明，模型在ATRW（r+i）数据集上平均精度的值为86.4%，高于原有的特征提取ViT和Resnet50_ibn模型，对不同数据分布和数据量具有良好泛化能力，适配野外可见光/红外多设备协同监测需求。本文所提方法为东北虎个体识别提供了兼具准确性与鲁棒性的技术方案。
- Abstract:
- To address challenges like difficult annotation of wild data and sample imbalance in Amur tiger individual re-identification, a collaborative framework based on the ATRW dataset was developed for unsupervised re-identification. The ViT self-attention mechanism was used to capture global long-distance dependent features of stripes. Combined with the coordinate attention mechanism, it enhanced spatial position analysis of stripes, making up for missing feature correlations caused by the locality of traditional CNNs. The Cluster Contrast mechanism was introduced to build a cluster-level memory dictionary. Momentum update balanced feature optimization rates of individuals with different sample sizes, alleviating feature biases from sample imbalance in unsupervised learning. Experiments show the model achieves an mAP of86.4% on the ATRW (r+i) dataset, outperforming the original feature extraction models (ViT and Resnet50_ibn). It exhibits good generalization ability across different data distributions and data volumes, and is suitable for the demand of collaborative monitoring with visible light/infrared multi-devices in the wild. The method proposed in this study provides a technical solution with both accuracy and robustness for Amur tiger individual identification.
- 参考文献/References:
-
[1] 国家林业和草原局, 农业农村部. 《国家重点保护野生动物名录》(2021年2月1日修订)[J]. 野生动物学报, 2021, 42(2): 605-640 National Forestry and Grassland Bureau, Ministry of Agriculture and Rural Affairs. List of national key protected wild animals (revised on February 1, 2021)[J]. Chinese journal of wildlife, 2021, 42(2): 605-640
[2] QI Jinzhe, HOLYOAK M, NING Yao, et al. Ecological thresholds and large carnivores conservation: implications for the Amur tiger and leopard in China[J]. Global ecology and conservation, 2020, 21: e00837
[3] 王凤昆, 李艳, 姜广顺. 中俄东北虎自然保护地建设进展[J]. 自然保护地, 2024, 4(4): 38-52 WANG Fengkun, LI Yan, JIANG Guangshun. Progress in the construction of Sino-Russia Amur Tiger(Panthera Tigris altaica) Protected Areas[J]. Natural protected areas, 2024, 4(4): 38-52
[4] 吴峰, 温佩颖, 唐志珍, 等. 东北虎豹国家公园植物DNA微条形码数据库[J]. 北京师范大学学报(自然科学版), 2023, 59(4): 623-628 WU Feng, WEN Peiying, TANG Zhizhen, et al. A DNA mini-barcode reference library for environmental DNA study in the Northeast Tiger and Leopard National Park[J]. Journal of Beijing Normal University (natural science edition), 2023, 59(4): 623-628
[5] 韦怡, 姜广顺. 虎豹及有蹄类猎物种群数量监测方法概述[J]. 生物多样性, 2022, 30(9): 129-146 WEI Yi, JIANG Guangshun. Overview of monitoring methods for tigers, leopards and ungulate prey[J]. Biodiversity science, 2022, 30(9): 129-146
[6] HE Fengping, LIU Dan, ZHANG Le, et al. Metagenomic analysis of captive Amur tiger faecal microbiome[J]. BMC veterinary research, 2018, 14(1): 379
[7] KERLEY L L. Using dogs for tiger conservation and research[J]. Integrative zoology, 2010, 5(4): 390-396
[8] TUIA D, KELLENBERGER B, BEERY S, et al. Perspectives in machine learning for wildlife conservation[J]. Nature communications, 2022, 13: 792
[9] YE Mang, CHEN Shuoyi, LI Chenyue, et al. Transformer for object re-identification: a survey[J]. International journal of computer vision, 2025, 133(5): 2410-2440
[10] YE Mang, SHEN Jianbing, LIN Gaojie, 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
[11] 张国印, 王传博, 高伟. 抗遮挡的行人多目标跟踪算法[J]. 智能系统学报, 2024, 19(5): 1248-1256 ZHANG Guoyin, WANG Chuanbo, GAO Wei. Pedestrian multiobject tracking algorithm with anti-occlusion[J]. CAAI transactions on intelligent systems, 2024, 19(5): 1248-1256
[12] 王路遥, 王凤随, 闫涛, 等. 结合多尺度特征与混淆学习的跨模态行人重识别[J]. 智能系统学报, 2024, 19(4): 898-908 WANG Luyao, WANG Fengsui, YAN Tao, et al. Cross-modal person re-identification combining multi-scale features and confusion learning[J]. CAAI transactions on intelligent systems, 2024, 19(4): 898-908
[13] 赵玲, 周桂红, 任力生. 基于牛脸和躯干综合信息的奶牛个体识别研究[J]. 河北农业大学学报, 2024, 47(2): 112-118 ZHAO Ling, ZHOU Guihong, REN Lisheng. Individual identification of dairy cows based on comprehensive face and trunk information[J]. Journal of Agricultural University of Hebei, 2024, 47(2): 112-118
[14] FREYTAG A, RODNER E, SIMON M, et al. Chimpanzee faces in the wild: log-euclidean CNNs for predicting identities and attributes of Primates[C]//Pattern Recognition. Cham: Springer, 2016: 51–63.
[15] 顾佳音, 刘辉, 姜广顺. 东北虎(Panthera tigris altaica)个体识别技术研究进展[J]. 野生动物, 2013, 34(4): 229-237, 248 GU Jiayin, LIU Hui, JIANG Guangshun. A review of potential techniques for indentifying individual Amur tigers (Panthera tigris altaica)[J]. Chinese wildlife, 2013, 34(4): 229-237, 248
[16] GLOVER J D, SUDDERICK Z R, SHIH B B, et al. The developmental basis of fingerprint pattern formation and variation[J]. Cell, 2023, 186(5): 940–956. e20.
[17] SHI Chunmei, LIU Dan, CUI Yonglu, et al. Amur tiger stripes: individual identification based on deep convolutional neural network[J]. Integrative zoology, 2020, 15(6): 461-470
[18] 马光凯, 张静, 戴文锐, 等. 基于Transformer的东北虎体侧条纹个体识别[J]. 野生动物学报, 2024, 45(4): 734-743 MA Guangkai, ZHANG Jing, DAI Wenrui, et al. Body stripes individual identification of Amur tigers based on Transformer[J]. Chinese journal of wildlife, 2024, 45(4): 734-743
[19] CHMIELA S, SAUCEDA H E, MULLER K, et al. Addressing the elephant in the room: uncertainties in physical predictions from machine-learned force fields[J]. Bulletin of the American physical society, 2020.
[20] LI Shuyuan, LI Jianguo, TANG Hanlin, et al. ATRW: a benchmark for Amur tiger re-identification in the wild[C]//Proceedings of the 28th ACM International Conference on Multimedia. Seattle: ACM, 2020: 2590–2598.
[21] PARIKH A, T?CKSTR?M O, DAS D, et al. A decomposable attention model for natural language inference[C]//Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2016: 2249–2255.
[22] DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An image is worth 16×16 words: transformers for image recognition at scale[EB/OL]. (2020-10-22)[2025-08-03]. https://arxiv.org/abs/2010.11929.
[23] HOU Qibin, ZHOU Daquan, FENG Jiashi. Coordinate attention for efficient mobile network design[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 13708–13717.
[24] DAI Zuozhuo, WANG Guangyuan, YUAN Weihao, et al. Cluster contrast forUnsupervised person re-identification[C]//Computer Vision–ACCV 2022. Cham: Springer, 2023: 319–337.
[25] WANG Xiaogang, DORETTO G, SEBASTIAN T, et al. Shape and appearance context modeling[C]//2007 IEEE 11th International Conference on Computer Vision. Rio de Janeiro: IEEE, 2007: 1–8.
[26] ZHENG Liang, SHEN Liyue, TIAN Lu, et al. Scalable person re-identification: a benchmark[C]//2015 IEEE International Conference on Computer Vision. Santiago: IEEE, 2016: 1116–1124.
- 相似文献/References:
-
[1]张国印,王传博,高伟.抗遮挡的行人多目标跟踪算法[J].智能系统学报,2024,19(5):1248.[doi:10.11992/tis.202307002]
ZHANG Guoyin,WANG Chuanbo,GAO Wei.Pedestrian multiobject tracking algorithm with anti-occlusion[J].CAAI Transactions on Intelligent Systems,2024,19():1248.[doi:10.11992/tis.202307002]
备注/Memo
收稿日期:2025-7-12。
基金项目:国家自然科学基金项目(32371864).
作者简介:赵亚凤，副教授，博士，主要研究方向为计算机视觉与模式识别、人工智能与智能控制、无线传感器网络。主持和参与国家自然科学基金、黑龙江省自然科学基金等项目10余项，获发明专利及实用新型专利授权10余项，发表学术论文20余篇，作为主编及副主编出版著作2部。E-mail：zyf@nefu.edu.cn。;于继超，硕士研究生，主要研究方向为计算机视觉。E-mail：jichaoyu@nefu.edu.cn。;王梓丞，副教授，博士后。黑龙江省光学学会会员，国际期刊《Photonics》特邀编辑，《半导体光电》编委。主要研究方向为高密度集成光学器件设计、智能光纤传感系统开发及多物理场耦合仿真技术。主持或参与省部级及横向项目10余项，获黑龙江省科学技术发明二等奖，指导学生获国家级、省部级竞赛奖项多项。获发明专利及软件著作授权12项。近5年第一作者/通信作者发表论文共13篇。E-mail：wangzicheng1992@163.com。
通讯作者:王梓丞. E-mail：wangzicheng1992@163.com
更新日期/Last Update:
1900-01-01