集成空间注意力和姿态估计的遮挡行人再辨识

杨静; 张灿龙; 李志欣; 唐艳平

doi:10.7544/issn1000-1239.20200949

集成空间注意力和姿态估计的遮挡行人再辨识

¹(广西多源信息挖掘与安全重点实验室(广西师范大学) 广西桂林 541004)
²(桂林电子科技大学计算机与信息安全学院广西桂林 541004) (jingy_7@163.com)

基金项目: 国家自然科学基金项目(61866004,61966004,61962007,61751213);广西多源信息挖掘与安全重点实验室系统性研究课题基金项目(20-A-03-01);广西自然科学基金项目(2018GXNSFDA281009,2019GXNSFDA245018,2018GXNSFDA294001);广西“八桂学者”创新研究团队项目(2018010768);广西研究生教育创新计划项目(JXXYYJSCXXM-2021-007)

详细信息

中图分类号: TP391
计量
- 文章访问数: 254
- HTML全文浏览量: 5
- PDF下载量: 135
出版历程
- 发布日期: 2022-06-30

Integrated Spatial Attention and Pose Estimation for Occluded Person Re-Identification

¹(Guangxi Key Laboratory of Multi-source Information Mining & Security (Guangxi Normal University), Guilin, Guangxi 541004)
²(School of Computer Science and Information Security, Guilin University of Electronic Technology, Guilin, Guangxi 541004)

Funds: This work was supported by the National Natural Science Foundation of China (61866004, 61966004, 61962007, 61751213), the Research Fund of Guangxi Key Laboratory of Multi-source Information Mining and Security (20-A-03-01), Guangxi Natural Science Foundation (2018GXNSFDA281009, 2019GXNSFDA245018, 2018GXNSFDA294001), Guangxi “Bagui Scholar” Teams for Innovation and Research Project (2018010768), and the Innovation Project of Guangxi Graduate Education (JXXYYJSCXXM-2021-007).

摘要

摘要: 自然场景下监控设备所拍摄的行人图片总是存在被各种障碍物遮挡的情况，因此遮挡是行人再辨识面临的一个很大的挑战.针对遮挡问题，提出了一种集成空间注意力和姿态估计(spatial attention and pose estimation, SAPE)的遮挡行人再辨识模型.为了同时兼顾全局特征和局部特征，实现特征的多细粒度表示，构建了多任务网络.通过空间注意力机制将感兴趣区域锚定到图像中未遮挡的空间语义信息，从全局结构模式中挖掘有助于再辨识的视觉知识；然后结合分块匹配的思想，将残差网络提取到的特征图水平均匀分割成若干块，通过局部特征的匹配增加辨识的细粒度；在此基础之上，改进姿态估计器去提取图像中行人的关键点信息，并与卷积神经网络抽取的特征图相融合，然后设置阈值去除掉遮挡区域，得到辨识性强的特征，以消除遮挡对再辨识结果的影响.在Occluded-DukeMTMC, Occluded-REID, Partial-REID这3个数据集上验证了SAPE模型的有效性，实验结果表明提出的针对遮挡的模型具有良好的效果.
- 深度学习 /
- 注意力机制 /
- 姿态估计 /
- 多任务网络 /
- 遮挡行人再辨识
Abstract: Since the pedestrian images taken by the monitoring equipment in natural scenes are always occluded by various obstacles, occlusions is a great challenge for person re-identification. For the above problems, a spatial attention and pose estimation (SAPE) is proposed. In order to give consideration to both global and local features, a multi-task network is constructed to realize multi-granularity representation of features. By means of spatial attention mechanism, the region of interest is directed to the spatial semantic information in the image, and the visual knowledge which is helpful for re-identification is mined from the global structural pattern. Then, combined with the idea of part matching, the feature map extracted from the residual network is evenly divided into several parts horizontally, and the identification granularity is increased by matching the local features. On this basis, the key information of pedestrians in the image extracted by the improved pose estimator is fused with the feature map extracted by the convolutional neural network, and the threshold is set to remove the occlusion area, and the features with strong identification are obtained, so as to eliminate the influence of occlusion on the re-identification results. We verify the effectiveness of the SAPE model on three datasets of Occluded-DukeMTMC, Occluded-REID and Partial-REID. The experimental results show that SAPE has achieved good experimental results.
- deep learning /
- attention mechanism /
- pose estimation /
- multitask network /
- occluded person re-identification

HTML全文

参考文献(0)

施引文献(15)

期刊类型引用(8)

1.	吴树芳，高梦蛟，朱杰. 融合兴趣主题矩阵和主题生命树的社交用户长短兴趣挖掘. 情报理论与实践. 2024(02): 161-169 . 百度学术
2.	赵容梅，孙思雨，鄢凡力，彭舰，琚生根. 基于对比学习的多兴趣感知序列推荐系统. 计算机研究与发展. 2024(07): 1730-1740 . 本站查看
3.	刘鑫. 内容过滤技术与挖掘算法的设计优化. 电子技术. 2024(05): 42-43 . 百度学术
4.	张文龙，孙福振，吴相帅，李鹏程，王绍卿. 基于反向延长增强的对抗生成网络推荐算法. 计算机应用研究. 2024(07): 2033-2038 . 百度学术
5.	陈万志，王军. 时间感知增强的动态图神经网络序列推荐算法. 计算机工程与应用. 2024(20): 142-152 . 百度学术
6.	谢鸿博. 前后端分离架构下基于图神经网络的社交网络关系挖掘. 信息技术与信息化. 2024(12): 104-107 . 百度学术
7.	耿杰，刘春丽，魏雪梅，程明月，袁昆，李洋，刘业政. 基于用户重购行为的产品推荐方法. 计算机研究与发展. 2023(08): 1795-1807 . 本站查看
8.	李乃文，王胜男. 融合用户属性的虚拟学术社区用户画像模型构建研究. 情报探索. 2022(10): 85-90 . 百度学术