1. College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310007, China 2. Jiangnan Social University, Suzhou 215124, China
An identity recognition system combing face recognition and person Re-identification (Re-id) was designed for identity recognition under real scenes. When the pedestrian ’s face was detected, the recognition accuracy of the system was significantly improved through the face recognition technology. Improved cloth-elimination shape-distillation (ICESD) module was used to solve the problem of cloth-changing of pedestrians. In a real scene with a large number of unknown pedestrians, the recognition accuracy is significantly improved through the threshold filtering recognition process of identifying the known pedestrian first and then the specific pedestrian category. The application tests and comparative tests in real scenes demonstrate that the proposed Re-id method achieves significant performance improvement compared with mainstream deep learning methods under cloth-changing scenarios, and face recognition technology can also significantly improve the overall recognition accuracy. In the meantime, the real-time analysis of the system shows that the system can be deployed under real scenes.
Fig.4Attention module of improved cloth-elimination shape-distillation module
Fig.5Face recognition model
指标
A/%
Rank-1
47.80
Rank-5
65.40
Rank-10
75.60
正样本正确划分率
82.50
无关样本正确划分率
99.98
Tab.1Performance of person Re-id model
模型
mAP
Rank-1
ResNet-50（Baseline）
10.7
13.3
Baseline+CESD3
26.7
44.4
Baseline+ICESD3
26.9
45.6
Baseline+CESD4
26.6
44.5
Baseline+ICESD4
26.9
45.6
Baseline+CESD3+ CESD4
27.0
46.7
Baseline+ICESD3+ ICESD4
27.7
47.8
Tab.2Component analysis of Re-id model
模型
mAP
Rank-1
ResNet-50[14]
10.7
13.3
SENet-50[24]
16.5
20.5
HACNN[17]
21.6
39.9
MuDeep[16]
22.7
40.2
PCB[11]
24.3
43.6
本研究所提模型
27.7
47.8
Tab.3Comparison of person Re-id
数据样本
数据样本
Ac/%
正样本
总正样本
58
带人脸正样本
69
不带人脸正样
44
无关样本
总无关样本
97
带人脸无关样本
97
不带人脸无关样本
97
总样本
总样本
89
带人脸数据
92
不带人脸数据
86
Tab.4Statisticoftestdataoutput
模型
P/M
FLOPs/G
行人重识别模型
46.6
6.0
ResNet-50[14]（行人）
24.0
4.1
人脸识别模型
24.0
1.6
ResNet-50[14]（人脸）
24.0
1.6
Tab.5Params and FLOPs of deep model
数据类型
数据样本
t/s1)
1）注：时间为10次测试平均值.
包含人脸数据
系统运行整体时间
4.69
导入模型时间
3.90
运算时间
0.79
不包含人脸数据
系统运行整体时间
4.52
导入模型时间
3.94
运算时间
0.58
Tab.6Computing time of recognition system
[1]
罗浩, 姜伟, 范星, 等 基于深度学习的行人重识别研究进展[J]. 自动化学报, 2019, 45 (11): 2032- 2049 LUO Hao, JIANG Wei, FAN Xing, et al A survey on deep learning based person re-identification[J]. Acta Automatica Sinica, 2019, 45 (11): 2032- 2049
[2]
李幼蛟, 卓力, 张菁, 等 行人再识别技术综述[J]. 自动化学报, 2018, 44 (9): 1554- 1568 LI You-jiao, ZHUO Li, ZHANG Jing, et al A survey of person re-identification[J]. Acta Automatica Sinica, 2018, 44 (9): 1554- 1568
[3]
QIAN X L, WANG W X, ZHANG L, et al. Long-term cloth-changing person re-identification [EB/OL]. [2020-05-26]. https://arxiv.org/abs/2005.12633.
[4]
JIN X, LAN C, ZENG W, et al. Style normalization and restitution for generalizable person re-identification[C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 3140-3149.
[5]
WEI L H, ZHANG S L, GAO W, et al. Person transfer GAN to bridge domain gap for person re-identification [C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 79-88.
[6]
KHATUN A, DENMAN S, SRIDHARAN, S, et al. End-to-end domain adaptive attention network for cross-domain person re-identification [EB/OL]. [2020-05-07]. https://arxiv.org/abs/2005.03222.
[7]
ZHAI Y P, LU S J. AD-Cluster: augmented discriminative clustering for domain adaptive person re-identification [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 9018-9027.
[8]
SCHROFF F, KALENICHENKO D, PHILBIN J. FaceNet: a unified embedding for face recognition and clustering [C]// Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition. Boston: IEEE, 2015: 815-823.
[9]
WEN Y D, ZHANG K P, LI Z F. A discriminative feature learning approach for deep face recognition [C]// Proceedings of the 2016 European Conference on Computer Vision. Amsterdam: Springer, 2016: 499-515.
[10]
ZHANG J, LIU L, XU C, et al. Hierarchical and efficient learning for person re-identification [EB/OL]. [2020-05-18]. https://arxiv.org/abs/2005.08812.
[11]
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 2016 European Conference on Computer Vision. Munich: Springer, 2018: 3387–3396.
[12]
LIAO S C, HU Y, ZHU X Y, et al. Person re-identification by local maximal occurrence representation and metric learning [C]// Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition. Boston: IEEE, 2015: 2197−2206.
[13]
TAGORE N K, SINGH A, MANCHE S, et al. Deep learning based person re-identification [EB/OL]. [2020-05-07]. https://arxiv.org/abs/2005.03293.
[14]
HE K M, ZHANG X, REN S, et al. Deep residual learning for image recognition [C]// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
[15]
ULYANOV D, VEDALDI A, LEMPITSKY V. Instance normalization: the missing ingredient for fast stylization [EB/OL]. [2016-06-27]. https://arxiv.org/abs/1607.08022.
[16]
QIAN X L, FU Y, XIANG T, et al Leader-based multi-scale attention deep architecture for person re-identification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42 (2): 371- 385
doi: 10.1109/TPAMI.2019.2928294
[17]
LI W, ZHU X, GONG S. Harmonious attention network for person re-identification [C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 2285–2294.
[18]
CHEN B, DENG W, HU J. Mixed high-order attention network for person re-identification [C]// Proceedings of the 2019 IEEE International Conference on Computer Vision. Seoul: IEEE, 2019: 371-381.
[19]
周勇, 王瀚正, 赵佳琦, 等. 基于可解释注意力部件模型的行人重识别方法[EB/OL]. [2020-08-23]. https://kns.cnki.net/kcms/detail/detail.aspx?dbcode=CAPJ&dbname=CAPJLAST&filename=MOTO20201105000&uniplatform=NZKPT&v=JCreBvAGjWgg9eTStxOX1MHBH7tyLrog82SWoZ97XdB1Id9qWj877hYBzM9evYAl. ZHOU Yong, WANG Han-zheng, ZHAO Jia-qi, et al. Interpretable Attention Part Model for Person Re-identification [EB/OL]. [2020-08-23]. https:/kns.cnki.net/kcms/detail/detail.aspx?dbcode=CAPJ&dbname=CAPJLAST&filename=MOTO20201105000&uniplatform=NZKPT&v=JCreBvAGjWgg9eTStxOX1MHBH7tyLrog82SWoZ97XdB1Id9qWj877hYBzM9evYAl.
[20]
ZHANG K, ZHANG Z, LI Z, et al Joint face detection and alignment using multitask cascaded convolutional networks[J]. IEEE Signal Processing Letters, 2016, 23: 1499- 1503
doi: 10.1109/LSP.2016.2603342
[21]
JIN X., LAN C, ZENG W, et al. Global distance distributions separation for unsupervised person re-identification [EB/OL]. [2020-06-01]. https://arxiv.org/pdf/2006.00752.pdf.
[22]
LIN Y T, XIE L X. Unsupervised person re-identification via softened similarity learning [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 3387-3396.
[23]
ZHENG L, SHEN L, TIAN L, et al. Scalable person re-identification: a benchmark [C]// Proceedings of the 2015 IEEE International Conference on Computer Vision. Santiago: IEEE, 2015: 1116–1124.
YANG Bing, WANG Xiao-hua, YANG Xin, HUANG Xiao-xi. Face recognition method based on HOG pyramid[J]. Journal of ZheJiang University (Engineering Science), 2014, 48(9): 1564-1569.
