Visual anomaly detection method for passenger facial images based on knowledge transferring
-
摘要: 基于深度学习的人脸识别技术以数据为驱动,对输入图像的质量要求较高。在铁路刷脸进/出站场景下,为滤除因各种因素导致的成像异常的人脸图像,提升人脸识别精度,文章研究人脸图像正常的特征分布,通过知识迁移,提出无须针对异常样本建模的人脸图像异常检测算法。理想情况下,该算法对人脸图像异常检测的ROC曲线下面积(AUROC,Aera Under Receiver Operating Characteristic)可达到0.979。实验结果表明,该算法在计算精度与运行成本的组合上具有较高的自由度,可实现不同场景、硬件条件下的算法适配,为优化旅客人脸识别的输入环节,提高各场景下的旅客人脸识别率提供了技术支撑。Abstract:As a data-driven technology, face recognition based on deep learning demands high-quality images as input.In order to filter out facal images with abnormalities caused by various factors in the scene of railway face recognition entering and exiting stations, and improve the accuracy of facial recognition, this paper studied the feature distribution of normal facial images, through knowledge transfer, and proposed a anomaly detection algorithm for facial images that did not require modeling abnormal samples. The AUROC of this algorithm for detecting facial images with abnormalities was 0.979. The experimental results show that the algorithm has a high degree of freedom in the combination of computational accuracy and operating costs, and can meet the algorithm adaptation in different scenarios and hardware conditions. It provides technical support for optimizing the input link of passenger face recognition and improving the passenger face recognition rate in various scenarios.
-
-
输入:人脸正常的特征分布$\mathbfcal{D}$,保留特征的百分比r 输出:特征选择后的子集$\mathbfcal{S}$ $\mathbfcal{S}\leftarrow \left\{\right\}$ $for\;i\;in\;range\left(r\right|\mathbfcal{D}\left|\right):$ ${{\boldsymbol{m}}}_{i}\leftarrow{\mathrm{argmax}}_{{\boldsymbol{m}}\in \mathbfcal{D}-\mathbfcal{S}}\;\;\;\;\;{\mathrm{min}}_{n\in \mathbfcal{S}}{\left|\right|{\boldsymbol{m}}-{\boldsymbol{n}}\left|\right|}_{2}$ $\mathbfcal{S}\leftarrow \mathbfcal{S}\bigcup \left\{{{\boldsymbol{m}}}_{i}\right\}$ $ end $ 表 1 人脸图像异常检测结果(有人脸分割)
特征百分比 $ r $ 模型 AUROC 运行速度/fps 1% ResNet-18 0.857 约450 ResNet-50 0.947 ResNet-100 0.979 10% ResNet-18 0.882 约50 ResNet-50 0.950 ResNet-100 0.975 25% ResNet-18 0.890 约20 ResNet-50 0.948 ResNet-100 0.971 50% ResNet-18 0.893 约10 ResNet-50 0.944 ResNet-100 0.969 表 2 人脸图像异常检测结果(无人脸分割)
特征百分比 $ r $ 模型 AUROC 运行速度/fps 1% ResNet-18 0.898 约450 ResNet-50 0.924 ResNet-100 0.938 10% ResNet-18 0.880 约50 ResNet-50 0.911 ResNet-100 0.929 25% ResNet-18 0.866 约20 ResNet-50 0.904 ResNet-100 0.923 50% ResNet-18 0.860 约10 ResNet-50 0.901 ResNet-100 0.921 -
[1] 戴琳琳,阎志远,景 辉. Insightface结合Faiss的高并发人脸识别技术研究 [J]. 铁路计算机应用,2020,29(10):16-20. DOI: 10.3969/j.issn.1005-8451.2020.10.004 [2] 贾成强,戴琳琳,徐海涛,等. 基于人脸识别技术的铁路实名制进站核验系统研究及设计 [J]. 铁路计算机应用,2018,27(7):49-53,63. DOI: 10.3969/j.issn.1005-8451.2018.07.012 [3] 蒋方玲,刘鹏程,周祥东. 人脸活体检测综述 [J]. 自动化学报,2021,47(8):1799-1821. DOI: 10.16383/j.aas.c180829 [4] 衣 帅,朱建生,景 辉. 铁路刷脸场景下基于MTCNN的人脸遮挡识别研究 [J]. 计算机仿真,2020,37(5):96-99. DOI: 10.3969/j.issn.1006-9348.2020.05.020 [5] Doosti B, Naha S, Mirbagheri M, et al. HOPE-Net: a graph-based model for hand-object pose estimation[C]//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 13-19 June, 2020, Seattle, USA. New York, USA: IEEE, 2020: 6608-6617.
[6] 吴佳丽,干宗良. 区域自适应的非均匀光照图像增强 [J]. 计算机技术与发展,2022,32(7):58-63,69. DOI: 10.3969/j.issn.1673-629X.2022.07.010 [7] 赵 月,王来花,王伟胜,等. 融合反向传播的无参考模糊图像质量评价 [J]. 计算机应用与软件,2022,39(9):248-254,306. DOI: 10.3969/j.issn.1000-386x.2022.09.037 [8] Bergmann P, Fauser M, Sattlegger D, et al. MVTec AD—A comprehensive real-world dataset for unsupervised anomaly detection[C]//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 15-20 June, 2019, Long Beach, USA. New York, USA: IEEE, 2019: 9584-9592.
[9] Menze B H, Jakab A, Bauer S, et al. The multimodal brain tumor image segmentation benchmark (BRATS) [J]. IEEE Transactions on Medical Imaging, 2015, 34(10): 1993-2024. DOI: 10.1109/TMI.2014.2377694
[10] Chandola V, Banerjee A, Kumar V. Anomaly detection: a survey [J]. ACM Computing Surveys, 2009, 41(3): 15.
[11] Schölkopf B, Williamson R, Smola A, et al. Support vector method for novelty detection[C]//Proceedings of the 12th International Conference on Neural Information Processing Systems, 29 November, 1999-4 December, 1999, Denver, USA. Cambridge, USA: MIT Press, 1999: 582-588.
[12] Ruff L, Vandermeulen R, Göernitz N, et al. Deep one-class classification[C]//Proceedings of the 35th International Conference on Machine Learning, 10-15 July, 2018, Stockholm, Sweden. New York, USA: PMLR, 2018: 4390-4399.
[13] Zhuang F Z, Qi Z Y, Duan K Y, et al. A comprehensive survey on transfer learning [J]. Proceedings of the IEEE, 2021, 109(1): 43-76. DOI: 10.1109/JPROC.2020.3004555
[14] Zhang W L, Li R J, Zeng T, et al. Deep model based transfer and multi-task learning for biological image analysis[C]//Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 10-13 August, 2015, Sydney, NSW, Australia. New York, USA: ACM, 2015: 1475-1484.
[15] Dai W Y, Yang Q, Xue G R, et al. Self-taught clustering[C]//Proceedings of the 25th International Conference on Machine Learning, 5-9 July, 2008, Helsinki, Finland. New York, USA: ACM, 2008: 200-207.
[16] Yan H L, Ding Y K, Li P H, et al. Mind the class weight bias: weighted maximum mean discrepancy for unsupervised domain adaptation[C]//Proceedings of 2017 Conference on Computer Vision and Pattern Recognition, 21-26 July, 2017, Honolulu, USA. New York, USA: IEEE, 2017: 2272-2281.
[17] 张永福,宋海林. 基于跳跃特征金字塔的域适应目标检测模型 [J]. 计算机技术与发展,2022,32(9):28-35. DOI: 10.3969/j.issn.1673-629X.2022.09.005 [18] Lee C H, Liu Z W, Wu L Y, et al. MaskGAN: towards diverse and interactive facial image manipulation[C]//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 13-19 June, 2020, Seattle, USA. New York, USA: IEEE, 2020: 5549-5558.
[19] Yu C Q, Wang J B, Peng C, et al. BiSeNet: bilateral segmentation network for real-time semantic segmentation[C]//Proceedings of the 15th European Conference on Computer Vision, 8-14 September, 2018, Munich, Germany. Cham: Springer, 2018: 334-349.
[20] Roth K, Pemula L, Zepeda J, et al. Towards total recall in industrial anomaly detection[C]//Proceedings of 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 18-24 June, 2022, New Orleans, USA. New York, USA: IEEE, 2022: 14318-14328.
[21] Deng J K, Guo J, Xue N N, et al. ArcFace: additive angular margin loss for deep face recognition[C]//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 15-20 June, 2019, Long Beach, USA. New York, USA: IEEE, 2019: 4690-4699.
[22] Sener O, Savarese S. Active learning for convolutional neural networks: a core-set approach[C]//Proceedings of the 6th International Conference on Learning Representations, 30 April-3 May, 2018, Vancouver, BC, Canada. OpenReview. net, 2018.
[23] Theodoridis S, Koutroumbas K. Pattern recognition[M]. 3rd ed. Amsterdam: Elsevier, 2006.
-
期刊类型引用(8)
1. 彭鹏,朱峰,李力,高振坤. 雅万高铁焊轨基地信息管理系统的开发及应用. 铁路计算机应用. 2022(04): 54-58 . 本站查看
2. 程智博,吴艳华,郑金子,赵正阳. 高速铁路更新改造与大修整治项目计划管理系统研究与实现. 铁路计算机应用. 2020(02): 1-5 . 本站查看
3. 苏尔慈,刘敏,李平,马小宁,李鑫. 基于铁路数据服务平台的铁路工务设备安全画像应用设计方案. 铁路计算机应用. 2020(06): 34-38 . 本站查看
4. 马建军,李平,马小宁,邵赛. 铁路一体化信息集成平台总体架构及关键技术研究. 中国铁道科学. 2020(05): 153-161 . 百度学术
5. 赵正阳,吴艳华,程智博,王云龙. 基于3D卷积神经网络的高铁轨道质量指数预测方法. 铁路计算机应用. 2020(12): 7-11+16 . 本站查看
6. 沈海燕,端嘉盈,王浩,徐晓磊,赵婉妤. 云物大智、区块链、CPS间的关系及在铁路领域研究综述. 铁路计算机应用. 2019(02): 1-6+11 . 本站查看
7. 曾志清. 试论铁路线路轨道工务维修养护技术. 建材与装饰. 2019(01): 280-281 . 百度学术
8. 王志华,冯文晖,米建设,付涛,刘忠海. 基于BIM的铁路房建信息管理平台方案研究. 铁路计算机应用. 2019(06): 64-68+77 . 本站查看
其他类型引用(4)