第1期 王成济，等：一种多层特征融合的人脸检测方法 ·145· 图像分割问题的算法框架尝试应用于人脸检测问 Recognition.Boston,MA,USA,2015:3431-3440 题。在前人的基础上本文探索了不同的特征串联方 [9]VIOLA P,JONES M.Rapid object detection using a boos- 法对人脸区域坐标回归的影响，通过实验发现并不 ted cascade of simple features[C]//Proceedings of the 2001 是特征组合得越多结果越好，本文使用p0ol4和 IEEE Computer Society Conference on Computer Vision pool5两个特征层的特征取得了很大的提升。在后 and Pattern Recognition.Kauai,HI,USA.2001.1:I-511-I- 518 处理阶段，本文通过比较分析不同的非极大值抑制 策略的性能，发现通常使用的不加权的非极大值抑 [10]LOWE D G.Distinctive image features from scale-invari- ant keypoints[J].International journal of computer vision, 制方法虽然高效，但会在一定程度上影响目标检测 2004,60(2):91-110. 方法的性能。本文在人脸区域分类问题和人脸区域 [11]DALAL N,TRIGGS B.Histograms of oriented gradients 内像素点坐标偏移量回归两个问题实际上是分开处 for human detection[C]//Proceedings of the 2005 IEEE 理，在今后的研究中如何发现并使用这两个问题之 Computer Society Conference on Computer Vision and 间的关联性是一个很重要的研究思路。本文虽然使 Pattern Recognition.San Diego,CA,USA,2005,1:886- 用加权得分的方法在一定程度上缓解了非极大值抑 893. 制方法检测算法的影响，但没有得出一般性的结 [12]OSUNA E,FREUND R,GIROSIT F.Training support 论，这个问题同样值得深入研究。 vector machines:an application to face detection[C]//Pro- 参考文献： ceedings of the 1997 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.San Juan, [1]ZAFEIRIOU S,ZHANG Cha,ZHANG Zhengyou.A sur- Argentina,1997:130-136. vey on face detection in the wild:past,present and future [13]FRIEDMAN J,HASTIE T,TIBSHIRANI R.Additive lo- [J].Computer vision and image understanding,2015,138: gistic regression:a statistical view of boosting(with dis- 1-24. cussion and a rejoinder by the authors)[J].The annals of [2]GIRSHICK R,DONAHUE J,DARRELL T,et al.Rich fea- statistics,2000,29(5):337-407. ture hierarchies for accurate object detection and semantic [14]ZITNICK C L,DOLLAR P.Edge boxes:locating object segmentation[C]//Proceedings of the IEEE Conference on proposals from edges[Cl//Proceedings of the 13th European Computer Vision and Pattern Recognition.Columbus,OH, Conference on Computer Vision.Zurich,Switzerland, USA,2014:580-587. 2014:391-405. [3]GIRSHICK R.Fast R-CNN[C]//Proceedings of the IEEE In- [15]UIJLINGS JRR,VAN DE SANDE K E A,GEVERS T, ternational Conference on Computer Vision.Santiago, et al.Selective search for object recognition[J].Internation- Chile.2015:1440-1448 al journal of computer vision,2013,104(2):154-171. [4]REN Shaoqing,HE Kaiming,GIRSHICK R,et al.Faster R- [16]LI Haoxiang,LIN Zhe,SHEN Xiaohui,et al.A convolu- CNN:towards real-time object detection with region pro- tional neural network cascade for face detection[Cl//Pro- posal networks[C]//Proceedings of the 28th International ceedings of the IEEE Conference on Computer Vision and Conference on Neural Information Processing Systems. Pattern Recognition.Boston,MA,USA,2015:5325-5334 Montreal,Canada.2015.1:91-99. [17]FARFADE S S,SABERIAN M J,LI Lijia.Multi-view [5]HUANG Lichao,YANG Yi,DENG Yafeng,et al.Dense- face detection using deep convolutional neural networks[C]// Box:unifying landmark localization with end to end object Proceedings of the 5th ACM on International Conference detection[J].arXiv preprint arXiv:1509.04874,2015. on Multimedia Retrieval.Shanghai,China,2015:643-650 [6]YU Jiahui,JIANG Yuning,WANG Zhangyang,et al.Unit- [18]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.Im- Box:An advanced object detection network[Cl//Proceed- ageNet classification with deep convolutional neural net- ings of the 2016 ACM on Multimedia Conference.Amster- works[C]//Proceedings of the 26th Annual Conference on dam,The Netherlands,2016:516-520. Neural Information Processing Systems 2012.Lake Tahoe, [7]SIMONYAN K,ZISSERMAN A.Very deep convolutional Nevada,USA,2012:1097-1105 networks for large-scale image recognition[C]//Proceedings [19]HE Kaiming,ZHANG Xiangyu,REN Shaoqing,et al. of the International Conference on Learning Representa- Deep residual learning for image recognition[C]//Proceed- tions.Oxford,USA.2015. ings of the IEEE Conference on Computer Vision and Pat- [8]LONG J,SHELHAMER E,DARRELL T.Fully convolu- tern Recognition.Las Vegas,NV,USA,2016:770-778. tional networks for semantic segmentation[C]//Proceedings [20]HARIHARAN B,ARBELAEZ P,GIRSHICK R,et al.Hy- of the IEEE Conference on Computer Vision and Pattern percolumns for object segmentation and fine-grained local-图像分割问题的算法框架尝试应用于人脸检测问 题。在前人的基础上本文探索了不同的特征串联方 法对人脸区域坐标回归的影响，通过实验发现并不 是特征组合得越多结果越好，本文使用 pool4 和 pool5 两个特征层的特征取得了很大的提升。在后 处理阶段，本文通过比较分析不同的非极大值抑制 策略的性能，发现通常使用的不加权的非极大值抑 制方法虽然高效，但会在一定程度上影响目标检测 方法的性能。本文在人脸区域分类问题和人脸区域 内像素点坐标偏移量回归两个问题实际上是分开处 理，在今后的研究中如何发现并使用这两个问题之 间的关联性是一个很重要的研究思路。本文虽然使 用加权得分的方法在一定程度上缓解了非极大值抑 制方法检测算法的影响，但没有得出一般性的结 论，这个问题同样值得深入研究。 参考文献： ZAFEIRIOU S, ZHANG Cha, ZHANG Zhengyou. A sur￾vey on face detection in the wild: past, present and future [J]. Computer vision and image understanding, 2015, 138: 1–24. [1] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich fea￾ture hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Columbus, OH, USA, 2014: 580–587. [2] GIRSHICK R. Fast R-CNN[C]//Proceedings of the IEEE In￾ternational Conference on Computer Vision. Santiago, Chile, 2015: 1440–1448. [3] REN Shaoqing, HE Kaiming, GIRSHICK R, et al. Faster R￾CNN: towards real-time object detection with region pro￾posal networks[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems. Montreal, Canada, 2015, 1: 91–99. [4] HUANG Lichao, YANG Yi, DENG Yafeng, et al. Dense￾Box: unifying landmark localization with end to end object detection[J]. arXiv preprint arXiv: 1509.04874, 2015. [5] YU Jiahui, JIANG Yuning, WANG Zhangyang, et al. Unit￾Box: An advanced object detection network[C]//Proceed￾ings of the 2016 ACM on Multimedia Conference. Amster￾dam, The Netherlands, 2016: 516–520. [6] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[C]//Proceedings of the International Conference on Learning Representa￾tions. Oxford, USA, 2015. [7] LONG J, SHELHAMER E, DARRELL T. Fully convolu￾tional networks for semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern [8] Recognition. Boston, MA, USA, 2015: 3431–3440. VIOLA P, JONES M. Rapid object detection using a boos￾ted cascade of simple features[C]//Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Kauai, HI, USA, 2001, 1: I-511–I- 518. [9] LOWE D G. Distinctive image features from scale-invari￾ant keypoints[J]. International journal of computer vision, 2004, 60(2): 91–110. [10] DALAL N, TRIGGS B. Histograms of oriented gradients for human detection[C]//Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. San Diego, CA, USA, 2005, 1: 886– 893. [11] OSUNA E, FREUND R, GIROSIT F. Training support vector machines: an application to face detection[C]//Pro￾ceedings of the 1997 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. San Juan, Argentina, 1997: 130–136. [12] FRIEDMAN J, HASTIE T, TIBSHIRANI R. Additive lo￾gistic regression: a statistical view of boosting (with dis￾cussion and a rejoinder by the authors)[J]. The annals of statistics, 2000, 29(5): 337–407. [13] ZITNICK C L, DOLLÁR P. Edge boxes: locating object proposals from edges[C]//Proceedings of the 13th European Conference on Computer Vision. Zurich, Switzerland, 2014: 391–405. [14] UIJLINGS J R R, VAN DE SANDE K E A, GEVERS T, et al. Selective search for object recognition[J]. Internation￾al journal of computer vision, 2013, 104(2): 154–171. [15] LI Haoxiang, LIN Zhe, SHEN Xiaohui, et al. A convolu￾tional neural network cascade for face detection[C]//Pro￾ceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Boston, MA, USA, 2015: 5325–5334. [16] FARFADE S S, SABERIAN M J, LI Lijia. Multi-view face detection using deep convolutional neural networks[C]// Proceedings of the 5th ACM on International Conference on Multimedia Retrieval. Shanghai, China, 2015: 643–650. [17] KRIZHEVSKY A, SUTSKEVER I, HINTON G E. Im￾ageNet classification with deep convolutional neural net￾works[C]//Proceedings of the 26th Annual Conference on Neural Information Processing Systems 2012. Lake Tahoe, Nevada, USA, 2012: 1097–1105. [18] HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]//Proceed￾ings of the IEEE Conference on Computer Vision and Pat￾tern Recognition. Las Vegas, NV, USA, 2016: 770–778. [19] HARIHARAN B, ARBELÁEZ P, GIRSHICK R, et al. Hy￾percolumns for object segmentation and fine-grained local- [20] 第 1 期 王成济，等：一种多层特征融合的人脸检测方法 ·145·