Please wait a minute...
浙江大学学报(工学版)
浙江大学学报(工学版)  2019, Vol. 53 Issue (7): 1374-1379    DOI: 10.3785/j.issn.1008-973X.2019.07.017
自动化技术、计算机技术     
基于级联形状回归的多视角人脸特征点定位
桑高丽(),王国滨,朱蓉,宋佳佳
嘉兴学院 数理与信息工程学院，浙江 嘉兴 314001
Multi-view facial landmark location method based on cascade shape regression
Gao-li SANG(),Guo-bin WANG,Rong ZHU,Jia-jia SONG
Department of Mathematics and Information Engineering, Jiaxing University, Jiaxing 314001, China
 全文: PDF(728 KB)   HTML
摘要：

针对当前人脸特征点定位精度低、且当人脸图像存在较大姿态变化时不能在同一模型框架下实现任意姿态人脸图像的面部特征点精确定位问题，提出基于级联形状回归的对姿态、遮挡都鲁棒的人脸特征点定位方法. 为了提高定位的准确度，提出按姿态偏转造成的遮挡程度对人脸区域进行分块，针对每一块分别训练形状估计回归器；为了能够在同一框架下实现任意姿态人脸的特征点定位，在特征点形状定义中引入特征点的可见/不可见属性；为了提高该算法的性能，在特征的计算方法和计算策略上分别进行改进. 在Multi-PIE、AFLW、COFW和300-W数据库上的实验结果表明，提出算法不但对姿态、遮挡具有很强的鲁棒性，而且在其他不可控因素影响下取得很好的效果.
关键词： 特征点定位;  级联形状回归;  区域划分;  可见/不可见属性;  多视角    
Abstract:

A new cascade shape regression based facial landmark location method was proposed in order to improve the low landmark location accuracy and perform landmark location with large pose variations in a unified frame. The face area was divided according to the degree of occlusion in order to improve the accuracy of landmark location, and the shape regression was separately trained for each block. The visible/invisible attribute was introduced to the landmark's definition in order to locate the landmarks of any pose under the unified frame. The calculation and the strategy of feature extraction were improved in order to guarantee the performance of the proposed algorithm. The experimental results on the Multi-PIE, AFLW, COFW and 300-W databases show that the proposed method not only has strong robustness to pose and occlusion, but also achieves good results under other uncontrollable factors.
Key words: landmark location    cascaded shape regression    regional division    visible/invisible attribute    multi-view
收稿日期: 2018-05-21 出版日期: 2019-06-25
CLC:  TP 391  
作者简介: 桑高丽（1986?），女，讲师，从事模式识别相关的研究. orcid.org/0000-0002-6567-1652. E-mail： g.sang@foxmail.com
服务　　
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章　　
桑高丽
王国滨
朱蓉
宋佳佳

引用本文:

桑高丽,王国滨,朱蓉,宋佳佳. 基于级联形状回归的多视角人脸特征点定位[J]. 浙江大学学报(工学版), 2019, 53(7): 1374-1379.

Gao-li SANG,Guo-bin WANG,Rong ZHU,Jia-jia SONG. Multi-view facial landmark location method based on cascade shape regression. Journal of ZheJiang University (Engineering Science), 2019, 53(7): 1374-1379.

链接本文:

http://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2019.07.017    或    http://www.zjujournals.com/eng/CN/Y2019/V53/I7/1374

图 1  级联形状回归过程示意图
图 2  基于改进级联形状回归的多视角人脸图像特征点定位算法框架
图 3  人脸区域划分及特征点位置、顺序图
方法 e
CPR[8] 6.89
GCPR 4.14
表 1  提出方法（GCPR）与对比方法在Multi-PIE数据库上可见特征点的平均定位误差
图 4  本文方法与基准方法在Multi-PIE数据库测试集上不同姿态下人脸特征点定位误差
图 5  本文算法在测试数据库上的部分实验结果
方法 e
可见点 可见+不可见
CPR[8] ? 8.55
PIFA[11] ? 6.52
GCPR 4.45 5.60
表 2  提出方法（GCPR）与对比方法在AFLW数据库上可见特征点与不可见特征点的平均定位误差
图 6  提出方法与对比方法在AFLW数据库测试集上的各特征点定位误差
方法 e
可见点 可见+不可见
CPR[8] － 9.25
RCPR[9] － 8.50
CRC[15] － 7.30
SDM[16] 6.69 7.70
TCDCN[17] － 8.05
GCPR 5.10 5.40
表 3  提出方法（GCPR）与对比方法在COFW数据库上可见特征点与不可见特征点的平均定位误差
方法 普通集 挑战集 合集
RCPR[9] 6.18 17.26 8.35
SDM[16] 5.57 15.4 7.50
TCDCN[17] 4.80 8.60 5.54
ECT[18] 4.68 8.69 5.47
GCPR 4.50 7.40 5.22
表 4  提出方法（GCPR）与对比方法在300-W数据库上特征点的平均定位误差
1 WANG N, GAO X, TAO D, et al Facial feature point detection: a comprehensive survey[J]. Neurocomputing, 2018, 275: 50- 65
doi: 10.1016/j.neucom.2017.05.013
2 DU C, WU Q, YANG J, et al. SVM based ASM for facial landmarks location [C] // 8th IEEE International Conference on Computer and Information Technology. Chongqing: IEEE, 2008: 321-326.
3 石正权, 赵启军, 陈虎 基于cpr和clm的多视角人脸特征点定位方法[J]. 计算机技术与发展, 2015, 25 (11): 1- 5
SHI Zheng-quan, ZHAO Qi-jun, CHEN Hu A multi-view facial landmark localization method based on CPR and CLM[J]. Computer Technology and Development, 2015, 25 (11): 1- 5
4 CAUNCE A, TAYLOR C J, COOTES T F. Improved 3D model search for facial feature location and pose estimation in 2D images [C] // Proceedings of BMVC. Aberystwyth: BMVA, 2010: 1-10.
5 YU X, HUANG J, ZHANG S, et al. Pose-free facial landmark fitting via optimized part mixtures and cascaded deformable shape model [C] // 2013 IEEE International Conference on Computer Vision (ICCV). Sydney: IEEE, 2013: 1944-1951.
6 JOURABLOO A, LIU X. Large-pose face alignsment via CNN-based dense 3D model fitting [C] // Proceedings of CVPR. Las Vegas: IEEE, 2016: 4188-4196.
7 OZUYSAL M, CALONDER M, LEPETIT V, et al Fast keypoint recognition using random ferns[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 32 (3): 448- 461
doi: 10.1109/TPAMI.2009.23
8 DOLLAR P, WELINDER P, PERONA P. Cascaded pose regression [C] // Proceedings of CVPR. San Francisco: IEEE, 2010: 1078-1085.
9 CAO X, WEI Y, WEN F, et al Face alignment by explicit shape regression[J]. International Journal of Computer Vision, 2014, 107 (2): 177- 190
doi: 10.1007/s11263-013-0667-3
10 CHAI X, WANG Q, ZHAO Y, et al Robust facial landmark detection based on initializing multiple poses[J]. International Journal of Advanced Robotic Systems, 2016, 13 (5): 1729881416662793
11 JOURABLOO A, LIU X. Pose-invariant 3D face alignment [C] // Proceedings of ICCV. Santiago: IEEE, 2015: 3694-3702.
12 WU J, TRIVEDI M M A two-stage head pose estimation framework and evaluation[J]. Pattern Recognition, 2008, 41 (3): 1138- 1158
doi: 10.1016/j.patcog.2007.07.017
13 KOESTINGER M, WOHLHART P, ROTH P M, et al. Annotated facial landmarks in the wild: a large-scale, real-world database for facial landmark localization [C] // Proceedings of ICCV. Barcelona: IEEE, 2011: 2144-2151.
14 SAGONAS C, TZIMIROPOULOS G, ZAFEIRIOU S, et al. 300 faces in-the-wild challenge: the first facial landmark localization challenge [C] // IEEE International Conference on Computer Vision Workshops. Washington: IEEE, 2013: 397–403.
15 FENG Z H, HUBER P, KITTLER J, et al Random cascaded-regression copse for robust facial landmark detection[J]. IEEE Signal Processing Letters, 2015, 22 (1): 76- 80
16 XIONG X, DE LA TORRE F. Supervised descent method and its applications to face alignment [C] // Proceedings of CVPR. Portland: IEEE, 2013: 532-539.
17 ZHANG Z, LUO P, LOY C C, et al Learning deep representation for face alignment with auxiliary attributes[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38 (5): 918- 930
doi: 10.1109/TPAMI.2015.2469286
[1] 高雪艳,潘安宁,杨扬. 基于图像混合特征的城市绿地遥感图像配准[J]. 浙江大学学报(工学版), 2019, 53(6): 1205-1217.
[2] 成翔昊,达飞鹏,汪亮. 基于融合约束局部模型的三维人脸特征点定位[J]. 浙江大学学报(工学版), 2019, 53(4): 770-776.
[3] 刘冬旭, 董红召. 共享自行车系统调度区域的分形树自平衡划分算法[J]. 浙江大学学报(工学版), 2018, 52(7): 1275-1283.
[4] 孙文财, 杨志发, 李世武, 徐艺, 郭梦竹, 魏学新. 面向驾驶员注视区域划分的DBSCAN-MMC方法[J]. 浙江大学学报(工学版), 2015, 49(8): 1455-1461.
[5] 何智翔, 丁晓青, 方驰, 文迪. 基于LBP和CCS-AdaBoost的多视角人脸检测[J]. J4, 2013, 47(4): 622-629.