考虑识别鲁棒性和虹膜颜色影响的瞳孔精准定位方法

doi:10.13232/j.cnki.jnju.2024.01.010

南京大学学报(自然科学版) ›› 2024, Vol. 60 ›› Issue (1): 97–105.doi: 10.13232/j.cnki.jnju.2024.01.010

考虑识别鲁棒性和虹膜颜色影响的瞳孔精准定位方法

罗亚波(), 李鑫

武汉理工大学机电工程学院，武汉，430070

收稿日期:2023-11-07 出版日期:2024-01-30 发布日期:2024-01-29
通讯作者: 罗亚波 E-mail:luoyabo1973@163.com
基金资助:
校企合作攻关项目(20222h0155)

Accurate pupil localization method considering recognition robustness and the influence of iris color

Yabo Luo(), Xin Li

School of Mechanical and Electronic Engineering，Wuhan University of Technology，Wuhan，430070，China

Received:2023-11-07 Online:2024-01-30 Published:2024-01-29
Contact: Yabo Luo E-mail:luoyabo1973@163.com

摘要/Abstract

摘要：

准确的瞳孔定位广泛应用于疲劳监控、注意力分析、凝视跟踪等领域.当前，对于瞳孔定位的研究，还存在两个难点问题：（1）瞳孔检测的精准率受到图像分辨率、照明、头部姿态的影响，因而自然条件下的定位精准度较低；（2）虹膜颜色影响定位精准率，但当前对于不同虹膜颜色的瞳孔定位方法的研究还不完善.针对以上问题，提出一种新的适用于包含全脸图像的瞳孔定位方法.所提方法无需训练，可直接用于瞳孔定位任务.方法核心是将表示局部径向对称性的自相似性分数，与根据瞳孔和周围区域之间的梯度信息计算得到的眼部区域梯度辐射分数相结合，取联合分数峰值坐标为瞳孔中心.在BioID数据集和GI4E数据集上评估本方法.在归一化误差 $e ≤ 0.05$ 的情况下，准确率分别为94.67% （BioID），97.09% （GI4E），在 $e ≤ 0.10$ 的情况下，准确率分别为99.47% （BioID），99.51% （GI4E）.所提方法在由低分辨率深色虹膜的人脸图像组成的自制数据集上准确率为98.66% （ $e ≤ 0.05$ ）和100% （ $e ≤ 0.10$ ），表明所提方法对于虹膜颜色有较好的鲁棒性.

关键词: 瞳孔定位, 图像处理, 自相似性, 图像梯度, 人因工程

Abstract:

Accurate pupil localization is widely used in fatigue monitoring，attention analysis，gaze tracking，and other fields. Currently，there are two difficult problems in the research of pupil localization. (1) The accuracy of pupil detection is affected by image resolution，illuminance，and head pose，and thus the accuracy of localization under natural conditions is relatively low.(2) Iris color affects the accuracy of localization，but the current research on pupil localization methods for different iris colors is not perfect. To address these two problems，this study proposes a new pupil localization method for whole?face images. The proposed method requires no training and is directly used for pupil localization tasks. The core of the method is to combine the self?similarity score representing the local radial symmetry，with the gradient radiation score of the eye area calculated based on the gradient information between the pupil and the surrounding area，then take the peak coordinate of the joint score as the pupil center. The approach was on the BioID，GI4E dataset. At a normalized error of $e ≤ 0.05$ ，the accuracy is 94.67% (BioID) and 97.09% (GI4E)，respectively. At a normalized error of $e ≤ 0.10$ ，the accuracy is 99.47% (BioID) and 99.51% (GI4E) ，respectively. The proposed approach on the self?made dataset composed of low?resolution dark iris facial images yields an accuracy of 98.66% ( $e ≤ 0.05$ ) and 100% ( $e ≤ 0.10$ )，indicating that the proposed approach has preferable robustness to iris color.

Key words: pupil localization, image processing, self?similarity, image gradient, human factors engineering

中图分类号:

TP391

罗亚波, 李鑫. 考虑识别鲁棒性和虹膜颜色影响的瞳孔精准定位方法[J]. 南京大学学报(自然科学版), 2024, 60(1): 97–105.

Yabo Luo, Xin Li. Accurate pupil localization method considering recognition robustness and the influence of iris color[J]. Journal of Nanjing University(Natural Sciences), 2024, 60(1): 97–105.

图/表 11

图1

图2

图3

图4

图5

图6

表1

表2

图7

图8

表3

本文方法在自制数据集上的准确率"

归一化误差e	$e ≤ 0.01$	$e ≤ 0.02$	$e ≤ 0.03$	$e ≤ 0.04$
准确率	65.44%	98.32%	98.32%	98.32%
归一化误差e	$e ≤ 0.05$	$e ≤ 0.06$	$e ≤ 0.07$
准确率	98.66%	99.33%	100%

表3

参考文献 25

1	Min?Allah N， Jan F， Alrashed S. Pupil detection schemes in human eye：A review. Multimedia Systems，2021，27(4)：753-777.
2	苟超，卓莹，王康,等. 眼动跟踪研究进展与展望. 自动化学报，2022，48(5)：1173-1192.
	Gou C， Zhuo Y， Wang K，et al. Research advances and prospects of eye tracking. Acta Automatica Sinica，2022，48(5)：1173-1192.
3	张宏薇，王仕洋，李宪龙,等. 基于Hough变换的瞳孔识别方法研究与实现. 液晶与显示，2016，31(6)：621-625.
	Zhang H W， Wang S Y， Li X L，et al. Research and implementation of pupil recognition based on Hough transform. Chinese Journal of Liquid Crystals and Displays，2016，31(6)：621-625.
4	Gu K K， Dong Y F， Zhou Z，et al. Real?time pupil detection based on contour tracking∥Current Trends in Computer Science and Mechanical Automation vol2. Warsaw，Poland：De Gruyter Open Poland，2022：31-40.
5	宋保全，杜文华，段能全,等. 非接触眼压计的人眼瞳孔定位算法研究. 电子测量技术，2022，45(3)：112-117.
	Song B Q， Du W H， Duan N Q，et al. Research on pupil location algorithm of non?contact tonometer. Electronic Measurement Technology，2022，45(3)：112-117.
6	王鹏，温宏韬，王世龙. 基于近眼红外图像的高精度瞳孔中心定位方法. 哈尔滨理工大学学报，2022，27(5)：38-46.
	Wang P， Wen H T， Wang S L. High?precision pupil center positioning method based on near?eye infrared image. Journal of Harbin University of Science and Technology，2022，27(5)：38-46.
7	Zhou W J， Lu X Y， Wang Y. A robust pupil localization via a novel parameter optimization strategy. Wireless Communications and Mobile Computing，2022，2022：2378911.
8	El Nahl W， Zaini H G， Zaini R H，et al. Robust and high accuracy algorithm for detection of pupil images. Computers，Materials & Continua，2022，73(1)：33-50.
9	Khan W， Hussain A， Kuru K，et al. Pupil localisation and eye centre estimation using machine learning and computer vision. Sensors，2020，20(13)：3785.
10	崔家礼，曹衡，张亚明,等. 一种复杂场景下的人眼检测算法. 北京航空航天大学学报，2021，47(1)：38-44.
	Cui J L， Cao H， Zhang Y M，et al. A human eye detection algorithm in complex scenarios. Journal of Beijing University of Aeronautics and Astronautics，2021，47(1)：38-44.
11	Gdoura A， Degünther M， Lorenz B，et al. Combining CNNs and Markov?like models for facial landmark detection with spatial consistency estimates. Journal of Imaging，2023，9(5)：104.
12	King D E. Dlib?ml：A machine learning toolkit. The Journal of Machine Learning Research，2009，10：1755-1758.
13	Maver J. Self?similarity and points of interest. IEEE Transactions on Pattern Analysis and Machine Intelligence，2010，32(7)：1211-1226.
14	BioID. The BioID face database. Nuremberg，Germany：BioID，2022.
15	Villanueva A， Ponz V， Sesma?Sanchez L，et al. Hybrid method based on topography for robust detection of iris center and eye corners. ACM Transactions on Multimedia Computing，Communications，and Applications，2013，9(4)：25.
16	Jesorsky O， Kirchberg K J， Frischholz R W. Robust face detection using the hausdorff distance∥Bigun J，Smeraldi F. Third international conference on audio? and video?based biometric person authentication. Springer Halmstad Sweden，2001：90-95.
17	Monforte P H B， Araujo G M， De Lima A A. Evaluation of a new kernel?based classifier in eye pupil detection∥The 17th IEEE International Conference on Machine Learning and Applications. Orlando，FL，USA：IEEE Press，2018：380-385.
18	Xia Y F， Yu H， Wang F Y. Accurate and robust eye center localization via fully convolutional networks. IEEE/CAA Journal of Automatica Sinica，2019，6(5)：1127-1138.
19	Dai L H， Liu J G， Ju Z J，et al. Iris center localization using energy map synthesis based on gradient and isophote. Journal of Intelligent & Fuzzy Systems：Applications in Engineering and Technology，2020，38(4)：4511-4523.
20	Ahmed N Y. Real?time accurate eye center localization for low?resolution grayscale images. Journal of Real?Time Image Processing，2021，18(1)：193-220.
21	Ahmad N， Yadav K S， Ahmed M，et al. An integrated approach for eye centre localization using deep networks and rectangular?intensity?gradient technique. Journal of King Saud University?Computer and Information Sciences，2022，34(9)：7153-7167.
22	Ahmed M， Laskar R H. Evaluation of accurate iris center and eye corner localization method in a facial image for gaze estimation. Multimedia Systems，2021，27(3)：429-448.
23	Li B. Eye centre localisation with hybrid regression framework. Electronics Letters，2018，54(13)：819-821.
24	Zhu J， Li M T， Yu Y Z，et al. A joint cascaded framework for simultaneous eye state，eye center，and gaze estimation∥2022 26th International Conference on Pattern Recognition. Montreal，Canada：IEEE Press，2022：770-776.
25	Gou C， Zhou Y C， Xiao Y，et al. Cascade learning for driver facial monitoring. IEEE Transactions on Intelligent Vehicles，2023，8(1)：404-412.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

算法	文献[17]	文献[18]	文献[19]	文献[20]	文献[21]	本文方法
e≤0.05	95.4%	94.4%	90.1%	91.7%	94.3%	94.67%
e≤0.10	99.6%	99.9%	98.8%	97.9%	98.4%	99.47%

算法	文献[22]	文献[23]	文献[20]	文献[24]	文献[25]	本文方法
e≤0.05	90.9%	95.4%	96.7%	98.5%	97.3%	97.09%
e≤0.10	97.3%	99.6%	98.7%	99.7%	99.8%	99.51%

[1]	孟浩, 刘强. 基于FPGA的卷积神经网络训练加速器设计[J]. 南京大学学报(自然科学版), 2021, 57(6): 1075-1082.
[2]	李康,谢宁,李旭,谭凯. 基于卷积神经网络和几何优化的统计染色体核型分析方法[J]. 南京大学学报(自然科学版), 2020, 56(1): 116-124.
[3]	林　銮，陆武萍，唐朝生，赵红崴，冷　挺，李胜杰. 基于计算机图像处理技术的松散砂性土微观结构定量分析方法[J]. 南京大学学报(自然科学版), 2018, 54(6): 1064-1074.
[4]	王云鹏，田洪舟，刘承智，周　政*. 强化气液喷射搅拌反应器传质特性研究[J]. 南京大学学报(自然科学版), 2017, 53(3): 598-.
[5]	孙馨喆1,2，文立1,2，杨武夷1,2*，张宇1,2. 基于时频图像处理的宽吻海豚声通讯信号自动检测方法[J]. 南京大学学报(自然科学版), 2015, 51(7): 16-.

考虑识别鲁棒性和虹膜颜色影响的瞳孔精准定位方法

Accurate pupil localization method considering recognition robustness and the influence of iris color

RichHTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 25

相关文章 5

Metrics

本文评价

推荐阅读 0