摘要
为了能够快速准确地定位人眼,提出了一种基于近红外图像的人眼定位算法。使用主动红外光结构照明来产生亮瞳效应并产生候选区域,对于一些由于眼镜反射等因素没有亮瞳的图像,需要对其进行对称变换来产生候选区域;利用SVM算法对瞳孔区域进行分类确认;利用瞳孔的梯度以及几何特征快速准确定位到其中心位置。实验结果表明,该方法具有较高的可靠性和实时性。
To realize accurate position for eyes fast in a near infrared image which has bright pupil effect, a new method of eyes position in infrared images is put forward. The active infrared for illumination is adopted to produce the bright pupil effect and generate the candidate region. However, the existence of eyeglasses and other reasons has a negative effect on selection of candidates. Regarding this the symmetry transform is exploited. Then SVM algorithm is then used to verify the pupil candidates. Finally, accurate pupil center location is found by means of gradients. Test proves that the method of the eyes position in infrared image has good effect and high precision.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第5期1717-1721,共5页
Computer Engineering and Design
基金
江苏省自然科学基金项目(BK2012397)
高等学校博士学科点专项科研基金项目(20123219120024)
江苏省社会安全图像与视频理解重点实验室(南京理工大学)基金项目(30920130122006)
关键词
近红外图像
人眼定位
亮瞳效应
对称变换
SVM算法
梯度特征
near infrared image
eye locatiom bright pupil effect
symmetry transform
SVM algorithm
gradient feature
