利用多光谱图像的伪造虹膜检测算法

An Algorithm for Fake Irises Detection Using Multi-Spectral Images

虹膜识别系统易受各种类型伪造虹膜的攻击,需要预先检测虹膜的活性.本文提出分别采集860 nm和480 nm波长的虹膜图像,根据活体人眼的特殊光谱特性,从图像中提取结膜血管变化数(RNCV)和纹理熵比(ERIT)特征.使用训练好的支持向量机(SVM)对特征向量进行分类,输出活性检测结果.在自建的伪造虹膜数据库上的测试结果表明,本文算法可以有效排除打印图像,人造眼,彩色隐形眼镜等各类伪造样本,能满足实时应用要求.

Iris liveness detection is a necessary module for iris recognition because iris recognition systems are easily attacked by fake irises.This paper proposes to prevent fake irises using iris images at 860 nm and 480 nm wavelengths.According to spectral characteristics of human eyes,relative number of conjunctival vessels（RNCV） and entropy ratio of iris textures（ERIT） are extracted.The final detection results are output after the trained support vector machine（SVM） is used to classify these feature vectors.Experimental results on the constructed fake iris database show that our algorithm can effectively exclude printed images,artificial eyes and colored contact lenses and the execution time can meet the requirements of the real-time applications.