Abstract: For the condition of spots on iris image, this paper proposes an iris location algorithm based on orientation gradient and Random Sample Consensus（RANSAC）. It extracts inner edge points of iris by orientation gradient according to some points in the pupil, and then locates the inner edge with RANSAC. The iris image is down-sampled. Two circles are fitted by circle differential operator on the left and right side of the pupil, and then the two circles are combined to one circle. It locates the outer edge of iris by using the parameters of the circle fast and precisely. The experimental results show that the proposed algorithm is better than traditional localization algorithms in the aspect of correct rate, speed and robustness.

Key words: iris recognition, iris location, orientation gradient, Random Sample Consensus（RANSAC）

摘要： 针对虹膜图像中有较多光斑的情况，提出一种基于有向梯度和随机抽样一致性（RANSAC）相结合的虹膜定位算法。该算法根据瞳孔内某点利用有向梯度提取内缘像素点，采用RANSAC定位虹膜内缘；下采样虹膜图像，利用圆差分算子在瞳孔左右两侧拟合出两个圆，进而合并为一个圆；根据圆的参数在虹膜图像中快速精确定位外缘。实验结果表明：该算法在正确率、定位速度和鲁棒性方面均优于传统的虹膜定位算法。

关键词: 虹膜识别, 虹膜定位, 有向梯度, 随机抽样一致性