基于改进粒子群算法及分区去噪的虹膜定位研究
2021-07-11刘辉朱文玉娄成芝张彦文唐闯陈希
刘辉 朱文玉 娄成芝 张彦文 唐闯 陈希



摘 要:针对亚洲人眼虹膜图像,结合人眼瞳孔的构成及特征,借助改进的粒子群寻优算法、图像反补增强相整合的途径,拟合定位虹膜内圆;参考虹膜外边界的特征,通过改进的粒子群寻优算法、分区域去噪法,有效定位虹膜外圆,并利用坐标变换法对定位后的图像作了归一化处理;最后,通过中科院虹膜图库及自建虹膜图库样本验证了方法的可行性,研究中较好的克服了图库中光斑的干扰,且能高效完成虹膜区域的切分。
关键词:虹膜定位;粒子群算法;分区域去噪
中图分类号:TP319.7 文献标识码:A 文章编号:1001-5922(2021)04-0091-05
Abstract:Aiming at the Asian iris image, combining the composition and characteristics of the human eye pupil, with the help of an improved particle swarm optimization algorithm and an integrated approach of image denial enhancement, fitting and positioning of the inner circle of the iris; with reference to the characteristics of the outer boundary of the iris, the improved particle swarm optimization algorithm and the sub-region denoising method are used to effectively locate the outer circle of the iris, and using the coordinate transformation method to normalize the image after positioning; finally, the feasibility of the method is verified by the Chinese Academy of Sciences iris library and self-built iris library samples. In the research, the interference of the light spot in the gallery is better overcome, and the segmentation of the iris area can be completed efficiently.
Key words:iris location; particle swarm optimization; denoising by region
0 引言
人眼圖像虹膜定位作为虹膜识别、虹膜医学辅助诊断等技术领域的基础,但由于图像采集过程中,人眼到镜头距离的变化、光照对瞳孔的刺激、个体差异等诸多因素的影响,使得后期出现识别率底,特征提取错误等现象。著名学者Daugman最先对虹膜分割法进行了定义,结合虹膜同圆环相近的形状,通过圆形检测匹配器法对图像进行分割[1]。Wildes学者首次通过霍夫变换(Hough Transformation) 来定位虹膜,对累加器数组的最大值参数组合进行计算,并作为虹膜内外圆参数[2]。Javed、Basit两位学者在亮度值分析的基础上,构建了全新的虹膜定位法,借助两大不同方式对虹膜内边缘数据进行运算,并借助对各方向梯度Max值进行寻找,从而得到虹膜外边缘[3]。粒子群算法是Kennedy和Eberhart博士提出的一种基于群智能的优化算法,最大特征即为该算法最初的收敛性较强,然而可能后期会出现局部最优的状况,收敛速度及精度会下降[4]。……
