一种基于极限学习机的多粒度虹膜识别算法

A multi-granularity iris recognition algorithm based on extreme learning machine

针对Gabor滤波器能精确捕获低频和中频纹理信息,而灰度共生矩阵法(Gray level co-occurrence matrix,GLCM)对高频纹理信息具有较好的获取能力,本文采用将GLCM与2D Gabor滤波器组相结合的方法对预处理后的虹膜图像进行特征提取,得到多粒度虹膜向量,有效避免了单一特征提取方法的缺陷。最后采用极限学习机(Extreme learning machine,ELM)对虹膜进行分类识别。实验结果表明,本文提出的基于ELM的多粒度虹膜识别算法在保证实时性的情况下能使识别精度高达99.86%,优于主流的虹膜识别算法。

The Gabor filter can capture the low and intermediate frequency texture information accurate- ly, and the gray level co-occurrence matrix （GLCM） has a better ahility to obtain the high frequency tex- ture information. In order to avoid the defect of the single feature extraction method, a multi-granularity extraction method based on 2D-Gabor filters and GLCM is proposed to generate a multi-feature vector. The recognition process uses the extreme learning machine （ELM）. Experimental results show that the proposed multi-granularity iris recognition algorithm based on ELM has a higher accuracy of 99.86% under the premise of real-time performance. It outperforms other mainstream iris recognition algorithms.