基于最小最大概率机的虹膜图像分类方法研究

为了提高虹膜图像分类的准确性和稳定性，提出了一种基于最小最大概率机的虹膜图像分类方法．该方法通过控制错分概率实现分类的最大化，将一般的二维分类问题扩展到虹膜特征的多维空间，并利用最小最大概率机的高维映射泛化特性，实现了不同核函数下的虹膜图像多维分类问题，具有分类准确率高、稳定性好的特点．通过虹膜图像库的实验验证表明，该方法在保持分类稳定性的同时，获得了径向基核函数高达98．55％的分类率，该分类率比最近特征线方法和相异度函数方法的分类率分别提高了4．47％和6．41％．

Study on Iris Image Classification Approach Based on Minimax Probability Machine

In order to improve the accuracy and stability of iris image classification, an iris image classification method was developed based on minimax probability machine. The classification maximization was realized through controlling minimal error of the classification probability and expanding two-dimension classification to multi-dimension iris feature space. The iris multi dimension classification problem with different kernel function is solved through mapping iris fea tures to high dimension space, which possesses the features of high classification rate and strong stability. Experiment validation has been carried out with iris image database, and the results show that the classification rate reaches to 98. 55 % for radial kernel function while keeping the stability of classification, and the classification rate is increased by 4. 47% and 6.41% respectively compared with the nearest feature line method and dissimilarity function method.