基于二叉树和SVM的指纹分类

Research of fingerprint classification combined by binary tree and SVM

为解决支持向量机(Support Vector Machine,SVM)进行指纹多类分类存在困难的问题,在应用二叉树理论的基础上,提出了一种新型的指纹分类方法.该算法首先使用二叉树进行多类决策,将原始分类数据分解成3个二类分类问题,然后利用SVM进行二类分类,使3个分类超平面得到优化.两者的有机结合,充分发挥了SVM在二类分类问题方面相对于其它方法的优势,从而使算法的推广能力有较大提高,总的分类正确率可达97.9%.实验结果证明,二叉树构造多类框架将指纹多类分类问题分解成3个二类分类器系统,不仅可以有效的提高指纹分类的效率,还充分发挥了SVM分类器解决二类分类问题的优势.

In order to solve the difficulties existed in fingerprint multi-classification for Support Vector Machine, this paper proposes a novel fingerprint classification method based on binary tree theory. This algorithm uses binary tree to construct the multi-class frame by decomposing the problem into three 2-class classification problems, then uses Support Vector Machine optimizing the three hyperplanes. The combination of the two exerted the superiority for 2-class classification of SVM over other algorithms completely, the generalization ability has improved greatly and the total accuracy for the new sample is 97.9%. Experimental results show that the fingerprint multi-class problem is divided into three 2-class classifier system by using binary tree to construct the multi-class frame, which not only can improve the efficiency of fingerprint classification but also exerts the superiorities of SVM classifier for two-class classification problem sufficiently.