张玉欣 孙浩

摘要：提出一种基于D-S证据理论和支持向量机相结合的多特征融合纸币真伪识别方法，解决了单一特征纸币真伪识别存在的低正确率和低稳定性问题。在对纸币进行图像处理的基础上，提取出纸币颜色、纹理、互相关系数等三类特征，以单特征的支持向量机初步识别结果作为独立证据计算基本信任函数，再利用D-S证据理论进行决策级融合，最后根据分类决策规则的门限值给出最终的纸币真伪识别结果。实验结果表明，50个测试样本的多特征融合识别正确率达100%，与单特征的纸币真伪识别相比，本文所述方法识别正确率高、稳定性更好。

关键词：纸币识别；D-S证据理论；支持向量机；特征提取；多特征融合

中图分类号：TP183 文献标识码：A 文章编号：1009-3044（2018）21-0226-04

Abstract： A multi-feature fusion method based on D-S evidence theory and SVM for identifying the authenticity of banknotes was proposed to solve the low accuracy and low stability of the single feature-based method. Firstly， on the basis of a series of image processing for paper money， three types of features such as color， texture and cross-correlation coefficient were extracted. Secondly， the paper money were identified according to each type of features utilizing SVM .then， using the results as independent evidence to compute the belief function， using D-S evidence theory to achieve multi-feature fusion of the decision level .Finally， the final identification results was given by the thresholds of classification decision rules. The experimental results show that the accuracy of multi-feature fusion method under 50 paper money test samples was 100% ， and the method in the paper has high accuracy and stability than single feature method.

Key words： identify the authenticity of banknotes； D-S evidence theory ； artificial neural network； feature extraction ； multi-feature fusion

1 引言

假幣的出现极大损害了人民群众的利益并且严重干扰了社会的金融秩序。常用的伪钞鉴别技术是通过磁性检验和光学检验等手段，对纸币的磁性安全线、磁性油墨、红外、紫外等防伪特征进行检测。但是一些专门针对纸币防伪特征的高仿币，却可以通过这些真伪检测。随着智能化无人收费系统的广泛应用，如何提高纸币真假的智能识别技术至关重要。针对单特征纸币真假识别的低准确率和低稳定性，提出一种D-S证据理论和支持向量机相结合的多特征纸币真假识别方法。首先通过图像处理提取出纸币颜色、纹理、互相关系数三类特征， 并分别通过独立的支持向量机进行真假分类识别。……