基于退化四元小波变换的纸币识别.doc

基于退化四元小波变换的纸币识别 盖杉1,2 罗立民1 （1.东南大学计算机科学与工程学院，南京，210096；2. 江苏省社会安全图像与视频理解重点实验室，南京，210094TP391.43　　　 文献标识码　A Banknote Recognition Based on Rotated Quaternion Wavelet Transform Gai Shan1,2 , Luo Limin1 ( 1. School of Computer Science and Engineering of Southeast University,Nanjing,210096,P.R.China； 2. Jiangsu Key Laboratory of Image and Video Understanding for Social Safety, Nanjing University of Science and Technology, Nanjing, 210094, P.R. China) Abstract: A new banknote classification method is proposed by using the phase concept of the reduced quaternion wavelet transform to improve the banknote recognition rate and feature extraction. The preprocessing of banknote which includes edge detection and slant correction is conducted, and decomposing the image by reduced quaternion wavelet. The statistical characteristics of the decomposition coefficients are used as feature of the banknote image for classification. Finally, the support vector machine is applied as classifier in the banknote classification system. The experimental results show that the proposed method can obtain better results compared with other conventional methods and satisfy the real-time requirements. Keywords : reduced quaternion wavelet transform; feature extraction; support vector machine; banknote recognition 引 言 纸币流通期短暂，需要不断地被清点、分类。金融系统每天需要对海量的纸币进行清分，分辨种类、鉴别真伪，并按新旧、污损程度进行分类。人工完成这一过程效率低，易出错，并且缺乏客观标准。自动化的纸币清分系统能够在纸币一次通过机器的情况下，根据纸币的图案与尺寸，对钞票的柔软性、肮脏、胶带、折角、多张重叠、破洞、裂缝与缺损等多种状况进行检测，最终实现对纸币的点算、鉴伪并按照面额、版本、正反面进行清分。其中纸币图像特征提取技术是纸币清分系统中最为重要的步骤。 目前提取纸币图像特征的常用方法主要有几何特征[1]、网格特征[2-3]、自由掩模[4]和类Haar小波特征。Hasan等人[5]针对污损纸币提出一种基于分量的特征提取算法，该算法在一定程度上克服了污损对纸币图像特征的影响。Ye等人[6]利用统计模型模拟纸币图像在流通中的退化过程，同时构造了特征提取模型。Gai等人[7]利用四元小波对纸币图像的统计特性进行分析，并建立了纸币图像的统计模型。 四元小波变换(Quaternion Wavelet Transform，简称QWT)[8]QWT已经广泛应用于图像理解与分析等领域[9-12]。然而，QWT构造准则中的乘法运算不具有交换性，同时也很难区分图像对角线方向的纹理特征。针对上述问题，本文提出一种新的基于退化四元小波变换(Reduced Quaternion Wavelet Transform，简称RQWT)RQWT捕捉纸币图像具有的纹理特征，然后提取出每个分解子带的能量和标准差作为特征向量，采用最小二乘支持向量机[13-14](LS-SVM