摘要
提出一种基于PCA的钞票识别算法。通过对经预处理后的钞票图像作感兴趣区域(ROI)切割,获得ROI图像;然后对ROI图像做K-L变换提取钞票图像的主成分,构造钞票的特征空间;利用训练样本在特征空间中的投影向量构建特征模板。通过计算待识别目标与特征模板的最小距离来完成识别。以第五套人民币作为实验对象进行实验,实验结果表明本文提出方法的有效性。
Algorithm of recognition on bank note was proposed.Using Image of bank note after preprocessing,ROI incision was performed.Taking ROI image as input object,principle component was extracted by K-L transformation.The feature space of bank note were then constructed.The feature template were constructed with projection of training samples on feature space.The recognition was perform according to the minimal distance between target input and feature templates.The approach was tested using a set of 400 Chinese RMB(version 5) bank note images.As high as possible accuracy was achieved for some of the bank notes,which clearly show the effectiveness of the algorithm.
出处
《微计算机信息》
2010年第7期202-204,共3页
Control & Automation
基金
广州市2007年重大科技工程(2007Z1-I0041)
关键词
PCA
K-L变换
特征空间
钞票识别
最小距离
PCA
K-L transformation
feature space
recognition of bank note
minimal distance