摘要
提出一种基于独立分量分析的钞票识别算法。先对钞票图像作感兴趣区域(ROI)切割;接着对ROI图像作标准化、白化预处理;然后采用基于负熵独立性判据的固定点方法(FastICA)对预处理后的ROI图像做ICA分离,提取独立基图像,进而获得钞票的特征空间,并构建特征模板;通过计算待识别目标与特征模板的距离实现识别。以第五套人民币作为实验对象进行实验,实验结果表明方法的有效性。
An algorithm of bank notes recognition based on independent component analysis (ICA) is proposed. First the segmentation of rectangle of interesting (ROI) on bank note image was made and followed with pre-processing of data standardization and whitening on ROI images. Then the negentropy independence criterion-based FastICA was adopted to perform ICA division on pre-processed ROI image to extract independent radical images, and further the feature space of bank note was acquired and the feature template was constructed. Bank notes rec- ognition is implemented by calculating the distance between the target to be recognised and the feature template. The approach was tested on a set of 600 Chinese RMB ( version 5 ) bank note images. The experimental results clearly shown the effectiveness of the algorithm proposed in the paper.
出处
《计算机应用与软件》
CSCD
2010年第8期244-247,共4页
Computer Applications and Software
基金
广州市2007年重大科技工程(2007Z1-I0041)
关键词
ICA
模式识别
最近邻分类
白化
独立性判据
ICA Pattern recognition Nearest classify Whitening Independent criterion