基于HSI空间和纹理的频谱度量的印鉴识别  

Seal Identification Based on HSI Space and Frequency Measurement of Texture

作　　者：孙红岩[1] 田树学[2] 张学东[3] 

机构地区：[1]辽宁科技大学软件学院,辽宁鞍山114051 [2]辽宁科技大学学生工作处,辽宁鞍山114051 [3]辽宁科技大学电子与信息工程学院,辽宁鞍山114051

出　　处：《沈阳师范大学学报（自然科学版）》2010年第2期221-224,共4页Journal of Shenyang Normal University:Natural Science Edition

基　　金：国家自然科学基金资助项目(607750366)

摘　　要：印鉴在东方国家和地区的应用非常广泛,尤其在我国的应用更为广泛,从而导致伪造印鉴的犯罪活动屡禁不止。为了实现大量印鉴的自动识别,已经提出了许多方法,这些方法主要根据边缘和模板匹配的方法实现的,其匹配过程需要调整印鉴的旋转角度。以印章图像为研究对象,采用HSI色彩空间来描述,它比RGB色彩空间更符合视觉特性,并且可以大大简化图像分析和处理的工作量。运用HSI色彩空间对印章图像颜色特征提取等一系列预处理,研究了多种识别方法后发现基于纹理特征的印鉴识别方法不但速度快,而且识别率也较高,并且方法简便。通过将极坐标和傅里叶变换结合,计算出印鉴图像纹理的频谱度量,来完成印鉴的特征提取,运用支持向量机分类器对印鉴进行识别,实验证明,方法具有良好的旋转不变纹理分析性能,提高了识别率。Seal is widely used in some oriental countries and areas,especially in China. Economic crimes caused by forgery seals are hard to be forbidden.Many methods have been put forward to realize the seal identification,most of these methods are based on stroke edge matching and template matching and need to adjust the rotation angle.It was found a fast,simple and a higher seal image（HSI） identification rate way to identify the seal characteristics based on texture identification after a lot of methods was studied by using a series of pre-processing including the use of HIS color space to extract the seal color image feature.In this experiment.Fourier transform is combined with polar transform to compute the spectral of the seal imprint used to achieve the characteristics of it,and support vector machine（SVM） classifier to implement final identification decision of seal imprint.Our experiments show that the method has good analyse capability of circumgyrating fixedness texture and could improve the identification rate.

关 键 词：HSI色彩空间 预处理 纹理 支持向量机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]