摘要
提出了一种基于改进的广义K-L变换的掌纹识别方法.在主成分分析思想的基础上,改进后的广义K-L变换把样本的类间信息也加入到算法的分类过程中,相对于广义K-L变换,识别效果得到了提高.将原始图像投影到一个由图像的类间散布矩阵的特征向量所构成较低维的特征子空间中,在这个特征子空间中,用欧式距离来对分属于不同手掌的掌纹图像进行分类.基于不同的特征长度和训练样本数,做了一系列的实验,最佳识别率达99 17%.
This paper put forward a novel palm print recognition method based on improved generalized K-L transform. With the principal components analysis as the basis concept, the improved generalized K-L transform also took the class specific information into consideration and improvements are achieved through amendments on generalized K-L transform. By means of improved generalized K-L transform, the original palm print images were projected into a lower dimensional feature sub-space spanned by the eigenvectors of the palm print between-class covariance matrix. In this feature sub-space, the palm prints from different palms were discriminated with the euclidean distance classifier. Various experiment schemes on feature length and training sample number are developed. A high recognition rate of 99.71% is achieved through the proposed method.
出处
《哈尔滨商业大学学报(自然科学版)》
CAS
2004年第6期659-661,665,共4页
Journal of Harbin University of Commerce:Natural Sciences Edition
基金
国家自然科学基金(90209020
60340460005)