基于Gabor小波和LLE的掌纹识别

为了更好地提取掌纹图像的非线性特征,文中提出一种基于Gabor小波变换和局部线性嵌入的掌纹识别算法。通过提取ROI进行光照和滤波预处理,之后进行Gabor小波变换,提取掌纹图像的多尺度特征,利用非线性的LLE算法提取主元,用最近邻方法进行分类。通过PolyU掌纹库进行验证,比较了预处理、不同ROI提取方法、LLE算法的参数对识别率的影响。实验表明,此方法相比于传统的线性降维算法以及单独的LLE算法在识别率上均有所提高。

Palmprint Recognition Based on Gabor Wavelets and LLE

An algorithm based on Gabor wavelet transform and palmprint recognition locally linear embedding (LLE) is proposed for better extraction of non-linear characteristics.Firstly,region of interest is extracted and preprocessed through filtering and suppression light.Then,the multi-scale characteristics of the palmprint image are obtained by wavelet transform of Gabor and the principal component is extracted by the non-linear LLE algorithm.Finally the nearest neighbor classifier is used to classify the palmprint.The experiment is based on the PolyU palmprint database and compares the effects on recognition rate of the preprocessing,different ROI extraction methods,and the parameters of the LLE algorithm.The experimental results show that this method achieves better recognition rate than both the traditional linear dimension reduction algorithm and the separate LLE algorithm.