【摘要】 为了在频域内对相位谱建立模型,运用相关滤波器方法进行掌纹识别。首先运用图像下抽样方法降低掌纹空间的维数,在低维图像上应用训练图像设计最小平均相关能量(MACE)滤波器,然后计算训练图像和测试图像的相关输出平面,根据峰与相关能量(PCE)比值指标进行掌纹匹配。运用PolyU掌纹图像库,对本文算法进行测试。实验结果表明,与传统子空间方法相比,本文方法的识别性能最优,识别率为100%,特征提取和匹配总时间小于0.5 s,具有快速、有效和易于实现等优点。更多还原

【Abstract】 In order to directly model phase information in frequency domain,the palmprint recognition method based on correlation filter is proposed.Image sub-sampling is first applied to reduce the palmprint space dimensionality.The minimum average correlation energy(MACE) filters based on the training set are designed.Then the training set and test set are correlated with the filters to produce correlation output.The peak-to-correlation energy(PCE) ratio as a match metric is calculated to match the palmprints.The algorithm is tested in PolyU palmprint database.The experiment results show that the recognition performance of the algorithm exceeds those of the traditional subspace methods.The recognition rate is 100%.The feature extraction and matching time is less than 0.5 s,so the proposed method is fast,efficient,and easy to realize.更多还原