移动互联网金融下签名认证静态特征提取技术

Static feature extraction technology of signature verification in mobile internet banking

在签名图像预处理研究的基础上,提出了通过提取图像形状特征、不变距特征以及基于Gabor滤波纹理方向特征而得到签名图像静态特征的方法,并通过基于稀疏表示的L1范数分类方法在提取的特征样本集上进行签名鉴别。实验结果表明,在相同10组样本的特征集下,稀疏分类最小残差法的平均FRR和平均FAR分别为9.25%和4.63%,明显低于经典KNN法的12.15%和8.67%,也明显低于经典SVM法的13.31%和7.26%。该文的研究成果达到了移动互联网金融业务的性能要求。

Based on the research on the preprocessing of signature image, a new research approach that the static characteristics of signature images are obtained by extracting the shape features of the image, the invari- ant moments features of the image and the features of texture direction based on the Gabor filter in the image is proposed. Signature verification is carried out on the set including 10 feature samples extracted by the L1 norm classification method based on sparse representation. The experimental results show that the average FRR and the average FAR of Min_ error method for the sparse classification are 9. 25% and 4.63% , respectively. They are significantly lower than 12. 15% and 8.67% , respectively for classical KNN method, and 13.31% and 7.26% , respectively for classical SVM method. The research results can meet the performance requirements of mobile internet banking.