应用非负矩阵分解和RBPNN模型的掌纹识别方法

Palmprint recognition methods using non-negative matrix factorization and RBPNN model

提出一种基于非负矩阵分解(NMF)和径向基概率神经网络的掌纹识别方法。NFM是一种有效的图像局部特征提取算法,用于图像分类时能得到较高的识别率。考虑PolyU掌纹图像数据库,应用NMF、局部NMF(LNMF)、稀疏NMF(SNMF)和具有稀疏度约束的NMF(NMFSC)算法分别对掌纹图像进行特征提取,并对提取到的局部特征基图像进行分析对比;在特征提取的基础上,应用径向基概率神经网络(RBPNN)模型对掌纹特征进行分类,分类结果表明了RBPNN模型对掌纹特征具有较好的识别能力。实验对比结果证明了基于RBPNN的NMF掌纹识别方法在掌纹识别中的有效性,具有一定的理论研究意义和实用性。

A palmprint recognition method based on Non-negative Matrix Factorization （NMF） and Radial Basis Probabilistic Neural Network（RBPNN） is proposed. NMF is an efficient local feature extraction algorithm of images, and it can obtain high recognition rate in image classification task. Considered PolyU palmprint image database, the palm features are extracted by using several algorithms, such as NMF, Local NMF（LNMF）, Sparse NMF（SNMF）, and NMF with Sparseness Constraints（NMFSC） et al. And these feature ba- sis images extracted are analyzed and compared. On the basis of feature extraction, the RBPNN classifier is utilized to classify palm- print features, and the classification results show that the RBPNN model has better palmpriut recognition property. Compared classifica- tion results obtained by different algorithms, it is clear to see that the palmprint recognition results based on RBPNN and NMF are in- deed efficient, and these algorithms behave certain theory research meaning and application in practice.