融合局部双向编码与深度学习的非接触式掌纹识别

Contactless Palmprint Recognition Based on Local Bidirectional Coding and Deep Learning

文章提出一种基于局部双向编码(Local Double-Orientation Code,LDOC)与深度学习的非接触式掌纹识别方法。该方法分别取Gabor滤波器的实数和虚数部分,并选用最大和次大响应进行编码作为深度神经网络的输入。在香港理工大学非接触式掌纹数据集和中国科学院自动化研究所非接触式掌纹数据集上进行实验。结果表明,所提出的基于局部双向编码与深度学习的非接触式掌纹识别方法取得了不错的结果,具有普适性。

The article proposes a contactless palmprint recognition method based on Local Double-Orientation Code(LDOC)and deep learning. The method takes the real and imaginary parts of the Gabor filter, respectively, and chooses the maximum and second largest responses for coding, which are used as the input of the deep neural network. Experiments are conducted on the contactless palmprint dataset of the Hong Kong Polytechnic University and the contactless palmprint dataset of the Institute of Automation, Chinese Academy of Sciences. The results show that the proposed contactless palmprint recognition method based on local bidirectional coding with deep learning achieves good results and is generalizable.