一种提高手机人民币图像真伪识别率的CNN框架

A CNN Framework for Improving the Recognition Rate of Authenticity on Mobile Phone RMB Images

论文为了解决手机人民币图像给人民币鉴伪任务带来的困难,构造了一个基于B-CNN的手机人民币图像鉴伪框架,该框架以带有提取手机人民币图像红色分量Lambda层的VGG16的block5的输出作为输入。将提取手机人民币红色分量的Lambda层加在VGG16网络的最前面,并用此时VGG16的block5的输出模拟双路搭建B-CNN网络。实验部分将通过两种不同的训练方法获取的论文提出的鉴伪框架和单一的VGG16、加了提取红色分量Lambda层的VGG16在手机人民币图像鉴伪识别上的方法进行了对比,实验表明在手机人民币图像上,论文提出的方法有更高的真伪识别性能。

In order to solve the difficulties brought by the mobile phone RMB image to the RMB authentication task,this paper constructs a mobile phone RMB image authentication framework based B-CNN.The frame takes as input the output of Block5 of VGG16 with Lambda layer for extracting red components.The Lambda layer extracting the red component of the mobile phone RMB is added to the front of the VGG16 network,and the output of the block5 of the VGG16 is used to simulate the two-way construction of the B-CNN network.The experimental part compares the authentication framework proposed by the papers obtained by two different training methods with the single VGG16 and the VGG16 with the red component Lambda layer on the mobile phone RMB image identification.Experiments show that on the mobile phone RMB image,the method proposed has higher recognition performance.