【摘要】 针对深度学习的指静脉识别算法在训练样本较少以及训练次数较低时算法识别率降低问题,提出基于改进卷积神经网络的指静脉识别算法。通过增加卷积层数并使用LeaKy ReLU作为激活函数,提高网络泛化能力;使用改进的池化模型降低网络特征维度;反向传播调整权值时,引入判别信息作为约束条件。实验结果表明,该算法准确率在训练样本较少以及训练次数较低时明显优于其它指静脉识别算法。更多还原

【Abstract】 Aiming at the problem that the recognition rate of finger vein recognition algorithm for deep learning is low when training samples and training times are insufficient,the algorithm of finger vein recognition based on improved convolutional neural network was proposed.The number of convolutions was increased and LeaKy ReLU was used as the activation function to improve the generalization ability of the network.The improved pooling model was used to reduce the network feature dimension.When the weights were adjusted in reverse,the discriminant information was introduced as a constraint condition.Experimental results show that the accuracy of the proposed algorithm is obviously superior to other finger vein recognition algorithms when training samples and training times are insufficient.更多还原