摘要
针对现有方法在指静脉纹理特征退化的情况下识别率低的问题,提出一种基于改进Xception网络和CBAM的识别算法。首先,选取Xception网络为基础网络,并添加跳跃结构以强化浅层纹理特征的传递和利用,调整网络深度以加快网络训练速度并防止过拟合;其次,改进卷积注意力模块(CBAM)的池化过程和通道压缩比,使网络更加注重纹理特征;最后,通过Softmax分类器对测试样本进行分类。实验结果表明,该算法相较于其他现有方法具有一定的优越性。
Aiming at the problem of low recognition rate of existing methods in the case of finger vein texture feature degradation,a recognition algorithm based on improved Xception network and CBAM is proposed.Firstly,select the Xception network as the basic network,and add skip connection to strengthen the transmission and utilization of shallow texture features,adjust the network depth to speed up the network training speed and prevent overfitting;secondly,improve the pooling process and channel compression ratio of the convolutional attention module(CBAM)to make the network pay more attention to texture features;finally,the test samples are classified through the Softmax classifier.The experimental results show that the algorithm has certain advantages compared with other existing methods.
作者
李治中
LI Zhi-zhong(School of Science and Technology,North China University of Technology,Beijing 100144,China)
出处
《电脑与信息技术》
2022年第1期4-7,13,共5页
Computer and Information Technology
基金
国家自然科学基金资助项目(项目编号:401053761706)。