一种应用于嵌入式设备的指印活性检测方法

A fingermark liveness detection method applied to embedded equipment

现有的指印活性检测方法因存在模型复杂、训练参数量大等问题,在嵌入式设备这种运算能力受限的场景中应用较为困难。为解决这一问题,提出一种应用于嵌入式设备的指印活性检测方法。该方法构建了一个轻量级的神经网络模型,在传统卷积神经网络模型的基础上,取消了全连接层,采用分通道的残差模块替代原有的卷积层,精简了网络结构,大幅度降低了模型参数量,缩短了模型运行时间。建立指印数据集,并用其进行实验分析,实验结果表明:笔者构建的轻量级神经网络模型在测试集上准确率为96.22%,相较于传统神经网络模型在指印活性检测方面准确率更高,参数量更少,对设备运算性能要求更低。

The existing fingermark liveness detection methods have some problems,such as complex model and large number of training parameters,so it is difficult to apply in scenarios where the computing power of embedded devices is limited.In order to solve this problem,a fingermark liveness detection method applied to embedded devices is proposed.Based on the traditional convolutional neural network,the lightweight neural network is built in this method,the sub-channel residual module is used to replaces the original convolutional network,the network structure is simplified,the number of network parameters are greatly reduced,and the model running time is shortened.The fingermark data set is established and used for experimental analysis.The experimental results show that the accuracy of the lightweight neural network is 96.22%in the test set.Compared with the traditional neural network,it has higher accuracy in fingermark liveness detection,smaller number of parameters and lower requirements for equipment operation performance.