论文部分内容阅读
针对隧道地质超前预报过程中GPR(Ground Penetrating Radar)线测图分类解释准确率不高的问题,本文在深度学习模型Lenet的基础上,根据GPR线测图的尺寸特点设计了Lenet-6。Lenet-6在Lenet低隐含层部分新增一个卷积层和一个子采样层。通过深入研究Lenet-6低隐含层,给出了卷积层核函数选取的一般准则。仿真实验结果表明:在相同的迭代次数下,Lenet-6比Lenet具有更高的分类准确率。本文模型可为制定隧道施工和开挖计划提供科学合理的依据。
In order to solve the problem that the classification accuracy of GPR (Ground Penetrating Radar) line survey is not high in the tunnel geological advance prediction, Lenet-6 is designed based on the depth learning model Lenet and the size characteristics of GPR line survey. Lenet-6 adds a convolutional layer and a sub-sampling layer to the Lenet Low Hidden Layer section. By studying the Lenet-6 low-hidden layer, the general criterion for choosing the kernel of convolutional layer is given. Simulation results show that Lenet-6 has higher classification accuracy than Lenet under the same number of iterations. This model can provide a scientific and reasonable basis for the development of tunnel construction and excavation plans.