岩溶隧道开挖时，采用探地雷达探明岩溶不良地质的空间位置、几何形态以及物质属性对于评定灾害危险等级、制定详细施工方案具有极为重要的意义。但由于隧道探测环境中的干扰、电磁衰减、散射以及岩溶地质复杂性，导致岩溶不良地质探测难以取得满意的效果。本项目拟从探地雷达信号处理研究入手，采用理论分析、模型试验、现场试验以及数值模拟等研究手段，首先针对探地雷达在隧道地质预报过程中“弱信号，强干扰”信号特点研究基于Shearlet变换的噪声压制及弱信号提取方法；然后系统研究探地雷达属性提取及属性定量化表示，建立典型岩溶不良地质与探地雷达回波信号属性参数的关系模型；最后基于深度学习理论研究探地雷达数据特征自动提取，建立基于探地雷达多属性融合的典型岩溶不良地质空间分布与物质属性非线性智能识别模型。该项目的实施对探地雷达信号噪声压制及弱信号提取方法是一完善，为探地雷达识别岩溶不良地质提供理论依据和技术指导。

During the excavation of karst tunnel, detection by Ground penetrating radar (GPR) of the spatial position, geometrical morphology and properties of the unfavorable geology in karst is of great importance for grade estimation of hazard and formulation of detailed construction method. But the interference, attenuation and scattering of electromagnetic wave and the complexity of karst in the tunnel make it difficult to obtain satisfied results of detecting unfavorable geology. The research starts with the signal processing of GPR by using theoretical analysis, model experiment, field test and numerical simulation. Firstly, we will research the method of noise suppressing and weak signal extraction based on Shearlet transformation according to the signal properties of GPR, which is weak but has strong interference, in geologic forecast for karst tunnel. Secondly, we will carry out systematic research of the extraction and quantitative representation of GPR attribute as well as the relation between the unfavorable geology in karst and the attribute of GPR echo. Finally, we will research the automatic extraction of the GPR data attributes based on deep-learning theory and establish the nonlinear intelligent discrimination model with multiattribute fusion of material properties and spatial distribution of the unfavorable geology in typical karst based on GPR. The implementation of the project will improve the method of noise suppressing and weak signal extraction of GPR and provide theory and technique support for detecting unfavorable geology in karst.