基于小波域KL变换的地质雷达信号处理技术研究

为了克服现有信号处理算法对地质雷达直耦波和噪声滤除的不足,基于KL变换和小波变换进行算法融合设计,提出一种适用于地质雷达信号滤波的小波域KL变换方法。采用电磁波时域有限差分法模拟雷达检测过程,并基于理想无噪声的雷达仿真信号设计验证实验,通过与KL变换方法、小波阈值去噪方法的对比,对小波域KL变换方法的滤波效果进行定量分析和评价。实验结果表明:小波域KL变换对于直耦波的辨识能力较强,用于地质雷达信号直耦波的去除可以取得理想的效果;在采用KL变换和小波变换滤除噪声时,去噪信号的信噪比分别为10.16和15.12,而小波域KL变换对应的结果为18.34,对于噪声的滤除具有更好的效果;同时,小波域KL变换滤波结果对小波函数和分解层数的敏感度较低,对于深部噪声信号的辨识能力亦较强。基于地质雷达实测数据的测试结果同样验证了小波域KL变换方法在实际工程应用中的良好性能。

Study on GPR Data Processing Technology Based on Wavelet Domain KL Transform method

The wavelet domain KL transform method is proposed in this paper, based on the fusion of KL transform and wavelet transform method, to overcome the shortcoming of the existed data processing methods in filtering of the GPR data. A finite difference time domain model is built to create the ideal GPR data with no noise, and after some noises are added to the ideal data, the wavelet domain KL transform method is tested with the contaminated GPR data, and the accuracy is quantitatively evaluated by comparing with the KL transform and wavelet thresholding methods. The results shows that the wavelet domain KL transform method has a strongly identification of the direct and coupled waves, as well as the noise. The SNR generated by KL transform and wavelet thresholding method is 10.16 and 15.12 respectively, while the SNR generated by the wavelet domain KL transform method is 18.34. Besides, a better identification of the deeper noise with a lower energy and a lower sensitivity to the wavelet function and decomposition level are proved by the filtering test. Another filtering test based on the measured GPR data also verifies the good performance of the wavelet domain KL transform method in engineering applications.