Ground Penetrating Radar Microwave Denoising Based on Improved K-SVD Dictionary Learning Method

The underground environment is complex, so the microwave emitted by the ground-penetrating radar (GPR) will be disturbed by the clutter and the stochastic noise, which will make the echo signal no longer meet the characteristics of stability and Gaussian. Therefore, the traditional attenuation methods are difficult to meet the requirements of GPR microwave denoising. In this paper, we introduced an improved K-singular value decomposition (K-SVD) dictionary learning method into the GPR microwave denoising. Replace the fast iterative shrinkage threshold (FISTA) method with the orthogonal matching pursuit (OMP) algorithm to obtain sparse encoding, and the SVD algorithm is used to update the dictionary. Finally, the smaller values in the sparse coefficients are removed to suppress the stochastic noise in the GPR data. The simulation results showed that the method proposed in this paper can effectively improve the signal-to-noise ratio (SNR) of the GPR microwave.

I. Introduction

Ground penetrating radar (GPR) is an important method to detect underground targets by emitting microwave, which has the advantages of high resolution, high efficiency, intuitive results, and non-destructive detection. Due to the complex underground environment and uneven distribution of under-ground media, the microwave is usually disturbed by random noise and clutter. So the microwave detection needs to remove the clutter and the stochastic noise. In general, there is no way to interpret the subsurface media without processing the microwave signals. Therefore, it is necessary to process the received microwave to improve the signal-to-noise ratio (SNR) [1]. The traditional denoising methods perform awful in the complex underground environment with high clutter. In order to improve the adaptive denoising performance of the GPR data in environment of stochastic clutter, dictionary learning and greedy algorithms are applied to GPR data denoising. The echo signals of the GPR data contain direct wave, clutter, random noise and the target signal, where the energy of the target signal accounts for only a small fraction of the energy of the echo signal. Therefore, processing the echo signal to remove the clutter and the random noise is significant for improving the SNR of the GPR microwave data.