Effective Clutter and Noise Elimination in GPR: A Hybrid PCA and Wavelet Packet Approach

One of the most efficient non-destructive geophysical techniques for finding subsurface objects is ground penetrating radar (GPR), which has enormous applications including landmine, cracks in bridge decks, railway ballasts and utility detection etc. However, challenges such as background noise and clutter can obscure shallowly buried objects, making accurate target signature extraction difficult. This work proposes a hybrid approach combining principal component analysis (PCA) and wavelet packet transform for effective clutter and noise elimination. PCA is employed to suppress clutter by leveraging its ability to identify dominant signal components associated with ground reflections, while wavelet packet transform is utilized for fine-grained denoising to remove residual random noise. The proposed methodology is evaluated using simulated datasets generated with gprMax, demonstrating superior performance compared to traditional techniques in terms of Peak Signal-to-Noise Ratio (PSNR) and image entropy. These results highlight the method's capability to enhance GPR image quality, ensuring reliable and accurate subsurface object detection.