Oil and Gas Pipeline GPR Image Denoising Based on Wavelet Enhanced Generative Adversarial Network Method

With the increasing demand for oil and gas resources in the whole society, the safety and reliability of pipelines cannot be ignored, and pipeline leakage anomaly detection becomes particularly important. Ground Penetrating Radar(GPR) is commonly used for detecting pipeline leakage anomalies. GPR images are widely used in oil and gas pipeline detection, underground target recognition, and other fields. However, due to the non-uniformity of the medium, clutter interference, and equipment noise, their image signal-to-noise ratio is low and target edges are blurred, resulting in a decrease in feature extraction accuracy. This article proposes a wavelet enhanced generative adversarial network (WavGAN) fusion denoising method, which combines wavelet and GAN to solve the contradiction between noise suppression and effective signal preservation in traditional methods. Compared to the single-resolution generation of conventional generators, this method can synchronously optimize the details of different frequency levels, and especially excels in preserving the highfrequency texture of weak reflection signals in GPR images. The experiment shows that compared with wavelet threshold denoising, non local mean (NLM) and GAN denoising, this method improves PSNR to 28.9 dB, which is 5% higher than other methods. This study provides a new approach for denoising GPR images.