Improving data-driven estimation of significant wave height through preliminary training on synthetic X-band radar sea clutter imagery

X-band marine radar captures the signal reflected from the sea surface. Theoretical studies indicate that the initial unfiltered signal contains meaningful information about wind wave parameters. Traditional methods of significant wave height (SWH) estimation rely on physical laws describing signal reflection from rough surfaces. However, recent studies suggest the feasibility of employing artificial neural networks (ANNs) for SWH approximation. Both classical and ANN based approaches necessitate costly in situ data. In this study, as a viable alternative, we propose generating synthetic radar images with specified wave parameters using Fourier-based approach and Pierson–Moskowitz wave spectrum. We generate synthetic images and use them for unsupervised learning approach to train a convolutional component of the reconstruction ANN. After that, we train the regression ANN based on the previous convolutional part to obtain SWH back from the synthetic images. Then, we apply preliminary trained weights for the regression model to train SWH approximation on the dataset of real sea clutter images. In this study, we demonstrate the increase in SWH estimation accuracy from radar images with preliminary training on synthetic data.