An integrated prediction method for sea clutter amplitude distribution in complex spatio-temporal scenarios

The characteristics of sea clutter undergo changes influenced by the spatial-temporal variations in marine meteorological conditions, exhibiting significant spatial-temporal non-uniformity. This severely impacts the detection capability of radar systems in different maritime domains. In this study, considering the description of sea clutter amplitude distribution relies on both distribution types and parameters, this paper proposes an integrated prediction method for sea clutter amplitude distribution based on multi-task parallel learning using deep learning technology. To address the issue of negative output values in parameter prediction, a loss function that suppresses negative values is introduced. This enables parallel prediction of sea clutter amplitude distribution types and parameters. Through training and testing on a dataset from the South China Sea comprising S-band sea clutter data and spatio-temporally corresponding marine meteorological parameter data, the results indicate that this approach can effectively predict sea clutter amplitude distribution types and parameters in spatio-temporal scenarios influenced by changing marine meteorological conditions. This method facilitates the prediction of sea clutter amplitude distribution characteristics in large-scale maritime regions with varying spatio-temporal conditions.