Deep Sigma Point Processes for RCS Modeling in Spaceborne SAR Imagery

This paper introduces a deep sigma-point process (DSPP) for radar cross-section (RCS) prediction in spaceborne synthetic aperture radar imagery. It evaluates the approach on a RADARSAT-2 dataset containing 208,191 verified ships. Built on a hierarchical Gaussian-process framework with Bayesian inference, DSPP produces predictive distributions rather than point estimates, allowing uncertainty to be represented. A Matérn kernel with automatic relevance determination ranks important radar, operational, and environmental features. On the test data, the authors report a 20.83% reduction in RMSE, a 25.89% increase in R-squared, and 44.4% reductions in both residual interquartile range and median absolute deviation versus linear-regression baselines.