Model Mismatch Effect in Coarse Resolution Heterogeneous Clutter Cells
Julie Ann Jackson · 2019
Passive radar imaging with today's signals of opportunity result in coarse range resolution images with heterogeneous clutter mixed with targets in a single resolution cell. Although land use land cover maps can aid in determining clutter composition, it is difficult to accurately model heterogeneous clutter in a coarse resolution cell. Furthermore, the target and its shadow will replace some amount of clutter in the target present hypothesis. Target detector performance prediction will be in error if signal model does not match the true physics of the situation. This paper characterizes how much variation occurs in receiver operating characteristic curves when the target present model does not account for a reduction in clutter due to the target and shadows. Results show the mismatched model over-predicts probability of detection by as much as 43% compared to the model that matches the scene physics. Examples assume complex random variables with Weibull magnitude and uniform phase, with complex Gaussian random variables as a special case.