CFAR Detection Strategies for Distributed Targets Under Conic Constraints

In this paper we deal with the problem of adaptive detection of mismatched mainlobe targets and/or sidelobe interfering signals that are distributed in range. To this end, we investigate the impact of modeling the actual useful signal as a vector belonging to a proper cone with axis the nominal steering vector as a means to improve the robustness of the decision rule in presence of mainlobe targets; similarly, in order to improve the rejection capabilities of the decision rule in presence of sidelobe interferers we study the effects of replacing the usual noise-only hypothesis with a noise-plus-interferers hypothesis where interferers belong to the complement of a cone with axis the nominal steering vector. At the design stage we resort to the two-step GLRT-based design procedure; to this end, we assume that a set of training data is available, namely data free of signal components, but sharing the same Gaussian distribution of the noise in the cells under test. Remarkably, proposed detectors possess the CFAR property under the noise-only hypothesis. The performance assessment, conducted by Monte Carlo simulation, is aimed at assessing the effectiveness of proposed solutions, also in comparison to existing ones.