Hybrid procedure for CFAR in non-Gaussian clutter

The paper deals with false alarm regulation in the presence of non-Gaussian clutter through a hybrid clutter-map/L-filtering technique. The authors propose a new system wherein the adaptive threshold is formed starting with a set of range cells, grouped in a map cell, whose returns are filtered through a nonlinear system, and subsequently processed on a scan-by-scan basis. It is shown that, with suitable choice of the coefficients of the L-filters, the scheme achieves CFAR in all cases where the clutter distribution is of location-scale type or can be forced into this class through suitable transformations. The authors also tackle the problem of system optimisation, showing that the minimum-variance estimators of the adaptive threshold are simply the linear combination of the corresponding optimum estimators of the clutter location and scale parameters. A thorough performance assessment is presented, assuming Weibull and lognormal clutter: it is shown that the new system largely outperforms all the other biparametric systems. Moreover, proper sample censoring allows the prevention of the self-masking effect from slow targets, whether point-like or range-spread, which persist in the same map cell for several scans.