Performance analysis of the OS family of CFAR schemes with incoherent integration of M-pulses in the presence of interferers

The author analyses the performance of the ordered statistics (OS) CFAR schemes for M-correlated sweeps when these processors operate in multiple-target environments. These schemes include the conventional OS scheme as well as its modified versions. Members of the OS family use local estimators based on ordered statistics and generate both a near and far range noise-level estimate. Local estimates are then combined through a mean-level (ML), a maximum (MX) or a minimum (MN) operation to estimate the final noise power level. Exact expressions are derived for the detection probabilities of these techniques under a Rayleigh fluctuating target model. The processor performance for that case is characterised by an integral equation, which is solved numerically to assess the radar system´s behaviour against the background conditions. It is shown that for a fixed M, the relative improvement, over the single sweep case, increases as the correlation between consecutive sweeps decreases. For the same parameter values, the ML-OS detector gives the best homogeneous performance. In multiple-target situations, the ML-OS scheme still has the best performance, given that the number of interferers in each half of the reference window is within its allowable values (R⩽N-K). On the other hand, the MN-OS scheme is preferable when a cluster of radar targets appears among the candidates of the reference window