Scan-by-scan averaging CFAR

The main shortcoming of currently working CFAR procedures based on the slidingwindow concept is a substantial performance degradation in the presence of nonhomogeneous background interference. Alternative procedures, based on clutter maps, exploit the intrinsic local homogeneity of the radar environment, thus yielding enhanced robustness and, at the same time, little loss of detectability with respect to optimum detectors. The paper analyses one such procedure and shows that it works well under quite general conditions; plots are given of detection probability against the average signal-to-noise ratio for several values of false alarm probability. The problem of the self-masking target is also investigated and the performance impairment due to such phenomena is assessed; plots of detection performance against the number of scans for which the target persists in the same map cell are given for several values of the system parameters.