Heterogeneous performance evaluation of sophisticated versions of CFAR detection schemes

Abstract

The detection of radar targets in a background, the statistical parameters of which are unknown and may not be stationary, can be effectively achieved through CFAR processors. The CA-CFAR scheme performs optimally for homogeneous and exponentially distributed clutter observations. However, it exhibits severe performance degradation in the presence of outlying target returns in the reference set or in regions of abrupt change in the background clutter power. The OS-CFAR processor has been proposed to solve both of these problems. Although this processor may treat target multiplicity quite well, it lacks effectiveness in preventing excessive false alarms during clutter power transitions. The TM-CFAR algorithm, which implements trimmed averaging after ordering, can be considered as a modified version of OS technique. By knowingly trimming the ordered samples, the TM detector may actually perform better than the OS processor. To simultaneously exploit the merits of CA, OS, and TM schemes, two combinations namely CAOS and CATM have been suggested. Each one of these versions optimizes good features of two CFAR detectors, depending on the characteristics of clutter and searched targets, with the goal of enhancing the detection performance under constant level of false alarm. It is realized by parallel operation of two standard types of CFARschemes. Our goal in this paper is to analyze these two developed versions in heterogeneous situations, to show to what extent they can improve the behavior of the conventional CFAR processors.