A major problem that occurs in constant false alarm rate (CFAR) schemes is presented by regions of nonhomogeneous clutter background. The situation occurs when the total noise power received in a single reference window does not follow the assumption of independent and identically distributed clutter in all reference window cells. Bayesian statistics provide a mathematical procedure for changing or updating the degree of belief about the clutter parameter in light of more recent information. A Bayesian CFAR (Bay-CFAR) processor is developed and analyzed. The Bay-CFAR processor exploits a priori knowledge of a nonhomogeneous clutter environment to considerably improve the detection performance relative to a classical cell averaging CFAR (CA-CFAR) processor. The performance improvement is demonstrated with a small reference window size that allows the processor to respond quickly to a rapidly changing clutter environment.
展开▼
