A low-loss CFAR algorithm for Weibull background is discussed. The two parameters (shape and scale) of the background statistics are estimated using a maximum-likelihood algorithm. A CFAR threshold based on parameters estimated in this way exhibits a smaller variance, and hence a smaller CFAR loss, than thresholds based on other estimation algorithms such as moments or order statistics. The analysis covers both complete and censored background sample sets.
No takes yet. Share an insight, caveat, or question.
Ravid et al. (1992) studied this question.