Performance of a new CFAR detector based on trimmed mean

Constant false alarm rate (CFAR) algorithms are used to detect radar targets in noise or clutter background whose mean power level is not known a priori. In this paper, a new CFAR detector (MTM) based on trimmed mean (TM) is proposed. It takes the mean value of two local noise power estimations created by leading and lagging windows which apply the TM method as a global noise power estimation, and it also uses the automatic censoring technique proposed by Y. He (1994). Under the Swerling II assumption, the analytic expressions of P_fa, P _d and ADT are derived for the detector. By comparison with other schemes, the results show that the detection performance of MTM is improved both in homogeneous background and in nonhomogeneous environment caused by strong interfering targets and clutter edges, particularly in dense interfering target situations, it is greatly improved, while the sample sorting time of MTM is less than half of that of OS