ملخص: The performance of the mean level detector (MLD) is considered when two correlated targets are present in the reference cells used in estimating the detection threshold. We derive the probabilities of detection and false-alarm, and study the performance of this detector in terms of the correlation coefficient, p, between the targets and their ratio, I/S. It is shown that for a weak interfering target, I/S [les] 1, the ‘capture effect’ is minimized when the targets are heavily correlated. For high signal-to-noise ratios (SNR) the probability of detection of this detector is comparable to that of an ideal detector. However, for a strong interfering target, I/S > 1, not only the probability of detection is seriously reduced, but also, a ‘threshold effect’ takes place when the targets are heavily correlated. In order to overcome this problem, we consider the performance of the censored mean level detector (CMLD) in which the largest reference samples are censored and the noise level estimate in the test cell is obtained by combining the remaining samples in the reference window. We derive asymptotic expressions for the probabilities of detection and false-alarm in the sense that the SNR of the interfering target, I, is assumed to be very large (I --> [is proportial to]). This assumption implies that the interfering target return is the highest ordered sample in the reference window and must be censored. We observe that this procedure achieves robust detection performance in the presence of a strong and a heavily correlated interfering target with the primary target.

لغة: إنجليزية