CFAR detection of targets under directional noise background

Abstract

For a radar or sonar system, target detection is a basic function of the system. In sensor array processing, target detection is facilitated by examining the beamformer output of the received signal. The conventional method integrates the beam power observations by simple summation or linear integration and no statistical information is used. However, if the noise environment is directional, this will result in non-CFAR (constant false alarm rate) detection performance. In this paper, we exploit generalized likelihood ratio test (GLRT) to design the detector. We divide the data sequence from all the sensors into different segments and calculate the power observation of each segment. Then the probability distributions of the power observation under both hypotheses are derived. The unknown parameters of the signal model can be estimated by moment estimation based on statistical properties of the observations. Compared with the conventional processor, the GLRT detector can normalize the background of the output test statistic so that it shows CFAR property. Under the noise with directionality, the new detector we developed can still work well while the conventional one will show a false result due to the directionality of the noise when the SNR is low. Simulation result shows that the GLRT detector can also be used for multiple targets situation.