Comparison of Two CFAR Detection Scheme for SAR Imagery

Abstract:
SAR is one of the most important microwave-based sensors. Its main advantages are the production of images of very high resolution and the possibility of all-weather operation. As to detecting or recognizing a target in SAR image, the algorithm must identify the target pixels and reject the clutter pixels as precise as possible. Many researches have been done on developing a detector which can keep constant false alarm rate (CFAR). In this paper two CFAR detectors will be introduced. One is based on a mixed distribution which is derived from the combination of a normalized Rayleigh and Weibull distribution. The other is based on the G distribution derived from Multiplicative Model which is useful to describe the extremely heterogeneous clutter. Then the performance of the above two detectors both on simulated data and real SAR imagery will be given. The real data are from MSTAR clutter data embed with different targets chips. Further discussion on design a CFAR detector on SAR imagery will present at last.