CFAR ship detection in SAR images based on lognormal mixture models

In this paper, we propose a new model, the lognormal mixture model (LMM), for characterizing the non-negative sea clutter in intensity/amplitude SAR images. By a change of variables, we show that the LMM is in fact equivalent to the Gaussian mixture model (GMM) in the log intensity/amplitude domain, and thus the parameters can be effectively estimated using the expectation-maximization (EM) method. Furthermore, we solve the threshold calculation problem by Newton´s method which enables a fast convergence. Accordingly, Constant False Alarm (CFAR) ship detection algorithm is designed using the LMM, and its effectiveness is demonstrated with SIR-C/X SAR data.