Performance Comparison of Statistical Models for Characterizing Sea Clutter and Ship CFAR Detection in SAR Images

A fundamental issue of maritime applications of synthetic aperture radar (SAR) data is the development of precise statistical models for clutter pixels. Several statistical models including the GK, K+R, andGAOhave been demonstrated to be promising for characterizing sea clutter in SAR images. This work is devoted to investigating the improvements in clutter fitting and ship detection performances by using the recently proposedGAO, compared to that using the GK and K+R. First, the solution uniqueness of parameter estimators by applying the “method of log cumulants” (MoLC) for theGAOis mathematically proven in the first time. Then, we assess the fitting performance of different models for sea surfaces with different wind speed conditions. Next, the constant false alarm rate (CFAR) detection performance of ships based on different models is compared by the indicators of CFAR loss and detection efficiency. Experiments performed on L-band ALOS-PALSAR SAR data verify the modeling capability of theGAOmodel for sea clutter. Moreover, several ship detection examples indicate the usefulness and potential of theGAOmodel for CFAR detection in practical applications.