A Coarse-to-Fine Approach for Ship Detection in SAR Image Based on CFAR Algorithm

Among ship detection methods for SAR image, constant false alarm rate (CFAR) is the most important one. However, several factors, such as detector parameter and distribution of ocean clutter, affect the performance of CFAR detection. This paper proposes a novel hierarchical complete and operational ship detection approach based on detector parameter estimation and clutter pixel replacement, which is considered a sequential coarse-to-fine elimination process of false alarms. First, a simple barycentric algorithm is adopted to estimate target-window size, and the morphology method is used to estimate false alarm rate for CFAR detector. Second, a clutter pixel replacement approach based on the statistical features of sea clutter is presented to obtain statistically independent, stationary, and Weibull distributed random data for CFAR detector to remove all false alarms. Experimental results of the detection methods on a SAR image dataset show that the proposed approach is effective in reducing false alarms and obtains a satisfactory ship detection performance.