Percentile-Based Estimator for Generalized Rayleigh CFAR Ship Detection in SAR Imagery

Synthetic Aperture Radar (SAR) imaging is a valuable tool for remote sensing, offering insights into complex environments. SAR images often, though, contain clutter from natural and man-made structures, making accurate target detection difficult. To overcome this challenge, we introduce, for complex Generalized Rayleigh (GR) environments, a Constant False Censoring Rate (CFCR) algorithm based on the Percentile Estimator (PCE) along with a Constant False Alarm Rate (CFAR) detection algorithm, yielding the PCE-CFAR detector. In doing this, based on the assumption of a homogenous middle half set of pixels within the Sliding Reference Window (SRW), we first propose closed-form expressions for the lower and upper censoring thresholds using the PCE. Then, upon censoring the undesirable data at both ends, we employ the remaining dataset to estimate the unknown distribution parameters via the same expressions of the PCE used in the censoring algorithm, and so the detection threshold closed-form expression. Finally, through a series of experimentations and comparisons with existing techniques, we showcase the effectiveness and resilience of the proposed ship target detector in real SAR images.