TABLE 14. THE CP SAR TARGET DETECTION METHODS. CATEGORY METHOD ADVANTAGES Mixed method Mixed method PSF + CFAR Proposed hybrid detection algorithm can effectively detect ships in CP SAR images with medium and low resolution PSF + CFAR Results suggest that the use of a FAR value based on the properties of the polarization ellipse and other polarimetric information has potential to improve the detection performance of a SAR system Mixed method PSF + CFAR Improves the ship-sea contrast (and the signal-clutter ratio) more than popular detectors, such as entropy and span Mixed method PSF + CFAR + machine learning Reduces false alarms while retaining targets with weak scattering signals (CONTINUED) DISADVANTAGES Expected to take sea state and ship-related measurements into account Difference between real and simulated HP SAR data is ignored APPLICATION Ship detection Iceberg detection DATA SOURCE Simulated RCM from RADARSAT-2 Simulated CTLR mode from RADARSAT-2 REFERENCE [40] [218] Ignores differences between real and simulated HP SAR data CFAR method needs complex calculation Size of guard filter is sensitive to that of a ship Ship detection Simulated HP SAR data from RADARSAT-2 and ALOS PALSAR Ship detection Simulated CP SAR data from Gaofen-3 quad-polarization SAR images Mixed method PSF + CFAR + visual attention Mixed method PSF + machine learning Proves the importance of polarimetric information in reducing false alarms and the feasibility of the visual attention model in CP SAR ship detection the selected features contain sufficient information to detection and identify Internal solitary waves and the methods based on them outperform traditional Wishart polarization clustering algorithm Mixed method PSF + machine learning Raney decomposition features slightly outperform FP backscatter coefficients CP could be more efficient in the detection of new planting season in tropical regions, considering lower power requirement than FP Mixed method Mixed method Single method Single method PSF + machine learning PSF + machine learning CFAR Extracts effective, less-correlated CP parameters for oil spill classification using the random forest classification algorithm Performs well in detecting ship targets Can reject azimuth ambiguities CP system provides better performance than a conventional dual-polarization and singlepolarization systems CFAR Proposed CFAR method based on the notch filter provides a promising technique for ship detection using HP SAR data Single method CFAR Allows for automatic and adaptive implementation of ship detection in complicated sea backgrounds in practical applications Long calculation time Ship detection Simulated CTLR images from RADARSAT-2 Polarization Scattering characteristics of internal solitary waves are not discussed Internal solitary wave detection Simulated CTLR mode from ALOS PALSAR [62] [66] [52] [204] Some variants of random forests and SVMs may yield overall accuracy below expectations Waterlogged rice field detection simulated CP SAR data from PALSAR-2 [146] Sample quality may affect training results, causing overfitting SVM algorithm is sensitive to training data Difference between real and simulated HP SAR data is ignored Difference between real and simulated HP SAR data is ignored Oil spill detection Ship detection Ship detection Ship detection RCM CP SAR data from RADATSAT-2 Simulated CTLR mode from AIRSAR Simulated CTLR mode from RADARSAT-2 Simulated CTLR mode from ALOS PALSAR and RADARSAT-2 Complicated calculation Ship detection Simulated CP SAR data from AIRSAR [67] [54] [83] [60] [61] (Continued ) SEPTEMBER 2022 IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE 141