Iceberg Detection in Dual-Polarized C-Band SAR Imagery by Segmentation and Nonparametric CFAR (SnP-CFAR)

We propose an unsupervised method for iceberg detection over sea ice-free waters. The algorithm is based on the segmentation and nonparametric constant false alarm rate (SnP-CFAR) approach. Unlike in parametric CFAR detection, our method, there no need to define target, guard, background areas explicitly. Instead, we apply pixels within each detected segment formed of nearby not included target segment. By using probability density function (PDF) estimates, also eliminate assuming a specific type PDF. compared results with operational Danish Meteorological Institute (DMI) Gamma-CFAR results. were evaluated against icebergs manually identified by Finnish (FMI) Ice analysts. Our exhibits reduced number alarms. present SAR channel-cross-correlation (CCC). CCC was able distinguish many true targets low However, seems miss some its main use would be confirming observations.