Contrast enhancement for an iceberg detector with a CFAR test using dual-polarized radar imagery

Abstract

The identification of small icebergs in SAR images is challenging especially when these are embedded in sea ice. In this work, a new detector is proposed based on dual-polarized incoherent SAR images. Small icebergs have a stronger cross polarization accompanied by a higher cross- over co-polarization ratio compared to sea ice in many cases. This is the rational at the base of the detector. The new detector is tested with dual-polarization ground detected Sentinel-1a Extra Wide swath images acquired over the time span of two months. The test area is on the East Coast of Greenland, where large numbers of icebergs occur. This work is mostly focused on the development of an appropriate Constant False Alarm Rate for the contrast enhanced image, that is based on a modification of the Cell-Averaging CFAR. A quantitative comparison of the proposed detector with an ordinary CA-CFAR using only the cross polarization channel is provided. We focus on grounded icebergs as test targets. The proposed methodology seems able to improve the contrast between icebergs and sea ice clutter up to 75 times. This returns an increased probability of detection and reduced probability of false alarms.