Fast estimation of ocean background reflectivity in synthetic aperture radar images

Abstract

Several ship detection algorithms have been developed over the years, where some of the most commonly used are the constant false alarm rate (CFAR) algorithms. There are challenges to the existing algorithms, both when it comes to processing time and difficult scene situations, such as heterogeneous ocean clutter and multiple targets. Existing algorithms that are equipped to handle situations where clutter edges and multiple targets are present, will require significantly increased processing time.

The goal of this thesis is to present a new fast method to estimate the underlying radar reflectivity from a speckled SAR image. The aim is to recover the mean intensity, which is a key parameter in statistical models of SAR intensity measurements over ocean. The estimation of the mean intensity should be fast and robust, in the sense that it handles heterogeneous clutter edges and the presence of multiple targets. The result from the presented algorithm is meant to be an input parameter for current ship detectors.

The steps of the intended algorithm is to take advantage of a nonuniform FFT (NFFT) to truncated SAR data, which will result in frequencies on a regular grid. Next, a lowpass filter will be applied in order to suppress speckle present in the image. Finally, an inverse transformation will be utilized and an estimated mean intensity can be recovered. Thus, a threshold can be determined based on this mean value. Because the NFFT did not provide the desired results, it was not possible to present a complete algorithm.