Oil spill segmentation in fused Synthetic Aperture Radar images

Abstract

Abstract-Synthetic Aperture Radar (SAR) satellite systems are very efficient in oil spill monitoring due to their capability to operate under all weather conditions. Systems such as the Envisat and RADARSAT have been used independently in many studies to detect oil spill. This paper presents an automatic feature based image registration and fusion algorithm for oil spill monitoring using SAR images. A range of metrics are used to evaluate the performance of the algorithm and to demonstrate the benefits of fusing SAR images of different modalities. The proposed framework has shown 45% improvement of the oil spill location when compared with the individual images before the fusion.

A Gaussian Process Regression Approach for Fusion of Remote Sensing Images for Oil Spill Segmentation

“…This phase computes the descriptors for the image region making it highly distinctive and invariant to illumination and angle changes. This is achieved by placing a Gaussian window over the regions and adding each sample to form orientation histograms, summarising the contents over an 8 × 8 sub region as shown in Fig.6 and 7 in [7].…”

“…Sensors in different bands of the electromagnetic spectrum have been applied, e.g. in hyperspectral and multi spectral bands, or Synthetic Aperture Radar (SAR) for oil spill detection in marine environment [1], [4]- [7]. However, SAR images are the most widely used for this purpose as they are not affected by local weather conditions and cloudiness [8].…”

“…The individual images need to be registered in space and time. A suitable fusion algorithm needs to be chosen that will efficiently bring together the complimentary information from the individual modalities [7]. Additionally, it is important to define the level of fusion since data fusion can be performed at different information levels including pixel, feature and decision level [16].…”

“…In [17], fusion of SAR and hyperspectral images (HSI) is performed at pixel level, although the fusion method used is not described in the paper, the approach focused on fusing images from different sensors (SAR and HSI) and not on multi-frequency SAR data. The works of [7], [18] explored fusion of multi-frequency (S&X-bands) and multi-resolution (C-band) SAR images by adopting the wavelet transform approach.…”

“…The rest of the paper is organised as follows. In section II, we present the proposed framework and a brief description of the registration is provided, with more detailed explanation of the process already discussed in our previous work [7]. Subsequently, GP is introduced and the proposed kernel design and fusion process are described.…”