A Backscattering-Suppression-Based Variational Level-Set Method for Segmentation of SAR Oil Slick Images

Oil Spill Detection in Terma-Side-Looking Airborne Radar Images Using Image Features and Region Segmentation

“…One example of this is the automatic estimation of seeds from the saliency map, as described in Section 3.2 and Section 3.3 . Unlike others, such as the works shown in [ 9 , 10 , 11 , 12 , 13 , 27 , 28 , 29 , 30 ], the proposed approach always uses full images of the mission as input without using pre-processing masks to eliminate both land and noise pixels. The processing of the full image complicates the detection process but helps achieve a high level of automation.…”

“…There are two ways in which to approach the oil-spill detection problem: studying the characterization of the slick by means of the multi-polarization features of SAR techniques, as occurs in [ 5 , 6 , 7 ], or using the brightness image obtained from the backscatter signal without considering the parameters of the processes of image acquisition and formation, as in [ 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 ]. Our work falls within the second category.…”

“…Our work falls within the second category. Most of the approaches based on image processing techniques focus on segmenting SAR images in order to identify the dark-spot region that represents the oil slick [ 8 , 9 , 10 , 11 , 12 ], on extracting features so as to recognize the oil slick from previously segmented images [ 13 , 14 , 15 ], or on classifying dark spots using machine learning techniques [ 16 ], statistical classifiers [ 17 , 18 , 19 ] or/and artificial neural networks [ 20 , 21 , 22 , 23 ]. State-of-the-art methods sometimes use a combination of several approaches, as in [ 24 ], and on other occasions, the image processing methods are combined with Geographic Information Systems (GIS), as in [ 25 ], to provide both the detection and precise location of marine spills.…”

“…Moreover, the methods are usually tested in an offline process and this may make the performance of the detection task in an onboard detection system difficult, because they are not always designed for input images with sufficient variability (i.e., the acquisition conditions are always chosen in the same way, signifying that the background produced is usually homogeneous or has a specific texture pattern). Another drawback of some of the state-of-the-art methods is the use of cropped images as inputs rather than using images representing a whole mission or full scan (i.e., an image represents a close-up of the oil spill with some background) [ 9 , 12 , 27 , 28 , 29 ]. The input images also frequently show only two types of targets (i.e., the sea and oil slick zones) [ 30 ] and when the image shows others, such as land, then pre-processing filters are manually configured to hide or eliminate the pixels in those zones [ 10 , 13 ].…”

“…In particular, the proposed method is based on the segmentation technique category, without considering the multi-polarization features of the radar. But unlike the works in [ 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 ], the input is a SLAR and not a SAR image. It is very difficult to find works in the state of the art that address oil-spill detection using SLAR images.…”