Satellite Oil Spill Detection Using Artificial Neural Networks

Exploratory Data Analysis of Synthetic Aperture Radar (SAR) Measurements to Distinguish the Sea Surface Expressions of Naturally-Occurring Oil Seeps from Human-Related Oil Spills in Campeche Bay (Gulf of Mexico)

“…Fiscella et al [17] suggest using the perimeter to area ratio (PtoA). Fiscella et al [17] and Singha et al [18] recommend using a dimensionless normalized perimeter to area ratio (PtoA.nor). While small PtoA.nor values are related to simple geometry, larger values come from oil slicks with more complex geometries [19].…”

“…These are separately calculated for the twelve radiometric-calibrated image products ( Table 1). As suggested by [18,21], an arithmetic mean (AVG) of all pixels of each oil slick is computed. Three other central tendency representations are also considered: median (MED), mode (MOD), and mid-mean (MDM).…”

“…COV originally involves the ratio between STD and AVG, but herein other pair-values are also given (Table 3). Different authors recommend exploring COV, e.g., [18,24] refer to it as power-to-mean ratio, whereas [20] describe it as the oil slick's homogeneity, and for [21] it depicts the oil slick's heterogeneity. However, Miranda [25] emphasizes that oil slicks with different spatial structures (i.e., with completely different pixel configuration) can have identical average or median values and the same standard deviation.…”

Synthetic Aperture Radar Image Segmentation with Reaction Diffusion Level Set Evolution Equation in an Active Contour Model

“…In the past decades, SAR image segmentation technology has been widely and intensively researched as an important step in information extraction and automatic interpretation, and a variety of methods have been proposed. These proposed methods can be classified into many categories: (1) clustering-based methods such as the fuzzy c-means clustering algorithm [1], (2) threshold-based methods, including the Otsu algorithm [2], (3) specific theory-based methods, e.g., artificial neural network and deep learning, which are often applied to SAR image segmentation under complicated scenarios [3,4], (4) super pixel-based methods such as the multi-kernel joint sparse graph and multi-feature ensemble models [5,6], and (5) level set evolution (LSE)-based methods such as the active contour model (ACM) [7]. Of these existing methods, ACMs, which were first proposed by Kass et al [8], have attracted increasing attention from researchers owing to its in a level set framework.…”

Oil spill GF-1 remote sensing image segmentation using an evolutionary feedforward neural network

“…Image segmentation is considered as an important task in the analysis, interpretation, and understanding of images and is also widely used for image processing purposes such as classification and object recognition [13]. Singha et al [14] described the development of a new approach to oil spill detection by employing two different artificial neural networks (ANN), used in sequence. Ganta et al [15] represented a method for segmenting oil spill regions in satellite images taken in broad daylight using illumination-reflectance based level set model.…”