Automated Oil Spill detection with ship borne radar

Nasser Mostafa Saleh
Egyptian Environmental Affairs Agency
30 Misr Helwan Zerae Road, Maadi, Cairo, Egypt
saleh@itc.nl

ABSTRACT

This research outlines a real-time, ship-borne radar facility, SHIRA, for the tracking of nearby oil spills during oil containment and cleaning operations at sea with high quality imagery.

SHIRA is an imaging digital X-band radar, developed by TNO for the measurement of oceanographic features. It was especially adapted for the imaging of oil slicks by implementing a number of improvements, based on the outcome of a previous oil detection experiment. The main aspects that make SHIRA suitable for the current purpose are its high sensitivity and dynamic range, and its ability to integrate a large number of images of the same scene, after correction for platform motions. A quantitative comparison has been made between various filters, which are able to reduce variance in homogeneous areas, preserve edges and lines, suppress point scatter, and preserve spatial variability, while avoiding artefacts. A Gamma Filter yields the best smoothing and despeckling. Moreover, it seems to keep the edges without blurring the minimum. Eight textures are applied, based on the co-occurrence matrix. These textures include mean, variance, homogeneity, contrast, dissimilarity, entropy, second moment, and correlation. Co-occurrence measures use a grey-tone spatial dependence matrix to calculate texture values. This is a matrix of relative frequencies with which pixel values occur in two neighbouring processing windows separated by a specified distance and direction. It shows the number of occurrences of the relationship between a pixel and its specified neighbour. The use of the correlation texture implies a scaling of the axes so that the features receive a unit variance. It prevents certain features from dominating the image because of their large value. Correlation texture performs best, in terms of the probability of oil occurrence because it allows clear discrimination between synthetic oil (oval-round shaped), natural oil (striped), water waves and water for data with high waves (dataset Prestige event, Spain) or with low-moderate waves (dataset with oil simulations in North Sea).

Non-Standardized Principal Component Analysis can discriminate between oil and water regardless of wind speed. To avoid the disadvantage of the use of linear combinations that lead to a high correlation of the resulting textures and loss of textural information, Principal Component Analysis has been applied. The first principal component of covariance matrix, i.e. the eigenvector with the largest Eigen value corresponds to an identifiable physical property, which is most important for interpretation of the patterns. The higher order Non-Standardized Principal Component Analysis and Correlation methods produce the same results in this case.

Last Update: 2005-04-5