Each year, ships and industries are damaging the delicate coastal ecosystem in many parts of the world, releasing oil or pollutants into rivers and coastal waters. The offshore environments are also polluted by mineral oil, mainly for the following reasons:
• tanker incidents at sea, where large amounts of oil are spilled into the sea;
• illegal oil discharges from vessels, during their “normal operating procedures” (i.e., oil dumped during cleaning operations);
• natural oil seepage.
After a tanker accident, the biggest problems encountered are the difficulty in obtaining an overall view of the affected area, getting a clear idea of the extent of the slick, and predicting the direction in which it will move. For natural oil spills and those caused by humans, it is necessary to plan a regular monitoring programme. The aerial surveys over large areas (e.g., the Mediterranean Sea) to check the presence of oil are limited to daylight and good weather conditions.
The Mediterranean Sea is characterized by an extensive marine traffic because it provides maritime access to the Middle East (and the Suez Canal), the Black Sea and Southern Europe. Much of this traffic is caused by oil tankers. This high level of traffic creates a high risk of pollution and even ecological disaster, made worse by the fact that the Mediterranean is a closed sea, and that the pollutants cannot be diluted over a larger area by ocean currents.
Oil pollution monitoring in the Mediterranean Sea is normally carried out by aircraft or ships. This is expensive, and is constrained by the limited availability of these resources. The satellite imagery can provide a significant contribution in this field, identifying probable spills over very large areas, then guiding aerial surveys for precise observations of specific locations.
In order to provide all-weather and global monitoring of such events, spaceborne Synthetic Aperture Radar (SAR) has been recognized as a cornerstone. This instrument offers the most effective means of monitoring oil pollution. Oil slicks appear as dark patches on SAR images because of the damping effect of the oil film to the sea waves. The sea appears less rough in these areas, and the backscattering is reduced. Hence, such an area would appear darker in a radar image.
Automatic detection of oil spill in SAR images has been a wide field of research in the last years, much effort being dedicated to the classification of oil spill candidates (i.e., dark patches in the SAR image). This thesis work belongs to the DESPOS project (Exploitation of COSMO-SkyMed system for Detection of Ships resPonsible of Oil Spills) that aims to define a technique for the detection of the responsible ships of oil spills exploiting the short revisit time of the COSMO-SkyMed (CSK) constellation to identify these oil spills and to correlate them to the detected ships at the same instant, after and/or before the passages of the CSK satellites, offering a tool to support control activities for inspection and monitoring aimed to identify illegal actions in which oil tankers are involved.
Many ocean surface signatures in SAR imagery are characterized by relatively low normalized radar cross section values. Distinguishing among these signatures objectively can be very difficult, especially with only the single-band and single-polarization SAR imagery. Employing a fractal spectral characterization of the sea surface, in particular by means of the Fractionally Integrated Autoregressive Moving Average (FARIMA) and Fractionally EXPonential (FEXP) models for the Power Spectral Density (PSD) as a wave number function, the proposed technique will be capable of detecting oil spills using conventional and innovative algorithms to recognize and discriminate these one from other phenomena (look-alikes). Only two parameters belonging to the models are necessary for the classification, making the algorithm faster with a very low computational load.
The thesis is organized as follows.
In the Chapter 1 is examined the sea surface, the rough surface scattering and look-alike phenomena.
In the Chapter 2 a short view about the COSMO-SkyMed system is made, in particular are described the instrument functions, the acquisition mode characteristics, the system performances and the images under test.
In the Chapter 3 are described the techniques about speckle noise reduction and the segmentation algorithm that is able to detect all the dark patches in the scene under test. How to recognize and discriminate these dark patches is described in the Chapter 4, in particular are introduced the FARIMA and FEXP model and the classification algorithm, including an analysis about the sea surface spectra, since that is the first step for the application of these models.
In the Chapter 5 are shown the experimental results of the techniques described applied to the images under test.
Eventually are introduced the conclusions and future works.