Oil Spill Detection Based On Polarimetric Features

In recent years, oil spill accidents on the sea surface take place frequently, which raise more and more public concern. Traditional monitoring means have many disadvantages, such as high cost, limited monitoring scope, easily affected by the weather, etc, making it difficult to meet the requirement of wide range, real-time monitoring. Synthetic aperture radar (SAR) is an active, microwave high-resolution sensor that provides valuable measurements at both day and night. SAR has proven to be a valuable tool for the detection of sea oil spill.Scientists have got some achievements in oil spill detection with single polarized SAR. Compared with single-polarized SAR, multi-polarimetric SAR images contain not only geometrical and backward scattering characteristics of the scattering targets, but also the polarization features of the scattering targets. Therefore, multi-polarimetric SAR images can better reflect the physical property of sea surface targets. The main results of this paper are summarized as follows:1. Two new parameters:P and the single bounce eigenvalue relative difference (SERD) are defined to detect oil slicks on the sea surface. The parameter P is able to reflect the proportions of mirror scattering in the total echo signal, while SERD is related to the roughness of sea surface. For oil-covered sea surface, mirror scattering account for a larger proportion, and the surface is smoother than oil-free surface. Therefore, we can discriminate slicks from sea background based on these two parameters.2. C-band polarimetric SAR performs much better than L-band in discriminating slicks from sea background. A systematic comparison of 10 polarimetric features was conducted using L-band and C-band full polarimetric SIR-C data. The data was acquired during the oil spill experiment in 1994 in the north sea. The experiment results show that:the polarimetric parameters of C band are more sensitive to the roughness of sea surface, and these parameters can also reflect the different scattering mechanism.3. The comparison among ten multi-polarization features were performed based on C-band SAR data. We found that Pedestal height, the copolarization correlation coefficient, conformity coefficient and P performs better than the other six parameters as they are in high slick-to-water contrast. These four parameters can be further used in the oil spill detection.4. An oil spill detection algorithm based on polarimetric features and artificial neural network is developed. Based on the above conclusions, pedestal height, the copolarization correlation coefficient, conformity coefficient, the new parameter P and Normalized Radar Cross Section(NRCS) are selected as criterions, and artificial neural network is used to discriminate oil spills. Experiments are conducted with two SAR images to initially explore the validity of this algorithm for oil spill detection. The experiments results show that the algorithm makes a good performance on discriminating slicks from sea background.