Oil Spills Detection In SAR Images Using Nonlinear Fuzzy Filter

Oils spills broach high degree of pollution into the “blue” bodies which are considered fatal for the water ecosystem. So these oil spills need to be spotted at right time to prevent this disaster pursue. Many techniques are very actively inculcated for the same. Synchronous Aperture Radars (SAR) which is a space borne technique is primarily used for this purpose. Techniques which were used a way back beard its own hiders as follows: (1) the distinguishion between the look-alikes and the oil spills did not meet the satisfying accuracy, (2) the desired precision of clarity in the images were not obtained, (3) the oil territory were not detected to a accurate topology. So considering into the hurdles faced by the previously used techniques, we propose a novel system based on a fuzzy control filtering approach. It uses adaptively varying membership functions and incorporating fuzzy associative memory (FAM) with conventional multilevel median filter (MLMF) to detect the oil spills in SAR images. It also preserves object boundaries and structures, while removing noise effectively in the region of heterogeneous physical properties. This is an attempt to enhance spatial resolution and sensitivity of SAR images for better visualization and analysis. The system minimises the output mean squared error by tuning the shape of the membership function. A parabolic membership function is used, for the first time, to adaptively fine tune the reduction of noise level in the tomograms. The performance of the system is tested using oil spill SAR images. The system restores images corrupted with speckle noises of different levels. High impulse noise is effectively eliminated without significant loss in the sharpness of the image features. System performance is evaluated visually as well as by computing quantitative metrics such as standard deviation error (SDE), root mean square error (RMSE), normalized mean square error (NMSE) and peak signal to noise ratio (PSNR). Numerical measures show fuzzy filters to outperform the convincing performance that is superior to the conventional MLMF method. Among the two membership functions, the parabolic funcion is found to be more effective in noise removal.