Retrieval of Oil-Water Mixing Ratios in Marine Oil Spills Using a GA-Optimized BP Neural Network

Estimating the oil-water mixing ratio (MR), defined as the proportion of oil in an oil-water mixture, is essential for effective oil spill treatment. This study presents a method for retrieving oil-water MR from fully polarized SAR images using a Genetic Algorithm (GA) optimized Back Propagation (BP) neural network (GA-BP) model. The proposed method utilizes polarimetric features extracted from fully polarized SAR images, specifically the mean scattering angle and the co-polarization power ratio, which exhibit a strong correlation with MR, as input to the model. To enhance the predictive accuracy of the BP neural network, the Genetic Algorithm is employed to optimize BP neural network. Experimental results show that the proposed method achieves high accuracy, highlighting the potential of fully polarized SAR data for MR inversion.