Oil Spill Detection With Polarimetric Features Based On Optimized Wavelet Neural Network

In recent years,offshore oil transportation and mining are more frequent,due to the increasing demands for oil resources.The frequent occurrence of marine oil spill accidents has seriously affected the marine ecological environment,and restricted the development of marine economy.Therefore,oil spill detection is of great significance for marine eco-environmental protection,offshore oil and gas resources development and marine economic development.For oil spill detection,the traditional methods have large limitations,such as high cost,small range and easily affected by the weather.Polarimetric Synthetic Aperture Radar system is a valuable sensor for oil spill detection in the field of remote sensing,because it is an active,highresolution microwave sensor that can work independent on the weather.Since 2006,deep learning has opened a new chapter,and has attracted great attention in the field of image processing.Deep learning has become a hot research topic.In this paper,sevral polarimetric SAR features were investigated and used for the detection of marine oil spill.The deep learning was introduced to improve the wavelet neural network.The main achievements are as follows:(1)Through optimizing the initial value of wavelet neural network,the network was largely improved.Two Radarsat-2 oil spill datasets were hired to verify the proposed method.Experimental results show that the optimized neural network is able to enhance the oilspill detection accuracy.(2)Through the study of the wavelet neural network and Sparse Autoencoders network of deep learning,a new deep wavelet neural network was proposed based on the above two architectures in this study.Multilayer SAE network was used to build deep wavelet neural network.Error back propagation theory was employed to train the network.Experimental results show that the network can effectively detect the marine oil spill from the water.