A SAR Oil Spill Image Recognition Method Based on DenseNet Convolutional Neural Network

Synthetic Aperture Radar (SAR) satellites can monitor oil spills at sea. Due to the small difference of SAR image features between oil film and oil-like film, a dense connection network model based on DenseNet convolution neural network is proposed in this paper. The model extracts multi-scale features of the image, and improves the ability of capturing subtle features of the image and the accuracy of classification and recognition before reusing the features of each convolution layer. After filtering and denoising, the original oil spill SAR image is input into CNN network. Then the feature of SAR image is extracted by using CNN model. Finally, the feature is classified by using Soft-max. Experiments using ERS-2 SAR image data prove the effectiveness of this method in identifying "oil slicks" and "oillike slicks " images.