Oil spills identification in SAR image based on Convolutional Neural Network

Synthetic Aperture Radar (SAR) satellites can be used to detect the oil film diffusion caused by marine oil spill accidents. The automatic identification of oil slicks and lookalikes oil slicks can provide an important basis for decision-making of oil spill accidents. An oil spill SAR image recognition method based on convolutional neural network is presented in this paper, which automatically extracts category features and avoids the non-standardity of manual extraction methods. The original oil spill SAR image is filtered and denoised, before it is input into the CNN network, and the feature extraction is performed on the SAR image using the CNN model. Finally, the features are classified by Soft-max. Experiments have been carried out using ERS-2 SAR image data, and the results of the identification demonstrate that the proposed method has high accuracy in identifying "oil slicks" and "look-alikes oil slicks" images.