Research On Intelligent Detection Of Ocean Oil Spills Using Spaceborne Synthetic Aperture Rada

Oil spill discharges from societal maritime activities like ships,oil rigs and other structures,leaking pipelines,as well as natural hydrocarbon seepage pose serious threats to marine ecosystems and fisheries.Satellite synthetic aperture radar(SAR)is a unique microwave instrument for marine oil spill monitoring,as it is not dependent on weather or illumination conditions.Existing SAR oil spill detection approaches are limited by algorithm complexity,imbalanced data sets,and uncertainties in selecting optimal features.To this end,we develop an intelligent SAR marine oil spill detection method based on deep learning model,using the Faster Region-based Convolutional Neural Network(Faster R-CNN)for oil spill location detection and the U-Net model for oil spill dark patch segmentation.This approach is capable of achieving oil spill detection and segmentation with reasonable accuracy on an entire SAR image.In order to enhance the contrast of oil spill areas in SAR images for visual interpretation and model training,we designed a pre-processing process,which includes speckle noise suppression,land masking,incidence angle correction and data normalization.A large oil spill detection dataset by combining priori knowledge of experts and environmental ancillary data and optical remote sensing images to accurately label 15,774 oil spill samples from 1,786 C-band Sentinel-1 and RADARSAT-2 vertical polarization SAR images is used to train,validate and test the Faster R-CNN model.3,000 oil spill samples were selected from the oil spill detection data set to train,validate and test the U-Net model.Our experimental results show that the proposed method exhibits good performance for oil spill detection and segmentation with wide swath SAR imagery.For oil spill detection,the precision and recall metrics are 89.23% and 89.14%,respectively,and for oil spill segmentation,the precision and recall metrics are 87.63%and 86.59% respectively.Combining the two models can realize marine oil spill location detection and segmentation on the whole SAR image.In addition,we use quasi-synchronous optical satellite images to validate the oil spill detection results of our proposed method,which demonstrates the reliability of our proposed method.Furthermore,the computer runtime for oil spill detection and segmentation is less than0.4 s for each full SAR image,using a workstation with NVIDIA Ge Force RTX 3090 GPU.This suggests that the present approach has potential for applications that require fast oil spill detection and segmentation from spaceborne SAR images.