<p><strong>Operational oil discharges from ships</strong>, also known as &#8220;bilge dumping,&#8221; have been identified as a major source of petroleum products entering our oceans, cumulatively exceeding the largest oil spills, such as the Exxon Valdez and Deepwater Horizon spills, even when considered over short time spans. However, we still don&#8217;t have a good estimate of</p><ul><li>How much oil is being discharged;</li>
<li>Where the discharge is happening;</li>
<li>Who the responsible vessels are.</li>
</ul><p>This makes it difficult to prevent and effectively respond to oil pollution that can damage our marine and coastal environments and economies that depend on them.</p><p>&#160;</p><p>In this presentation we will share SkyTruth&#8217;s recent work to address these gaps using machine learning tools to detect oil pollution events and identify the responsible vessels when possible. We use a convolutional neural network (CNN) in a ResNet-34 architecture to perform <strong>pixel segmentation</strong> on all incoming <strong>Sentinel-1 synthetic aperture radar</strong> (SAR) imagery to classify slicks. Despite the satellites&#8217; incomplete oceanic coverage, we have been detecting an average of <strong>135 vessel slicks per month</strong>, and have identified several geographic hotspots where oily discharges are occurring regularly. For the images that capture a vessel in the act of discharging oil, we rely on an <strong>Automatic Identification System</strong> (AIS) database to extract details about the ships, including vessel type and flag state. We will share our experience</p><ul><li>Making sufficient training data from inherently sparse satellite image datasets;</li>
<li>Building a computer vision model using PyTorch and fastai;</li>
<li>Fully automating the process in the Amazon Web Services (AWS) cloud.</li>
</ul><p>The application has been running continuously since August 2020, has processed over 380,000 Sentinel-1 images, and has populated a database with more than 1100 high-confidence slicks from vessels. We will be discussing <strong>preliminary results</strong> from this dataset and remaining challenges to be overcome.</p><p>&#160;</p><p>Our objective in making this information and the underlying code, models, and training data <strong>freely available to the public</strong> and governments around the world is to enable public pressure campaigns to improve the prevention of and response to pollution events. Learn more at https://skytruth.org/bilge-dumping/</p>
Terahertz optical machine learning for object recognition
