Publications

The oil spill has a detrimental effect on the environment due to its pollution and long-term damage to sea wildlife. As the facility ages, the pipeline leak may increase as integrity reduces due to corrosion or erosion and worsens by minimal maintenance activity. To detect the oil leak, some assessments in the United States statistically found that leak detection system (LDS) effectiveness is less than 20% based on Aloqaily and Arafat (2018). Probably, LDS might not always give a satisfactory result to detect leaks and oil spills and may need to rely on other manual surveillance. Nevertheless, due to limited personnel and the large area of interest, oil spill usually goes undetected until local people and fishermen report it. In an oil spill case, having an early notification is crucial to limiting the leakage and improving mitigation time. To put it in perspective, one of the largest oil spills is the Deepwater Horizon, with an estimation of oil discharged around 4.1 – 4.9 million bbls, and legal fees cost up to 61.6 billion dollars. Looking at this number, we can estimate how important it is to stop oil spills at the very initial of occurrence to minimize environmental damage. This paper aims to exhibit a new approach in oil spill detection using deep convolutional neural networks and transfer learning. We develop an “artificial eye” to automatically classify the surrounding image and identify external manifestations to detect oil spills. We offer a concept upon how we leverage artificial intelligence to automatically classify a stream of the picture, whether it is an oil spill or not. Furthermore, we introduce an IoT and drone technology concept to maximize it to survey the pipeline path regularly. The image captured by these devices is then fed through a deep learning classifier model that decides whether the leak is present or not. By utilizing this technology, we hope to create automatic early notification if leakage occurs so that the oil spill combat team can cure the problem as fast as possible before the leak expands further.