群知能と進化的計算の国際ジャーナル

概要

Effective Oil Spill Monitoring Approach over the Gulf of Oman by Using Advanced Machine Learning and Data Mining Tools

Aishah Saeed Jumah Alabdouli, Muhammed Sirajul Huda Kalathingal, Shaher Bano Mirza*, Fouad Lamghari Ridouane

Oil spills negatively impact the environment by endangering marine ecosystems, and coastal surroundings. The environmental damage from an oil spill by a tanker, pipeline, or offshore rig can be devastating almost immediately and can last for decades. Consequently, the purpose of this study is to detect oil spills in the Gulf of Oman. To find oil spills, Sentinel-2 spectral imageries are used. Sentinel-2 divides the image into N grids and uses Sentinel-2 band ratio for mapping oil spills to execute instance segmentation using a Yolov7 to accomplish oil spill detection in a single step. In our experiment, the trained Yolov7 instant segmentation model was able to produce exceptionally accurate intersection over union results, correctly identifying 91% of the actual oil spill. These results explain the potential of artificial intelligence and significant impact that can be acquired on environment.