Online Oil Spill Monitoring Based upon a Shore-Based Hyperspectral Imaging (HIS) System

Oil spills cause serious harm to aquatic ecosystems, so rapid and accurate monitoring is needed to take timely action. Traditional methods usually use satellite remote sensing, buoy sensors, or unmanned aerial vehicles (UAVs). Satellite remote sensing is affected by resolution, transit time, and weather conditions, so it is impossible to monitor oil spills in small-scale and inclement weather conditions continuously and accurately. Buoy monitoring is a point-based measurement, unable to obtain the oil spill distribution, UAVs are limited by endurance time, and cannot conduct long-term monitoring. Here shore-based hyperspectral scanning image technology is proposed, which combines hyperspectral image and remote sensing technology with high resolution and spectral extraction capabilities. This method provides long-term, high-frequency, and real-time monitoring, forecasting, and early warning. A shore-based hyperspectral imaging device was developed to complete the actual field data acquisition and model verification in Three Gorges of China. The results show the precision of oil spill recognition is 84%, and the oil spill recall rate is 89% using the developed weighted CatBoost oil spill classification. The water recognition is close to 100%. The recall rate is also near 100%. The average relative error of oil spill thickness based on the multi-scale continuous wavelet transform model is 23.1%, which shows that the method is effective and can be extended in the oil-spill monitoring for ports, factories, and pipelines.