Context and objectives
In the last decade, intensive research efforts based on remote sensing methods have been dedicated to oil spills in marine environments. Various sensors, methods and spectral characteristics have been deemed to be appropriate for sensing oil pollution. However, the vast majority of scientific papers published on this topic refer to accidents happening in open sea, either at oil extraction platforms or transportation tankers. Oil spills in ports are particular cases of oil pollution in water environments that call for specific monitoring measures. Apart from the ecological threats they pose, the proximity to human activities and the financial loss induced by disturbed port activities add to the need for immediate action. Furthermore, this type of spills, although happening with much higher frequency than the ones at open sea, has not been thoroughly investigated in the literature. The SWIPE project has been dedicated to identify sensors and methods that could be safely employed for oil spill detection in port environments on mobile platforms such as drones and ships.
Project outcome
The SWIPE activities started by exploring established methods and data types in order to test their applicability to port environments. The oil spills happening in ports have various characteristics that differentiate them from the ones at open sea: 1) the spilled oil is almost always processed, not crude oil; 2) the oil layer is thin, unlike the crude oil layers spreading around exploitation platforms when accidents occur; 3) the ports are crowded and animated environments where the presence of objects and the human activities impede the usage of low resolution (satellite) imagery. Following the initial tests and given these specific traits, it was concluded that SWIR-based solutions have limited applicability in ports, and the detection or monitoring should be performed at local scale based on mobile platforms (UAVs or ships) carrying lightweight sensors such that potential accidents do not endanger people present in the area and port infrastructure.
A physical model has been proposed to estimate the oil thickness and volume based on optical reflectance. It has been determined that the visible spectral region can be used to this goal when the oil thickness is larger than 200µm. The method has been validated on RGB images and attained high accuracy for hydraulic oil, but lower for other oil types. A custom deep learning model has been also built to infer the extent of oil spills. The model showed good results for hydraulic, lubrication and fuel oils when applied to RGB videos. However, for oil thickness lower than 500µm, the model does not achieve satisfactory results. Many times, the spilled oil in port environments has much lower thickness. Therefore, a new direction has been explored: the use of ultraviolet (UV) cameras.
UV data were indicated by available literature as useful when investigating thin oil layers. However, they were not yet analysed in the specific port conditions. Moreover, most available solutions use active systems based on UV lasers that first excite the oil compounds, then sense the response in the visible spectral range. The active systems are not ideal in port environments due to their size and weight. SWIPE focused, thus, on investigating the feasibility of lightweight passive UV sensors to distinguish between oil and water areas. An UV camera sensitive at wavelengths lower than 400nm has been employed in a two-fold approach: i) controlled experiments at VITO premises, and ii) real data acquisition from a vessel active in oil spill cleaning at the port of Antwerp. In both cases, comparisons to RGB cameras were performed.
The controlled experiments revealed that the UV cameras have better oil-water discrimination capabilities than the RGB cameras in different scenarios (different oil types, viewing angles, and atmospheric conditions). Moreover, the UV images are much less affected by bottom effects and they are able to observe the oil spills from any viewing angle, different from the RGB cameras whose capabilities proved to be strongly dependent on the specific data acquisition conditions. After the completion of the controlled experiments, the camera has been mounted on the vessel PROGRESS (owned and operated by Brabo Cleaning Company) at the Port of Antwerp-Bruges and it acquired valuable images over the course of three months (March-May 2024). On the one hand, the acquired data allowed for defining an automated procedure to adapt camera settings to the environmental conditions and, on the other hand, confirmed the superior capability of the UV camera to distinguish oil and water areas, in comparison to the RGB camera used for comparison. SWIPE was the first project to successfully prove the feasibility of employing lightweight passive UV cameras for oil spill detection in port environments.
Societal and environmental relevance
The advancement of oil spill detection and monitoring methods in port environments has positive implications in multiple directions: financially – by reducing the intervention times and consequently the financial impact of the disruption in the port activities; environmentally – by limiting the harmful impact of the spill on the ecosystem; societally – by reducing the exposure of humans present at the port to harmful oil compounds that were shown to affect human health over time.
Outreach
SWIPE: Detecting oil spills in ports
SWIPE: Détecter les rejets d’hydrocarbures dans les zones portuaires
Project leader(s): | port of Antwerp-Bruges | |||||||
Belgian partner(s) |
|
|||||||
Location: |
Country:
Region:
|
|||||||
Related presentations: | ||||||||
Website: |