Abstract
The oil pollution of seas is increasing, especially in local areas, such as ports, roadsteads of the vessels, and bunkering zones. Today, methods of monitoring seawater are costly and applicable only in the case of big ecology disasters. The development of an operative and reasonable project for monitoring the sea surface for oil slick detection is described in this article using drones equipped with optical sensing and artificial intelligence. The monitoring system is implemented in the form of separate hard and soft frameworks (HSFWs) that combine monitoring methods, hardware, and software. Three frameworks are combined to fulfill the entire monitoring mission. HSFW1 performs the function of autonomous monitoring of thin oil slicks on the sea surface, using computer vision with AI elements for detection, segmentation, and classification of thin slicks. HSFW2 is based on the use of laser-induced fluorescence (LIF) to identify types of oil products that form a slick or that are in a dissolved state, as well as measure their concentration in solution. HSFW3 is designed for autonomous navigation and drone movement control. This article describes AI elements and hardware complexes of the three separate frameworks designed to solve the problems with monitoring slicks of oil products on the sea surface and oil products dissolved in seawater. The results of testing the HSFWs for the detection of pollution caused by marine fuel slicks are described.
Highlights
The volume of sea pollution by oil products as a result of sea transport and the operation of technical facilities is constantly increasing
HSFW1 and HSFW3 were tested in the city
HSFW1 testing was conducted while monitoring actual marine fuel slicks that occurred during bunkering
Summary
The volume of sea pollution by oil products as a result of sea transport and the operation of technical facilities is constantly increasing. In the case of thin slicks formed by different types of fuel, it is important to determine with a high probability that this spot is oil pollution, i.e., that the observed slick is not a biogenic slick or is not caused by hydrophysical processes in the ocean (internal waves, upwelling, etc.) [12,13,14,15] To solve this problem, the deep learning procedure was carried out for real images of marine fuel slicks, which were detected at different times in the coastal zone of Peter the Great Bay. The second stage of adaptation of the computer vision method for monitoring oil slicks on the sea surface is the development of an artificial intelligence (AI) element for autonomous segmentation and classification of slicks.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
