Mathematical methods of oil and petroleum products spill identification based on application of unmanned aerial systems

The study provides a theoretical justification of mathematical methods applied to the system intended to identify the accident-prone situations arisen in oil pipelines through monitoring with unmanned aerial systems. As a mathematical algorithm for visual identification, a neural network was used that identified three groups of events: oil and petroleum products spill, presence of smoke and combustion products suspensions, flame combustion. The study reviews the principle used to create learning datasets for each of the three groups, using a map of the area. The discovery process is based on the processing of images on digital photos given the lighting conditions, coordinates of every image, and the flight height of the air vehicle. These restrictions allow reducing the payload weight and using air vehicles with standard equipment. Mathematical processing of an image is through the onboard microprocessor integrated into the unmanned aircraft for real-time monitoring without obligatory transfer of photo materials to the ground control station.