A program for recognizing oil spill on the sea surface, based on an artificial intelligence element, was developed and tested on the example of real oil pollution in Peter the Great Bay for use on an unmanned aerial vehicle. The feature of the spectra of broadband radiation ascending from the sea surface is analyzed. It is concluded that the method of recording the spectra of the ascending radiation can be used to detect heavy oil fractions on the sea surface. A software algorithm for the formation of datasets of spectra of induced fluorescence of sea water containing various dissolved grades of petroleum products has been developed and tested. A machine learning procedure has been carried out to create a program element for classifying the type of oil hydrocarbons dissolved in seawater.
Limits of detection of concentrations for dissolved samples of Medium fuel oil and Marine gas oil were measured by two frequency of laser induced fluorescence with 266 and 400 wave length and 100 fs pulses duration. Dynamics of fluorescence spectrums of different types of oil products was investigated.