Laser Method for Oil Pollution Classification on Earth's Surface Using Neural Network Algorithm

• Main Authors: A. D. Shteingart, M. L. Belov, Yu. V. Fedotov, V. A. Gorodnichev, D. L. Gotalskii, O. A. Chernavskaya
• Format: Article
• Language: Russian
• Published: MGTU im. N.È. Baumana 2014-01-01
• Series: Nauka i Obrazovanie
• Subjects: laser; classification; fluorescence spectra; oil pollution; earth's surface; neural net-work algorithm
• Online Access: http://technomag.edu.ru/jour/article/view/684

<p>Today an environmental control of oil pollution of water and terrestrial surfaces in the course of production and transportation of oil and oil products is a challenge.</p><p>One of the remote system options to monitor oil pollution on a terrestrial surface is the laser fluorescent monitoring system of oil pollution from the aircraft. As the fluorescence spectra of oils and oil products differ from the fluorescence spectra of elements of a terrestrial landscape, there is a potential opportunity to find and classify oil pollution through a record and analysis of the fluorescence spectrum form at terrestrial surface site under study.</p><p>The principle of laser fluorimeter operation for monitoring of oil pollution is based on radiation of the studied site of a terrestrial surface by laser in the ultra-violet range and record of fluorescent radiation spectrum (or fluorescent radiation in several narrow spectral ranges).</p><p>The problem of oil pollution classification on the terrestrial surface using a laser fluorimeter with terrestrial surface radiation at the wavelength of 266 nm and record of the laserinduced fluorescent radiation in several spectral ranges was investigated earlier in a number of works. However, presently developed algorithms for classification of oil pollution are heuristic and strictly are unproven.</p><p>The work studies the method capabilities to classify oil pollution on the terrestrial surface using a record of laser radiation intensity in five spectral ranges and a neural network for data processing of measurements. The method allows four group classifications: terrestrial surface (uncontaminated oil products), flood on a terrestrial surface of the light cleared oil products, spill of heavy oil products, and spill of crude oil.</p><p>Work results of the neural network show that to solve a problem of the oil products classification on a terrestrial surface the developed neural network can ensure probability of the correct classification of oil products at least 99,32%.</p>