Current systems for monitoring leaks from petrol pipes can detect large leaks only, and their sensitivity limit is about 1% of the whole petrol pipe’s capacity. In this paper, a problem of remote detection of small leaks (less than 1%) from petrol pipes was considered. One of possible variations of such a system is a monitoring system of oil pollution at the earth surface along the petrol pipe. In this paper experimentally obtained data such as fluorescence spectra of oil products (crude oil, light-end oil products, heavy oil products), various earth surfaces (soil, vegetation, water, asphalt) and oil products spilled over various earth's surface were used for the excitation wavelength of 266 nm. It was shown that use of the laser method based on detection of fluorescence radiation within three narrow spectral bands and a neural network algorithm of measured data processing allowed one to detect oil pollution on the earth surface with a probability of correct classification close to 1 and low probability of false alarm.