Use of Pattern Recognition Techniques for Typing and Identification of Oil Spills.

摘要

The infrared spectra of oils have many sharp peaks and good identification results from reading just the amplitudes of these peaks. Indeed, about fifteen properly chosen peaks will usually suffice. The Coast Guard contracted with Rice University to set up identification procedures based on ultraviolet fluorescence spectra of oils, and to determine whether or not techniques used for infrared analysis could be applied to interpretation of fluorescence spectra. The basic result is that simply reading peak amplitudes of fluorescence spectra will not suffice to give good identification. The same conclusion holds if one also reads the amplitudes at the valleys in between the peaks. However, by assigning a sharpness factor to valleys and peaks, one can extract sufficient information to result in very reliable identification. An algorithm for doing this has been devised and implemented in the form of a computer program. A second purpose of the contract was to test the independence of the infrared and fluorescence spectra. A conservative formula is given based on the limited information currently available.