Support Vector Machines for Classification of Aggregates by Means of IR-Spectra

The increasing physical and technical demands placed on construction materials, especially as they are being used more and more up to the limits of their mechanical strength, has led the aggregates industry to search for more efficient methods of quality control. Information from theoretical work on rock spectra in near-infrared and mid-infrared light as well as achievements gained in signal processing can all be used to improve quality control in an economically acceptable manner. As engineering properties of aggregates are to a great extent determined by the petrological composition of the rock aggregates, the question is, whether a statistical classification rule for identification of rock aggregates can be developed. However, the classification of rocks is complicated by the fact that the optical behavior of minerals forming the rock often appears muted. In addition, minor constituents may dominate the spectrum. Furthermore, the relevant spectra form high dimensional data that are extremely difficult to analyze statistically, especially when curves are very similar. In this paper, support vector machines for classification of rock spectra are investigated, since they are appropriate in classifying highly dimensional data such as spectra.