Sistema de clasificación de materias primas mediante espectroscopía óptica de imagen

Abstract

A non-intrusive, non-contact, real-time system for the detection of non desirable material in raw material chains on industry environments is presented. Through a spatial optical spectroscopic technique, scene line spectrographs based on the Visible-Near Infrared (Vis-NIR) reflectance of the material under study, are obtained. To reach a representative spectral fingerprint, the large amount of data is compressed using Principal Component Analysis (PCA) fast algorithm prior to the classification made by a Neural Network. The technique has been successfully checked on the tobacco industry. However, instead of tobacco leaves, other materials can be discriminated or classified using this technique.