Industrial defect discrimination applying infrared imaging spectroscopy and artificial neural networks
EstadísticasView Usage Statistics
Full recordShow full item record
AuthorGarcía Allende, Pilar Beatriz; Conde Portilla, Olga María; Madruga Saavedra, Francisco Javier; Cubillas de Cos, Ana María; López Higuera, José Miguel
A non-intrusive infrared sensor for the detection of spurious elements in an industrial raw material chain has been developed. The system is an extension to the whole near infrared range of the spectrum of a previously designed system based on the Vis-NIR range (400 - 1000 nm). It incorporates a hyperspectral imaging spectrograph able to register simultaneously the NIR reflected spectrum of the material under study along all the points of an image line. The working material has been different tobacco leaf blends mixed with typical spurious elements of this field such as plastics, cardboards, etc. Spurious elements are discriminated automatically by an artificial neural network able to perform the classification with a high degree of accuracy. Due to the high amount of information involved in the process, Principal Component Analysis is first applied to perform data redundancy removal. By means of the extension to the whole NIR range of the spectrum, from 1000 to 2400 nm, the characterization of the material under test is highly improved. The developed technique could be applied to the classification and discrimination of other materials, and, as a consequence of its non-contact operation it is particularly suitable for food quality control.
Enlace a la publicación
Collections to which it belong
- D50 Congresos