If, on one hand, the visual appreciation of wines color is of the greatest importance, on the other hand, its objective characterisation is essential. For more than a century, many methods have been proposed. The technique we describe here is new under two main aspects:
- first, it allows a quantitative traitment of the whole spectrum, including both major, classical peaks and the residual part (part of the spectrum remaining after elimination of major peaks), unexploitable until now. The spectrum is decomposed in a vector whose the number of scalars components is computed in such a way that the loss of information resulting of this decomposition is determined in advance;
- secondly, the vector representing the whole spectrum is used in an original way for classifications and multivariate comparisons.
We applied this technique to classification of 213 identified wines of Bordeaux area and compared our results to usual methods. Beyond this application, one can see that any wine can be defined, from now, by a vector whose components are of any kind: spectral, chromatographic, analytic, organoleptic. The great number of possible components permits us to eliminate the less discriminant ones and to select, iteratively, the more discriminant ones and to improve, by the way, classifications and comparisons.The measurements and their computerized treatment are so fast that we can speak of on-line results.
AttachmentsNo supporting information for this article