Publication: Rapid Grape Composition Prediction using Non-Destructive UV/VIS Diffuse Reflectance Spectroscopy
All || By Area || By Year| Title | Rapid Grape Composition Prediction using Non-Destructive UV/VIS Diffuse Reflectance Spectroscopy | Authors/Editors* | Fadock M. N., Reynolds A., Brown R. B. |
|---|---|
| Where published* | ASEV Eastern Conference 2010 |
| How published* | Proceedings |
| Year* | 2010 |
| Volume | |
| Number | |
| Pages | |
| Publisher | |
| Keywords | |
| Link | |
| Abstract |
Inexpensive UV/VIS reflectance spectroscopy can replace NIR transmission spectroscopy for grape composition prediction. This work presents the results of rapid grape composition prediction from a non-destructive UV/VIS diffuse reflectance spectroscopy instrument. Weekly grape samples of Syrah, Cabernet Sauvignon, and Cabernet Franc were collected for 9 weeks in 2008, and 12 weeks in 2009. Analysis consisted of UV/VIS diffuse reflectance spectroscopy followed by the chemical assay of total soluble solids (°Brix), pH, titratable acidity (TA), phenols, and anthocyanins. Partial least squares regression (PLS) prediction models were generated using a variety of preprocessing methods. A second set of models was generated using a genetic algorithm (GA) wrapper for variable selection to identify the most informative wavelengths for PLS regression. All models were compared using root-mean-square error of prediction (RMSEP) and R2 to evaluate the most robust year-to-year calibration. The relationships between PLS loadings, GA selected wavelengths, and the absorption spectra of various assay components are detailed. Preliminary results indicate that UV/VIS spectroscopy is able to quantify BRIX, pH, TA, and anthocyanins in red grapes. Sources of error, improvement in device construction, and its application to real-time harvest monitoring are also discussed. |
Back to page 24 of list