Tecniche analitiche non distruttive e metodiche chemometriche applicate alla valutazione qualitativa di nocciola e castagna
Author(s)
Moscetti, Roberto
Date Issued
April 4, 2013
Type
Doctoral Thesis
Abstract
The purpose of the research was to study the quality characteristics of
hazelnut (Corylus avellana L.) and chestnut (Castanea sativa Miller) fruit, and
the development of a discriminant function able to distinguish fruits by quality
classes. The feasibility of a rapid, on-line detection system was evaluated.
Measurements were performed using a spectrophotometer for Vis/NIR
(Visible/Near Infrared) spectra assays. Data were subjected to chemometric
computations. An iterative Linear Discriminant Analysis algorithm was used to
choose the best combination of pre-treatments and a relatively small set of
wavelengths to correctly classify the samples.
The best result for a properly detection of hazelnut-fruit flaws was equal to
5.2% total error (5.4% false negative, 5.0% false positive). The optimal features
were the wavelengths at 564 nm, 600 nm, 1223 nm, 1283 nm and 1338 nm.
A total error of 8.4% (16.8 false negative, 0.0% false positive) was
obtained for classification of chestnut fruit infested by larvae (Cydia splendana
Hb., Cydia fagiglandana Zel., Pammene fasciana L. e Curculio elephas Gyll.).
The best discriminant performance was computed using the wavelengths at 1582
nm, 1900 nm and 1964 nm.
The results demonstrated the feasibility of Vis/NIR spectroscopy to
correctly classify hazelnut flaws and chestnut affected by infestations, in a rapid
on-line system.
Additional information
Dottorato di ricerca in Meccanica agraria
File(s)![Thumbnail Image]()
Name
rmoscetti_tesid.pdf
Size
3.16 MB
Format
Adobe PDF
Checksum (MD5)
98dd8fba1c53c5208ccd944980691865