| Sumario: | Food fraud is a serious concern for the food industry and consumers. A common
fraud is mixing fruit cultivars with similar appearance but significant differences in quality and
sensory characteristics, and, hence, different prices. Detecting these abnormalities by visual
inspection is challenging when all fruits appear similarly. It is then required tools capable of
separating the fruits by detecting some internal properties, such as those based on spectral
information.
In this study, two loquat cultivars were used: ‘Algerie’, a traditional sweet cultivar, and
‘Xirlero’, a cultivar with good production but slightly astringent. Both are harvested during the
same period and have similar external features but differ in sensory characteristics and price.
Samples corresponding to 300 ‘Xirlero’ and 259 ‘Algerie’ loquats were selected. Hyperspectral
images were acquired in the range 450 – 1000 nm, and the mean spectra of each loquat were
extracted. The spectra collected were divided into a training set (70%) and an independent test
set (30%). Three models were built to classify the two varieties: Partial Least Squares Discriminant
Analysis, Support Vector Machine, and Extra Trees Classifier. All models achieved accuracy
above 85%, indicating that hyperspectral imaging is a promising technology for distinguishing
between very similar cultivars of fruits.
|
|---|