Non-destructive identification of defects and classification of Hass avocado fruits with the use of a hyperspectral image
Laen...
Kuupäev
2022
Kättesaadav alates
Ajakirja pealkiri
Ajakirja ISSN
Köite pealkiri
Kirjastaja
Estonian University of Life Sciences
Abstrakt
Sensory analysis and instrumental analytical methods are used in determining the
maturity and quality monitoring of avocado fruits, which are labor-intensive and do not allow the
determination of fruit quality in real time. The use of hyperspectral imaging (HSI) methods in the
range of 400–1,000 nm and of the multivariate analysis was demonstrated for a non-destructive
grading of Hass avocado fruits into quality classes according to the number of hidden defects.
Using the sensory analysis, avocado fruits were separated into quality classes according to the
number of defects after being stored for 10 days. Development of a classification model included
several steps: image recording and analysis using the ANOVA and PCA method, image
segmentation (selection of ROI), pre-processing (SNV-correction, centering), selection of a
multivariate classification method (PLS-DA, SIMCA) and a spectral range, model verification.
The analysis of hyperspectral images of avocado fruits has detected spectral regions with the
maximal variance responsible for the change of the content of pigments and moisture within the
avocado fruit exocarp. Comparison of PLS-DA and SIMCA models on the basis of best accuracy
and test-validation results was carried out. Comparison of models showed SIMCA model as the
most efficient model for fruit classification into quality classes depending on the number of
hidden defects. The implementation of the developed approach as a digital avocado fruit sorting
system at different stages of the product life cycle is proposed.
Kirjeldus
Received: January 15th, 2022 ; Accepted: April 19th, 2022 ; Published: May 2nd, 2022 ; Correspondence: Metlenkin.DA@rea.ru
Märksõnad
HSI, sensory analysis, chlorophyll, multivariate analysis, PCA, PLS-DA, SIMCA, articles