Nondestructive maturity determination in green tomatoes using a handheld visible and near infrared instrument

Abstract

A nondestructive method based on visible and near infrared spectroscopy, was investigated for determining the maturity of green tomatoes at harvest. The interactance spectrum of tomato fruit at the firm green stage were measured in less than 5s with a handheld spectrometer (400–1000nm) utilizing an optical design with a 0° angle between the incident illumination and the detection view. Results showed that the predominant change in the interactance spectra in green tomatoes useful for predicting changing maturity levels occurred in the 600–750nm portion of the 400–1000nm region, typically associated with chlorophyll. Variety-specific Bayesian classification models and a joint variety ‘global’ classification model were developed to predict tomato maturity after 7d of storage in a 20°C ripening room using canonical discriminant analysis techniques applied to the interactance spectra from 600nm to 750nm. Variety specific models correctly identified 75–85% of immature tomatoes and 82–86% of mature green tomatoes in internal cross-validation, however external validation performance decreased when applied to predict maturities in a cultivar external to the training set. False positive rates of these models in identifying immature and mature green tomatoes varied from 3% to 40% and 0% to 31% respectively. A ‘global’ model, trained on two cultivars showed more stability and correctly predicted 71% of immature and 85% of mature green tomatoes, with false positive error rates of 13% and 22%, respectively, in internal cross-validation of both varieties. This handheld system showed good potential as a rapid, nondestructive technique to aid tomato production managers in the identification of immature green tomatoes at harvest and could be a valuable tool in delivering more flavorful fruit to consumers by reducing the amount of immature fruit harvested by workers.