Derivation of a Blueberry Ripeness Index with a View to a Low-Cost, Handheld Optical Sensing Device for Supporting Harvest Decisions

Abstract

Listen

Translate article

Abstract. Berry fruit farming and marketing have widely increased in recent years in marginal areas of Italy, especially in response to consumers’ growing interest in adhering to a bioactive and health-protecting diet. The ripening stage of berry fruits, on which their quality and nutraceutical attributes depend, is typically estimated by growers through visual inspection based on the growers’ experience. Low-cost, handheld, user-friendly devices could assist growers in the field in determining the optimal harvest date in accordance with the desired ripeness of berries. In order to explore the technical feasibility of such a system, this study focused on defining a simple ripeness index for blueberries by identifying the main spectral changes accompanying their ripening process, with special attention paid to the last and most relevant stages. With this aim, spectral measurements in the range of 445 to 970 nm were carried out on Vaccinium corymbosum (cv. Brigitta) during two different growing seasons. Spectra were acquired on 942 individual berries in the field at different dates. Measurements were accompanied by weekly samplings, with picked fruits divided into four ripeness grades according to commercial growers’ classification. A principal component analysis of fruit spectra highlighted that two main wavebands (680 and 740 nm) can maximize the differences between fully ripe samples and those close to ripeness or unripe. Hence, spectral values at 680, 740, and 850 nm (the latter being an additional normalization waveband) were used to create a blueberry ripeness index (BRI) as a linear combination of two spectral ratios: S1 = log(I680/I850) and S2 = (I740/I850). The definition of specific ripeness thresholds for the BRI according to more or less selective criteria was illustrated in an application example, and the ripeness classification capability was then assessed on a separate validation set of 471 berries. When applying a less selective threshold approach, the BRI correctly classified as ripe 85% of manually graded fully ripe berries, whereas 13% of close but not yet ripe validation samples were misclassified as fully ripe and ready to harvest. Comparatively, when a more demanding ripeness threshold was applied, the amount of nearly ripe berries misclassified as ripe decreased to 8%, but the amount of fully ripe berries not identified as ripe rose to 25%. In both cases, none of the unripe validation samples was erroneously classified as ripe fruit. These results, which were obtained with a BRI defined by spectral measurements at just three discrete wavelengths, point to the feasibility of a simple, microcontroller-based, handheld optical device able to implement the BRI to quickly assess the ripeness of sets of berries during the last and most delicate stages of the ripening process.