Abstract

Rapid, non-destructive fruit sorting techniques are increasingly being adopted to ensure that producers, industry, and consumers receive products that meet their quality requirements. Quality attributes typically used to assess fruit ripeness include soluble solids content (SSC) and flesh firmness (FF). In this study, hyperspectral imaging operating at 400–1000 nm (Vis/NIR) was adopted to evaluate the ripeness degree of ‘Hayward’ kiwifruit. Partial least square (PLS) regression models were developed to estimate SSC and FF, while two different types of PLS discriminant analysis (PLS-DA) were used to classify samples according to three repining classes (defined on the base of SCC and FF values). To reduce the computation complexity, and simplify the calibration models, two variable selection methods (genetic algorithm GA, and variable importance in projection VIP) were adopted. For SSC, the prediction R 2 values ranged from 0.85 (RMSE = 1.10 °Brix) to 0.94 (RMSE = 0.73 °Brix), and for FF from 0.82 (RMSE = 14.51 N) to 0.92 (RMSE = 9.87 N). Classification sensitivity reached values of 97% and 93%, for the model considering the SCC and FF classes, respectively. Prediction and classification performances remained substantially unchanged by reducing the number of wavelengths. Therefore, hyperspectral imaging appears to be suitable for prediction of kiwi quality attributes and their classification. • Hyperspectral imaging to assess kiwi ripening stage rapidly and non-destructively. • GA and VIP methods to select the variables (wavelengths). • PLS regression models to predict quality attributes SSC and FF. • PLS-DA and soft PLS-DA to classify fruit by predefined SSC and FF classes.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call