Hyperspectral image analysis is a fast and non-destructive technique that is being used to measure quality attributes of food products. This research investigated the feasibility of predicting internal quality attributes, such as Total Soluble Solids (TSS), pH, Titratable Acidity (TA), and maturity index (TSS/TA); and external quality attributes such as color components (L*, a*, b*) as well as Color Index (CI) of Valencia orange fruit using hyperspectral reflectance imaging in the range of 400-1000 nm.
Oranges were scanned by the system in order to build full models for predicting quality attributes using partial least squares regression. Optimal wavelengths were identified using the regression coefficients from full models, which were used to build simplified models by multiple linear regression. The coefficient of determination of prediction (R2p) and the Standard Error of Prediction (SEP) were used to measure the performance of the models obtained.
Full models for internal quality attributes had low performance (R2p<0.3, SEP>50%). Full models for external quality attributes presented a high performance for L* (R2p=0.898, SEP=19%), a* (R2p=0.952, SEP=13%), b* (R2p=0.922, SEP=20%), and CI (R2p=0.972, SEP=12%). The simplified models presented similar performance to those obtained for external quality attributes.
Hyperspectral reflectance imaging has potential for predicting color of oranges in an objective and noncontact way.