Multiple regression models and Computer Vision Systems to predict antioxidant activity and total phenols in pigmented carrots

The relationships between colour parameters obtained by a Computer Vision System (CVS) and both antioxidant activity (AA) and total phenol contents (TP) on coloured carrots were expressed as multivariate models obtained by multiple linear regression. The AA and TP predicted by the proposed models showed a good correlation with the real AA (R2 = 0.97, P ⩽ 0.001) and TP (R2 = 0.94, P ⩽ 0.001) measurements on the data set including internal and external parts of carrots. The predictions on the data set including only the internal (unevenly pigmented) parts of the carrots exhibited lower determination coefficients (R2 = 0.93 for AA and R2 = 0.86 for TP, P ⩽ 0.001). The effectiveness of the models was checked also on the colour information provided by a colorimeter whose measures proved to be more sensitive to the uneven pigmentation of the carrots. Finally, the proposed models were able to successfully estimate the AA and the TP contents of pigmented carrots when applied to colours measured by the CVS.

► We developed a Computer Vision System for quality evaluation of coloured carrots.
► Predictive models of antioxidant activity and total phenol in carrots were developed.
► Computer Vision System outperforms colorimeter in colour evaluation.
► Colour predicts antioxidant activity and total phenol contents.