RESUMEN
Prediction of dry matter content in whole potatoes is a desired capability in the processing industry. Accurate prediction of dry matter content may greatly reduce waste quantities and improve utilization of the raw material through sorting, hence also reducing the processing cost. The following study demonstrates the use of a low resolution, high speed NIR interactance instrument combined with partial least square regression for prediction of dry matter content in whole unpeeled potatoes. Three different measuring configurations were investigated: (1) off-line measurements with contact between the potato and the light collection tube; (2) off-line measurements without contact between the potato and the light collection tube; and (3) on-line measurements of the potatoes. The offline contact measurements gave a prediction performance of R(2)=0.89 and RMSECV=1.19. Similar prediction performance were obtained from the off-line non-contact measurements (R(2)=0.89, RMSECV=1.23). Significantly better (p=0.038) prediction performance (R(2)=0.92, RMSECV=1.06) was obtained with the on-line measuring configuration, thus showing the possibilities of using the instrument for on-line measurements. In addition it was shown that the dry matter distribution across the individual tuber could be predicted by the model obtained.
