The current study shows that the use of a high speed NIR interactance instrument with a low resolution spectrophotometer measuring 15 wavelengths is a feasible approach for on-line prediction of dry matter content in potatoes. The extraction of spectra from the background could be fully automated, and the dry matter content could be rapidly and accurately predicted on-line for individual potatoes. The model obtained from on-line measurements was significantly better than the model obtained from the off-line non-contact configuration and as good as the ones obtained from measurements in the off-line contact configuration. This was ascribed to a more representative sampling when measuring in the on-line configuration. Even dry matter gradients along the longitudinal axis of individual potatoes are reflected in the NIR spectra. The current measurement configuration only allows for measurement of one row of single potatoes, but it is expected that further instrumental development could increase expected sample throughput. The development of a user friendly instrument with the necessary hardware adapted to suit industrial environments is the natural next step in this development.
Acknowledgment