Fast ingredient quantification in multigrain flour mixes using hyperspectral imaging.

In the past decade, hyperspectral imaging has emerged and matured into one of the most powerful and rapidly growing methods for non-destructive food quality analysis and control. This paper demonstrates fast and automatic quantification of food ingredients with hyperspectral imaging in the visible and near-infrared range in combination with chemometrics and image processing techniques. The application of ingredient quantification in bread flour recipes requires high spatial resolution in addition to spectral discrimination power. Our results show that automatic and accurate quantification of all ingredients can be done, reaching pixel discrimination accuracies above 90% and ingredient quantification errors within the required 1% absolute error in weight. The classification accuracy obtained using 15 wavebands on the test images is around 15% higher than what was obtained with colour imaging. • Accurate ingredient discrimination (>90%) with VNIR hyperspectral imaging. • Accurate seed count based on spatial post-processing of classified image. • Accurate quantification of weight in non-seed ingredients. • Increased discrimination accuracy versus color imaging. [ABSTRACT FROM AUTHOR]