Gelatin, a valuable commodity in food processing, pharmaceuticals and photography, is produced by boiling the connective tissues, bones and skins of animals. To be able to predict the quality of the resulting gelatin, a number of parameters, such as percentage of fat, protein, water and mineral content, are measured in the raw bones. We evaluate in this paper whether hyperspectral imaging can perform the required fast and accurate prediction of these parameters based on the spectral response of bone samples. This would allow replacing the time-consuming chemical analysis. The spectral response of nine different bone batches in the 600–1000 nm range (Vis-NIR) is correlated by means of Partial Least Square regression with the measured parameters. Our results show that high prediction accuracy can be obtained for all measured parameters based on the Vis-NIR spectral response. We can then conclude that hyperspectral imaging is a promising metric for the estimation of these chemical characteristics.
Keywords: Vis-NIR spectral response, chemical characteristics, regression