Under CopyrightGruna, R.R.GrunaVieth, K.-U.K.-U.ViethMichelsburg, M.M.MichelsburgPuente Leon, F.F.Puente Leon2022-03-112.2.20112010https://publica.fraunhofer.de/handle/publica/36817110.24406/publica-fhg-368171Perfect quality control of food becomes more and more a matter of course. Foreign materialslike pieces of rocks, insects, plastics, carton etc. have to be identified and sorted out. For many years, near-infrared spectroscopy has been a widely-used analytical offline method for quality inspection for small samples in the food and agriculture industry. Up to now, high-speed automatic sorting machines work usually on the basis of visual inspection, not near-infrared inspection. This article describes how an in-line food sorting system can be developed on the basis of hyperspectral imaging data. We focus on analyzing the vast amount of data to yield minimum band selection with optimal classification results required for an industrial sorter.ennear-infrared hyperspectral imagingfood sortingquality inspectionfeature selectionband selectionGermanykcm004670conference paper