|Abstract||In this work, a method for the quality detection of the in-shell hazelnuts, based on the low field NMR, has been proposed. The aim of the work is to develop an in-line classification system able to detect the hidden defects of the hazelnuts. After an analysis of the hazelnut oil, carried out in order to verify the applicability of the NMR techniques and to determine some configuration parameters, the influence factors that affect these measurements in presence of solid sample instead of liquids have been analyzed. Then, the measurement algorithms were defined.
The proposed classification procedure is based on the CPMG sequence and the analysis of the transverse relaxation decay. The procedure includes three different steps in which different features are detected: moisture content, kernel development and mold development. These quality parameters have been evaluated .analyzing the maximum amplitude and the second echo peak of the CPMG signal, and the T2 distribution of the relaxation decay. In order to assure high repeatability and low execution time, special attention has been put in the definition of the data processing. Finally, the realized measurement system has been characterized in terms of classification performance. In this phase, because of the reduced size of the test sample (especially for the hazelnuts with defects) a resampling method, the bootstrap, was used. [edited by Author]||it_IT