Grain sample quality assessment using Intechn and Unscrambler platforms

M. Mladenov, Ts. Draganova, R. Tsenkova

Abstract: A comparative analysis of the results obtained in quality assessment of maize grain samples using the INTECHN and Unscrambler platforms are presented in this paper. The INTECHN platform is developed within the frame of the research project “Development of Intelligent Technologies for Assessment of Quality and Safety of Food Agricultural Products”, founded by the Bulgarian National Science Fund. The sample elements are divided in nine quality groups according to their surface features, which are related to the surface color and surface texture. The assessment of these features is accomplished using an analysis of the object reflectance spectra. Three different INTECHN approaches are applied for feature extraction from spectra and for data dimensionality reduction: principal component analysis and combinations of two kinds of wavelet analyses and principal component analysis. Three classifiers, based on radial basis elements, are used for classification of the maize grain sample elements. The principal component analysis and the three Unscrambler classifiers (Linear discriminant analysis, Soft independent modeling of class analogy and Support vector machine) are used as referent tools. The validation, training and testing errors of the two platforms are evaluated and compared.