Classification of drug tablets using hyperspectral imaging and wavelength selection with a GAWLS method modified for classification


    loading  Checking for direct PDF access through Ovid

Abstract

Right drug tablets must be brought to the right places. We apply hyperspectral imaging, which can measure infrared spectra at many points on a two-dimensional plane, to classify tablets correctly. The k-nearest neighbor algorithm (kNN) is employed to classify tablets using a database including their spectra and true classes. Although classification accuracy is not 100%, we can correctly classify tablets overall, since spectra at many points are measured with spectroscopy and misclassification at some points does not have much influence on the final tablet classification result. In addition, we propose a wavelength selection method for classification. Genetic algorithm-based wavelength selection is applied to classification and combined with kNN, and thus, not wavelengths but wavelength-regions can be selected in classification problems. Through a case study, we confirmed that the proposed method could classify three kinds of tablets correctly and select appropriate wavelength-regions.

    loading  Loading Related Articles