Terahertz Time-domain Spectroscopy (THz-TDS) for classification of blueberries according to their maturity

Abstract

ABSTRACT Non-destructive determination of blueberry compound using spectral detection method is still a challenge due to the spectral THZ variation caused by abundant biological variations, such as geographic origins and harvest seasons. In order to investigate the potential of Terahertz time-domain spectroscopy to classify fruit maturity states, terahertz spectra (0.5-10 THz) of 4 states of blueberry maturity were examined. The acquired data matrices were submitted to the application of MATLAB 2019b Classification Learner by using 24 classifier models. 84.3 is the highest accuracy, obtained by the Fine Gaussian SVM Algorithm Model with a 0.35 Kernel Scale and a Multiclass Method One vs One. The coefficients for this application of PCA are PC1 (79.9%) and PC2 (20.1%). It was concluded that the combined processing and classification of images obtained from Terahertz time-domain spectroscopy and using Machine learning algorithms can be used to classify the different maturity states of blueberries.