Identification Of Grape Fruits Based On Fourier Transform Infrared Spectroscopy

Grape is one of the most important cash crops in Yunnan Province of China, with its sweet taste and high nutritional value are loved by the public. Grape has rich variety and abundant resources; usually it is very difficult to distinguish by their appearance. In this dissertation, Fourier transform infrared(FTIR) spectroscopy combined with statistical analysis was used to identify different grapevine cultivars seedlings.The infrared spectra of grapevine cultivars seedlings band mainly composed of the bands of protein, polysaccharides and cellulose; the infrared spectra of different grapevine cultivars seedlings have little difference. The infrared spectra of five varieties were similar on the whole, only minor difference were exhibited in the range of 1800~750 cm-1. The second derivative spectra of 1800~750 cm-1 were selected to perform principle component analysis and hierarchical cluster analysis; the first three principle components had the cumulative contribution rate of 94.9%, and yield classification accuracy of 100% in the principal component analysis; whereas hierarchical cluster analysis accuracy is 96%, both could identify five different grapevine cultivars seedlings.Fourier transform infrared spectroscopy combined with morlet wavelet principal component analysis and partial least squares discriminant analysis(PLS- DA) were used to study Red Globe and Autumn Royal. The original spectra are very similar,only in the range of 1800~750 cm-1 with tiny differences. The spectra in this range were selected to perform wavelet transform. The decomposition level 7 of morlet wavelet and original spectra were used to perform principal component analysis, and the one-dimensional continuous wavelet transform for the 20-scale was produced. The results were applied to the partial least square discriminate analysis. Results show that the morlet wavelet principal component analysis and PLS-DA can distinguish two kinds of grape, morlet wavelet principal component analysis provides about 100% of classification accuracy; and when the latent variables is nine with the highest accuracy,PLS-DA yields 100% classification accuracy both for Red Globe and Autumn Royal.Our study indicated that FTIR spectroscopy combined with statistical analysis could be used to identify different grapevine cultivars seedlings, can supply a fast and accurate method to discriminate different grapevine cultivars seedlings.