摘要
In order to analyze matcha green tea powder grade classification and predict the protein content of matcha with near infrared spectra quickly, 240 samples of 3 matcha brands were collected for Vis/NIR spectroscopy on 325-1075 nm using a field spectroradiometer. The spectral data was processed by chemometrics which was integrated with partial least squares (PLS), principal component analysis (PCA) and back propagation neural network (BPNN) models. 180 samples (60 with each brand) were selected to build training model, the remaining 60 samples (20 with each brand) were applied as the prediction set. Firstly, PLS models were developed with comparison of different spectral pre-processing by smoothing, standard normal variant (SNV), multiplicative scatter correction (MSC), first and second derivative. The best PLS models were obtained by SNV both for the grade classification and protein measurement. Secondly, the selected principal components (PCs) from pre-processed spectra by SNV or original spectra were used as the inputs of back propagation neural networks (BPNN) models. The prediction results showed that PCA-BPNN models with original spectra were better than PLS models. The recognition ratio of 100% was achieved in validation set for matcha samples of three different brands. Moreover, an excellent precision was obtained in validation set for predicting the protein content, resulting in correlation coefficient (R), RMSEP and bias of 0.954, 1.114 and -0.123, respectively. The overall results indicated that Vis/NIR combined with PCA-BPNN was successfully applied for the grade classification and protein content determination in matcha.
摘要文本来自PubMed、OpenAlex或Europe PMC,版权归作者和出版方所有。