Classification and Prediction by LF NMR

The classification analysis and firmness prediction of sweet corn subjected to different treatments were investigated to develop the potential application of low field nuclear magnetic resonance for food quality control. Principal component analysis (PCA) was firstly used for exploiting the invisible changes of the internal characteristics of blanched sweet corn. Then, two classification methods involving regression analysis in combination with linear discriminant analysis (LDA), which is based on data compression technology (PCA and partial least squares, PLS), were compared. The PCA score plots clearly showed that the samples varied according to the temperature of the treatment. For the goal of classification, the principal components extracted from the PLS analysis were more useful than those obtained from PCA. Cross validation was helpful to determine the appropriate number of principal components. In fact, it proved that PLS combined with LDA yielded the highest success classification rate (94.3%), and PLS also performed well in firmness prediction of processed sweet corn.