NIRS Identification of Maturity of Huanghua Pears with Sparse Principal Component Analysis

Five Huanghua pears were collected from each of 12 pear trees in a same fruit yard in each of 4 definite dates in August, giving totally 240 pear samples. Spectral data was collected with each sample. Characteristic informations related to maturity in spectra were extracted by SPCA and PCA, and explanations were made. Discriminant models for maturity of the Huanghua pears were built by SPCA and PCA in combination with artificial neural network (ANN). As shown in the loading vector diagrams, it was found that:① SPCA was effective to extract specific characteristics related to maturity from the spectra. The seven sparse principle components were found to reflect separately informations about saccharides content, moisture content, pigmentation, hardness and so on of pears; ② Total recognition in prediction of maturity by models built by SPCA-ANN were attained to 93.33%, which is higher than 91.67% the total recognition attained by PCA-ANN.