Prediction of pear sugar content based on near infrared spectroscopy

The purpose of this study is to study the nondestructive detection technology of pear sugar content based on near-infrared spectroscopy, and to determine the effect of combining different eigenvector extraction methods and model building method on the detection accuracy. First, the near-infrared spectroscopic data and the soluble solid content were collected from 150 pear samples by doing the experiments. The NIR spectral data for pear samples ranged from 833 to 2500nm. Sample sets were divided by using the Kennard-Stone method. Then, the characteristic vectors were extracted by the Principal Component Analysis (PCA) method and the Successive Projections Algorithm (SPA), yielding 14 principal components and 13 characteristic wavelengths, respectively. Finally, ELM prediction models were developed based on the full-spectrum, the PCA-extracted principal components, and the feature variables extracted from the SPA, respectively. The comparison shows the most suitable nondestructive detection model for the pear samples sugar content. The results show that the prediction of the SPA-ELM model is optimal when the sugar content of the pear samples was tested by Nondestructive Testing.