Soluble solids content and firmness non-destructive inspection and varieties discrimination of apples based on visible near-infrared hyperspectral imaging

Abstract

For apples, soluble solids content (SSC) and firmness are two very important internal quality attributes, which play key roles in postharvest quality classification. Visible-near infrared hyper spectral imaging techniques have potentials for nondestructive inspection of apples’ internal qualities, which will be used for the SSC and firmness non-destructive inspection and varieties discrimination of apples. Spectrums of 396 apples from four varieties were extracted. There were 264 apples used for calibration and the remaining 132 apples for prediction. After collecting hyper spectral images of calibration apples, the sugar meter and firmness tester was used to measure SSC and firmness’s reference values of each calibration apples. Then the principal component analysis (PCA) was used to extract effective wavelengths of calibration apples. The reference values and effective wavelengths’ reflectance of apples can be used to set linear regression models based on partial least squares (PLS). Once the prediction model was established, in order to get the SSC and firmness’s predicted values of apples, it only need achieve the effective wavelengths’ reflectance of apples combing with the program of MATLAB. Finally, the SSC and firmness values of prediction set were used as independent variables of the Linear Discriminant Analysis (LDA) to realize varieties discrimination. The correlation coefficients were 0.9127 for SSC and 0.9608 for firmness values of prediction models. The accuracy of varieties discrimination was 96.97%. The results indicated that the methods to SSC and firmness non-destructive inspection and varieties discrimination of apples based on Vis–NIR hyper spectral imaging was reliable and feasible.