Nondestructive quality assessment of longans using near infrared hyperspectral imaging

Abstract

Near infrared hyperspectral imaging (NIR-HSI) is a method that can be used to evaluate quality of fruit nondestructively. The objective of this research was to study the feasibility of NIR-HSI reflectance mode, within the wavelength of 935-1720 nm, for predicting quality of longans. The two important factors chosen were: total soluble solids (TSS) and moisture content (MC). Each longan was assessed by first measuring its spectral data then measuring its TSS and MC to establish calibration models using multiple linear regression (MLR) compared with partial least squares regression (PLSR). Original spectra of longans gave the optimum results by PLSR for developing the models with correlation coefficients (R_p) of 0.76 for TSS and 0.88 for MC as well as root mean square error of predictions (RMSEP) of 0.42% and 0.45% respectively. By image processing, the predictive images from the models for TSS and MC were created based on color scales. They showed different colors of longans related to the level of TSS and MC and the deviation in levels in different parts of each longan by the predictive image. The results showed it could be used for grading fruit giving NIR-HSI potential to be developed in on-line systems.