Rapid and Non-destructive Determination of Moisture Content of Peanut Kernels Using Hyperspectral Imaging Technique

Abstract

Moisture content (MC) is a fundamental and very important quality indicator of peanut, which has significant influence on the overall quality of peanut in the process of storage. This study aimed to investigate the potential of hyperspectral imaging technique in the spectral range I (400–1000 nm) and spectral range II (1000–2500 nm) for predicting the MC of peanut kernels non-destructively. Hyperspectral images were obtained, and the corresponding spectral data was extracted. The calibration models were built between the extracted spectral data and the measured MC using partial least squares regression (PLSR) analysis. The established PLSR models using the full wavelengths showed good performance with determination coefficient (R 2 p) of 0.908 and 0.906, and root mean square errors by prediction (RMSEP) of 0.063 and 0.063 %, respectively. Optimal wavelengths were then selected based on the regression coefficients of the established PLSR model. The simplified PLSR models established only using identified optimal wavelengths also showed good performance with R 2 p of 0.910 and 0.900 and RMSEP of 0.061 and 0.060 %, respectively. The best PLSR model, established only using six optimal wavelengths (409, 508, 590, 663, 924, and 974 nm) selected from the spectral range I, was used to shift the spectrum of each pixel into its MC value for visualizing the distribution map of MC in peanut kernels. The results demonstrated that hyperspectral imaging technique in tandem with chemometrics analysis has the potential for rapid and non-destructive prediction of MC in peanut kernels.