Rapid and Non-destructive measurement of moisture content of peanut (Arachis hypogaea L.) kernel using a near-infrared hyperspectral imaging technique

Abstract

This study used a rapid and non-destructive way of determining and predicting the moisture content (MC) of peanut kernels using hyperspectral imaging in the near-infrared region (900–1700 nm). Using partial least square regression (PLSR), spectral data from the peanut kernel hyperspectral images were extracted to predict MC. The MC PLSR model displayed good performance with determination coefficient of calibration (R2c), validation (R2v) and prediction (R2p) of 0.9309, 0.9083 and 0.9368, respectively. Also, the root-mean-square error of calibration (RMSEC), cross-validation (RMSEV), and prediction (RMSEP) of 1.6978, 1.9571, and 1.8715, respectively, were achieved. Optimization was done by selecting wavelengths with the highest absolute weighted regression coefficients; on this basis, 20 significant wavelengths were identified for further analysis. These wavelengths were used to build an optimized regression model which resulted in R2c of 0.9357, R2v of 0.9133, and R2p of 0.9445 as well as RMSEC, RMSEV, and RMSEP of 1.6822, 1.8316 and 1.9519, respectively. The optimized model has applied to the peanut kernel hyperspectral images in a pixel-wise manner obtaining peanut kernel moisture content distribution map. Results show promising potential of the hyperspectral imaging system in the near-infrared region combined with partial least square regression (PLSR) for rapid and non- destructive prediction of moisture content of peanut kernels.