Application of a portable near-infrared spectrometer for rapid, non-destructive evaluation of moisture content in Para rubber timber

Predictive models of the moisture content of Para rubber timber were developed and validated using near-infrared (NIR) spectroscopy. A portable NIR spectrometer connected to a smartphone was developed to facilitate real-time moisture content determination of Para rubber timber. An android application was created to control the NIR spectrometer in acquiring, displaying, and processing the spectral data on the smartphone. Three predictive models for determining the moisture content of Para rubber timber were developed using partial least squares regression in the range 0–89% on a dry basis (%db), < 30% db, and > 30% db. The predictive model for moisture content in the range 0–89%db had a correlation coefficient of cross validation (Rcv) of 0.98 and a root-mean-square error of cross validation (RMSECV) of 5.48%. The predictive model for moisture content in the range less than 30%db had values for Rcv and RMSECV of 0.84 and 2.05% db, respectively. Lastly, the predictive model of moisture content in the range greater than 30% db had values for Rcv and RMSECV of 0.85 and 4.57% db, respectively. The performance of the predictive equation for determining the moisture content in Para rubber timber in the range less than 30% db was validated using unknown samples, which were not used in the model development, assigned as an external prediction set, which resulted in a root-mean-square error of prediction (RMSEP) of 1.44% db. The portable NIR spectrometer connected to a smartphone through an android application is a suitable system for practical application in sawn timber factories because it is cheap and conveniently portable. In addition, the calibration model should be updated and more samples added to cover the variation in growing season and growth area for greater robustness and accuracy of the predictive models.