Monitoring the drying process of glutinous rice using hyperspectral imaging coupled with multivariate analysis

Hyperspectral imaging (HSI) is an optical technology that harnesses the combination of imaging and spectroscopy techniques for quantifying and qualifying the physical, chemical, and biological properties of materials. The visible-near infrared (Vis-NIR) HSI system with a spectral range of 450–1000 nm was employed in this study to monitor the moisture content of glutinous rice during hot-air drying. The partial least squares method was used for the model calibration based on the full spectral range. The redundant wavelength was removed and the wavelength features that are strongly associated with the moisture content of glutinous rice were chosen using the competitive adaptive reweighted sampling algorithm (CARS). Using the full spectral band, the calibrated partial least square regression (PLSR) model had an accuracy of 0.9106 and 0.8010 for R²C and R²CV with an RMSEC and RMSECV of 1.1446 and 1.7086 respectively. The model prediction performance shows an accuracy (R²P) of 0.9206 with RMSEP of 1.1410. Using CARS algorithm resulted in the selection of 21 optimal wavelengths that are highly associated with the moisture content of glutinous rice during the drying process and the developed PLS model based on the selected wavelength gave an increased model accuracy with R2CV and RMSECV of 0.9176 and 1.0986 respectively. Therefore, the HSI coupled with multivariate analysis (CARS+PLS) is a potential technique for monitoring the drying process of glutinous rice on both small and industrial scales.

Text 113076.docx - Submitted Version Restricted to Repository staff only Download (263kB) | Request a copy

Actions (login required)

View Item