A COMBINE-MOUNTED NIR SPECTROSCOPY-BASED SENSOR FOR SINGLE RICE KERNEL PROTEIN CONTENT MEASUREMENT

Abstract

It is useful to develop the monitoring technology of spatial variability of the rice quality when a combine harvests within fields. The moisture content and the protein content of kernels are important information for the unhulled rice. In addition, it is necessary to research the nondestructive measurement of single kernels in order to separate the rice depending on the quality of kernel for the future. Therefore near-infrared (NIR) spectroscopy was selected in this study. The NIR transmittance spectra of single kernels of unhulled rice (rough rice) were measured just after the harvest. The moisture content of these kernels was determined by oven method, and the reference protein content was obtained by determination of total nitrogen using a CN coder. The measured NIR transmittance spectra were applied the multiple linear regression, the principle component regression, and the partial least squares regression in order to identify the regression models. As the results, the best coefficient of determination (r2) and the standard error of prediction (SEP) of validation set were r2 = 0.76 and SEP = 2.60 for the moisture content, r2 = 0.22 and SEP = 0.53 for the protein content. To make a kind of possibility for this technique monitoring on combine, more improved calibration model has to be developed.