Abstract
We developed a near-infrared (NIR) sensor mountable on head-feeding combines for measuring rice protein in real time while harvesting. The developed sensor employed reflectance optic instead of transmission one which usually is used since : 1) it operates under heavy duty condition such as vibration and dust, 2) it measures high-moisture, low-fluidity paddy, and 3) low light transmission absorbed by husk was anticipated. The light source was a tungsten halogen lamp, and a diffusion cylinder was installed so that uniform light would illuminate the sample. The detector could measure the spectrum from 740nm to 1140nm with a post-dispersive grating spectrophotometer using a diffraction grating. We made a calibration curve based on brown rice protein from the spectrum of rough rice examined in a laboratory. The calibration curve accuracy was r=0.87 and SECV (Standard Error of Cross-Validation)=0.47%. In the measurement method adopted, the sensor loaded the rough rice into a wide sample chamber by natural gravity and analyzed the loaded grain in the bottom using a reflected near-infrared signal. The developed sensor was able to measure the protein content of brown rice from the spectra of rough rice under severe conditions, such as the high-vibration, high-dust environment of harvesting. In addition to the protein content, the rice weight and moisture content could be displayed on the monitoring terminal in real time. The accuracy of measuring protein content in the field examination was r=0.65 and SEP (Standard Error of Prediction)=0.22%. The SEP was far better than SECV of the calibration ; protein content fell in a narrow range in this field test. So we concluded that the actual accuracy would be the same as the calibration.
