應用近紅外光技術檢測單粒白米含水率

Abstract

本研究之目的為建立白米內部品質在近紅外光譜短波段(750-1100nm)之檢測模式，並探討應用於快速檢測動態單粒白米之內部品質之可行性。本研究之白米內部品質以水分為指標。 本研究以連續波段之近紅外光譜儀進行靜態單粒白米含水率之檢測，並應用小型光纖光譜儀(S1024DW，Ocean)和實驗室型分光光譜儀(FOSS NIRS 6500)兩種儀器進行光譜量測工作。小型光纖光譜儀反射式實驗中(642-1217nm)，利用部份最小平方差迴歸(PLSR)所得之迴歸式之判定係數R2 = 0.966，標準校正誤差(SEC) = 0.899。實驗室型分光光譜儀反射式實驗中(全波段400-2500nm)，以原始光譜進行部分最小平方差迴歸(PLSR)所得到之檢量線判定係數R2 = 0.981，標準校正誤差(SEC) = 0.543。小型光纖光譜儀穿透式實驗中(部份波段642-1217nm)以部分最小平方差迴歸(PLSR)所得到之含水率檢量線判定係數R2 = 0.968，標準校正誤差(SEC) = 0.878。 單粒白米動態檢測系統係應用近紅外光多光譜影像系統在910、960和1014nmn三個波長之組合來檢測移動中單粒白米樣本之含水率。本系統以部分最小平方差迴歸(PLSR)所得之含水率檢量線之判定係數R2 = 0.651，標準校正誤差(SEC) = 2.518。 應用近紅外光光譜預測單粒白米含水率之效果以靜態方式以全波段反射式光纖所得到之效果為較佳。近紅外光多光譜影像系統在動態檢測含水率之結果還有改善空間，但確具有運用於生產線上檢測動態單粒白米內部品質之可行性。

The objective of this study was to evaluate the feasibility of applying near infrared technology for establishing regression model to detect internal qualities of single rice kernels. The short near infrared wavelength range from 750 nm to 1100 nm was used to develop an on-line moisture content detecting system for single rice kernels. NIR spectrophotometers of S1024DW (Ocean Optics) and FOSS NIRS 6500 (Foss NIRSystems) with specially designed fiber-optics were used to measure the NIR spectra of stationary rice kernels. The partial least squares regression (PLSR) was adopted for predicting rice moisture content by analyzing reflectance spectra and transmittance spectra. The coefficient of determination (R2) and standard error of calibration (SEC) for S1024DW reflectance spectra, FOSS NIRS 6500 reflectance spectra and S1024DW transmittance spectra were respectively 0.966 & 0.899, 0.981& 0.543, and 0.968 & 0.878. The near infrared multi-spectral imaging system with detection wavelength range at 910 nm, 960 nm and 1014 nm was adopted for on-line moisture content measurement of running single rice kernels. The coefficient of determination (R2) and standard error of calibration (SEC) obtained by PLSR were respectively 0.651 and 2.518. The predictions of moisture content of stationary single rice kernels are better in the system as the reflection spectra were adopted instead of transmittance spectra at near infrared wavelength range from 750nm to 2500nm. Further improvement of the near infrared multi-spectral imaging system developed in this study is required for future implementing this system in the on-line rice milling industry.