本研究的目的在探討近紅外線分光光度計偵測稻穀精米程度的可行性。第一部份嘗試以含水率的多寡預測精米程度的大小,第二部份為了消除雜訊的影響,以白米與糙米的近紅外線反射光譜相除以偵測精米程度。利用多重線性迴歸(MLR)與部份最小平方法(PLSR)來建立校正線,希望能迅速且正確的偵測精米程度的大小,以確保食味品質與碾米廠的收益。 實驗結果顯示以含水率為指標預測精米程度並不可行,而以白米與糙米之近紅外線反射光譜相除,利用MLR模式預測時,選取1408、1940、2144、2216與2428nm等五個波長,其相關係數達到0.937,校正標準偏差與預測標準偏差分別為1.155與1.728。以PLSR模式預測時,選取七個因子數,其相關係數達到0.964,校正標準偏差與預測標準偏差分別為1.003與1.245。PLSR模式預測雖然有較高的預測精度,但若以模式的簡單性及分光光度計之成本為考量,則建議採用MLR預測模式。
The object of this research is to determinate the possibility of detecting the degree of milling of rice by a near-infrared spectrophotometer. First the degree of milling was tried to predict by measuring moisture content of rice. Second, the measured spectra of milled rice was devided by the spectra of brown rice to avoid the effect of noise, and to find out the degree of milling. By the multiple linear regression and partial least squares regression to determinate the calibration curve, these provided for detecting the degree of milling rapidly, and ensured rice quality and good milling profit. Experimental Results showed that it was not satisfactory for measuring moisture content to detect degree of milling, but it was very good for using the modified spectra. Five wavelengths(1408, 1940, 2144, 2216 and 2428nm) were selected by MLR for calibration curve with correlation coefficient, the standard error of calibration(SEC) and standard error of prediction(SEP) respectively, 0.937, 1.115 and 1.728. Seven factors were selected by PLSR for calibration with correlation coefficient, the standard error of calibration(SEC) and standard error of prediction(SEP) respectively, 0.964, 1.003 and 1.245. The performance of calibration curve by PLSR was better than that by MLR, but the prediction from MLR equation has the practical application due to the lower cost of NIR spectrophotometer.