資料處理對于稻穀食味主要成分之近紅外線校正線之影響

Data Processing Affecting the Nir Calibration Curves of Major Constituents of Rough Rice Taste

李汪盛(Wang-Sheng Li) ； 蕭介宗(Jai-Tsung Shaw)

《農業機械學刊》 5卷4期 (1996/12) Pp. 19-34

https://doi.org/10.30062/JAM.199612.0003

摘要

本研究之目的為探討均勻的樣本挑選、一次差分、二次差分與波長選擇對於整粒稻穀與粉碎稻穀之含水量、蛋白質、直鏈澱粉、脂肪酸度之近紅外線校正線之影響。結果顯示，整粒稻穀與粉碎稻穀含水率及蛋白質近紅外線校正線均以二次差分之校正線較佳。
整粒稻穀含水率校正線所選定之波長為1152、1276、1328、1912 nm，相關係數r為0.992，驗證標準偏差(SEP)為0.346% w.b.。粉碎稻穀含水率校正線所選定之波長為1560、1912、2344 nm，相關係數r為0.981，驗證標準偏差為0.416% w.b.。整粒稻穀蛋白質校正線所選定之波長為1228、1456、1636、169C、1756、2220、2360 nm，相關係數r為0.876，驗證標準偏差為0.665% d.b.。粉碎稻穀蛋白質校正線所選定之波長為1356、1628、2056、2184 nm，相關係數r為0.931，驗證標準偏差為0.396% d.b.。均勻樣本之挑選及配合一次差分或二次差分可以有效提昇含水率及蛋白質校正線之預測準確度。
有關直鏈澱粉及脂肪酸度的校正線，可能因為不均勻顆粒，品種間的變異，加上化學分析之差異等綜合影響，雖用上述相同的資料處理仍未能達到理想的校正線及預測結果，尚待繼續研究。

關鍵字

近紅外線 ； 校正線 ； 稻穀 ； 一次差分 ； 二次差分

並列摘要

The study is to evaluate the effect of data processing including the uniform sample selection, first and second differences, and wavelength selection on the moisture content, protein content, amylose content and fat acidity of whole kernel and powdered rough rice for NIR calibrations. The results show the calibration curves of moisture content and protein content for whole kernel and powdered rough rice will be better with second difference treatment.
For moisture content, the calibration curve of whole kernel rough rice has a 0.992 correlation coefficient with a 0.335%w.b. SEP at 1152, 1276, 1328 and 1912 nm wavelength. For powdered rough rice, the calibration curve of moisture content has a 0.981 correlation coefficient with a 0.416%w.b. SEP at 1560, 1912 and 2344 nm. For protein content, the calibration curve of whole kernel rough rice has a 0.876 correlation coefficient with a 0.665%d.b. SEP at selected wavelengths 1228, 1456, 1636, 1692, 1756, 2220 and 2360 nm, but for powdered rough rice, the correlation coefficient and SEP will be respectively 0.931 and 0.396%d.b. at 1356, 1628, 2056 and 2184 nm wavelength. Therefore, uniform sample selection, first and second differences are able to improve the prediction accuracy of calibration curves of moisture and protein content.
For the calibration of amylose content and fat acidity, some combined factors may affect the correlation coefficient and standard errors of performance. These factors could be a non-uniform particle size in NIR testing, variation in sample species, and poor duplication in chemical analysis. Even after the same data processing mentioned above, the result still remains not quite acceptable. Therefore, further study may be required for better calibration and prediction.

並列關鍵字

Near-infrared ； Calibration curve ； Rough rice ； First difference ； Second difference