近紅外線光譜分析檢測生乳乳脂含量之研究

Detecting the Content of Fat in Raw Milk Using Near Infrared Spectroscopy

連振昌(Chen-Chang Lien) ； 萬一怒(Ye-Nu Wan) ； 陳煥南(Huan-Nan Chen)

《畜產研究》 35卷2期 (2002/06) Pp. 113-126

摘要

本研究目的在使用近紅外線光譜分析技術，以不同的光譜修正技巧、數學處理方法及統計迴歸，探討量測生乳中乳脂含量的近紅外線光譜檢量線，達到快速及準確檢測生乳中乳脂含量；研究採集泌乳牛各分房前乳，以波長400～2500 nm近紅外線光譜進行掃瞄，由多重線性迴歸得到的檢量線其校正標準偏差SEC為0.0529，決定係數R^2為0.9982，預測標準偏差SEP為0.049；而由修正部分最小平方迴歸得到的檢量線其校正標準偏差SEC為0.0496，決定係數R^2為0.9984，預測標準偏差SEP為0.060；經過光譜修正及微分處理後的近紅外線光譜可得到的較小的校正偏差SEC及較高的決定係數R^2。

關鍵字

近紅外線 ； 生乳 ； 乳脂含量

並列摘要

A near-infrared spectral calibration equation developed from various spectral corrections, mathematical processing, and statistical regression has been proposed as a guidline for rapidly and accurately detecting the content of fat in raw milk. Near infrared of a wavelength ranging from 400 to 2500 nm was used to measure the content of fat in quarter foremilk collected from dairy cows. A calibration equation derived with multiple-linear regression resulted in a standard error calibration of 0.0529 (SEC), a decision coefficient of 0.9982 (R^2) and a standard error prediction of 0.049 (SEP). A calibration equation derived with modified partial least squares regression resulted in a standard error calibration of 0.0496 (SEC) , a decision coefficient of 0.09984 (R^2) and a standard error prediction of 0.060 (SEP). Application of appropriate spectral corrections and mathematical methods can reduce standard error calibration SEC and promote the decision coefficient R^2 significantly. Thus, a promising calibration equation was developed and its performance was validated using expertise inspection results.

並列關鍵字

Near infrared ； Raw milk ； Fat content