手持式近紅外線光譜儀分析盤固乾草及燕麥乾草成分之研究

芻料品質對於畜牧產品的品質及產量影響甚鉅，而芻料的品質評定包含了外觀、嗜口性及化學成分。傳統的芻料成分需要仰賴相當繁複的實驗室流程去檢測其含量。近年使用近紅外線光分析儀來測定物質中的化學成分越來越普遍，已被廣泛運用在農業產品的品質檢測上，其具有非破壞性及快速檢測的優點。本研究使用近紅外光譜建立台灣盤固乾草及燕麥乾草的化學成分檢量線，並撰寫圖形化使用者介面(GUI, Graphic user interface)提供現場快速檢測牧草成分。本研究的樣本共有70個盤固乾草及80個燕麥乾草，由畜試所恆春分所提供。取樣後分為未處理之段狀樣本及經過打磨作業的粉狀樣本，以探討不同樣本形態的近紅外光光譜表現。每個樣本會進行一次掃描行程，每個掃描行程會對樣本五個不同點取樣得到五條光譜，在分析時所有光譜資料會隨機分為校正組與測試組。本研究使用波長為1600nm-2400nm，解析度為8nm的microPHAZIER手持式光譜儀取得盤固乾草及燕麥乾草之光譜，經前處理後以部分最小平方迴歸(Partial Least Squares Regression, PLSR)及多元線性迴歸(Multiple Linear Regression, MLR)與粗蛋白(CP)、中洗纖維(NDF)、酸洗纖維(ADF)三種由實驗室得到的化學成分含量建立檢量線。結果顯示，段狀盤固乾草CP、NDF、ADF最佳判斷係數R2為0.79、0.51及0.51，RMSEP分別為0.97、1.45及2.20；粉狀盤固乾草CP、NDF、ADF最佳判斷係數R2為0.95、0.77及0.80，RMSEP分別為0.48、1.14、1.58。段狀燕麥乾草CP、NDF、ADF最佳判斷係數R2為0.48、0.31及0.46，RMSEP分別為1.94、3.26及2.54。粉狀燕麥乾草CP、NDF、ADF最佳判斷係數R2為0.94、0.68及0.68，RMSEP分別為0.58、1.21及1.77。

關鍵字

牧草成分；盤固草；燕麥；近紅外線光譜； PLSR ； MLR

並列摘要

Hay quality is an important factor that affects yield and quality of animal products. It also influences its market price. Therefore, it is crucial to have quantitative value on hay quality. Conventionally hay quality is judged manually by its appearance, smell, and chemical composition that needs tedious laboratory analysis. Spectroscopy in near infrared are techniques that have been broadly studied in agricultural field. Spectroscopy contain many advantages, such as, simple preparation for samples, non-destructive on samples, and quick detection. This study used near-infrared spectroscopy to calibrate a chemical composition for Pangola and Oat hay. Seventy Pangola and Eighty Oat hay for NIR analysis. The all Pangola hay samples were randomly collected from Kenting, a southern town in Taiwan. They were sampled by a motor driven drill cutter which made the length of hay straw and leaves in between 1~3 cm. Each sample was scanned by a monochrometer (microPHAZER) to record its reflective spectrum at 1600~2400 nm. Among these sample data, 2/3 of them were used as calibration set while the rest were used as test set. For spectroscopy to predict composition, PLSR (Partial least square regression) and MLR (Multiple linear regression) models were tested to predict CP (Crude protein), NDF (Neutral detergent fiber) and ADF (Acid detergent fiber). At Pangola hay of cut type, results showed the best model had R2 (coefficient of determination) on prediction and measurement on test set of 0.79, 0.50 and 0.52 for CP, NDF and ADF respectively and with RMSEP (root mean squared error of prediction) value of 0.97, 1.63, and 2.20. At Pangola hay of powder type, results showed the best model had R2 on test set of 0.95, 0.77 and 0.80 for CP, NDF and ADF respectively and with RMSEP value of 0.48, 1.14, and 1.58. The oat hay of cut type that had R2 on test set of 0.48, 0.31, and 0.46 for CP, NDF and ADF respectively and with RMSEP value of 1.94, 3.26, and 2.54. At oat hay of powder type that had R2 on test set of 0.94, 0.68, and 0.68 for CP, NDF and ADF respectively and with RMSEP value of 0.58, 1.21, and 1.77.