近紅外線光譜用於肉品分類與新鮮度之鑑別

Meat Classification and Freshness Determination Using Near Infrared Spectroscopy

周震煌(Chen-Huang Chou)

指導教授 ： 蕭介宗

國立臺灣大學/生物資源暨農學院/生物產業機電工程學研究所/博士(2010年)

https://doi.org/10.6342/NTU.2010.00208

摘要

研究以Bran-Luebbe InfraAlyzer 500光譜分析儀之外接雙光束用來量測絞碎豬里脊肉的蒸餾水萃取液在波長範圍600-1368 nm之吸收值，配合量測之pH值、揮發性鹽基態氮(VBN)和生菌數(APC)三個新鮮度指標來建立迴歸模式。120個絞碎豬小里脊肉樣本儲存在7oC溫度共5天，部份最小平方回歸(PLSR)模式所預測的VBN值與實際量測的VBN值的相關係數為0.83，而對200組絞碎小里脊肉樣本的新鮮度判定，平均鑑別率是90.5%。量測45個黑毛豬小里脊肉、30個白毛豬小里脊肉和30個牛肉里脊肉的蒸餾水萃取液樣本，選取在652-668 nm及804-836 nm (每4 nm取一個)和1320 nm共15個光譜值，以主成分分析(PCA)所建立的分類校正模式，在5%顯著性水準(p<0.05)，分辨25個黑毛豬、15個白毛豬和15個牛肉共有55個驗證組樣本，鑑別率為83.8%。在雞腿距離骨柄5 cm處，另用FOSS NIRSystem 6500近紅外線光譜儀的外接光纖探針量測45國產冷藏與45進口冷凍再解凍雞腿之光譜吸收值(波長範圍400-2200nm)，以Unscrambler軟體的PLSR法，搭配虛擬回歸技術(DRT)，在不同波段、分類規範與各種變數量建立9種校正模式，國產冷藏與進口冷凍再解凍雞腿之平均鑑別率從76.7% 到93.3%。

關鍵字

可見光/近紅外線光譜儀 ； 萃取液 ； 新鮮度 ； 分類

並列摘要

A dual-beam Bran-Lubbe InfraAlyzer 500 spectrophotometer was used to measure the absorbance of distilled water filtrate of 120 ground pork tenderloin samples at wavelength 600-1368 nm region, and to construct three fresh prediction models of pH values, volatile basic nitrogen (VBN) values and aerobic plate counts (APC) during 5 days storage at 7oC. The correlation coefficient of the measured and the predicted VBN values based on the partial least square regression (PLSR) model was 0.83. Using 200 ground tenderloin samples for validation, the average accuracy was 90.5%. The absorbance of 15 selected wavelengths from sample extracts of 45 black-hair hogs, 30 white-hair hogs and 30 beefs in the range of 652-668 nm and 804-836 nm (with 4 nm interval) and 1320 nm, the developed principal component analysis (PCA) calibration model was used to validate samples of 25 black-hair hogs, 15 white-hair hogs and 15 beefs, and had 83.6% classification rate at 5% significance level. From 400 to 2200 nm with 2 nm interval, an extension fiber optic probe of FOSS NIRSystem 6500 was used to measure the absorbance of chicken thighs at 5 cm from the end of Tibia bone. The measured data were used to establish 9 calibration models by PLSR of Unscrambler software with dummy regression techniques under different wavelength ranges and criteria with different variables. Using the established models for distinguishing the local chilled storage from the imported frozen-then-thaw chicken thighs, the average discrimination rate varied from 76.7% to 93.3%.

並列關鍵字

Visible/Near-infrared spectrophotometer ； Extracts ； Freshness ； Classification

參考文獻

7. 周震煌、周瑞仁、蕭介宗。2007。以可見光/近紅外線光譜值鑑別國產冷藏與進口冷凍再解凍雞腿。農業機械學刊16(2): 1-11。

6. 周震煌、劉昌群、蕭介宗。2005。以機械視覺與可見光/近紅外光應用於畜肉之分類。農業機械學刊14(3):39-49。

16. 劉昌群、蕭介宗、彭敬益、洪梅珠、沈明來。2005。近紅外線光譜值的波長選擇對水稻品種鑑別的影響。農業機械學刊14(2):27-38。

14. 蕭介宗、駱秋英、陳力騏、劉昌群、洪淑玲、蘇淑禎。2004。豬肉鮮度指標與近紅外線光譜值相關性之探討。農業機械學刊13(4):27-36。

11. 張鴻文、蕭介宗、李盛銘、洪梅珠。2000。以近紅外線及影像技術鑑別水稻品種。農業機械學刊9(4): 1-16。