螢光光譜影像檢測技術於萵苣受大腸桿菌污染之應用

本研究以螢光光譜影像技術進行受大腸桿菌污染的生食用紅蘿蔓萵苣之檢測與分析，達到快速且非破壞檢測生鮮萵苣受污染之目的。本研究建立之螢光影像系統包括暗箱、影像擷取系統與波長365 nm的UV-A激發光源。研究中以影像擷取程式，自動取得450～700 nm的螢光影像，並撰寫影像處理程式取得相對螢光強度，轉換成螢光強度曲線圖。實驗以馬氏距離計算遭受與未遭受大腸桿菌污染之螢光強度變化，在波長550 nm時最為明顯。以高濃度染菌樣本並挑選波長480 nm、520 nm、540 nm、550 nm的螢光強度資料進行PCA（principal component analysis）分析，其投影的三維空間可將染菌與未染菌之樣本明顯分群。以同筆資料建立SIMCA（soft independent modeling classification analogy）模型，預測結果顯示受大腸桿菌污染之判別正確率達100%。最後將光譜影像資訊與實際污染大腸桿菌之菌數進行MLR（Multiple linear regression）迴歸分析，在搭配波長560 nm、540 nm和550 nm時，r_c可達0.8056，標準誤差SEC=1.959 log CFU，驗證組r_v=0.7369，SEV=2.294 log CFU。結果顯示螢光光譜影像技術對於受大腸桿菌污染的紅蘿蔓萵苣能有效檢驗，可應用於實務確保食品安全。

Keywords

螢光；高光譜螢光影像；大腸桿菌；萵苣；食品安全

Parallel abstracts

This research presents a rapid and nondestructive technique using hyperspectral fluorescence imaging for Escherichia coli contamination inspection on fresh-to-eat red romance lettuces. A hyperspectral fluorescence imaging system including darken chamber, image acquisition module and xenon light source with band-pass filter at 365 nm for excitation light has been successfully developed. The hyperspectral fluorescence images at 450 nm to 700 nm were acquired automatically by the image acquisition program. Afterwards, the image processing program was also developed to obtain good quality of hyperspectral images and convert into relative fluorescence intensity spectra. Fluorescence emission band at 550 nm showed the greatest potential for detection of E. coli contamination by the result of Mahalanobis distance. We found that using the hyperspectral image data in specific wavelengths at 480 nm, 520 nm, 540 nm and 550 nm could improve the distinguishability of PCA (principal component analysis); meanwhile, the contaminated lettuce samples with higher E. coli concentrations could be apparently distinguished. In addition, SIMCA (soft independent modeling classification analogy) analysis was performed with same data setting as PCA. The result of SIMCA indicated the 100% prediction accuracy of E. coli contaminated lettuces. A statistical method of MLR (multiple linear regression) was conducted to build the prediction models for number of E. coli. The best MLR result showed r_c = 0.8056, SEC = 1.959 (log CFU) for calibration, and r_v = 0.7369, SEC = 2.294 (log CFU) for validation. This study demonstrated that hyperspectral fluorescence imaging technique is a potential way for inspecting pathogenic contamination on fresh vegetables to assure food safety.