Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/51390
Title: | 螢光光譜影像檢測技術於萵苣受大腸桿菌污染之應用 Detecting Escherichia coli Contamination on Fresh-to-eat Lettuce Using Hyperspectral Fluorescence Imaging |
Authors: | Ruey-Chih Lee 李睿芝 |
Advisor: | 陳世銘(Suming Chen) |
Keyword: | 螢光,螢光光譜影像,大腸桿菌,萵苣,食品安全, Fluorescence,Hyperspectral Fluorescence Imaging,Escherichia coli,Lettuce,Food Safety, |
Publication Year : | 2016 |
Degree: | 碩士 |
Abstract: | 近年來大量食品衛生事件以及因微生物污染而造成食物中毒的事件時有所聞,使消費大眾對於食品安全除了感到擔憂之外,也引發對於食安議題的高度關切。本研究以檢驗食品中一般衛生指標菌之大腸桿菌的污染問題為研究重點。由於傳統的微生物檢測方法操作步驟繁複、檢測時程冗長,無法即時判斷食品的衛生狀況,因此尋找快速微生物檢測之替代方法備受矚目。本研究將結合螢光光譜與影像資訊,應用於受大腸桿菌污染的生食用蔬菜之檢測。不同於螢光光譜之單點測量,光譜影像比前者多出二維的空間資訊,可量測到染菌前後之蔬菜樣本的螢光強度變化,並將人工培養之已知污染濃度與影像上擷取檢測面積換算成實際染菌量,達到一個快速、非破壞性且實用性高的檢測方法。
本研究成功建立高光譜螢光影像系統,進行受大腸桿菌污染的紅蘿蔓萵苣之螢光光譜影像量測與分析。系統硬體包括影像擷取室、影像擷取系統與波長365 nm的UV-A激發光源。分析方法包含螢光光譜影像擷取程式,可自動取得波長450 nm至700 nm的螢光影像並儲存影像,利用自行開發影像處理程式取得相對螢光強度與其空間資訊,並予以轉換成螢光強度曲線圖以利後續分析。 本研究以馬氏距離計算出在波長550 nm時,未受大腸桿菌污染與污染樣本,在此波長下的螢光強度變化最為明顯。在PCA (Principal Component Analysis) 與SIMCA (Soft Independent Modeling of Class Analogy) 定性分析將受大腸桿菌污染與未受污染之萵苣進行分類。選擇高濃度染菌樣本,並僅挑選出波長480 nm、520 nm、540 nm、550 nm的螢光強度資料進行PCA分析,將螢光光譜影像轉換並投影至三維空間可將兩類樣本明顯分群,累計變異量達99.88%。以同筆資料建立SIMCA模型,預測結果顯示受大腸桿菌污染之判別正確率達100%,而未受大腸桿菌污染類別之判別正確率提高至近67%。結果顯示選擇特定的特徵波長能有效提高定性分析之預測能力。有關定量分析方面,將光譜影像資訊與實際污染大腸桿菌之菌數進行PLSR (Partial Least Squares Regression) 和MLR (Multiple Linear Regression) 之迴歸分析。以螢光光譜影像預測大腸桿菌數PLSR校正組相關係數rc=0.8087、標準誤差SEC=1.945,驗證組rv=0.7155、SEV=2.372。MLR在挑選特徵螢光發射光波長為560 nm、540 nm和550 nm時有最佳檢量模式結果,校正組相關係數rc可達0.8056,標準誤差SEC=1.959,驗證組rv=0.7369, SEV=2.294,上述誤差單位皆為log CFU。結果顯示兩種檢量模型相關係數皆在0.7以上,但在感染低濃度大腸桿菌時,因螢光反應微弱使預測能力較為有限。綜合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具有可行性,若未來能將此技術發展成熟並應用於更多蔬菜之檢測,將大幅提升其應用性。 Food safety sensing on fresh-to-eat fruits and vegetables is an important part for agricultural products. In recent years, the number of foodborne illness outbreaks related to fresh agro-products is increasing year by year. In addition, food safety issues have drawn the public great concern, especially contamination of Escherichia coli has the potential risk to cause human infection. However, the procedures for inspecting Escherichia coli contamination using current standard laboratory methods are time-consuming and labor-intensive. These procedures are unable to determine the sanitation status of contaminated food in real time. Therefore, a rapid and non-destructive method for contamination inspection of fresh-to-eat vegetables is necessary for health of human being. This study successfully developed a hyperspectral fluorescence imaging system (HFIS) using xenon light source with band-pass filter at 365 nm for detection of Escherichia coli contaminants on red romaine lettuce. The software developed in this study consisted of HFIS control program and image processing programs; and hyperspectral fluorescence images were enhanced to exhibit fluorescence intensity. The result showed that fluorescence emission band at 550 nm are the best wavelength for the detection of E. coli contamination by calculating the Mahalanobis distance. The Principal Component Analysis (PCA) was performed using the hyperspectral image data in four specific wavelengths at 480 nm, 520 nm, 540 nm and 550 nm, which were determined by Mahalanobis distance and previous study. The contaminated lettuce samples with higher E. coli concentrations could be apparently distinguished by PCA, and the cumulative variance of first three PCs was more than 99%. Besides, Soft Independent Modeling of Class Analogy (SIMCA) analysis was also conducted in this study using the same data setting as PCA. The SIMCA showed that the prediction accuracy of E. coli contaminated lettuces was 100%, while 67% for non-E. coli contaminated samples. The results indicated that prediction ability of qualitative analysis could be improved by selecting only a few specific wavelengths instead of using the entire spectrum range. Regarding quantitative analysis, the prediction models were developed by using Partial Least Squares Regression (PLSR) and Multiple Linear Regression (MLR). The results of best PLSR model showed rc= 0.8087, SEC=1.945 (log CFU) for calibration while rv=0.7155, SEV=2.372 (log CFU) for validation. In addition, MLR analysis which at the fluorescence emission bands 560 nm, 540 nm and 550 nm, gave the results of rc=0.8056, SEC=1.959 (log CFU) for calibration, and rv=0.7369, SEC=2.294 (log CFU) for validation. This research demonstrated the feasibility and potential of implementing fluorescence imaging techniques for detection of E. coli contamination on lettuce vegetable. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/51390 |
Fulltext Rights: | 有償授權 |
Appears in Collections: | 生物機電工程學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-105-1.pdf Restricted Access | 3.55 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.