【作者】 李桂峰；

【导师】 刘兴华；

【作者基本信息】 西北农林科技大学 ， 食品科学， 2008， 博士

【摘要】 苹果是世界四大水果之一，我国是世界第一大苹果生产国，2006年生产总量为2530万吨左右，但年出口量仅占生产总量的1．5%左右，这与苹果第一生产大国的地位极不相称。制约我国苹果出口的主要因素是果实品质差、产业化水平低、检测技术落后和评价标准没有与国际接轨。目前，苹果贮藏中果肉褐变发生非常严重，褐变检测存在破坏样品、操作复杂、耗时长、成本高和无法实现在线检测等不足。因此，以研究苹果果肉褐变机理为基础，探索基于近红外光谱快速无损检测果肉褐变方法，并建立性能稳定的数学模型，是一项具有重要理论意义和实际应用价值的工作。本文以红富士苹果为研究对象，以褐变的发生机理和无损检测技术为研究目标，分别从苹果的果肉褐变机理、近红外光谱响应特性、褐变度无损检测影响因素、近红外光谱检测匹配参数等方面进行系统研究，并在此基础上建立苹果果肉褐变的快速无损检测模型。研究取得一下主要结果：（1）通过对苹果果肉发生褐变机理的研究，全面分析了酶褐变、美拉德反应和抗坏血酸氧化反应在苹果不同贮藏期的发生情况。结果发现，在不同贮藏阶段，苹果果肉发生褐变的机理不同，在贮藏前期（80d前），褐变的主要机理是糖类和氨基酸发生的美拉德反应，在贮藏后期（80d后），褐变的主要机理是PPO、POD所催化的酶褐变。抗坏血酸的氧化反应也有发生，但由于苹果中抗坏血酸的含量极少，因此，抗坏血酸氧化不是引起褐变的主要因素。（2）用透射电镜观察不同褐变程度苹果果肉细胞超微结构的变化，结果发现果肉褐变与果实衰老细胞超微结构的变化有相似之处，出现细胞壁变形、质壁分离、膜结构解体、叶绿体转变为淀粉颗粒、线粒体内嵴和膜破裂等。随着褐变程度加重，胞间连丝出现密度降低、数量减少、排列整齐度降低，严重时出现断裂现象。液泡、质体、线粒体等细胞器和细胞质膜的超微结构解体及细胞区域化结构破坏的发生滞后于果肉褐变。因此，细胞区域化结构破坏导致的酶褐变并不是引起果肉组织褐变的唯一途径，美拉德反应是初期褐变的一种主要形式。（3）通过对不同的测试距离、测定温度、测定部位、表面颜色和贮藏时间等对苹果近红外光谱响应特性影响的研究表明，检测部位、测量距离、贮藏时间对苹果光谱响应特性的影响显著，零距离下在果实的赤道处采集光谱，可以减小误差；果实温度和表面颜色对近红外光谱没有显著影响，在实际检测中温度和表面颜色的影响可以忽略不计。（4）对比分析了仪器的光谱采集参数对苹果近红外光谱响应特性的影响，确定了光谱仪最优参数为：仪器信号能量为5V，扫描分辨率为8cm^-1，扫描次数为64次。这样的优化参数既能满足苹果品质实际检测要求，又能简化操作、降低成本、提高检测效率。（5）基于欧氏距离的聚类分析方法判断和剔除异常样品，得到最有效代表性建模样品为170个，所建定性模型的误判率由原来的53%降为29%；分析不同光谱预处理方法对苹果果肉褐变定性预测模型的影响，SNV结合SD预处理后所建定性模型的预测效果最好，误判率降低到22%；得出苹果果肉褐变度检测有效光谱波段范围10500cm^-1～6960cm^-1、6760cm^-1～5300cm^-1、5100cm^-1～4500cm^-1，在此区域内所建定性模型误判率由原来全波段模型的22%降低到19%；比较标准算法和因子化法所建定性模型的预测精度，因子化模型预测准确率高，对建模集和检验集的误判率分别为16%和19%。（6）提出苹果果肉褐变度的定量无损检测，该方法不但可以检测苹果果肉的褐变程度，而且可以实现对果肉褐变的预测和预警。通过有效剔除理化检测异常样品和光谱异常样品，再用逐步剔除相似样品的方法，选择代表性建模样品，用PLS结合MSC建立苹果果肉褐变定量检测模型，模型的RMSECV为0．082，RMSEP为0．084，交叉验证决定系数R²为0．871，外部验证决定系数R²为0．853。表明所建模型具有较高的预测准确度，可对未知样品进行实际预测。（7）探明与苹果果肉褐变相关的特征指纹波数为8822cm^-1、7085cm^-1、7000cm^-1、6694cm^-1、5800cm^-1、5322cm^-1、4650cm^-1，并用特征指纹波数建立用于苹果果肉褐变检测MLR模型，模型的RMSECV为0．077，RMSEP为0．079，交叉验证决定系数R²为0．908，外部验证决定系数R²为0．878。该方法能够简化检测操作，提高检测效率，为苹果果肉褐变提供了快速、直观、简便、可行的检测方法。更多还原

【Abstract】 Apple is one of the four major fruits.China is the first main producing country in theworld and the total production of apple is 25 300 000 t in 2006.But the export amount is only1.5 percent of total production for its poor quality,lower industrialization standard dragglingdetection methods and isolate evaluation standard which is not consistent with the firstcountry of apple total yield.At present,there is very severity of the apple flesh browning inthe storage time,and the routine analytical methods used for fruit internal qualitymeasurement are destructive,expensive,complex,labor consume and off line by nature.It isvery urgent to find mechanism and mathematic models used to nondestructively measurementapple internal quality.Therefore,the research establishes the mathematic models fornondestructive measurement flesh browning of apple using near-infrared spectroscopy basedon browning mechanism,which it has the important theoretical meaning and applied value.Choosing red‘Fuji’apple as materials and taking aim at browning mechanism andnondestructive detection techniques,the paper studies browning mechanism,near-infraredspectrum characteristic,affecting factors of detection and the optimal working parameters.Anoptimal nondestructive detection model for apple flesh browning is proposed.The mainresults of the dissertation involved:（1） The mechanism of apple flesh browning was investigated in several stages.Theenzymatic browning,maillard reaction and AsA oxidizing reaction were analyzed in thedifferent storage time of apples.It would be concluded that apple flesh browning is inducedby maillard reaction before 80 days and by enzymatic browning after 80 days.AsA oxidizingreaction was very subordinate owing to AsA minim in apples.（2） The cell microstructural changes of the different lrowning apple flesh degrees wereobserved with JEFM-1230 transmission electron microscopy.Results show that cellmicrostructural changes were similar between browning and senescence,for example,cellwall distortion,incipient plasmolysis,membrance disaggregation,chloroplast translatingstarch granules,and mitochondrion membrance breaking,and so on.With browningaggravation,plasmodesmata density and number descend,and structure became chaos orrupture.Disaggregation of cell organs （vacuole,plastid,mitochondrion） and cytoplasm membrances and destructin of cell region structure lagged flesh browning.Therefore,enzymatic browning owing to destructin of cell region structure was not exclusive reason,butmaillard reaction was essence of earlier browning.（3） The near-infrared spectrum measurement factors including testing distance,sampletemperature,sampling position,surface colour and storage time were researched.Testingdistance,sampling position and storage time had significant influence on apple near-infraredspectrum,but sample temperature and surface colour did not have significant influence.（4） The instrument parameters of the effects on spectrum measurement were analyzedand the best matching parameters used to FT-NIR measurement were as the following:instrument signal energy was 5V,scan resolution was 8 cm^-1 and scan number is 64.It wouldbe concluded that the calibration models used the best matching parameters will give the bestprediction accuracy.The optimal working parameters could not only satisfy with applequality detection,but also simplify operation,reduce cost,improve detective efficency.（5） Outlier and similar sample diagnosis and eliminating algorithms based on clusteranalysis of Euclidean distance were put forward,and established qualitative model with 170effective representative samples.Eorror identical ratio of the model declined from 53% to29%.The paper analyzed the different spectral pretreatment methods influenced on the applebrowning qualitative model.The results indicated that the SNV+SD pretreatment was the bestmethod with 22% eorror identical ratio.The different selected spectral range was used tooptimize the calibration models.The best calibration models of apple browning was at thewavenumber of 10500 cm^-1～6960cm^-1、6760 cm^-1～5300 cm^-1、5100 cm^-1～4500 cm^-1 with19% eorror identical ratio.Comparative analysis on the apple browning qualitative model ofstandard methord and factorization method was researched and factorization method was thebest algorithm.The eorror identical ratios of the best calibration model of browning were16% on calibration set and 19% on validation set,respectively.（6） The apple browning quatitative nondestructive measurement method was put forward.The methord could not only measure browning degrees of apple,but also predict and warn thefruit inner browning.Choosing representative modeling samples by eliminating outlier andsimilar sample,the calibration models used to non-destructive measurement of the fleshbrowning were developed by partial least square （PLS） regression combined withmultiplicative scatter correction （MSC）.The best calibration model of apple samples gaveRMSECV 0.082,RMSEP 0.084,the cross validate R² 0.871,and the outer validate R² 0.853.The results revealed that the model prediction was more accurate and could accurately predictunknown samples.（7） The experiment proved up the characteristic fingerprints of apple flesh browning 8822 cm^-1,7085 cm^-1,7000 cm^-1,6694 cm^-1,5800 cm^-1,5322 cm^-1,4650 cm^-1,and thenondestructive measurement model was established by using MLR regression based on thefingerprints.The model can predict apple flesh browning accurately with RMSECV0.077,RMSEP 0.079,the cross validate R² 0.908,and the outer validate R² 0.878.Theprecision of the mathematical model optimized by MLR can obviously improved thehigh-correlation coefficient of prediction model and the low prediction standard error.A newmethod that could evaluate apple flesh browning rapidly,visually,simply and feasiblywas developed.更多还原