【作者】 李顺峰；

【导师】 刘兴华；

【作者基本信息】 西北农林科技大学 ， 食品科学， 2012， 博士

【摘要】 苹果霉心病又称心腐病，主要危害苹果果实，果实受害从心室开始，逐渐向外扩展霉烂。由于霉心病发生在果实内部（心室周围），如果不把果实切开，很难从外表看出该果实是否被霉心病菌所侵染，它不仅影响果实感官品质、降低商品价格，而且也可能对食用者健康造成危害；如果病害果流入消费者手中，对果商信誉、销售市场都会产生不利影响。苹果发生霉心病病害时，果实发生了哪些生理反应还未见报道。因此，有必要对苹果霉心病引起的果肉生理变化进行研究，并建立一种无损、可靠的方法来检测苹果霉心病。本文以红富士苹果为试材，对苹果霉心病病原菌、果肉生理变化及其近红外无损检测技术进行了研究，以期为苹果霉心病的防治和无损检测提供依据和参考。研究取得以下主要结果：（1）经过对霉心病苹果的调查发现，苹果霉心病主要发生在果实横径在70mm～85mm、果实纵径在60mm～70mm、果形指数为0.80～0.90的果实，且圆形或近圆形的果实发病率高。（2）通过对病害组织的分离、培养和致病性试验，确定苹果霉心病主要由链格孢[Alternaria alternata （Fr.） Keissl.]、粉红单端孢[Trichothecium roseum （Bull.） Link.]和/或链格孢和粉红单端孢混合侵染所引起。（3）通过对苹果霉心病果肉生理变化的研究，发现不同类型的霉心病对果肉生理变化的影响不同。病原菌侵染苹果发生霉心型病害时，果肉总酚含量和AsA含量显著上升。病原菌侵染苹果发生干心腐型病害时，果肉中SOD、CAT、PPO、APX、PAL、β-1,3-葡聚糖酶和几丁质酶的活性与总酚含量、超氧阴离子产生速率显著升高，蛋白含量显著下降。病原菌侵染苹果发生湿心腐型病害时，正常部果肉POD活性和总酚含量显著上升；病部果肉细胞膜透性、MDA含量与β-1,3-葡聚糖酶和几丁质酶活性显著升高，APX活性与总酚、AsA、H₂O₂、SSC和蛋白含量显著降低。（4）通过对波段范围的选取，经不同光谱预处理方法[原始光谱未处理、多元散射校正（MSC）、标准正态变量变换（SNV）、矢量归一化（VN）、最大归一化（MN）、面积归一化（AN）、9点平滑、一阶导数（9点平滑）]处理后提取到的主成分所建立的判别函数对健康苹果和霉心病苹果的判别效果研究发现，在全波段范围(12000cm^-1～4000cm^-1)内，经VN预处理后所建判别函数的正确判别率最高，为89.9%；经MN预处理所建判别函数的正确判别率最低，为73.3%；其他预处理方法所建判别函数的判别效果介于二者之间。（5）在全波段范围(12000cm^-1～4000cm^-1)内，光谱经VN预处理后提取主成分（20个主成分）所建健康苹果和霉心病苹果的Fisher判别函数判别性能稳定，校正集正确判别率为89.9%，对未知苹果样本（验证集）的正确判别率为87.8%，可以基本满足苹果霉心病无损检测的要求，说明近红外漫反射光谱技术应用于健康苹果和霉心病苹果的检测是可行的。（6）经OPUS-QUANT2定量分析软件的自动优化功能对建模条件进行自动优化，最优建模参数为：光谱经VN预处理，光谱波段范围为12000cm^-1～5446.2cm^-1和4601.5cm^-1～4246.6cm^-1，主成分维数为10时所建苹果霉心病发病程度偏最小二乘（PLS）模型的交叉验证均方根差（RMSECV）最小，为0.123，此时交叉验证决定系数（R²C V）为0.5172，优于全波段范围所建PLS模型。（7）光谱经9点平滑遗传算法（GA）筛选波段和VN预处理GA筛选波段所建苹果霉心病发病程度PLS模型最优，R²_CV与验证集预测决定系数（R²P）分别为0.6623与0.6328和0.6597与0.6466，它们对应的RMSECV与验证集预测均方根差（RMSEP）比较接近且较低，分别为0.1035与0.0929和0.1041与0.0944，GA筛选建模参数所建PLS模型优于OPUS优选建模参数所建模型。（8）通过预测偏差准则剔除32个异常样本后，光谱经9点平滑预处理GA筛选波段所建PLS模型校正集的R²_CV为0.7809，未知样本（验证集）R²P为0.6902，RMSECV（0.0846）和RMSEP（0.0860）均较低且比较接近；通过预测偏差准则剔除44个异常样本后，光谱经VN预处理GA筛选波段所建PLS模型校正集R²_CV为0.7730，未知样本（验证集）R²P为0.7105，RMSECV（0.0784）和RMSEP（0.0817）均较低，且比较接近，说明所建的PLS预测模型是稳健的，可以用来对未知苹果样本的发病程度进行预测。VN预处理GA筛选波段比9点平滑预处理GA筛选波段所建的优化PLS模型对苹果霉心病发病程度有较好的预测效果。更多还原

【Abstract】 Moldy core （MC） and core rot （CR）, pathogenic disorders associated with the coreregion of fruits, are among the most important postharvest diseases of apple （Malus domestica）and are mainly found in susceptible varieties that have open calyx tubes. Apple fruits werefirst infected in the seed locules, then gradually expanded into fruit flesh and finally resultedin fruit rot. MC and CR of apple fruits are undetectable unless the fruits are cut open or bitteninto because they occurred in the internal of apple fruits （surround seed locules）, they can notonly affect fruit quality, cause lower commodity prices, but also may cause damage toconsumers’ health, and if the diseased fruits were delivered into the table of consumers, it willhave a negative impact on both producers and consumers, such as loss of consumers trust andloyalty or even lost the market. However, there is no related report on the physiologicalresponses of apple fruits with MC and CR. Therefore, it is necessary to study the fruitphysiological responses caused by apple MC and CR, and to find or establish a reliable,nondestructive method to detecting apple MC and CR fruits.With the “Fuji” apple （Malus domestica cv. Fuji） as materials, this study investigated thecasual agents of apple moldy core, physiological responses of fruit flesh, and nondestructivedetection of apple moldy core and core rot using near infrared spectroscopy. The main resultswere exhibited as follows:（1） After investigation the relationship between MC and CR and apple fruits, the resultsshowed that apple MC and CR occur mainly in fruit diameter is70mm～85mm, fruitheight is60mm～70mm, fruit-shape index is0.80～0.90, and high apple MC and CRincidence rate occurs in circular or near circular apple fruits.（2） In order to clear the casual agents of apple MC and CR, the diseases through isolation,cultivation and pathogenicity experiment, found that the main pathogen caused apple MC andCR are Alternaria alternata, Trichothecium roseum, and/or the combination of Alternariaalternata and Trichothecium roseum.（3） The fruit flesh physiological responses of apple MC and CR was studied, the results indicated that different types of apple MC and CR have different effects on the physiologicalindices of apple fruits flesh. Total phenolic content and ascorbate （AsA） content increasedsignificantly when pathogen infecting apple fruits caused MC. When pathogen infecting applefruits caused dry CR （DCR）, Superoxide dismutase （SOD）, catalase （CAT）, polyphenoloxidase （PPO）, ascorbate peroxidase （APX）, phenylalanine ammonia-lyase （PAL）,β-1,3-glucanase and chitinase activities, and total phenolic content, superoxide radicalproduction of fruit flesh increased significantly; protein content decreased markedly. Whenpathogen infecting apple fruits caused wet CR （WCR）, peroxidase （POD） activity and totalphenolic content were increased significantly in asymptomatic tissue of WCR （asympWCR）fruit flesh; in symptomatic tissue of WCR （sympWCR） fruit flesh, membrane permeability,malondialdehyde （MDA） content, β-1,3-glucanase and chitinase activities were increasedsignificantly, thus, APX activity, and the contents of total phenolic, AsA, hydrogen peroxide（H2O2）, soluble solid content （SSC）, protein content were decreased significantly.（4） Select effective wavenumber range, apple MC and CR discriminant efficiency ofFisher function based on several spectra preprocessing methods （non-preprocess,multiplicative scatter correction （MSC）, standard normal variate transformation （SNV）, vectornormalization （VN）, maximum normalization （MN）, area normalization （AN）, smoothingwith a segment of9, first derivation with a segment of9） and principal component analysis（PCA） were investigated. In the whole wavenumber range (12000cm^-1～4000cm^-1), thehighest correct discriminant rate （89.9%） of Fisher discriminant function was based on VNpreprocessing, while the lowest correct discriminant rate （73.3） was based on MNpreprocessing, and the correct discriminant efficiency of Fisher discriminant function built byother preprocessing methods were between the upper two.（5） In whole wavenumber range (12000cm^-1～4000cm^-1), the correct discriminantefficiency of Fisher discriminant function based on VN preprocessing and extracting principalcomponents （PCs,20PCs） was stabe, correct discriminant rate for calibration set is of89.9%,for test set is of87.8%. It can basically meet the requirements of non-destructive testing ofapple MC and CR fruits. These results could indicate that near infrared diffuse reflectancespectroscopy applied to apple MC and CR detection is feasible.（6） By using automatic optimization function of OPUS-QUANT2quantitative analysissoftwarefor modeling contidions, the best partial least square （PLS） model built for relativedecay area of apple MC and CR were based on VN preprocess. The optimized parameters forPLS model were as follows: effective wavenumbers are of12000cm^-1～5446.2cm^-1and4601.5cm^-1～4246.6cm^-1; PCs rank of10. Under these conditions, root mean square error ofcross-validation （RMSECV） is the lowest （0.123）, the cross-validation determination (R2 0.5172, it is better than PLS model build in whole wavenumber range.（7） The best PLS model of decay area for apple MC and CR were built with the variablesselected by genetic algorithm （GA） after smoothing with a segment of9and VNpreprocessing. TheR2CVand prediction determination （R²P） for the PLS model built by GAselected variables after the both two preprocessing methods were0.6623and0.6328,0.6597and0.6466, respectively. Their corresponding RMSECV and root mean square errors ofprediction （RMSEP） were0.1035and0.0929,0.1041and0.0944, respectively. The resultsindicated that the PLS model buid by GA selected variables were better than the parametersoptimized by OPUS.（8） Eliminating32outliers by criterion of prediction difference, theR²_CVof PLS modelbuild by GA selected wavenumbers after smoothing with a segment of9spectrapreprocessing for calibration set is0.7809, RMSECV is of0.0846; theR2Pfor prediction setis0.6902, RMSEP is of0.0860. Eliminating44outliers by criterion of prediction difference,theR²_CVof PLS model build by GA selected wavenumbers after VN spectra preprocessingfor calibration set is0.7730, RMSECV is of0.0784; theR²_Pfor prediction set is0.7105,RMSEP is of0.0817. The above results indicated that both PLS models were robust, and canbe used to predicting decay area of apple MC and CRfruits, and the PLS model build by GAselected wavenumbers with VN preprocessing has a better prediction effect on decay area ofapple MC and CR fruits than GA selected wavenumbers with smoothing （a segment of9）preprocessing.更多还原