【作者】 李东华；

【导师】 纪淑娟；

【作者基本信息】 沈阳农业大学 ， 食品科学与工程， 2009， 博士

【摘要】 南果梨属秋子梨系,为辽宁省鞍山、海城一带特产优质水果,已有上百年栽培历史,享有“梨中皇后”的美誉,南果梨香气浓郁,汁液丰富,酸甜适中,深受消费者的喜爱。目前,南果梨品质的分级工作基本上仍靠人工挑选完成,依据的指标主要是大小、颜色、表面有无病斑等外部特征,这种人工分级的劳动量大、生产效率低,无法对南果梨的内在品质进行分级,不利于优质南果梨的产业化发展。近红外光谱分析技术是近十几年迅速发展的一门绿色分析技术,它具有快速、准确、无损伤检测的优点,正越来越广泛的应用于水果内在品质的无损检测。本研究通过对影响南果梨近红外光谱分析的仪器参数设置、样品状态和不同数据处理方法等影响因素的系统研究,确定了南果梨近红外无损检测分析的适宜参数,在此基础上建立了南果梨内在品质的近红外无损检测模型,并对建立的南果梨品质模型进行未知样品的预测验证和适用性的研究,为南果梨产业的自动化分级提供技术支持。具体的研究结果如下:（1）南果梨近红外透射光谱的适宜扫描方式为以果柄-果蒂为轴1 80°转动扫描2次、适宜的光谱扫描次数为单测点扫描一次、扫描参数的设定只要满足待测样品光谱曲线和参照物光谱曲线均在特征峰处相切于征服线即可,光谱测定的最佳环境条件为在完全遮光的房间内进行样品光谱的采集。（2）南果梨的果皮和表面颜色对其近红外无损检测分析的影响不大,而尺寸大小和样品温度的影响较大,为了提高模型的准确度,校正温度和大小对模型性能的影响,试验确定选用不同温度、尺寸大小适中的样品进行模型的建立;不同成熟度的南果梨样品所建立的糖、酸度模型的预测能力有差异,结果表明成熟度较高的样品所建模型的稳定性较好。（3）适宜的光谱导数预处理方法为二阶导数法、光谱平滑预处理方法为20点Savitzky-Golay平滑法、糖度和酸度的化学定标最佳参数分别为可溶性固形物含量和总酸度。（4）根据确定的适宜的建模参数,通过不同模型优化方法的使用,本试验最终建立了鞍山、海城两个产区统一的南果梨糖度、酸度的近红外无损检测模型。建模结果为:南果梨糖、酸度的近红外定标数学模型的校正集相关系数分别为0.935、0.911,校正均方根误差分别为0.212、0.018。（5）利用建立的鞍山、海城两产区的混合模型对不同贮藏期样品进行预测,确定南果梨糖、酸度近红外无损检测定标模型适用于预测常温贮藏12d内和冷藏贮藏4个月内样品的糖、酸含量;模型的动态适用性分析结果表明建立的模型适用于预测不同年份的一定时期内鞍山、海城产区南果梨的糖、酸含量。（6）试验利用PLS法建立了鞍山、海城两产区混合的硬度模型,最终建立定标硬度模型的结果为:校正集相关系数为0.970,校正集均方根误差为0.124;对模型的适用性分析结果表明所建立南果梨近红外硬度模型适用于预测硬度范围在2～15kg/cm²内的样品硬度。（7）采用光密度差法检测南果梨内部褐变的结果表明:不同褐变程度的南果梨透射光谱△OD_{（712.92nm～672.93nm）}值的大小顺序为:0级＜1级＜2级＜3级,黑心越严重的南果梨其对应的光密度差值越小;最终确定正常和褐变两个级别南果梨的判别标准为:△OD（（712.92nm～672.93nm））值小于或等于0.305时,南果梨为褐变梨,△OD_{（712.92nm～672.93nm）}值大于0.305时,南果梨梨为好梨;根据这个分级标准,100个褐变梨中有2个被误判成好梨,误判率为2%,30个正常梨中有2个被误判成轻微褐变梨,误判率为6.67%。（8）采用褐变面积法检测南果梨内部褐变的结果表明:利用3个波段进行南果梨褐变面积的分段建模,结果表明在全波段下建立的褐变面积定标模型的性能最优,其校正相关系数为0.965,校正均方根误差为0.016,是较理想的定标模型;应用130个预测集样品对该定标模型的预测性能进行评价,预测相关系数为0.811,预测均方根误差为0.059,预测值与真实值的散点图表明在褐变面积比值小于0.1时,正常与轻微褐变的南果梨有误判现象的存在。更多还原

【Abstract】 The‘Nanguo’ pear is a kind of high-quality fruit belong to Pyrus ussuriensis Maxim, which is a specialty and widely cultivated in Anshan and Haicheng districts of Liaoning province for hundreds of years.Aroma of the fruit is strong,the juice is abundant,the quality and taste are very good,it could satisfy the demand of consumer,and therefore it is honored with‘the queen of pears’.For the moment,the grading of quality on‘Nanguo’ pears still depends on manpower,and the grading index mainly are sample sizes,color,scabs and so on. This method has some disadvantages,such as the quantum of labor is great,the rate of production is low,also the method is not fit for the detecting of internal quality and industrialized development of‘Nanguo’ pears.Near infrared spectroscopy nondestructive detecting technique is a green analytical technique,which has developed faster in resent years, has fast,accurate and nondestructive merits.It can evaluate internal quality and external quality of fruit simultaneously without destroying samples.The factors（spectrometer parameter setting,sample conditions,spectral pretreatment methods and modeling methods） that influenced quality analysis of‘Nanguo’ pears by using NIRS analytical technique were studied in this research.Based on the proper parameters which were determined by factor analyzing,the nondestructive detecting models for the internal quality of pears（SSC、TA、firmness、brown heart） were established,then the predictive ability and applicability of models were analyzed.This study would provide supports to automation grading for‘Nanguo’ pears.Mainly research results were as follows:（1）Proper scanning conditions which were used for establishing NIRS nondestructive detecting models were confirmed.The results of test indicated that proper scanning times was 1 times to one point;proper scanning manner was 2 times by 180°running to several points; the enactment of scanning parameters simply satisfied the condition that the spectral line of samples and the spectral line of contrast were tangent in conquest line;it can obtain the greatest prediction performance with the spectral data which were determined in an aphotic room absolutely.（2）The influence degree of peer and color to NIR process for‘Nanguo’ pears were not prominence,whereas the degree of sample sizes and temperature were prominence.;In order to avoiding the disadvantages of sizes and temperature,different temperature and medium-sized pears were chosen to established calibration models;Disparity analysis results of models with different maturity degree samples demonstrated that the predictive ability of model with high maturity degree samples was better than the other two which were established with lower maturity degree samples.（3）The results of test indicated that second derivative method was proper derivative methods.Savitzky-Golay smoothing with 20 points was proper smoothing means.The contents of soluble solids and total acidity were proper chemic indices.（4）Based on proper modeling parameters and different optimization methods,uniform models were established,which were used for predicting SSC and TA content of‘Nanguo’ pears produced in Anshan and Haicheng districts at Liaoning province.For the results of SSC and TA models,correlation coefficient of calibration was 0.935 and 0.911;RMSEC was 0.212 and 0.018.（5）This trail utilized above two models to predict SSC and TA of‘Nanguo’ pears samples which were hoarding in different storage periods during room temperature and refrigeration,respectively,the predictive ability and applicable range of models were ascertained.The results of this test indicated that the NIRS nondestructive detecting models can be used for predicting SSC and TA of‘Nanguo’ pears samples which were in 12 days on the storage conditions of room temperature and in 4 months on the cold storage conditions.At the same time,the dynamic application of models to unknown samples from different years were completed;It showed that models were fit for predicting SSC and TA content of pears that produced in Anshan and Haicheng districts during different periods.（6）Based on the PLS method,uniform firmness model was established for‘Nanguo’ pears produced in Anshan and Haicheng districts of Liaoning province.For the results of firmness model,correlation coefficient of calibration and RMSEC were 0.970 and 0.124;the optimal predictive scope of firmness model was 2-15kg/cm2.（7）The brown results detecting by optical density difference methods showed that the value measurement order of△OD_{（712.92nm～672.93nm）} were:0 degree＜1 degree＜2 degree＜3 degree,and the discipline was that the more brown heart areas,the lower the value of△OO_{（712.92nm～672.93nm）};At last the assessment standard was instituted,when the value was high than 0.305,the pear was a good one,yet when the value was lower than or equal to 0.305,the pear was a brown one.According to the standard,2 pears from 100 brown pears were misjudged（false rate 2%） to good pears,2 pears from 30 good pears were misjudged to brown pears（false rate 6.67%）.（8）The brown results detecting by brown areas methods showed that brown model with samples spectrum which were gathered under all wavelength band was better than the others, For the results of brown areas model,correlation coefficient of calibration and RMSEC were 0.965 and 0.016;The estimated performance of the brown model was that correlation coefficient of prediction and RMSEP were 0.811 and 0.059;When the area ratio was lower than 0.1,the misjudgments probability would raise.更多还原