【摘要】：石榴在中國已有2000多年的種植歷史,目前中國石榴的栽種面積約為175萬余畝,位居世界第一。但在石榴的栽種及貯藏過程中,因為石榴品種繁多且在貯藏保鮮方面缺少有效快速的監(jiān)控方法,僅依靠人力還無法使得石榴的生產(chǎn)貯藏達到標準化要求。近紅外光譜技術(shù)以其簡單、快速、無損、無公害等優(yōu)點,被廣泛應用于水果品種的定性判別及品質(zhì)的定量檢測中。目前近紅外技術(shù)分析對象主要為薄皮水果,而對于石榴等厚皮水果的研究極少。本次研究以陜西臨潼石榴為研究對象,采用漫反射光譜,結(jié)合化學計量學,建立不同品種石榴和石榴理化品質(zhì)、酚類物質(zhì)含量的近紅外模型,為石榴品種、品質(zhì)檢測提供軟件支持,促進石榴產(chǎn)業(yè)的快速發(fā)展。主要研究內(nèi)容和結(jié)果如下:(1)采集“一串鈴”、“大凈皮甜”、“黃皮甜”石榴的近紅外光譜,經(jīng)MSC預處理后,分別采用PCA+MLP神經(jīng)網(wǎng)絡法、PCA+Fisher線性判別法、PLS-DA判別法進行模型的建立。通過對三種判別模型的比較,PLS-DA判別結(jié)果優(yōu)于其他兩種,PLS-DA判別分析法在全波段內(nèi)對驗證集三個石榴品種的正確識別率分別為97.30%、96.55%、96.77%。為進一步提高PLS-DA模型的判別率,采用“載重法”選取特征波段,優(yōu)化后模型對驗證集的正確識別率分別達到97.30%、100.00%、100.00%。表明經(jīng)波段優(yōu)化后的PLS-DA模型可以實現(xiàn)對石榴品種的鑒別且效果令人滿意。(2)為減小品種差異對近紅外光譜質(zhì)量品質(zhì)檢測模型的影響,以三個不同品種的石榴為研究對象,采用PLS法,建立單一品種和混合品種的石榴pH值的近紅外檢測模型。通過比較可知混合三個品種所建立的模型取得較好的預測效果。校正集和驗證集的相關(guān)系數(shù)均大于0.900。因此,采用三個混合品種的校正集所建立的模型可以實現(xiàn)對石榴pH值的快速、無損、準確測定。這一結(jié)論進一步推廣到其他質(zhì)量指標的測定。三個品種混合作為校正集,結(jié)合CARS-PLS波段篩選,TA、SSC和成熟度模型驗證集的R~2分別為0.903、0.930、0.853;RMSEP分別為0.019%、0.244°Brix、2.142。(3)采集整個石榴的近紅外光譜,采用PLS法建立籽粒中多酚和黃酮含量的近紅外檢測模型,所建立的PLS模型校正集和驗證集的相關(guān)系數(shù)均小于0.850,所建立的模型擬合和預測能力較差,預測精度有待于進一步研究。利用整個果實所采集的光譜進行籽�；ㄉ蘸繖z測時,所建立的PLS模型的預測效果良好,校正集R~2為0.881,RMSEC為1.318 mg/100g;驗證集R~2為0.863,RMSEP為1.266 mg/100g。
[Abstract]:Pomegranate has been planted in China for more than 2000 years. At present, the planting area of pomegranate in China is about 1.75 million mu, ranking first in the world. However, in the process of pomegranate planting and storage, because of the variety of pomegranate and the lack of effective and rapid monitoring methods in storage and preservation, the production and storage of pomegranate can not meet the requirements of standardization by manpower alone. Near infrared spectroscopy (NIR) has been widely used in qualitative discrimination and quantitative quality detection of fruit varieties because of its simple, rapid, nondestructive, pollution-free and other advantages. At present, the analysis object of near-infrared technology is mainly thin-skinned fruit, but the research on pomegranate and other thick-skinned fruit is very few. In this study, the near infrared model of physical and chemical quality and phenolic content of different varieties of pomegranate and pomegranate was established by using diffused reflectance spectroscopy and chemometrics to establish the near infrared model of pomegranate and pomegranate content in Lintong, Shaanxi Province. Quality testing provides software support to promote the rapid development of pomegranate industry. The main research contents and results are as follows: (1) the near infrared spectra of "a string of bells", "big skin sweet" and "yellow skin sweet" pomegranate were collected. After pretreatment with MSC, PCA MLP neural network method and PCA Fisher linear discriminant method were used respectively. PLS-DA discriminant method is used to establish the model. By comparing the three discriminant models, the PLS-DA discriminant results are better than the other two. The correct recognition rates of the three pomegranate varieties in the whole band by PLS-DA discriminant analysis are 9730%, 96.55% and 96.77%, respectively. In order to further improve the discrimination rate of PLS-DA model, the "load method" is used to select the feature band. The correct recognition rate of the optimized model to the verification set is 97.30%, 100.00% and 100.00%, respectively. The results show that the optimized PLS-DA model can identify pomegranate varieties with satisfactory results. (2) in order to reduce the influence of variety differences on the quality detection model of near infrared spectroscopy, Taking three different pomegranate varieties as the research object, the near infrared detection model of pomegranate pH value of single variety and mixed variety was established by PLS method. Through the comparison, it can be seen that the model established by mixing the three varieties has achieved better prediction results. The correlation coefficients of correction set and verification set are both greater than 0.900. Therefore, the model established by using the correction set of three mixed varieties can realize the rapid, nondestructive and accurate determination of the pH value of pomegranate. This conclusion is further extended to the determination of other quality indexes. Combined with CARS-PLS band screening, the R 鈮，

本文編號：2494371