摘要
在采后冷藏过程中,李果实很容易发生褐变,这是影响其品质的重要因素之一。有关李果实褐变的传统检验手段绝大多数为破坏性检验,且主观性强、一致性差。为此,使用了近红外光谱的方法来实现对李果实褐变和非褐变的无损、快速鉴别。采集4 000~12 500 cm^(-1)波长范围内的124个李果实样品(褐变样品70个,非褐变样品54个)的近红外漫反射光谱,基于主成分分析的马氏距离判别分析和反向传播人工神经网络定性鉴别模型,通过比较和考察上述模型对褐变样品和非褐变样品识别的准确程度,筛选出能够有效鉴别李果实褐变的新方法。结果表明:在对样品全波段光谱数据做主成分分析后,以前10主成分得分作为输入变量所建立起来的马氏距离判别分析和反向传播人工神经网络模型均能够对李果实褐变与否进行有效识别,且后者判别效果更佳,其校正集和预测集的判别正确率分别为100%和97.56%,对非褐变样品和褐变样品的判别正确率分别达到100%和98.57%。因此,采用近红外光谱分析技术并结合化学计量学方法能够对李果实是否褐变进行快速、无损、有效的鉴别。
Flesh browning mostly happens in plum fruit during the post-harvest storage period ,which is an important factor af-fecting the storage quality of plum fruits .Traditional methods used to discriminate plum browning involve the destruction of the intact fruit ,which are highly subjective and error-prone .Therefore ,the near-infrared (NIR) spectroscopy technique was applied to achieve rapid and non-destructive identification of plum browning and non-browning in this paper .The near infrared diffuse reflectance spectroscopy of 124 plum samples were collected in the band number of 4 000~12 500 cm ^-1 .These samples were classified into two groups ,browning (n=70) and non-browning (n=54) .In order to find a new way to effectively discriminate plum fruits with flesh browning ,three qualitative identification methods :the qualitative test ,Mahalanobis distances discriminate analysis (DA) and Back Propagation-artificial neural networks (BP-ANN) were used to compare their capacity of recognizing browning plums and non-browning oneswhile the last two approaches were based on the principal component analysis (PCA) method .These results showed that DA and BP-ANN could be used to conctruct effective classification models for identifying plum browning ,and the first ten principal components extracted from original spectra were applied as input variables to build DA and BP-ANN models .The optimal method was obtained with BP-ANN ,which gained an accuracy of 100% for calibration set and 97 .56% for prediction set ,and the identification accuracy rate reached 100% and 98 .57% for non-browning samples and browning ones ,respectively .It could be concluded that NIR spectroscopy technique combined with chemometrics methods has great potential to recognize plums of browning and non-browning rapidly ,non-destructively and effectively .
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2016年第7期2089-2093,共5页
Spectroscopy and Spectral Analysis
基金
国家自然科学基金项目(31201430)
河北省自然科学基金项目(C2013201113
C2015204182)
公益性(农业)科研专项项目(201303075)
河北省科技计划项目(14225503D)资助
关键词
李果实
褐变
反向传播人工神经网络
马氏距离判别分析
近红外光谱
Plum
browning
Back Propagation-artificial neural networks
Mahalanobis distances discriminate analysis
Near-in-frared spectroscopy