《苹果可溶性固形物NIR线性模型建立》-毕业论文设计（学术）.docVIP

本科生毕业论文 苹果可溶性固形物NIR线性模型建立 院 系 　 化学化工学院 专 业　　 食品科学与工程 　 ? ? 摘 要 本文以红富士苹果为样品，研究可见／近红外光谱技术进行苹果糖度的无损检测方法。采集苹果的漫透射光谱（500-1010nm），采用杠杆值（阈值为0.136）与学生残差 （阈值为3）检验判别方法剔除异常样本，考察了不同的光谱信号校正方法对光谱噪声消除的效果，并对处理后光谱进行变量优选，分别建立 PLS 和 SMLR，PCR 苹果糖度的校正模型，以RPD值评价模型稳健性，比较可得SMLR最优，其R=0.940，RMSEC=0.483，RMSEP=0.545，RPD=2.400。结果表明：在诸如苹果糖度这一类农产品品质综合指标的光谱检测中，应用SMLR建模进行定量分析是完全可行的。 关键词：可见/近红外光谱；无损检测；苹果；糖度 ? ABSTRACT In this paper, Fuji apple samples were used to study the visible / near infrared(Vis/NIR) spectroscopy for nondestructive testing method for souble solids content. Apples diffuse reflectance spectra (500-1010nm) were collected by portable Vis/NIR spectrometer. The performance of different signal correction methods were evaluated after removing abnormal samples according to the Studentized Residual(3) and Leverage(0.136) test methods. PLS, SMLR and PCR calibration models were constructed to predict the SSC of Fuji apple which are based on the selection of spectral variables. RPD value was used to evaluate model robustness, compare the best available SMLR, the R = 0.940, RMSEC = 0.483, RMSEP = 0.545, RPD = 2.400. The results showed that: in the souble solids content of this type of agricultural products such as Apples comprehensive index of the quality of spectral detection, the application of SMLR model for quantitative analysis is feasible. Key words：Visible / near infrared spectroscopy; Non-destructive measuring ; Apple; Souble solids content(SSC) 缩略词 PLS，Partial Least Squares，偏最小二乘法 MLR，Multiple Linear Regression，多元线性回归法 MSC，Multiplicative Scatter Correction，多元散射校正 PRESS，Prediction Residual Error Sum of Squares，预测残差平方误差总和 R，Correlation Coefficient，相关系数 RMSEC，the Root Mean Square Error of Calibration，校正均方根偏差 RMSECV，the Root Mean Square Error of Cross Validation，内部交叉验证均方根偏差 RMSEP，the Root Mean Square Error of Predication，预测均方根偏差 SMLR，Stepwise Multiple Linear Regression，逐步多元线性回归法 SSC，Soluble Solid Content，可溶性固形物 SNV，Standard Normal Variety，标准正态变量变化 目录 T