光谱学与光谱分析, 2009, 29 (7): 1768, 网络出版: 2010-05-26
近红外光谱预测猕猴桃硬度模型的简化研究
Study of Simplification of Prediction Model for Kiwifruit Firmness Using Near Infrared Spectroscopy
近红外光谱 猕猴桃硬度 净分析物预处理 偏最小二乘 Near infrared(NIR) spectroscopy Kiwifruit firmness Net analyte preprocessing (NAP) Partial least square(PLS)
摘要
为简化猕猴桃硬度的预测模型, 利用标准正态变量变换对猕猴桃1 000~2 500 nm近红外光谱进行预处理, 在优选建模波段和采用净分析物预处理(NAP)降低建模主因子数两个方面简化猕猴桃硬度偏最小二乘(PLS)模型。 结果表明, 优选5 189~5 370 cm-1, 4 549~4 620 cm-1, 6 049~6 230 cm-1, 6 999~7 730 cm-1, 6 249~6 614 cm-1等5个波段进行建模, NAP/PLS模型性能最佳, 主因子数为5, 校正集相关系数R2和均方根误差RMSECV分别为0.819 41和0.701 77, 预测集相关系数R2和均方根误差RMSEP为0.780 67和0.882 71。 与简化前的PLS模型相比, 模型不仅更加简洁, 而且预测能力和精度均有所提高。
Abstract
To simplify the prediction model of kiwifruit firmness, SNV was used to preprocess the near infrared(NIR)spectra (1 000-2 500 nm)of kiwifruit. PLS model simplification by optimizing spectral intervals and decreasing the number of factors through net analyte preprocessing(NAP)was carried out. Results showed that the performance of NAP/PLS model is the best. It was achieved with 5 factors in five wavenumber ranges(5 189-5 370, 4 549-4 620, 6 049-6 230, 6 999-7 730, and 6 249-6 614 cm-1). The optimal model was achieved with R2=0.819 41 and RMSECV=0.701 77 in the calibration set and R2=0.780 67 and RMSEP=0.882 71 in the prediction set. This indicates that the model not only may efficiently simplify PLS model, but also may improve precision and predictive ability.
吕强, 汤明杰, 赵杰文, 蔡健荣, 陈全胜. 近红外光谱预测猕猴桃硬度模型的简化研究[J]. 光谱学与光谱分析, 2009, 29(7): 1768. Lv Qiang, TANG Ming-jie, ZHAO Jie-wen, CAI Jian-rong, CHEN Quan-sheng. Study of Simplification of Prediction Model for Kiwifruit Firmness Using Near Infrared Spectroscopy[J]. Spectroscopy and Spectral Analysis, 2009, 29(7): 1768.