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

作　　者：吕强[1] 汤明杰[1] 赵杰文[1] 蔡健荣[1] 陈全胜[1] 

机构地区：[1]江苏大学食品与生物工程学院,江苏镇江212013

出　　处：《光谱学与光谱分析》2009年第7期1768-1771,共4页Spectroscopy and Spectral Analysis

基　　金：国家高技术研究与发展计划项目(2006AA10Z263);国家自然科学基金项目(30771243)资助

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。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 R^2 =0. 819 41 and RMSECV=0. 701 77 in the calibration set and R^2 =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.

关 键 词：近红外光谱 猕猴桃硬度 净分析物预处理 偏最小二乘 

分 类 号：S663.4[农业科学—果树学]