梨枣糖度无损检测建模分析——基于高光谱成像技术

Nondestructive Testing Modeling Analysis of Sugar Content in Pear Jujube——Based on Hyperspectral Imaging Technique

以梨枣为对象,利用去噪和去基后的全光谱数据及提取的42个近似系数,分别建立相应的PLS和PCR模型。结果分析表明:用近似系数所建的PLS模型校正集决定系数Rc(0.931)和PCR模型Rc(0.882)分别比用全光谱所建的PLS模型Rc(0.875)和PCR模型Rc(0.858)要高;PLS模型校正集方差RMSEC(0.986)、预测集方差RMSEP(1.159)和PCR模型校正集方差RMSEC(1.048)、预测集方差RMSEP(1.322)分别要比全光谱PLS模型校正集方差RMSEC(0.731)、预测集方差RMSEP(1.270)和PCR模型校正集方差RMSEC(0.958)、预测集方差RMSEP(1.361)的差值更为接近。这说明,应用近似系数所建模型较稳定。

In Huping jujube for the study. Using the denoising and to base after full spectrum data and extract the 42 approximate coefficient model corresponding PLS and PCR respectively. Results show that：PLS model is built by using approximate coefficient correction set decision coefficient of Rc （0. 931 ） and PCR models Rc （O. 882 ）, respectively, than with a full spectrum PLS model is built by Rc （0. 875 ） and PCR models Rc （0. 858 ） ; PLS model calibration set variance RMSEC （O. 986） ,prediction set variance RMSEP （ 1. 159 ） and PCR model calibration set variance RM- SEC （1. 048 ） ,prediction set variance RMSEP （1. 322）compared with the full spectrum PLS model calibration set vafiance RMSEC （O. 731 ） , prediction set variance RMSEP （ 1. 270 ） and PCR model calibration set variance RMSEC （0. 958 ） ,prediction set variance RMSEP （ 1. 361 ） difference value more close to. Explain application of approximate coefficient of the model is stable.