摘要
[目的]获得精度高、鲁棒性强的草莓近红外光谱糖度检测模型。[方法]利用K-S(Kennard-Stone)方法划分样本集,并用小波滤噪法对草莓1000~2500nm近红外光谱进行预处理,最后用偏最小二乘法(PLS)和区间偏最小二乘法(iPLS)分别建立预测模型。[结果]采用区间偏最小二乘法将光谱划分为20个子区间,利用其中的第16个子区间建立的糖度模型效果最佳,其校正时的相关系数Rc和校正均方根误差RMSEC分别为0.9355和0.259,预测时的相关系数邱和预测均方根误差RMSEP分别为0.9202和0.305。[结论]用小波滤噪和联合区间偏最小二乘法所建立的草莓糖度模型不仅能有效地减少建模所用的变量数,缩短运算时间,而且预测能力和精度均得到提高。
[Objective] The research aimed to obtain the testing model of sugar content of near infrared spectrum in strawberry with high accuracy and
strong robustness. [Method] The K-S (Kennard-Stone) method was used to divide the sample set and the wavelet noise filtering method was used to pretreat the near infrared spectrum at 1 000 - 500 nm in shawberry, at last the partial least squares(PLS)and interval partial least squares(iPIS)were used to set up the prediction model resp.. [Result] The spectrum was divided to 20 subinterval with the interval partial least squares and the effect of sugar content model established by their 16 subinterval was optimum.The correlation coefficient Re in correction and mot mean square enor of correction RMSEC were 0.935 5 and 0.259 resp.and the correlation coefficient Rp in forecast and the mot mean square error of forecast were 0.920 2 and 0.305 resp. [Conclusion]The strawberry sugar content model established by wavelet noise filtering method and interval partial least squares not only could decrease the
variable number of modeling effectively and shorten the operation time, but also could improve the prediction ability and precision.
出处
《安徽农业科学》
CAS
北大核心
2009年第12期5752-5754,共3页
Journal of Anhui Agricultural Sciences
基金
国家863高科技项目(2008AA10Z2)
国家自然基金资助项目(30671199)
江苏省自然科学基金资助项目(BK2006707-1)
关键词
近红外光谱
草莓
糖度
区间偏最小二乘法
NIR spectroscopy
Strawberry
Sugu degree
interval partial least square