摘要
小麦水分含量是评估小麦品质的重要指标,近红外光谱技术可以同时、快速、无损的对小麦水分含量进行检测与定量分析。基于模型集群分析(MPA)思想,结合引导软阈值算法(BOSS)对光谱变量进行选择,通过子模型回归系数得到变量权重,采用加权引导采样(WBS)逐步校正优化变量权重,收缩变量空间,选取交叉验证均方根误差(RMSECV)较小的子集为最优变量集,以此建立回归预测模型。结果显示,与全光谱模型相比,利用BOSS算法选择的特征变量建模,可以将预测均方根误差(RMSEP)由0.4717下降到0.2249,预测精度提高了52.3%,极大程度简化了模型,提高了模型预测能力。
Moisture content is an important index for evaluating wheat quality;moisture content in wheat was tested and quantitatively analyzed by near infrared spectroscopy simultaneously,rapidly and nondestructively.According to model population analysis(MPA),the wavelength variable was selected which is combined with Bootstrapping Soft Shrinkage(BOSS).Variable weights were obtained by the regression coefficients of sub-models.Variable space was shrunk through stepwise updating of the variable weights based on weighted bootstrap sampling(WBS).The optimal variable set with the lowest root mean squared error of cross-validation(RMSECV)was selected.Based on this,a regression prediction model is established.The result shows that the predicted root mean square error(RMSEP)of BOSS-PLS model is reduced from 0.4717 to 0.2249,which is compared with full spectrum model.The prediction accuracy has been improved 52.3%.The model has been greatly simplified and the prediction ability has been greatly improved.
作者
孙大明
宦克为
SUN Da-ming;HUAN Ke-wei(School of Science,Changchun University of Science and Technology,Changchun 130022)
出处
《长春理工大学学报(自然科学版)》
2020年第5期1-6,共6页
Journal of Changchun University of Science and Technology(Natural Science Edition)
基金
吉林省科技发展计划项目(20190701024GH)。
关键词
小麦水分
近红外光谱
模型集群分析
变量选择
moisture content of wheat
near-infrared spectroscopy
model population analysis
variable selection