烤烟叶片含水量的光谱检测模型研究  

Study on Spectral Monitoring Model of Tobacco Water Content

作　　者：李玉鹏[1] 凌智钢[1] 苟正贵 魏晓楠[1] 唐延林[1] 

机构地区：[1]贵州大学理学院,贵阳550025 [2]贵州省福泉市烟草公司,贵州福泉550500

出　　处：《中国农学通报》2013年第34期158-161,共4页Chinese Agricultural Science Bulletin

基　　金：国家自然科学基金"基于温度和施氮量的水稻品质遥感监测模型研究"(41061039);国家自然科学基金"烤烟理化参数的光谱监测机理与方法研究"(11164004);黔南州烟草公司技术开发基金"黔南烤烟良好农业规范体系建设及其应用研究"(201205)

摘　　要：为研究烤烟叶片中含水量随烘烤时间变化的问题,通过烘烤试验探索烤烟叶片含水量的可见-近红外光谱监测模型。针对光谱原始数据和预处理数据,利用交叉验证法采用偏最小二乘法建立回归模型,发现经过平滑处理的数据利用偏最小二乘法(PLS)可以较好地检测烤烟叶片含水量。基于全部波长建立回归模型,训练集r=0.9771,RMSEC=0.0742;交叉验证结果r=0.9573,RMSEC=0.1009;预测集r=0.9683,RMSEC=0.0862。结果表明,基于全部波长原始数据平滑处理的PLS模型预测烤烟叶片含水量是可行的。In order to study the change of water content of tobacco leaf with their drying time, the author explored an effective V-NIR spectral monitoring model of the water content of tobacco leaf through drying experiment in this article. Contraposing the primitive spectral data to the preprocessed spectral data, the author built a regression model through cross validation of partial least squares （PLS）. It had been found that the PLS model of smoothing data cloud monitor the water content of tobacco leaf very well. In the regression model basing on all wavelengths, there were the regression coefficient r=0.9771, RMSE=0.0742 in training set, r=0.9573, RMSE=0.1009 in cross validation set, and r=0.9683, RMSE=0.0862 in prediction set. The experiment results showed that it was feasible to monitor the water content of tobacco leaf by the PLS model basing on the smoothing primitive spectral data of all wavelengths.

关 键 词：可见-近红外光谱 烤烟叶片 含水量 偏最小二乘法(PLS) 

分 类 号：S127[农业科学—农业基础科学]