近红外光谱法快速测定欧美杨戊聚糖含量  

Rapid Determination of Pentosan Content in Populus×euramericanaby Near Infrared Spectroscopy

作　　者：刘栋梁[1] 刘胜[1] 

机构地区：[1]北京林业大学理学院,北京100083

出　　处：《西北林学院学报》2013年第5期167-171,共5页Journal of Northwest Forestry University

基　　金：国家自然科学基金项目(61179034)

摘　　要：利用欧美杨近红外光谱数据和戊聚糖含量数据,采用一种多模型建模方法,实现了对欧美杨戊聚糖含量的快速预测。预测值与试验值的相关系数为0.935 0,均方差系数为0.336 7,预测结果的平均相对误差为0.012 8。为进一步减小预测误差,对光谱数据进行了一种函数变换,将函数变换方法与多模型方法结合重新建模,预测值与试验值的相关系数增大为0.957 4,均方差系数减小为0.245 8,预测结果的平均相对误差减小为0.008 2。文中方法可以尝试应用于欧美杨其他化学成分含量的测定。A multi-model method was adopted by using near infrared spectral data of Populus X euramerica na and pentosan content to achieve a rapidly prediction of pentosan content. The correlation coefficient be tween the predicted value and the experiment value was 0. 935 0. The mean square error of prediction was 0. 336 7. The mean relative error of the result of prediction was 0. 012 8. To further reduce the prediction error, a function transformation method was conducted to the spectral data. A new model was established through the combination of function transformation method and multi-model method. The correlation coef- ficient between the predicted value and the experiment value increased to 0. 957 4. The mean square error of prediction decreased to 0. 245 8. The mean relative error of the result of prediction decreased to 0. 008 2. It is demonstrated that the method in this paper could be used to determine the contents of some other chemical components in Populus X euramericana, ide theoretical guide and basis for development and inno- vation of contemporary gardens.

关 键 词：近红外光谱 戊聚糖 欧美杨 多模型 

分 类 号：S792.119[农业科学—林木遗传育种]