多元回归与BP神经网络在烟草叶片SPAD反演时的建模精度比较

Comparison of Modeling Accuracy Between Multiple Regression and BP Neural Network in SPAD Inversion of Tobacco Leaves

为比较多元线性回归和BP神经网络在烟草叶片SPAD反演建模时的效果,采集烟草冠层的多光谱影像并提取植被指数,分别构建多元线性回归和BP神经网络反演回归模型。结果表明,基于提取出的24种植被指数所构建的反演模型中,以BP神经网络的回归反演效果最好,模型的R2达到0.85,RMSE为2.21;采用多元线性回归方法所构建的模型的R2仅为0.51,RMSE为1.52。该研究的结果说明,在进行烟草叶片SPAD监测时,可以采用BP神经网络构建反演模型。

In order to compare the effects of multiple linear regression and BP neural network in the SPAD inversion modeling of tobacco leaves,the multispectral image of the tobacco canopy was collected and the vegetation index was extracted.Then,using multiple linear regression and BP neural network algorithm,the inverse regression model was constructed respectively.The results showed that among themodels,the R2 of the model built by BP neural network was 0.85,and the RMSE was 2.21.By contrast,the R2 of the model constructed by the multiple linear regression method was only 0.51,and the RMSE was 1.52.The regression inversion precision of BP neural network was the better.The results of this study show that the BP neural network can be used to construct the SPAD inversion modeling of tobacco leaves.