基于无人机多光谱遥感的矮林芳樟叶片精油产量反演

芳樟(Cinnamomum camphora(Linn.)Presl)精油在林业经济发展中具有巨大市场潜力，多光谱遥感产量预测是高效反演芳樟精油产量的新方式。本研究以矮林芳樟收获期精油产量为研究对象，利用无人机多光谱遥感技术，筛选敏感植被指数作为输入变量，以地面同步观测的精油产量作为输出变量，采用支持向量机(Support vector machine, SVM)、随机森林(Random forest, RF)和反向传播神经网络(Back propagation neural network, BPNN)3种机器学习方法构建矮林芳樟精油产量预测模型。结果表明，修改型土壤调节植被指数(MSAVI)、优化土壤调节植被指数(OSAVI)、重归一化植被指数(RDVI)、土壤调整植被指数(SAVI)和非线性植被指数(NLI)对矮林芳樟精油产量呈现较高敏感性，其相关系数R分别为0.765 1、0.813 1、0.771 1、0.779 4、0.818 3。SVM、RF、BPNN 3种机器学习方法构建的矮林芳樟精油产量预测模型训练集的决定系数R²分别为0.723、0.853、0...

Inversion of Leaf Essential Oil Yield of Cinnamomum camphora Based on UAV Multi-spectral Remote Sensing

Cinnamomum camphora(Linn.) Presl essential oil has great market potential in the development of forestry economy. Multi-spectral remote sensing yield prediction is a new way to efficiently invert C.camphora essential oil. The yield of essential oil in the harvest period of C.camphora was taken as the research object. Using UAV multispectral remote sensing technology, the sensitive vegetation index was selected as the input variable, and the essential oil yield of ground synchronous observation was taken as the output variable. Three machine learning methods, support vector machine (SVM), random forest (RF) and back propagation neural network (BPNN), were used to construct the estimation model of essential oil yield of C.camphora. The results showed that modified soil adjusted vegetation index (MSAVI), optimized soil adjusted vegetation index (OSAVI), renormalized difference vegetation index (RDVI), soil adjusted vegetation index (SAVI) and nonlinear vegetation index (NLI) were highly sensitive to the essential oil yield of C.camphora, and the correlation coefficients R were 0.7651, 0.8131, 0.7711,0.7794 and 0.8183, respectively. The yield prediction models for essential oil of C.camphora were constructed by using three machine learning methods, SVM, RF, and BPNN. In the training set, the coefficients of determination R2 were 0.723, 0.853 and 0.770, respectively; the root mean square errors (RMSE) were 11.649kg/hm2, 9.179kg/hm2 and 10.484kg/hm2, respectively; the mean relative errors (MRE) were 7.204%, 10.808% and 7.181%, respectively. In the validation set, the R2 of validation set were 0.688, 0.869 and 0.732, respectively; RMSE were 7.951kg/hm2, 5.809kg/hm2, 8.483kg/hm2; MRE were 6.914%, 5.545%, 7.999%, respectively. Through the comprehensive comparison, with MSAVI, OSAVI, RDVI, SAVI, NLI as input data, the prediction model of C.camphora essential oil yield based on RF method achieved the highest accuracy. The research can provide a theoretical basis for improving the prediction accuracy of essential oil yield of C.camphora leaves based on UAV multi-spectral remote sensing and provide technical support for rapid monitoring of large area economic plant growth.