基于无人机多光谱影像与机器学习算法的棉花冠层叶绿素含量估算研究OA北大核心CSTPCD

[Objective]This study aims to monitor the chlorophyll content of cotton leaves by utilizing unmanned aerial vehicle(UAV)-based multispectral technology.[Methods]This study utilized multispectral images of cotton canopies obtained by UAV in southern Xinjiang and employed seven different machine learning methods to estimate the canopy chlorophyll content during the flowering and boll-setting stage which is the critical growth period of cotton.The seven methods include linear regression(LR)-based methods,i.e.,simple linear regression,partial least squares regression(PLSR),ridge regression(RR),least absolute shrinkage and selection operator(LASSO)regression,support vector regression(SVR),K-nearest neighbors regression(KNNR),and random forest regression(RFR).[Results]The results showed that compared with the LR method,the RFR,SVR and KNNR can improve the accuracy of prediction model of chlorophyll content in cotton canopies,especially the RFR algorithm,which had the coefficient of determination of 0.742,root mean square error of 1.158 mg L-1,residual predictive deviation of 1.969 with the validation dataset.[Conclusion]The use of UAV-based multispectral images with the RFR machine learning method,exhibits the most outstanding performance to estimate the chlorophyll content of cotton leaves and provide essential technical support for precision management in cotton field.