| In recent years,remote sensing technology has developed rapidly.Because of its dynamic,timely and accurate advantages,remote sensing has become an effective means to obtain biochemical parameters of crops.In particular,ground close-range remote sensing and low-altitude remote sensing data are widely used in agricultural monitoring to achieve accurate management.Rice is one of the most important food crops in China,even in the world.Accurate and timely access to rice growth is of great significance to the balance of food supply and demand,agricultural policy formulation and macroeconomic planning.Leaf area index(LAI)is closely related to pigment content,carbon cycle,biomass and phenology of crops.It is an important agronomic parameter to characterize crop growth,canopy structure and crop yield prediction.Therefore,the accurate prediction of rice LAI by remote sensing can provide timely and accurate agricultural information for governments,which also establish foundation for rice yield prediction.In this paper,rice in Hubei and Hainan provinces was taken as the research object.LAI estimation model of multi-varieties hybrid rice was established based on two kinds of different scaled remote sensing data,ground hyperspectral data and UAV multispectral data.The main conclusions include:(1)The relationship between Rice Canopy Spectral and leaf area index(LAI)was analyzed.The hyperspectral data and multispectral data were significantly correlated with LAI in near infrared band and red-edge band.(2)On the ground platform,compare and analyze six methods of Vegetation Index,red edge parameter,Partial Least Squares Regression based on spectral characteristics,Multiple Stepwise Linear Regression,Stepwise Regression of deflection angle based spectral retrieval and Support Vector Machine Regression to estimate rice LAI.The stepwise regression method of partial angle spectral retrieval has the highest accuracy(R~2=0.70,CV=25.2%).The optimal model of each method is composed of red-edge band,which shows that red-edge band plays an important role in estimation of rice LAI.(3)On the UAV platform,compare and analyze of five methods of Vegetation Index,Partial Least Squares Regression,Multiple Stepwise Linear Regression,stepwise regression of deflection angle,Support Vector Machine Regression for rice LAI estimation accuracy,Support Vector Machine Regression method for LAI estimation accuracy is the highest(R~2=0.77,CV=19.2%).(4)Comparing the models of different platforms with different methods,the accuracy of different platforms is from high to low:MCA12>MCA6>ASD;among all models,the accuracy of Support Vector Machine Regression and deflection angle stepwise regression method is higher.Combine the UAV optimal model SVR with the ground optimal model DABSR to establish the rice LAI inversion model.Compared with the UAV model,the accuracy of the combined model is improved less.Therefore,MCA12 can be selected as the data source for LAI inversion,which requires less data and is easy to obtain. |
