Research On Method Of Estimating Oilseed Rape Growth Parameters Based On UAV Borne Remote Sensing Data

Oilseed rape,as one of the oil crops with largest sowing area in China,has a wide range of distribution,its growth and development as well as formation of yield are greatly affected by nitrogen.The in-time,rapid,accurate monitoring of oilseed rape under different nitrogen application levels in test area has important and significant research meaning to production management and yield prediction.The growth situation of oilseed rape can be comprehensively reflected by concentration of leaf chlorophyll?Chl?,carotenoid?Car?and flavonoids?Flav?as well as height?H?,leaf area index?LAI?and so on.Unmanned aerial vehicle remote sensing technology,as one of the frontier methods of agricultural monitoring,is a highlight of precision agriculturual research and application which can make up for many shortcomings of traditional remote sensing technology by the advantages of low cost,multi-angle,high resolution,flexibility,simplicity,safety and easy operation.Based on the object of oilseed rape,this study established some inversion models of physiological and biochemical parameters at different nitrogen application levels and growth stage from unmanned aerial vehicle image and 5 bands multi-spectral data obtained by the UAV platform as well as the data measured in the ground.According to the above research contents,and the main results are as follows:?1?Vegetation index inversion model of oilseed rape biochemical parameters?NBI,Chl,Flav?and its influencing factors analysis.6 typical vegetation indexes?RVI?MCARI1?TCARI?OSAVI?TCARI/OSAVI?NDRE?were extracted based on mixed pixels and pure-rape pixels,and we analyzed the correlation between vegetation indices obtained by this two methods and NBI,Chl,Flav.The result indicated that the correlation was improved to a certain extent after the background factors such as soil and shadow were eliminated in the early growth stage,and there was little difference in budding stage.At all growth stages,the pearson correlation coefficient of 6 vegetation indexes and NBI were more than 0.79,Chl were more than 0.62,and Flav were more than 0.56.The determinant coefficient R² of the modeling set and verification set were higher,RMSE were smaller,and RPD were larger.Vegetation indices calculated by two methods are taken as independent variables respectively,and the LAI optimal inversion models were established for each growth stage using 5 kinds of regression analysis?linearity,exponent,logarithm,power,and polynomial?,and the results reveal that the model accuracy build on vegetation index of pure-rape pixels were better.Based on vegetation index of pure-rape pixels,the R²_cal>0.65,R² _val>0.75,RMSEC<2.46,RMSEP<3.19,RPD>1.51 of NBI optimal prediction model;the R² _cal>0.58,R² _val>0.51,RMSEC<1.99,RMSEP<2.51,RPD>1.47 of Chl optimal prediction model;the R²_cal>0.54,R² _val>0.69,RMSEC<0.07,RMSEP<0.09,RPD>1.42 of Flav optimal prediction model.The optimal regression models were mainly linear function,quadratic polynomial and exponential function,and most models choose NDRE as the optimal vegetation index.?2?Plant height estimation of oilseed rape based on UAV digital image.A complete process of rape plant height extraction based on UAV platform was established on digital surface model?DSM?data which generated from UAV point cloud data.The results of plant height extraction at different growth stages were excellent(R²_cal and R² _val>0.70,RMSEC and RMSEP<0.07m,RPD>2.0)and the best prediction accuracy of plant height was in the stage of ten-leaf(R²_cal=0.92,R²_val=0.97,RMSEC=0.02m,RMSEP=0.02m,RPD=4.1),which reveal that the whole process based on UAV digital image is a promising method to extract the plant height of oilseed rape,and provide a new method for measuring plant height.?3?Inversion of Rape LAI by Plant Height and Vegetation Index.6 vegetation indexes include NDVI,GNDVI,DVI,NLI,SAVI,Datt2,plant height and LAI had the highly significant positive correlation?r>0.74 in six leaf stage,r>0.80 in ten leaf stage,wintering and budding stage?.The LAI optimal inversion models were established using stepwise regression with 6 vegetation indices,6 vegetation indices and plant height taking as independent variables,respectively,and LAI as a dependent variable.The results reveal that the accuracy of LAI inversion is improved to a certain extent by adding plant height in the six leaf,ten leaf and wintering stage,and the addition of plant height showed no difference on the inversion accuracy of LAI in the budding stage.So the multiple collinearity and saturation between vegetation indices could be reduced to a certain extent and improved the accuracy in the LAI estimation based on multiple vegetation indexes by adding plant height in the early stage of growth.