Remote Estimation Of Rice Growth Parameter And Yield With Unmanned Aerial Vehicle Multispectral Data

Rice growth parameters refer to the complicated rice traits which temporally and spatially vary with rice growth and can reflect rice growth status,such as chlorophyll content,leaf area index(LAI)and above ground biomass(AGB).Timely and accurate measurement of rice growth parameters can monitor rice growth and predict rice yield at early growth stage.Traditionally,the measurement of rice growth parameters is by manual method in situ,which is time-consuming and labor-consuming.Moreover,the manual method may have destructive effects on rice plant and lose the final yield information.In the last decade,the development and steep rise of unmanned aerial vehicles(UAVs)has provided a novel platform for remote sensing,and makes it possible to acquire data of unprecedented spatial,spectral,and temporal resolution.Taking the advantages of real-time,fast,non-destructive and unaffected by weather condition,UAV remote sensing has great potential for retrieving rice growth parameters and yield at region scale.This study was supported by National High-tech Research and Development Program(863 Program)(2013AA102401)and National Key Research and Development Project(2016YFD0101105).Several studies have been conducted to estimate rice growth parameters and yield using UAV multispectral data in Hainan province and Hubei province respectively during 2015 to 2019,and the findings of this study include:(1)Vegetation index(VI)is one of spectral features of UAV image and has been widely used.Therefore,the relationship between VI and rice growth parameters was firstly analyzed to develop the estimation model of rice growth parameters based on VI under different nitrogen treatment and in different rice cultivars respectively.Results showed that the rice growth parameters could be accurately estimated in different growth conditions with VI and machine learning.(2)This study developed a new extraction method of image texture based on the Flourier transform with consideration of the row pattern of rice field.The Flourier texture inferred the width change of rice rows and thus directly related to rice growth.Flourier texture had stronger correlation with rice growth parameters.Results showed that the integration of Flourier texture and VI improved the estimation accuracy of LAI,canopy chlorophyll content and AGB.It can be inferred that the integration of texture and spectral feature may increase estimation accuracy of rice growth parameters based on UAV data.(3)Since UAV multispectral data can be used to estimate rice growth parameters in different rice growth conditions,UAV multispectral data was also used to estimate rice yield.Moreover,this study discussed the effects of rice panicle,rice cultivar and growth duration on yield estimation.Results showed that VI performed better in yield estimation than Flourier texture.However,the yield estimation accuracy based on VI decreased in heading stage,caused by the uneven emergence of rice panicle.In this study,a new approach which integrated VI and abundance information obtained from spectral mixture analysis was developed to improve the estimation accuracy of rice yield at heading stage.Besides,the single stage VI weakly correlated with the yield of different rice cultivars and was not able to estimate yield in different rice cultivars.By contrast,the multi-temporal VIs can be used to estimate yield in different rice cultivars.In addition,the change of rice growth duration may decrease the estimation accuracy of grain yield by using multi-temporal VIs.Therefore,the phenological stage of multitemporal VIs used in estimation model should be adjusted,when rice growth duration changed.