Rice Yield Estimation Based Satellite And Unmanned Aerial Vehicle(UAV)Multi-Sources Remote Sensing Data

Estimating crop yield timely and accurately is crucial for food reserves security in a country.Rice is one of the most important and widely grown crops in china,some agricultural information about rice such as yield is vital to formulate national agricultural policies and regulate food prices.Satellite remote sensing for agriculture has the advantages of non-destructive,non-invasive,fast and cost-efficient and it has been usually applied to estimate crop yield.With the development of UAV technology,all kinds of remote sensing sensors mounted on UAV which can acquire high spatial-temporal resolution remotely sensed data on demand and estimate crop yield accurately in parcel-level.In terms of satellite remote sensing estimation rice yield,we coupling trectangular hyperbolic photosynthetic experience model and photosynthetic mechanism model to establish a new daily GPP model,which takes satellite remote sensing data,meteorological data and vegetation types data as input parameters,simulating regional-scale crop canopy GPP accurately and efficiently.Besides,the conversion coefficient of carbon to biomass,harvest index and moisture content are included in this model.The dynamic harvest index contained in the model makes up for the shortcomings such as estimation error caused by fixed harvest index in traditional crop yield estimation model.Rice yield are estimated in Jiangsu Province from 2004 to 2014 by using this method,the results show that the average yield accuracy in Jiangsu Province from 2004 to 2014 are greater than 96%and achieve the goal of predicte rice yield accurately at regional scale.In the aspect of remote sensing rice yield estimation with UAV,we propose the concept of "relative spectral index" to eliminate the external conditions such as different atmospheric conditions,different illumination conditions and background values in different time-series vegetation indices which have negative influence in yield estimation.We estimate rice yield based on relative spectral variables with multiple growth stages in field scale and pixel scale.The results show that the multiple-growth-stage model RNDSI[808,744]at jointing stage,RNDSI[880,712]at booting stage and RNDSI[808,744]at filling stage is the optimal rice yield estimation model in field scale,and the mean absolute percentage error of estimated rice yield is 3.00%.RNDSI[784,635]at tillerling stage,RNDSI[807,744]at jointing stage,RNDSI[784,712]at booting stage and RNDSI[816,736]at heading stage is the optimal rice yield estimation model in pixel scale,and the mean absolute percentage error of estimated rice yield is 4.31%.So,rice yield can be estimatied accurately based the relative spectral variables both in field scale and pixle scale.