Research On The Estimation Of Citrus Yield Based On The Spectral Data By ASD Field Spectrometer

This research investigated the main growth stages of citrus,booting stage, frondesce stage,dormancy stage, maturity stage. Two analytical methods was applied in this paper, one was analysis method of multivariate statistical and stepwise regression,which was used to establish the relationship between canopy spectral or vegetation indices and citrus production; the other analysis technology was based on citrus canopy spectral variables, which was used to screened the right spectral variables agronomic parameters to estimate citrus yield.lt indicated that hyperspectral remote sensing technology was able to estimate citrus yield.The main results were as follows:1.The models of yield estimation in single phase were investigated by filtering out the best spectral variable in different growth stages.The fitting of exponential equation,quadratic and cubic equations was better than other equationgs for the best yield estimation model.The precision of cubic equations is best in the booting stages with the variable of NDVI_green_p.2.1n the four models of multitemporal yield estimation, the coefficient of determination of models with the group of booting and maturity stages was better than other groups of two growth stages,the coefficient of determination of models with booting stage,frondesce stage and dormancy stage is better, the coefficient of determination of models with the group of the four stages is better than foregoing multitemporal models.The bset accuracy model is the model with spectral variable of booting and frondesce stage,the coefficient of determination of that model is0.898.3.Among four compound spectrum estimation model of spectral variable -agronomy parameters-yield,the most precisions model is the group of booting-frondesce compound model,the coefficient of determination was0.964.In the model of best spectral variable-agronomy parameters-yield,the most precisions model is also the group of booting-frondesce compound model.the coefficient of determination was0.908.But the RSEM of the fomer model is1.362,less than the latter model’s RSEM(2.471).So,the best model in compound spectrum estimation model is best spectral variable-agronomy parameters-yield in booting-frondesce stage.4.Comparing the above model with the model based on (NDVI)-chlorophyll content-yield,the coefficient of determination of models in article far outweigh this model.It illustrate that models in article is better.But in terms of model accuracy, the model based on (NDVI)-chlorophyll content-yield is better.Considering the number of sample and complexity of model,the accuracy of this model is also good.