Yield Forecast Of Ginkgo Leaf Using UAV Remote Sensing

Along with medicinal value and the healthcare function of the Ginkgo leaf gain much more recognit ion,the plant scale of leaf used Ginkgo is constantly expanding.This essay use the technology method to finish the leaf product ion predict ion quickly and accurately,which can be useful for field crop precise management and industrial scale and technology service policies formulat ion.In this paper,we use the UAV equipped with mult ispectral camera and design the reasonable flight path for it to get the remote sensing image of Ginkgo biloba field,remote sensing images are preprocessed by image mosaic,radiometric calibrat ion,point cloud correct ion,image registration and clipping.In this paper,we do the research on leaf production predict ion for the reginal Ginkgo bilo ba.We make the correlat ion analysis between leaf production and different growth stage,vegetation index.The mathemat ics regression models of simple linear,mult iple linear and SVR method are established to predict.By comparison,compound linear model based on the NDVI index from July and September has the excellent prediction performance(R2=0.834,RMSE=0.0944kg/m2,MRE=11.91%).For the leaf production predict ion before harvest which need leads,simple linear mathemat ics regression model of July NDVI index is effect ive(R2=0.75,RMSE=0.1093 kg/m2,MRE=14.67%).Meanwhile,the research show the UAV flight height has no signifi cant impact on the predict ion results.For leaf production predict ion of the reginal Ginkgo bilo ba,the data was acquired when the UAV flight height was 60 m in June and August,and the good leaf product ion prediction for the reginal Ginkgo bilo ba could be realized by using NDVI mult ivariate linear model.Meanwhile,we do the research on leaf production predictio n of the single Ginkgo bilo ba.watershed algorithm,local maximum,object-oriented mult i-scale image segmentation method are used to segment the crown of single Ginkgo biloba,crown area,point cloud height and vegetation index are extracted separately from the remote sensing images of the segmented crown to establish mathematics regression simple linear,mult iple linear and SVR model with leaf production predict ion of the single Ginkgo biloba.An d the research result show the watershed algorithm is the best method to segment the crown of single Ginkgo biloba,and the mat hemat ics regression model based on the volume of single Ginkgo bilo ba has the highest accuracy(R2=0.848,RMSE=27.83 g,MRE=44.29%).We removed the single Ginkgo bilo ba sample which leaf production is less than 50 g to optimize the model,and this act ion can improve the accuracy of model(R2=0.89,RMSE=20.11 g,MRE=13.61%).For leaf production predict ion of the single Ginkgo biloba,the data was acquired when the UAV flight height was 20 m in August,the volume parameters of single plant were extracted by dividing the crown of Ginkgo biloba by watershed algorithm,and the good leaf product ion prediction of the single Ginkgo bilo ba could be realized by using the simple linear model.The results of this research show that UAV Remote Sensing Technology can be used to finish the leaf production predict ion for Ginkgo biloba which can support to fix the procurement price and formulate local agricultural policies.Leaf product ion prediction for the single Ginkgo biloba provide the new method to finish the further phenot ypic study of large quant it ies of crops which can support to realize the cult ivat ion of highly-yielding Ginkgo biloba and the precise management of field crops.