Citrus Canopy Characteristic Extraction And Yield Estimation Based On UAV Oblique Photography

Citrus is an important industrial pillar in southern rural of China,which is of great significance for farmers to get rid of poverty and rural revitalization.As perennial evergreen fruit tree,citrus canopy characteristics have important indicative significance for tree vigor,and can further analyze its relationship with growth status,physiological indexes and production.Therefore,rapid and accurate acquisition of canopy characteristics of citrus tree,such as tree height,canopy projection area and canopy volume,is conducive to not only the accurate variable application of fertilizer,water and pesticide but also the efficient accurate estimation of yield,and promote the popularization and application of precision agricultural technology in citrus industry.Affected by the regional distribution of citrus industry and the natural growth pattern of canopy,the accuracy and efficiency of manual measurement and ground measurement cannot meet the corresponding needs.With the development of unmanned aerial vehicle(UAV)technology,especially oblique photography and automatic navigation technology,it shows important development potential and application value in the field of precision agriculture.In this study,the oblique photography images of citrus orchards were obtained with the optimized acquisition parameters.After 3D modeling,the Canopy Height Model of single tree was obtained,then the tree height,canopy projection area and canopy volume were extracted from CHM.The canopy characteristics extraction method of Navelina navel orange tree based on UAV oblique photography was established,which proved the potential of UAV oblique photography for citrus canopy characteristics extraction,and the obtained data were used for production prediction.This study provides theoretical basis and technical support for rapid and accurate extraction of canopy characteristics and production prediction of citrus.The main contents and conclusions of this study are as follows:(1)Study on data acquisition parameters optimization of UAV oblique photography and generation of canopy point cloud.Based on the actual needs,multi-direction mode was selected for image acquisition in the study.Taking the cube as the object target,the different combinations of flight height,lens oblique angle,and overlap degree were set.Considering the two factors of image acquisition efficiency and volume extraction accuracy,the image acquisition parameters were set as follows:flight height was 20m,oblique angle of lens was 45°,longitudinal overlap was 75%and lateral overlap was 70%.Based on the above image acquisition parameters,the oblique image acquisition of Navelina navel orange orchard was completed.After three-dimensional reconstruction,the point cloud of the study area was obtained.After denoising,point cloud classification and point cloud segmentation,the CHM of single tree was obtained,which provides basic data for further canopy characteristics extraction.(2)Study on canopy characteristic extraction based on UAV oblique photography.The tree height was extracted directed from the obtained CHM of single tree,and the canopy projection area was obtained by calculating the area of convex hull of the point set obtained by projecting CHM of single tree onto the horizontal plane.And three volume extraction methods(Convex-hull algorithm,Trapezoidal projection summation method and Cone volume accumlation method)were used to extract the canopy volume of Navelina navel orange trees,and the extraction accuracy of the three algorithms was compared by using the determination coefficient(R~2),root mean square error(RMSE),mean absolute deviation(MAD)and absolute mean relative error(AMRE).The R~2between the extracted and measured volumes of the three algorithms were 0.8487,0.7422,0.7967;RMSE were 1.4966,0.8807,0.5809;MAD were 1.3437,0.7845,0.4744;AMRE were 64.20%,37.30%,23.28%,respectively.And the RMSE,MAD and AMRE of the Cone volume accumulation method of canopy volume were the smallest.Therefore,Cone volume accumlation method was selected as the optimal method for canopy volume extraction.Based on the mannual calculation results of canopy characteristics in the second phenological drop(SPD)period,the R~2between the three extracted characteristics and the mannual calculation results were 0.7082,0.7858,0.7967 and RMSE were 0.3365 m,0.5574 m~2,0.5809 m~3respectively,which proved the application of UAV oblique photography technology to the canopy characteristics extraction of Navelina navel orange tree is feasible.Based on the above method,the tree height,canopy projection area and canopy volume in turning stage(TU)period and mature period were extracted.(3)Study on production prediction technology of Navelina navel orange tree based on canopy characteristics.By analyzing the relationship between production and canopy characteristics at different phenological periods,the Second Phenological Drop(SPD)period was selected as the optimal production prediction phenological period.The optimal production prediction model of Navelina navel orange tree based on canopy volume in SPD was developed using Levenberg-Marquardt(LM)algorithm with R~2=0.7961,RMSE=11.85,MAD=9.89 and AMRE=14.25%.The R~2,RMSE,MAD and AMRE of the sample plants in the validation set were 0.9133,5.04,3.76 and 4.43%respectively,which proved the applicability of the model in production practice.The results show that it was feasible and effective to obtain the canopy characteristics by UAV oblique photography and then apply it to the production prediction of Navelina navel orange tree.