The estimation of wheat yield combined with UAV canopy spectral and volumetric data

Abstract

AbstractEstimating wheat yield accurately is crucial for efficient agricultural management. While canopy spectral information is widely used for this purpose, the incorporation of canopy volumetric features (CVFs) remains underexplored. This study bridges this gap by utilizing unmanned aerial vehicle (UAV) multispectral imaging to capture images and elevation data of wheat at key developmental stages—gestation and flowering stages. We innovatively leveraged the elevation differences between these stages to calculate canopy height, develop a novel CVF, and refine the wheat yield prediction model across various wheat varieties, nitrogen fertilizer levels, and planting densities. The integration of canopy volume information significantly enhanced the accuracy of our yield prediction model, as evidenced by an R2 of 0.8380, an RMSE of 313.3 kg/ha, and an nRMSE of 11.33%. This approach not only yielded more precise estimates than models relying solely on spectral data but also introduced a novel dimension to wheat yield estimation methodologies. Our findings suggest that incorporating canopy volume characteristics can substantially optimize wheat yield prediction models, presenting a groundbreaking perspective for agricultural yield estimation.