Above-ground biomass estimation and yield prediction in potato by using UAV-based RGB and hyperspectral imaging

Rapid and accurate biomass and yield estimation facilitates efficient plant phenotyping and site-specific crop management. A low altitude unmanned aerial vehicle (UAV) was used to acquire RGB and hyperspectral imaging data for a potato crop canopy at two growth stages to estimate the above-ground biomass and predict crop yield. Field experiments included six cultivars and multiple treatments of nitrogen, potassium, and mixed compound fertilisers. Crop height was estimated using the difference between digital surface model and digital elevation models derived from RGB imagery. Combining with two narrow-band vegetation indices selected by the RReliefF feature selection algorithm. Random Forest regression models demonstrated high prediction accuracy for both fresh and dry above-ground biomass, with a coefficient of determination (r(2)) > 0.90. Crop yield was predicted using four narrowband vegetation indices and crop height (r(2) = 0.63) with imagery data obtained 90 days after planting. A Partial Least Squares regression model based on the full wavelength spectra demonstrated improved yield prediction (r(2 )= 0.81). This study demonstrated the merits of UAV-based RGB and hyperspectral imaging for estimating the above-ground biomass and yield of potato crops, which can be used to assist in site-specific crop management.