Coupling of machine learning methods to improve estimation of ground coverage from unmanned aerial vehicle (UAV) imagery for high-throughput phenotyping of crops

Ground coverage (GC) allows monitoring of crop growth and development is normally estimated as the ratio vegetation to total pixels from nadir images captured by visible-spectrum (RGB) cameras. The accuracy GC can be significantly impacted effect ‘mixed pixels’, which related spatial resolution imagery determined flight altitude, camera characteristics (fine vs coarse textures). In this study, a two-step machine learning method was developed improve wheat (Triticum aestivum L.) coarse-resolution RGB an unmanned aerial vehicle (UAV) at higher altitudes. classification tree-based per-pixel segmentation (PPS) first used segment fine-resolution reference into background pixels. their segmented were degraded target resolution. These then generate training dataset for regression model establish sub-pixel (SPC) method. newly proposed (i.e. PPS-SPC) evaluated with six synthetic four real UAV image sets (SISs RISs, respectively) different resolutions. Overall, results demonstrated that PPS-SPC obtained in both SISs RISs comparing PPS method, root mean squared errors (RMSE) less than 6% relative RMSE (RRMSE) 11% SISs, 5% RRMSE 35% RISs. potentially applied plant breeding precision agriculture balance requirement height limited battery life operation time.