Plant height is an essential trait to evaluate in grain sorghum, being positively associated with potential grain yield. Standard manual measures of plant height for large breeding trials are labour-intensive and time-consuming. Due to potential field access issue and the remote nature of breeding trials, Unmanned Aerial vehicles (UAVs) are well-suited to measure plant height if the ground surface can be referenced. In this study, we compared existing algorithms with a new method for estimating plant height for a sorghum breeding trial. Images were captured by a RGB camera mounted on an UAV before emergence and near maturity to generate digital surface models (DSMs). Two existing methods (‘point cloud’ and ‘reference ground’) and a new method (‘self-calibration’) were used to estimate ground level and plant height at the plot level. The self-calibration method required manual measurements of the actual plant height in a sample of plots (fewer than 30), which could be completed during the 30-min flight time. UAV-derived plant heights from each method were compared to manual measurements. The self-calibration method had the best performance (R2=0.63; RMSE=0.07m; repeatability=0.74), with similar repeatability to manual measurement (0.78). The point cloud and reference ground methods had lower repeatabilities (0.34 and 0.38, respectively). For the self-calibration method, we tested different sampling strategies to balance accuracy and the workload of manual measurements, finding that a sample of 30–40 plots from the1440 total could obtain precision similar to manual measurement of the entire trial. The self-calibration method offers a pragmatic, robust and universal approach to high throughput phenotyping of plot plant height with UAV surveys.