Estimation of Above-Ground Biomass of Winter Wheat Based on Consumer-Grade Multi-Spectral UAV

One of the problems optical remote sensing crop above-ground biomass (AGB) is that vegetation indices (VIs) often saturate from middle to late growth stages. This study focuses on combining VIs acquired by a consumer-grade multiple-spectral UAV and machine learning regression techniques (i) determine optimal time window for AGB estimation winter wheat (ii) combination multi-spectral algorithms. UAV-based data manually measured wheat, under five nitrogen rates, were obtained jointing stage until 25 days after flowering in growing season 2020/2021. Forty-four used linear (LR), partial least squares (PLSR), random forest (RF) models this study. Results LR showed heading was most suitable prediction, with R2 values varying 0.48 0.93. Three PLSR based different datasets performed differently estimating training dataset (R2 = 0.74~0.92, RMSE 0.95~2.87 t/ha, MAE 0.75~2.18 RPD 2.00~3.67) validation 0.50~0.75, 1.56~2.57 1.44~2.05 1.45~1.89). Compared models, performance RF more stable prediction 0.95~0.97, 0.58~1.08 0.46~0.89 3.95~6.35) 0.83~0.93, 0.93~2.34 0.72~2.01 1.36~3.79). Monitoring prior found be effective than post-flowering. Moreover, demonstrates it feasible estimate multiple stages which overcomes saturation problem using individual VI-based models.