Estimation of maize LCC and FVC for full growth periods based on UAV remote sensing and ensemble learning

Fractional vegetation cover (FVC) and leaf chlorophyll content (LCC) are essential parameters that reflect crop growth. An empirical approach to estimating LCC and FVC involves analyzing vegetation index (VI) derived from crop canopy reflectance and integrating them with statistical regression methods. However, when estimating crop parameters throughout full growth periods, conventional single models frequently exhibit significant accuracy variations across different growth stages. To overcome this limitation, we introduced a novel phenology-adaptive framework using unmanned aerial vehicle (UAV) remote sensing and advanced ensemble learning (Stacking, Bagging, Blending) for maize monitoring. By synergistically integrating VI and texture features (TF) from UAV imagery as input variables, this study conducted the first systematic comparison of three ensemble models for estimating LCC and FVC. The estimation accuracy of VI, TF, and combined VI + TF inputs was rigorously evaluated. Notably, the Stacking model with VI + TF inputs achieved breakthrough performance: LCC (R 2 = 0.945, root mean squared error (RMSE) = 3.701 Soil-Plant Analysis Development (SPAD) units, mean absolute error (MAE) = 2.968 SPAD units) and FVC (R 2 = 0.645, RMSE = 0.045, MAE = 0.036), demonstrating >20% accuracy gain in early growth stages. This research successfully applied spectral, texture features, and ensemble learning to achieve high-precision estimation of LCC and FVC, providing a methodological reference for high-performance crop trait parameter estimation.