Estimation of Cotton Nitrogen Content Based on Multi-Angle Hyperspectral Data and Machine Learning Models

For crop growth monitoring and agricultural management, it is important to use hyperspectral remote sensing techniques estimate canopy nitrogen content in a timely accurate manner. The traditional nadir method has limited ability assess the trophic state of cotton shoots, which not conducive high-precision inversion, whereas multi-angle can effectively extract canopy’s physicochemical information. However, spectral information affected by variety factors, frequently causes shifts band associated with uptake, lowers estimation accuracy. capacity index aerial concentration (ANC) was therefore investigated this work under various observation zenith angles (VZAs), Relief−F employed select best weight for ANC that insensitive VZA. Therefore, study, explored different VZAs, Relief-F algorithm used optimize optimal angle (AINI) VZAs calculated using expression (R530 − R704)/(R1412 + R704). results show correlation between chosen study stronger than off-nadir observations, coefficients Photochemical Reflectance Index (PRI), AINI, are highest when VZA −20° −50° (r = 0.866 0.893, respectively). Compared vegetation index, AINI had > 0.84), performance backscatter direction estimated be better forward-scatter direction. At same time, model indices PRI combined machine learning achieve accuracy, prediction accuracy random forest (RF) R2 0.98 RMSE 0.590. This shows VZAs. Simultaneously, backscattered revealed more estimation. findings encourage observations nutrient estimation, will also help improve ground-based satellite sensors.