Estimation of Mango Canopy Nitrogen Content in Complex Backgrounds Using UAV Hyperspectral Imaging: A Fusion Method of Adaptive Pixel Weighting and Ensemble Learning

The hyperspectral data of mango canopies collected by UAVs are susceptible to interference from complex environmental factors. This study proposes a spectral feature selection method based on adaptive pixel weight optimization (SWA) to effectively extract key spectral signals from complex backgrounds, thereby enhancing the accuracy of the nitrogen content estimation model for mango canopies. Additionally, ensemble learning methods were introduced to further improve the prediction accuracy of nitrogen content. The results demonstrate that the SWA method significantly enhanced the correlation between spectral data and nitrogen content, increasing the correlation coefficient for reflectance in the 400-900 nm wavelength range to 0.5. After SWA denoising, the ensemble learning models outperformed traditional machine learning models, with the RF+SVR model showing the best performance under the stacking strategy, achieving an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> value of 0.95 on the test set and reducing the root mean square error (RMSE) to 0.45 mg/g. These results indicate that this method has high predictive accuracy and generalization capability for estimating nitrogen content in mango canopies.