基于无人机遥感与机器学习的水稻叶绿素反演方法研究OA

To address the challenges of high coupling of rice canopy spectral information by drone remote sensing and the vulnerability of retrieved features to background noise interference,this paper proposes a scheme based on scale analysis and feature optimization.The study introduced Continuous Wavelet Transform(CWT)to decompose high-dimensional spectra on multiple scales;compared multiple feature engineering methods to extract sensitive scales,and constructed a feature subset with high contribution and low redundancy;and systematically evaluated the prediction performance of multiple machine learning models.The results show that this method significantly compresses the feature dimensions,and the screened features can maintain high inversion accuracy and stability;the built integrated learning model has certain advantages compared with a single machine learning model,and the model determination coefficient(R2)is up to 0.786,providing a reliable technical path for accurate monitoring of rice chlorophyll content.