“…In general, researches are based on expert knowledge, correlation analysis, and feature selection to dimensionality reduction and remove redundant band information. For example, Li et al 11 and Ma et al 12 used canopy spectral features and leaf area index (LAI) at different growth stages to optimized quantitative remote sensing inversion models, then obtained the best feature parameters and realized high precision inversion. Previous studies mainly focused on the analysis of a single class of crops based on the average spectral data for feature analysis, which resulted in the weakening of the minor spectral differences between various crops and the lack of significant features of the mixed planting area of multiple crops.…”