Fig. 1. DF algorithm flow chart

Fig. 2. Particle swarm deep forest model model overall prediction method

Fig. 3. PSO-DF algorithm flow chart

Fig. 4. Experimental plot design

Fig. 5. Unmanned rice hyperspectral data collection

Fig. 6. Canopy spectral reflectance of rice at the filling stage with different levels of nitrogen application

Fig. 7. The bands in the spectrum with high correlation coefficients with rice nitrogen data

Fig. 8. The inverse fit of each model for the nitrogen content of rice leaves on the modeling set

Fig. 9. The inverse fit of each model for the nitrogen content of rice leaves on the validation set

Fig. 10. Inversion results of each model for nitrogen content of rice leaves