Fig. 1. The original reflectance spectra of desert grassland plants

Fig. 2. Obfuscation matrix of random forest model Note: the off-diagonal bubbles represent misjudgment, and the size of the bubbles represents the number of judgment. The larger the number of samples, the larger the bubble

Fig. 3. Variale importance of random forest classification model Note: The blue column in the figure is the importance of the variable. The larger the value of the variable, the stronger the importance of the variable. The yellow column is the Gini coefficient. The higher the coefficient, the better the classification cut

Fig. 4. Obfuscation matrix of SVM model

Fig. 5. KNN model error graph

Fig. 6. Obfuscation matrix of KNN model

Fig. 7. Original reflectance spectra

Table 1. Types and geographical positions of the grassland communities

Table 2. Instrument parameters and requirements

Table 3. Plant directory

Table 4. Spectral Index

Table 5. Gamma and cost