【Aim】 As one of the typical alien invasive species in China, the large scale invasion of Eichhornia crassipes has caused serious damage to the aquatic ecosystem. At present, the research accuracy of remote sensing monitoring methods of E. crassipes in different habitats is different. In this study, different classification methods were compared to screen out suitable classification methods for E. crassipes in the southern region of China. 【Method】 Based on Sentinel-2 and Landsat8 OLI multispectral images, maximum likelihood and support vector machine supervised classification, decision tree classification and vegetation index threshold classification methods were selected to classify the E. crassipes of five reservoirs in Hainan Province. The classification accuracy of different methods was evaluated according to the visual results of UAV optical images. 【Result】 The results showed that the classification accuracy of the decision tree based on the time phase characteristics of E. crassipes was the highest, and the overall accuracy was more than 90%. In the classification method based on spectral features, the user accuracy of maximum likelihood supervised classification is 77.88%, the mapping accuracy is 72.44%, and the user accuracy and producer accuracy of support vector machine classification are 87.00% and 84.48%, respectively. 【Conclusion】 The accuracy of decision tree classification based on time phase and spectral features is higher than that of supervised classification only based on spectral features. The simple vegetation index threshold method is difficult to distinguish the different habitats of E. crassipes. The results of this study can provide scientific basis for remote sensing monitoring and early warning of E. crassipes in southern China.