In order to understand the spatial distribution and composition of green spaces in major cities of Yunnan Province,the Wenshan,Jinghong,Mangshi,Shangri-La,and Yuxi are selected as the study areas.Based on Sentinel-2A satellite remote sensing images,features including spectrum,texture,vegetation index and terrain characteristics are extracted and.The random forest algorithm is used to classify the four types of green spaces:park green space,protective green space,affiliated green space,and production green space.The results show:1)Among the factors participating in the classification of four urban green spaces types,elevation exhibits the highest contribution.2)The mapping accuracy(PA)and user accuracy(UA)of green space classification in the five main urban areas surpass the accuracy of non-green space,with an overall accuracy(OA)and harmonic average F1 accuracy both exceeding 84%,and a Kappa value of 0.75.3)The classification accuracy of the four green spaces types in Jinghong City main urban area and Shangri-La main urban area is superior to that of Mangshi Main urban area,Wenshan main urban area and Yuxi main urban area.4)The total green space area in the five main urban areas of Wenshan,Jinghong,Mangshi,Shangri-La and Yuxi amounts to 6.22,20.81,2.52,6.65 and 8.58 km²,respectively.The random forest algorithm effectively classifies urban green spaces with high accuracy,thereby providing substantial technical support for green space resources planning and ecological environment protection management in Yunnan Province.