Based on Google Earth Engine (GEE), Landsat-5 (TM), Landsat-7 (ETM) and Landsat-8 (OLI) data were used as remote sensing data sources to classify land use types from 1985 to 2020 with the random forest classification method. The fractal theory method was used to calculate land use fractal dimension and stability index, and the complexity and stability of the spatial structure of land use types were quantitatively analyzed in two stages of 1985—2005 and 2005—2020. The results showed that: 1) From 1985 to 2020, cultivated land gradually decreased, grassland changed little, water gradually expanded, with the highest growth rate in 2015—2020, bare land changed more irregularly, increasing first and then gradually decreasing, and buildings grew exponentially. 2) Fractal dimension of the spatial structure of land use types generally showed a decreasing trend, indicating a decrease in its complexity. 3) Stability index of the spatial structure of land use types tended to increase in general, indicating it tended to be stable. In conclusion, with the implementation of environmental protection policies, the spatial structure of land use types in the study area has been gradually improved, and natural resources such as water and grassland have been gradually enriched.