Land use and land cover change reflects the evolution of urban spatial structure and reflects the state of urban development. In order to study the law of land use change in Wuhan city, the optimized recurrent neural network model is used to classify the remote sensing images in 2018, 2020 and 2023. Characteristics of its area change and spatial change have been analyzed in combination with land type transfer matrix. It is showed that the optimized recurrent neural network performs well in processing remote sensing image data. The test accuracy in three periods is 93.13%, 93.22% and 91.78% respectively, and the Kappa coefficients are all over 0.85, so the classification accuracy is good. In addition, during the study period, the water areas and water conservancy facilities in Wuhan city are mainly transferred to cultivated land, while the cultivated land is mainly transferred to towns and industrial and mining land. There is a similar transfer-in relationship between forest land and cultivated land, and the unused land is transferred to cultivated land, towns and industrial and mining land. The research results reveal land use changes in Wuhan city from 2018 to 2023. It can provide valuable references for urban land management and planning.