<p>China has entered the late stage of urbanization, where cities of various sizes have been initiating the restoration and construction of urban ecosystems, playing a crucial role in enhancing urban ecological functions. As the earliest city in China to complete urbanization, Shenzhen has implemented a series of urban ecological restoration and construction projects over the past two decades, significantly improving urban ecological quality. In this study, we focused on the Bao’an District in Shenzhen, utilizing high spatial resolution remote sensing images obtained from 2008 to 2023, employing six machine learning algorithms and the landscape security pattern analysis model, to systematically analyze the changes in urban greenspaces at the patch and landscape scales over the 15-year period and their impacts on the regional ecological security patterns. The analytical results indicated that, among the six machine learning models, the U-Net + model demonstrated the highest performance in extracting urban greenspaces. Secondly, the study period witnessed an increase in both the area and connectivity of urban greenspaces in the study area. Finally, the regional ecological security patterns across the study area have been significantly restored, with increases in both the low-security level and the high-security level area coverage of urban greenspaces. The improvements in the ecological qualities of the study area is primarily attributed to the designation of urban ecological redlines, the implementation of urban ecological restoration and construction projects, and the transformation of urban economic development. This study holds significant value for systematically assessing the changes in urban greenspaces of Chinese cities during the late stage of population urbanization. In addition, the above findings also provide practical guidance for optimizing urban greenspace patterns and enhancing urban ecological security.</p>