Deep learning in urban green space extraction in remote sensing: a comprehensive systematic review

Urban green spaces play indispensable roles in various aspects, including the urban soil environment, thermal environment, and human environment. Accurately capturing characteristics such as the spatial distribution of urban green spaces is essential for effective planning, construction, and assessment of green space, and it represents a focal point in urban ecological research. Leveraging its remarkable self-learning capability, deep learning has emerged as a prominent research direction for extracting urban green space features. However, given the diverse range of deep learning methods and the unique features of urban green spaces, the next focus will be on how to exploit more effectively and apply the advantages of deep learning to advance the survey of urban green spaces. Therefore, this study comprehensively reviews a large body of relevant literature, categorizing deep learning methods applied to urban green space extraction into supervised, unsupervised, and semi-supervised learning approaches. It systematically summarizes the technical foundations, research progress, and applicable public datasets in this field, provides a reasoned outlook on future directions, and aims to provide geographers and urban planners with a comprehensive and scientifically grounded reference.