Research Progress on Urban Built-up Area Identification in High-Resolution Remote Sensing Images Using Deep Learning

Against the backdrop of rapid urbanization, the urban built-up area, as the core component of urban development, requires accurate identification for urban planning, land use monitoring, and sustainable development assessment. How-ever, the complex internal features, significant scale variations, and blurred boundaries within high-resolution remote sensing images often lead to insuffi-cient accuracy with traditional identification methods. Deep learning, leverag-ing its powerful capabilities for automatic feature learning and semantic extrac-tion, has become the core technology for identifying urban built-up areas in high-resolution remote sensing images. This paper provides a comprehensive review of research on high-resolution remote sensing image-based urban built-up area identification from a deep learning perspective: Firstly, it clarifies the connotation of urban built-up areas and summarizes their identification characteristics and challenges. Secondly, it categorizes the commonly used mul-ti-source data types for identification, including high-resolution remote sensing imagery, multispectral data, and LiDAR data. Thirdly, it summarizes methodo-logical advancements in built-up area identification using deep learn-ing-based semantic segmentation models (e.g., CNN, FCN, U-Net, Transformer). Finally, it analyzes the current problems and challenges in the research and proposes prospects and suggestions for future development directions.