Fine-Scale Assessment of Urban Canopy and Green-Cover Composition Across Functional–Morphological Scene Types Using UAV Orthophotos and SAM3

Coarse urban green-space assessments often merge tree canopy, low-lying vegetation, and impervious surfaces within large spatial units, failing to resolve fine-scale greening composition in heterogeneous built environments. To address this gap, we evaluated whether centimetre-scale unmanned aerial vehicle (UAV) imagery coupled with SAM3 concept segmentation can quantify sub-hectare urban green structure across functional–morphological scene types within a representative mountainous, ecological medium-sized city in East China characterized by compact built-up spaces constrained by valley geography. Using 746 sub-hectare analytical tiles, we extracted canopy, visible non-canopy green, and canopy-share metrics, applying source-fixed-effects models with source-clustered inference to compare scene types. Primary scene type driven by urban morphological forms accounted for substantial variation in total green cover (35.6%) and tree canopy (17.5%), with planned residential areas exhibiting over twice the total green cover of compact residential areas and industrial/compact residential scenes presenting the lowest coverage. These results reveal that fine-scale greening in compact, valley-constrained ecological cities is highly heterogeneous and strongly governed by morphological layout, demonstrating that combining ultra-high-resolution imagery with vision foundation models provides a scalable tool for identifying micro-greening deficits and guiding context-sensitive urban renewal in such urban contexts.