Identification and Analysis of Urban Functional Zones Based on Multi-Source Geospatial Data—A Case Study of Central Kunming

Abstract

Urban functional zone (UFZ) identification supports refined territorial spatial governance and urban remote sensing. It faces the dual challenges of spectral confusion and functional mixing, which single-source remote sensing cannot resolve at high accuracy. Taking central Kunming as a study case, this paper integrates high-resolution remote-sensing imagery (HRI), points of interest (POIs), building footprints, and a digital elevation model (DEM) into a multi-dimensional feature system that combines spectral–textural features, POI functional semantics, building-morphology constraints, and topographic indicators. Mean shift object-based segmentation and a random forest are used for supervised UFZ classification, while an independent test set and ablation experiments quantify model performance and each feature group’s contribution. The optimal model reaches an overall accuracy (OA) of 82.22% and a Kappa of 0.7721, indicating reliable performance. Leave-one-out ablation shows that POI kernel density contributes most to identifying commercial and public zones, building morphology mainly improves industrial-zone accuracy, and terrain plays an auxiliary role in delineating ecological green-space boundaries. Landscape metrics and the standard deviational ellipse (SDE) are then applied to the classification results to examine functional spatial differentiation. Bounded by terrain and the Dianchi ecological barrier, central Kunming forms a composite pattern of one primary core, two secondary clusters, and interwoven ecological corridors, with clear differences among classes in extension direction, centroid location, and inter-district composition. The proposed multi-source fusion framework offers reliable methodological support for fine-scale UFZ mapping and refined territorial governance.