Urban Park green spaces serve as essential ecological foundations within urban ecosystems, with their spatial habitat structure playing a pivotal role in sustaining biodiversity, delivering ecosystem services, and improving landscape connectivity. However, existing habitat classification methods often suffer from limited identification accuracy, insufficient ecological indicative capacity, and poor alignment with practical management applications. Addressing these challenges, this study proposes a multi-source remote sensing-based habitat identification approach tailored to urban park green spaces, integrating both technical pathways and ecological applicability. Using six representative urban parks in Shanghai as case studies, the method integrates spectral index analysis derived from multispectral imagery, canopy structure extraction from LiDAR data, and manual interpretation of orthophotos. This approach aims to develop a hierarchical and multi-scale habitat classification system that comprehensively encompasses features from land cover to vegetation structure. A total of 14 typical urban habitat types were identified, with an overall classification accuracy of 0.843 and a Kappa coefficient of 0.830, indicating strong consistency in classification and ecological relevance. Further analysis revealed significant differences in habitat composition and spatial configuration among the parks, reflecting their unique landscape structure characteristics. This study not only achieves technical integration and innovation in urban remote sensing classification workflows but also establishes interpretative links between habitat types and ecological processes. The proposed approach provides a spatially explicit pathway to support urban ecological management. It can be applied to various fields, including urban ecological monitoring, assessing the functions of green spaces, and planning habitat optimization. It provides a case-based contribution to the adaptation of remote sensing ecological methodologies in urban contexts.