Integrating remote sensing and geospatial big data for urban land use mapping: A review

Abstract

Remote Sensing (RS) has been used in urban mapping for a long time; however, the complexity and diversity of urban functional patterns are difficult to be captured by RS only. Emerging Geospatial Big Data (GBD) are considered as the supplement to RS data, and help to contribute to our understanding of urban lands from physical aspects (i.e., urban land cover) to socioeconomic aspects (i.e., urban land use). Integrating RS and GBD could be an effective way to combine physical and socioeconomic aspects with great potential for high-quality urban land use classification. In this study, we reviewed the existing literature and focused on the state-of-the-art and perspective of the urban land use categorization by integrating RS and GBD. Specifically, the commonly used RS features (e.g., spectral, textural, temporal, and spatial features) and GBD features (e.g., spatial, temporal, semantic, and sequence features) were identified and analyzed in urban land use classification. The integration strategies for RS and GBD features were categorized into feature-level integration (FI) and decision-level integration (DI). To be more specific, the FI method integrates the RS and GBD features and classifies urban land use types using the integrated feature sets; the DI method processes RS and GBD independently and then merges the classification results based on decision rules. We also discussed other critical issues, including analysis unit setting, parcel segmentation, parcel labeling of land use types, and data integration. Our findings provide a retrospect of different features from RS and GBD, strategies of RS and GBD integration, and their pros and cons, which could help to define the framework for future urban land use mapping and better support urban planning, urban environment assessment, urban disaster monitoring and urban traffic analysis.