With diversification of data acquisitions and associated technologies, there is steady accumulation of multi-source spatial data, prompting more research on effective and fast data integration to provide information and technical support for geospatial research and applications. In this paper, procedures for integrating multi-source data are referred to as data fusion. Research developments are reviewed by considering them as classic and extended types of methods, respectively. To review the progress in classic methods for data fusion and related data processing, this paper clarifies some of the relatively well-established methods using a typology of object-and field-based models of spatial data. For extended methods, the paper discusses multi-point geostatistics, statistical-mechanistic methods, multi-scale signal analysis and reconstruction, and information-theoretic strategies. Some of the issues, such as scale mismatch, semantic inconsistency, and the temporal dimension, are discussed in the hope of better supporting the fusion of multi-sourced and heterogeneous data. Further research will enhance theoretical foundations of geographic information to enrich methodologies for spatial data and their analyses, and help add to the applicability and value of spatial information.