Maritime surveillance of coastal regions involves processing of data from a large number of heterogeneous surveillance sources. The generation of effective Maritime Domain Awareness requires that the tracks from these sources be fused. The vast majority of track pairs in an area of interest do not represent the same vessel. These fusion candidate pairs can often be rejected based on their geo-feasibility. This paper presents the geo-feasibility concept and its role in track-to-track fusion support in an extremely heterogeneous maritime surveillance environment.