Fusion of GPS and image data for accurate geocoding of street-level fisheye images

Geospatial tools and techniques are becoming more important for land surveyors to do their off-location inspections of the urban areas. Accurate geocoded street-level images are the base of these tools. For these applications, an error of 2.5 meters is tolerable. However, the geographic coordinates provided by GPS have error up to 10 meters. In this paper we propose an automatic method to improve the accuracy of geocoding of street-level images by registering them to the accurate geocoded reference image, which is the satellite image. The proposed technique uses an unconstrained nonlinear optimization method to find local optimal solutions by matching high-level features and their relative locations. A global optimization method is then employed over all of the local solutions by applying a geometric constraint. We used our algorithm for correcting the geographic information of more than 2500 fisheye images and show that the proposed algorithm can achieve an average error of 1.19 meters along both x and y directions.