Automating Image Map Generation Using Satellite Images and Aerial Photos

Image maps generated using satellite images and aerial photos are extremely useful. The map can provide detailed topographic, land-use and land-cover information on earth. With the growing number of satellite and aerial photo data, a technique which can generate/update maps automatically from satellite images and aerial photos is necessary. This paper presents a technique developed at PCI Geomatics which can generate/update maps automatically using satellite image or aerial photo data. The technique uses the ephemeris data and a geometric model for satellite data, and GPS/INS data together with the collinearity equations for aerial photos. In addition, a chip database containing a set of Ground Control Points (GCPs) image chips was used to collect GCPs automatically. To generate GCPs, a hierarchical multi-level correlation procedure was applied to match image chip stored inside an image database with raw images. This matching process includes the popularly used algorithms: sub-image pyramid, geo-transformation, resampling, correlation and least square matching. The automatically generated GCPs are then passed through a statistical filter to remove blunders, i.e., mismatched chips. Using the resulting GCPs, a refined, precise and reliable geometric model can then be obtained to generate the orthorectified images. An automatic mosaicking process, which can search the optimal cutlines, perform colour balancing and hot spot removal, was then used to generate a seamless mosaic. All the described processes and features in this paper have been fully integrated and implemented into some PCI software products, such as OrthoEngine and Geocomp-n systems.