- English |
Buying Optical Satellite Imagery - continued page 4
8. Collection Capacity
Project Example: Ortho Mosaics
Most satellite imagery comes georeferenced but not orthorectified. Orthorectification is a process that corrects inherent distortions in the optics and viewing geometry. The process incorporates a satellite orbital model, a digital elevation model (DEM) and optionally photo-identifiable ground control points (GCPs). Orthorectification is the most reliable way to correctly georeference all points in an image.
In addition to improving the absolute accuracy of the imagery, orthorectification also improves the relative spatial match across adjacent scenes and strips. However, if the source scenes have different or opposite look angles, some spatial mismatch may be unavoidable, especially in high-relief areas.
If an area of interest (AOI) comprises more than one scene/strip, the imagery also can be tonally balanced. Tonal balancing matches colors that otherwise would vary across adjacent scenes due to a variety of factors, including atmospheric conditions and vegetation seasonality across multiple imagery acquisition dates. In cases of extreme seasonality differences, it may not be possible to create an entirely seamless tonal match.
Multiple overlappping scenes can be used to output a tonally balanced ortho mosaic with cloud patching
Typically, tonal balancing only is applied to land areas, not water, as the appearance of water features, especially saltwater, can vary greatly across multiple dates/scenes. If a visually pleasing or realistic picture is desired, water areas can be manually manipulated with software tools such as Photoshop, but image intelligence, such as shallow water depth, would be lost.
When creating orthomosaics, limiting the number of scenes facilitates processing and, if applicable, helps reduce the total number of optimal Global Positioning System (GPS) ground control points (GCPs). In the United States, public-domain products, such as imagery from the National Agriculture Imagery Program (NAIP), can be used as control to improve the ortho accuracy of imagery with lower native accuracy. Outside of the United States, if GPS GCPs aren’t available, ideally the next-generation imaging sensors with higher native accuracy, such as GeoEye-1, WorldView-1 and WorldView-2, should be favored.
Consideration should be given to the DEM being used. In the United States, 10-meter (1/3 arc) National Elevation Dataset imagery is the best public-domain option. However, there are also high-quality commercial DEM options. Outside the United States, commercial and/or local government DEM options are available. However, if DEM selection is limited to public-domain options, despite the resolution difference, 90-meter (3 arc) Shuttle Radar Topography Mission data will likely yield better results than a 30-meter (1 arc) ASTER GDEM, especially in areas with high cloud cover and/or low contrast (snow, dense vegetation, etc). Also, imagery that has already had a DEM applied to it, including the Standard format from Digital- Globe, isn’t suitable to be orthorectified—the Ortho Ready format should be selected instead.
To the extent possible, limiting the seasonality differences across strips helps improve tonal balancing. If an AOI comprises multiple scenes, some satellite operators will crop out part of the overlap area to reduce the file size. From an image processing perspective, however, especially where imagery needs to be cloud-patched, the overlap is desirable and should be available at no extra charge, as long as it’s requested when the order is placed.