Orthorectification—a subtopic of georeferencing—is the process of converting images into a form suitable for maps by removing sensor, satellite/aircraft motion and terrain-related geometric distortions from raw imagery. This is one of the main processing steps for evaluating remote sensing data. Orthorectified data sets are required for most applications involving thematic image analysis, especially when using the image data in Geographic Information Systems (GIS), for data fusion and analysis of data from different sources or seasons, when overlaying images with existing data sets and maps, or using them for evaluations like change detection and map updating. With the increasing geometric resolution of modern satellite cameras, requirements for the geometric accuracy of orthorectification also increase. IMF develops techniques and procedures for orthorectification of optical data acquired from a wide variety of sensor systems; these can also be used in operational processing chains.
Today various methods and models of different complexity exist for orthorectification. At IMF the technique of “Direct Georeferencing and Rational Polynomial Functions” is mainly in use. Direct Georeferencing (DG) is based on a Line-of-Sight (LoS) model which makes extensive use of on-board measurements from Star Tracker and inertial measurement systems, often combined using Kalman filtering. High precision attitude and orbit determination systems (position and velocity) are also employed, like GPS (Global Positioning System), DORIS (Doppler Orbitography and Radiopositioning Integrated by Satellite), and in the future GALILEO, representing next-generation technology. The interior orientation—the sensor model and the sensor mounting (boresight angles)—are taken into account by using look direction vectors derived from laboratory and/or in-flight geometric calibration. The Universal Sensor Model (USM)—realized by Rational Polynomial Functions (RPF)—transforms object space coordinates to image space coordinates, where the exterior and interior orientations are implicitly encoded in the form of rational polynomial coefficients (RPC) using third order polynomials for nominator and denominator (80 coefficients). For both methods terrain displacements are taken into account using digital elevation models. The reconstructed 3D object coordinates are transformed to any map projection system (support for about 30 different map projection systems including geodetic datum transformations is available), where image resampling is performed. By using the information provided by Ground Control Points (GCP)—manually measured or automatically extracted from reference images—model parameters can be refined in order to improve the geometric accuracy of the orthoimages..