Geometric correction of satellite and aerial imagery to remove terrain-induced distortion, producing map-accurate orthoimages suitable for GIS analysis and measurement.
Orthorectification is the process of removing geometric distortions from satellite or aerial imagery caused by sensor orientation, terrain relief, and Earth’s curvature. Raw satellite images contain systematic positional errors, objects on hilltops appear shifted toward the sensor, while valleys are displaced away. Those distortions make raw imagery unreliable for accurate measurement, mapping, and multi-temporal analysis.
The orthorectification process uses a sensor model (describing the camera geometry and orbit) combined with a Digital Elevation Model (DEM) and, optionally, Ground Control Points (GCPs) to mathematically project every pixel to its correct planimetric position on the Earth’s surface. The result is an orthoimage, a geometrically corrected product where distances, angles, and areas can be measured directly, just like on a map.
Sentient’s orthorectification pipeline processes imagery from more than 150 satellite sources with sub-meter positional accuracy, enabling seamless mosaicking of multi-date, multi-sensor imagery into consistent basemaps.
The relationship between image coordinates and ground coordinates is established through a sensor model. This can be a rigorous physical model (describing the satellite orbit, attitude, and optics) or an empirical Rational Polynomial Coefficient (RPC) model. RPCs are provided as metadata with most commercial satellite imagery and define the mapping between 3D ground coordinates (latitude, longitude, elevation) and 2D image coordinates.
A Digital Elevation Model provides terrain height at every ground location. The DEM is critical for correcting relief displacement, the apparent shift of elevated objects toward the sensor. Higher-resolution DEMs (Copernicus DEM at 30 m, ALOS World 3D at 5 m) produce more accurate orthorectification, especially in mountainous terrain. The DEM must be co-registered to the same reference frame as the target coordinate system.
GCPs are identifiable features with known ground coordinates, used to refine the sensor model and correct residual biases in the satellite’s reported position and attitude. Modern high-resolution satellites (WorldView, Pléiades) reach sufficient accuracy with RPCs alone, but GCPs can sharpen absolute accuracy from ~5 m CE90 to under 1 m CE90 in challenging scenarios.
Each output pixel is back-projected through the corrected sensor model to find its corresponding location in the raw image. The pixel value is then interpolated using cubic convolution (preferred for visual quality) or nearest neighbour (preferred for thematic data). The output is a georeferenced orthoimage in the target projection, typically UTM or Web Mercator.
Create seamless, geometrically accurate basemaps from multi-date satellite imagery for GIS and web mapping platforms.
Ensure pixel-level alignment between multi-temporal images so detected changes reflect real ground changes, not geometric artefacts.
Enable accurate measurement of distances, areas, and perimeters directly from imagery for cadastral, agricultural, and urban-planning applications.
Combine adjacent satellite scenes into seamless mosaics by ensuring consistent geometry across overlapping strips and orbits.
