Image processing
Image processing is based on state-of-the-art vision software technology. It connects raw images by identifying a massive amount of corresponding points (the AAT phase) and uses this substantial amount of information to accurately register the position and orientation of the aerial images (the BBA phase). During the next phase, powerful calculations allow for the creation of dense point clouds. Based on these points, digital surface models (DSMs) and orthophoto mosaics are generated. Geo-referencing the products to your reference system of choice is typically carried out by introducing just a few control points manually.
Automatic aerial triangulation is based on algorithms that detect feature points between images to create a 'spider web' of connected points between the images. State-of-the-art vision technology uses powerful algorithms to create numerous points in order to generate a gigantic over-dimensioned system of equations to solve the external calibration parameters - which represent the position and orientation of the images - and internal calibration parameters - which represent the focus distance, principal point and distortion parameters of the camera.
The solver is called BBA (Bundle Block Adjustment) but differs from its conventional photogrammetry counterpart in being able to not only solve but also optimize the calibration data based on a highly over-dimensioned system. As such, highly accurate external and internal calibration data are generated to form the basis for geospatial data products with high spatial accuracy. Learn more
Based on the points retrieved during the AAT phase, the initial sparse point cloud can be constructed. The points are accurate but the DSM resolution is limited in areas with few features. By using calibration imagery and advanced image-matching techniques, an increasing number of points can be detected (depending on resources and time) to generate very dense (up to pixel level) point clouds.
The Digital surface model or DSM is the actual surface connecting the point cloud. The simplest version is referred to as a TIN (Triangulated Irregular Network), which is a DSM consisting of faces (triangles). Orthophoto mosaics are created by projecting the texture of the images onto the DSM.