| dc.description.abstract | This report describes a technique for generating a digital terrain model of a given scene. The technique is
based on first estimating the position and distance to parallelograms placed in various parts of the scene
by image processing. This estimation is performed by extracting the parallelograms by a region growing
procedure, determining its line segments, and applying a priori knowledge of its physical size for estimating
the distance. Next, these positions are used as input to a triangulizing algorithm. The resulting planes are in
turn used for determining the distance between a point in the scene corresponding to a given pixel and the
camera. The distance between the camera and points outside the planes are determined by extrapolation,
and are thus making the distance image of the scene complete.
In the report we have applied the technique on two different scenes. The experiments demonstrate that the
technique generates (terrain models and) distance images of acceptable accuracy. | en_GB |