Matching strategies for DSMs extraction in urban area from High Resolution Satellite Imagery

At present, thanks to the very high resolution and the good radiometric quality of the imagery ac-quired by very high resolution satellites such as Ikonos, WorldView-1, GeoEye-1, it seems possible to generate Digital Surface Models (DSMs) at a 2-3 meter accuracy level, which are comparable to the one of DSMs derived from middle scale aerial imagery. Nonetheless, DSMs accuracy level is strictly related to image orientation and matching strategy. Recently a facility for the image matching has been included in the SISAR package developed at Geodesy and Geomatic Area - University of Rome “La Sapienza”; SISAR is a scientific software able to carry out the image orientation with different models for several (very) high resolution sen-sors and to generate the RPCs based on its own physical models. The new SISAR image matching algorithm is based on affine transformations which drive the Fea-ture Based Matching and the Area Based Matching; several affine transformations are used at dif-ferent heights, estimated starting from some ground points at corresponding heights whose image coordinates are computed from suitably generated RPCs. In such way the image matching has been combined with image orientation in order to increase the accuracy and the reliability of DSMs in urban area. In order to assess the accuracy of the algorithm some tests were carried out, using a stereo pair of Augusta coastal zone (Siracusa, Sicily) acquired from WorldView-1 and a stereo pair acquired by GeoEye-1 over Rome; as reference data, a DSM obtained from LiDAR surveys and a 3D vector cartography at 1:2000 scale have been employed respectively