AUTOMATIC SELECTION OF CONTROL-POINTS FROM SHADOW STRUCTURES

The purpose of this paper is to describe a procedure for the automatic selection of control-points in remote-sensing images of high-relief terrains for alignment with a reference map. This problem has been found to be of strategic importance whenever remote sensing images have to be integrated in a Geographic Information System (GIS) and processed in real time. The procedure described here is based on the recognition of shadow structures in the satellite image and on their comparison with the computer-generated shadows obtained from the Digital Terrain Model (DTM) of the region. The procedure was developed for a Landsat TM image of the Aurina Valley (in the Pusteresi Alps) with the DTM obtained from an IGM (Istituto Geografico Militare) 1:25 000 reference map, but with minor changes it can be extended to other remote-sensing images. Comparison of the shadow structures is performed by similarity evaluation of a simplified model of their shapes described by means of inertia ellipses. Each pair of shadow structures, recognized as similar and meeting a number of positional constraints, yields a pair of corresponding points whose coordinates provide input values for determining the parameters in the transformation of the input image into a planimetrically corrected one. The performance and robustness of the method and its boundary applicability are assessed. An example is given in which the automatically-determined control-points are directly inserted in a warping function, with reasonably good results.