The accuracy of the conventional system correction of AVHRR imagery using ephemeris data is not sufficient for some applications which demand high accuracy in geometry. The correction using GCPs can meet such demands. But it sometimes occurs that the adequate number of GCPs can not be acquired because of clouds, etc.. This paper introduces the new method of geometric correction of AVHRR imagery which utilizes coliniality condition as well as ephemeris data. The method is adaptable, however many number of GCPs are obtained.
An extension is given for general closed surfaces based on the earlier reconstruction method for a closed convex surface. It fits an object surface with a polyhedral model using outlines taken from multiple projections of the object. Initially, the model is defined as a tetrahedron, whose center of gravity is matched to that of the object. Next, it either moves each vertex of the model by calculating the most appropriate shift vector or adds a new vertex to the model. The two procedures are repeated till the number of vertices of polyhedral model reaches a given value. The result of experiments using photographs of an object shows a relative error of projection fitness of around 3%.