On methods of improvement of the identification quality of graphical objects in geometric correlation methods

Methods of geometric correlation allow one to identify objects that are invariant under translation, scaling, and rotation. Theoretically, these methods must not depend on the shape of objects recognized. However, their application in real conditions has revealed several factors that affect the recognition quality including the dependence on the shape, which has to be taken into account. In this paper, the reasons for raising such factors and methods for eliminating their effect are considered.