This work will explore the application of a set of coordinate transformations to the derivation of orthographic projections of objects from grey-scale video camera images, and the use of those projections, which are viewpoint-invariant, as features for 3D object recognition. Most of the research in this area has been aimed at reconstruction of 3D object models or figure interpretations from projections that were initially given, or on object recognition from line drawings or shading information. This effort could lead to a useful new method for extracting invariant features for object recognition. The limitations of this method, in terms of requirements and constraints on the kinds of objects for which it can be effective, will be explored in this study.
