Recognition of a given polyhedral object in a single perspective image using a vanishing pseudo invariant

In this paper we use a pseudo perspective invariant for detecting a given three dimensional object in a single perspective view of a scene. It is a matrix determinant that vanishes when a 2D pattern is matched with the currusponding 3D one. This approach allows to handle very simple expressions and dous not require iterative solution of non linear equations. The method has been tested in several conditions, using either synthetic or real images. The proposed method in advantageous in terms of processing time and is particulary fit to be implemented in real-time application, such as mobile robots. The possibility of recognizing an object from a single perspective view, without calibrating camera, appears very attractive.