Aspect and Perspective Invariance Based Three-Dimension Object Representation

In this paper a representation method based on aspect of 3D object and geometrical invariant is presented. The method makes use of constrained 3D structures with the property of perspective invariance to represent aspects of 3D object. By using this representation, the machine vision system could avoid the computing of object pose, camera calibration and feature correspondence in the indexing phase of recognizing 3D object from a single view, therefore, realizing the strategy of indexing before matching and resulting in high real-time recognition efficiency. Firstly, the basic concepts of perspective invariance and constrained structure are discussed. And then, a general way to use such constrained structures to represent 3D object is given and also an application of it. At last, the weakness of this representation method is pointed out as well as further research work.