Recognition of 3D objects in various capturing conditions using appearance manifolds

In this paper, we propose novel techniques of constructing appearance manifolds in eigen-space for robust recognition of 3D objects. Four models are presented in this paper: Constant Covariance method, Point Interpolation method, Covariance Interpolation method, and comparison with the well known Parametric Eigenspace method. The developed application is used to recognize 3D objects from images with various capturing conditions. Experimental results show that the proposed methods could accurately recognize 3D objects even from images which are influenced by geometric distortions and quality degradation effects.