Robust representation and recognition of free-form objects

Higher degree implicit polynomials can be used conveniently for describing 2-D curves and 3-D surfaces. Implicit polynomials are known to possess few Euclidean and affine invariants. However, the instability of fitting algorithms and errors in the estimation of invariants due to occlusion are some of the major issues concerning their usefulness. We proposed a tensor based approach for obtaining the invariants and pose estimation using higher degree implicit polynomials. We demonstrated that the relative values and signs of the invariants from higher degree polynomials are not sensitive to small levels of occlusion and it can be used for recognition and pose estimation of free-form objects.