Three-dimensional object recognition using vector wavelets

In this work we present a new technique based on object boundary analysis for recognizing three-dimensional objects via a vector wavelet analysis. Vector wavelets are employed as they possess such important properties as orthogonality, symmetry, and compact support, which cannot be achieved simultaneously with scalar wavelet analysis. The model proposed in this paper can distinguish objects with certain freedom of movement in three-dimensional space and does not assume that the shape sequence is a stationary process. The proposed modeling algorithm is invariant to changes in scaling, shifting, rotation, and starting point. Extensive experimental results are provided to establish the model's effectiveness.