A new fourier description based on areas (AFD) and its applications in object recognition
Zhongfei Mark Zhang, Renjle Jing, Weikang Gu · 2005
On the basis of some invariant relations of the projection theory, a new Fourier Descriptor, the Area Fourier Descriptor (AFD),based on areas of the objects' silhouettes is presented, and the concept of multi-dimensional vector space is introduced. In order to speed up searching procedures in the model base for matching, we apply a new coding method to generate a Curve-Corner-Code (CCC), which can be used for any planar silhouettes. Taking the coordinate vectors in the multidimensional space introduced by AFD as the main features, we developed a new algorithm which can not only tolerate any translations, rotations, scale changing and any affine distortion, but also require less computation. Because of the limited terms of AFD, the algorithm is insensitive to truncation errors and can also be extended to recognize 3D general cylindrical objects.