Development of multiresolution hierarchical trees to non-cooperative target recognition

In this paper, problem of efficient representation of large database of target radar cross section is investigated in order to minimize memory requirements and recognition search time, using a tree structured hierarchical wavelet representation. Synthetic RCS of large aircrafts, in the HF-VHF bands, are used as experimental data. Hierarchical trees are built using wavelet multiresolution representation and K-means clustering algorithm. Criteria used to define these hierarchical trees are described and the obtained performances are presented.