3D target recognition based on fast skeleton extraction

Aiming at the IR 3D target recognition question, a method based on fast skeleton extraction is proposed. Based on the topology structure of target reflected by the skeleton, the correspondence between different targets' structural portions are established. Within the corresponding curve segments, curves are matched based on curve alignment, and the similarity is evaluated according to the sum of all the segments' cost. This method performs well in the present of commonly occurring visual transformations. Moreover, incorrect correspondence introduced by curve match's drawback, that is, curve matching tends to match local parts, and ignores the global view of how these parts are spatial arranged with respect to each other, can be avoided. Experimental results show that this algorithm has a rather good effect on 3D target recognition.