GROUP THEORY BASED RECOGNITION OF SYMMETRIC POLYHEDRONS

摘要

Theory of group of symmetry properties is used as a basis for construction of the proper classification system destined for object recognition purposes in robotic vision. The whole theory is similar to classification of crystals and it may be reduced to the very simple set of rules using the algebraic groups of point transformations. The advantages of presented method is invariance of symmetry features with respect to changes of dimmensions and positions of objects including their rotations in object space. The whole scene is scanned for normals to small patches. Projection of the normals on the Gauss sphere enables using cluster analysis for finding normals corresponding to facettes of the studied object. Then symmetry analysis is used for object classification.