Three-dimensional Objects Recognition Using Curvature Co-occurrence Matrix

Traditional shape characteristic representations have some difficulties to distinguish three-dimensional objects with slight different shape characteristic.In this paper,spatial curvature is utilized to describe the inherent characteristic of spatial curved surface,and then it is combined with the conception of traditional co-occurrence matrix.A new shape description approach named curvature co-occurrence matrix is proposed.Normalization process to curvature co-occurrence matrix and some characteristic invariants independent of the translation,scaling and rotation transforms are illustrated.Compared with traditional methods,experiments indicate a better classification rate to objects with abundant shape characteristic.