Abstract: In this paper, we propose a new method to represent the 2D shape and to recognize the object. In previous approaches, the boundary information is generally employed to recognize the 2D object. The points extracted from the polygonal approximation of the boundary are considered as the features and the point pattern matching is used. The methods, however, have the difficulty in the recognition of changeable objects. Thus we obtain the morphological skeleton and the distance transform and use them in the 2D object recognition. The matching method has the invariant features (rotation, translation, and scaling) and is a class of stochastic approach. And then the proposed method is employed effectively for the recognition of the changeful shaped object, e.g., tools like a wrench and a stripper or marks of the different shape and size.!15
展开▼
