Rearch On Plane Shape Description And Recognition

The interests in pattern recognition has been increased during past tens years. Pattern recognition mainly refers to the description and analysis of material and spirit. We always have to do image preprocessing of silencing of noise, and then extract a set of figures or non-figures to describe the subject we want to process. Based on the description we analysis the very subject. In the thesis we deal with the important branch of pattern recognition — the description and recognition of 2D shape.Shape recognition extends their application into character recognition, target track, medical diagnosis, intelligent robot, exploration, mapping and architectural design. Such as in intelligent robot, the profile of object could be describing in 2D shape. And the object could be recognized and quantifiably tested by comparing the real profile and the expected shape in pattern recognition.In real world, shape has many deformations, such as zoom, rotation, shear and so on. And there is a great deal of uncertain noises in the boundary of shape, which make the shape recognition even more difficult. Traditional method for shape recognition always can't satisfy various recognition problems well. In this thesis, we describe the boundary of shape by continuity and non-continuity method, and propose shape recognition methods based on the descriptors. The experiment results show the methods are effective.Shapes of the same kind correspond to a set of continuous curves. The curves form curve equivalence classes on an equivalence relation. In this thesis, we use curve equivalence classes to describe shapes. Shapes of different kinds are mapped to curve equivalence classes. This shape descriptor is invariant to translation, scale, and rotation. Based on this descriptor. We define a distance between curve equivalence classes and the distance reflects the similarity between the shapes. By this distance we can figure out whether they are similar. This method for shape recognition has immunity to noise and small perturbations.A novel approach, curve tree, is proposed for curve representation and...