Shape Recognition Based On Invariant Moment

Shape recognition is one of important application in the computer version field. It is widely used in many practical industries, such as preventing forgery of trademark, the recognition of character symbol and industry parts. And the identification of aircraft and attacking target for the national defense also makes use of the shape recognition.This paper carries forward some classic theories of the shape recognition, which can be divided into two major groups: shape recognitions based on the overall geometrical characters and that based on the local geometrical characters. The former theories include invariant moment, Fourier and wavelet descriptor, which utilize the overall information of image or shape to recognize. The Hough transform, neural network and deformation template, which belong to the later group, are used to settle the problem with shelter and deformation. In addition, the simplification of shape also plays an important role in the shape recognition areas.Among the above methods, my paper utilizes the invariant moment, which contains the statistic information of the image. The concept of invariant moment was firstly proposed by Hu in 1962. Later, the theory of the moment developed quickly and various invariant moments were constructed, such as complex moment, radial moment, Zernike moment and wavelet moment. For the fast computing speed of the invariant discrete moment, this paper proposes an invariant moment for vector graphics which is based on the discrete moment. The algorithm for the vector graphics is used to recognize the shapes of ship parts. Through the experiment, 26 shapes of ship parts were correctly recognized. The result indicates that the algorithm about the invariant moment for the vector graphics is feasible. Based on the recognizing algorithm, this paper provides another algorithm for the matching of the shapes. The matching algorithm produces a new graphics from synthesizing other two similar graphs, whose boundaries coincide. At last, the algorithms have been used in the shipbuilding industry.