Commonsensible recognition functions and action functions are very important for the intelligent robot, which can coexist with human being. In this paper, we pay our attention to printed character recognition of a moving robot comprising the sense of sight and suggest a revision method that takes advantage of a shape characteristic for a string of printed character with artificial distortion and malformation by movement of a viewpoint. In this method, revision of a principal axis of a character string, horizontal revision on the basis of unfilled space between characters, and perpendicular revision with a local approximation are performed. And the learning algorithm that the robot acts properly in any situation is also proposed.
展开▼
