Intelligent and Robust Method for Shape Recognition

Cuevas Tello J.C., Núñez Varela I., Puente Montejano C., Botello Aragón M.A.
International Conference on Artificial Intelligence and Applications - IASTED, Austria., 2004.


A method for two-dimensional shape recognition is described in this paper as is an extension to 3D. First the object is scanned using laser with the help of a hardware system designed inside. Then, the objects center of mass is calculated, with this calculation the method is able to obtain a unique signature or model of the object. This signature is compared with the existent signatures, using a correlation coefficient, a knowledge base and an inference process, in order to decide the objects shape. Actually the method has been implemented as a Windows application that simulates it. It is also able to distinguish between objects with an almost exact signature, for example a star compared with a pentagon, furthermore, the objects position does not matter, neither do the discontinuities, nor the rotation grade that it has. This method provides a simple way to object recognition using faster and easier algorithms, furthermore it does not require expensive equipment, like video cameras, it is very flexible and robust, since it is not necessary to have exactly the same objects in the knowledge base. And finally, it can be trained for storing more objects in its knowledge base.