Abstract: | The principal goal of this project is to determine the likely positions of small complex features i.e. eyes, noses and mouths from intensity image data. There are many image processing and pattern recognition systems, traditional and non-traditional. Most of them are memory consuming and need elaborate and expensive hardware. There is, however, a very big interest in these systems not only because humans need a definite improvement of pictorial information for their interpretation of image density data, but also because today's robots, have to be able to perceive their environment reliably and this has been proven feasible with the aid of such systems. A specific area, that of facial recognition, was examined in this project, using artificial neural nets instead of the traditional pattern recognition methods. Three pre-processing techniques have been used in order to ensure that the produced results are reliable. If the net topology is the optimum, and the preprocessing technique produces consistent and concise data, the net can recognise and identify the position of facial characteristics successfully.
|