|
|
|
|
The fractal embedding method (in french) Publications and works (in french) Research laboratoty (in french)
|
The Artificial Vision and Biophysics Group of Polytech' Marseille is researching and developing new dynamic complex data heuristic methods for image and signal analysis and vision applications. Our group succeeded numerous technological transfers into industry that were always guided by the need to solve a visual perception problem, with a real time constraint. The neural networks and fractal embedding based architectures we are working to improve and optimize allowed us to find robust solutions to various problems, like fire forest detection or eye movements analysis. Different companies were created or increased their affairs by marketing the products issued from our research group: Uratek (Artifical Vision), Synapsys (Biomedical), T2M Automation (Video security) and RMS-signal (Signal processing). At the beginning we were specialized in the domain of non linear and complex dynamic systems, and had to deal with experimental methods and dynamic image processing tools. This experience has driven us to discover the fractal embedding method that is a revolutionnary concept, with technical aspects on which we took a patent in 1997 (CNRS patent on a signal processing method). We published the method in different papers like the american "Optical Engineering" or the french "Signal Processing". We use today this method to create artifical neurons in a dynamic way with a powerful calculation time economy. The practical interest of our dynamic neural networks, that we built with the fractal embedding method, is to make realistic learning artifical vision brains with simple PC-based computers, and to leave out utopy their feasibility. |
