A new synthesis procedure of cellular optimal linear associative memories for robot vision systems

Abstract

A design procedure of discrete-time cellular neural networks (DTCNN) to be used as associative memories for robot vision is presented The choice of cellular neural networks is motivated by their architecture, suitable for storing images, and their locally connected structure which is effective for the hardware implementation of the designed memories. In particular, taking into account the constraints dictated by the discrete-time cellular neural networks structure in this paper a design procedure of DTCNNs, which also enables memories to recognize correctly the event of superimposition of tools, is developed. To this purpose, a cellular associative memory which behaves as an optimal linear associative memory (OLAM) is synthesized. The performances of the designed network are investigated and its behaviour as an optimal linear associative memory is confirmed by means of an example of recognition of superimposed tools handled by a robot in an assembly line.