Use of neural networks on parallel architecture for the classification of plane figures
Alberto L. Geraci, Giovanna A. Fargione · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1995
The use of neural networks for recognizing and measuring objects is proposed. A review of the state-of-the-art of neural network systems and communicating sequential process formalizing was made, in order to select network type and topology. After a standard image acquisition step, the developed system prototype was implemented on parallel architecture, based on Transputer network following the same distribution of the nodes, that constitute the activation functions typical of all types of neural network. The realization of the prototype was carried out in C++ language,l for its portability to different platforms. Then, the developed algorithm was optimized by the translation into the original OCCAM language. The study constitutes the first effort of a research activity, intending to link recognition and measuring of objects to vectorial imagery, using different image acquisition systems (including those working in the infrared band) and the developed system prototype, with software dedicated to industrial drawing applications.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.