Dynamics of noisy images recognition by neural system on the basis of large optical-electronic vector-matrix multiplier

摘要

Convergence analysis of a large-size optical-electronic neural net is performed within the framework of numerical model (on the basis of optoelectronic vector-matrix multiplier) to recognise the medium-size images. Main restoration parameters are investigated for two classes of noisy binary images: non-contoured images ("pictures") and contoured ones (letters of Roman alphabet). Influence of the neural net optoelectronic components on the restoration process is also under study.