O.V. Esikov, E.M. Suharev, A.V. Altuhov, E.A. Tarasov

One of the problems encountered in the processing of data for monitoring complex systems and facilities is the task of automated image processing and object recognition. Typically, this problem can be solved with a certain frequency for a long period of time under severe time constraints on decision making. This fact contributes to widespread implementation in practice of methods and algorithms for automated processing of graphical information and recognition of objects for monitoring. This task is difficult and complex, with multiple options for solutions. Proposed for the automated processing of graphics information to apply a set of methods and algorithms of spatial image processing, allocation of contours of objects. Formalized mathematical model for determining the composition and consistency of their application for the image that comes to the input of the system. In general, this task belongs to a class of ill-posed problems. To detect the contours of objects allocated and to address the possibility of invariance to foreshortening invited to use the apparatus of artificial neural networks. On the basis of pilot test results suggested to use a neural network perceptron type. To create and debug the training set, in order to reduce the possibility of the effect of cross-associations are invited to use the network Hamming. Experimental verification of the developed techniques, models and algorithms proved their efficiency and the possibility of practical application.