A.N. Ulyanov

Application of technology of neural networks in machine vision systems is promising, but the practical application of technology developers of vision systems face a number of practical points are not obvious, the account which will allow both systems to reduce development time and increase the efficiency of classification. This article describes the features of the main stages of neural network classification, in particular the coding of the inputs and the normalization of raw data, are some strategies for building sub-systems of classification. In conclusion, the technique of interpreting responses of neural networks for the class of problems in which one image can be objects of different classes.