Research on the development of algorithms for the use of neural networks associated with video surveillance systems on the identification of the desired objects today go to a whole new technological level. Categorization of classes-objects of research, determination of actual zones of fixation, integration of functions-cloud computing, analysis of incoming data and their evaluation determine the relevance in the creation of a trainable model that allows to solve a wide range of problems when working with the existing and the flow of incoming data in real time both at the state and at the level of transnational companies.