Gateway for:

Member Countries

Neural Network Methods for Disasters Prediction


Development of Hardware-Software Facilities of Neural Network Methods Support for Information Processing with the Purpose of Prediction of Technically Caused and Disasters

Tech Area / Field

  • INF-COM/High Performance Computing and Networking/Information and Communications

3 Approved without Funding

Registration date

Leading Institute
Research Center "Module", Russia, Moscow

Supporting institutes

  • TsNII Kometa, Russia, Moscow\nNII Kulon, Russia, Moscow


  • TZN Forschungs- und Entwicklungszentrum Unterlüß GmbH, Germany, Unterlöss\nSamsung Electronics - Representative office in Russia, Russia, Moscow

Project summary

The purpose of the present project is the development of hardware-software using neural network methods of image processing for objects extraction and recognition aimed at usage in automatic systems for observation of ecological situation and for prediction of technically caused and natural disasters.

Technical approach

Realization of the project is carried out in three directions:

- system questions - development of the conception of creation of the system registering and predicting technically caused and natural disasters,
- theoretical research - it is aimed at creation of neural network methods of information processing, enhancing of the methods of artificial neural networks (ANN) training, creation of practical methods of ANN training, creation of training and testing data bases,
- development of functional algorithms and hardware-software realizing neural network technologies of information processing with the purpose of pattern recognition.

Hardware-software is created on the basis of the specialized neuroprocessor of RC "Module" design manufactured by Samsung (South Korea).

The methodological scientific base of this project is the usage of neural network technology for solution of the problems of recognition. Here ANN are considered as:

1) A structure described by some functional equations generating complete systems of functions.

It permits to consider not only classic systems (complete) of functions but also to naturally cover such a new direction as multiple-scaled analysis where complete systems are assigned not on the whole numerical axis but only on its certain parts.

Such an approach permits to diminish the requirements for computer facilities realizing the data processing algorithms (without any loss in problem solution accuracy) or to increase the volume of calculations with the help of the computer facilities at hand, i.e. to process large data arrays.

2) Structures with the help of which it is possible to find the approximation of any other neural network.

It permits to use the already existing ANN modules and to reduce the time of their training.

Besides, RC "Module" has up-to-date computer facilities and technologies necessary for scientific research, development and small-scale production of different kinds of electronic equipment and for integrated circuits design.

The SUN high-productive servers and work stations of RISC-architecture, the latest models of IBM and APPLE personal computers, different network and communication equipment together with MICROSOFT and NOVELL operation and network systems allowed to create in RC "Module" its own network. The global communication network INTERNET is accessible.

The CADENCE computer-aided design system realized on the RC "Module" facilities permits to design digital, analogue, analogue-digital modules of practically any complexity and to design semicustom chips and programmable logical integrated circuits.

The IMS measuring equipment at hand automatically tests the complex electronic instruments including integrated circuits and multichip modules.

Anticipated results

The results of the project are:

- hardware-software demo complex providing data processing with the purpose of object detection and recognition under conditions of technically caused and natural disasters;
- methodology of information (TV and infra-red (IR) images) processing using neural network technologies;
- creation of hardware neurocalculators of wide application which can be sold well on the neurocomputer market as an independent commercial product;

The project theoretical results, algorithms and neuroaccelerators can be used in different branches of engineering connected with signal processing.

Potential role of collaborators

Joint usage of the development results and their commercial promotion.