The purpose of the study is to generalize to a certain extent the experience gained from solving topical binary pattern recognition problems for images of various nature using the majority committee method. According to this approach, the conclusion regarding the class an object considered belongs to is drawn by counting the votes of individual committee members "for" or "against" the object’s belonging to a particular class. The proposed method of solving recognition problems has been laid at the heart of modern decision-making systems and is similar to the work of experts, where collective judgments are integrated and used to obtain objective results. The work has yielded the following results: eight alternative variants of majority committees for the problem of fire recognition based on a set of color-brightness features have been produced and studied, and it has been shown how a committee can solve the problem of recognizing the aircraft types based on invariant moments using a direct propagation ANN. The committee elements include metrics and artificial neural networks with new activation functions based on parabolic functions and their combination with linear sections. The proposed solutions are easy to implement, which makes them preferable for use in aircraft onboard systems with limited computational resources. The conducted experimental studies have demonstrated good performance in terms of accuracy in solving the problems of recognizing fire processes and detecting rigid objects. The main result is the demonstration of the possibility and expediency of solving the problems of recognizing objects of various nature in real time directly onboard of aircraft with limited computational resources. The application area of the results is remote sensing for the purpose of detecting fires in forest areas and the presence of aircraft in the observed space.
DOI: 10.24160/1993-6982-2026-4-153-162
Download the article (PDF) from the Vestnik Moskovskogo Energeticheskogo Instituta website (registration required): https://vestnik.mpei.ru/index.php/vestnik/article/view/1556
At the Intelligent Control Laboratory of the Program Systems Institute of the RAS website: https://icontrol.psiras.ru/publications/
eLibrary: https://elibrary.ru/item.asp?id=91971873
Khachumov M. V., Emelyanova Yu. G., Yarussky I. M., Dvorkina E. V. Binary Classification by Majority Committees Based on Artificial Neural Networks with Parabolic Activation Functions // Vestnik Moskovskogo Energeticheskogo Instituta (Bulletin of MPEI), № 4, 2026, pp. 153–162.