Search academic texts

Registration Number

0225U004186, (0124U004250) , R & D reports

Title

Methods and means of active and passive recognition of mines based on deep neural networks

popup.stage_title

Розробка програмного забезпечення дла активного та пасивного розпізнавання мін з металічним та пластиковим покриттям в грунтах різної структури на основі глибокої осциляторної нейронної мережі.

Head

Peleshchak Roman M., Доктор фізико-математичних наук

Registration Date

24-11-2025

Organization

Lviv Polytechnic National University

popup.description1

The purpose of the project is to develop: • the architecture of an oscillatory neural network of different dimensions and its learning algorithm for recognizing multispectral input images at a high speed of their arrival; • mine detection technologies based on the synthesis of near-surface penetrating radar and an oscillatory neural network operating in resonance mode, which will increase the amplitude of the useful signal in the "signal/noise" ratio.

popup.description2

Within the framework of the project, a set of tasks aimed at developing efficient methods for landmine detection using neural networks was successfully completed. Existing approaches to active and passive landmine detection were analyzed. Data on methods utilizing radar signals and magnetic sensors combined with neural network algorithms were collected and systematized, highlighting their advantages and limitations. A mathematical model of a resonant neural network with oscillatory neurons was developed. The model accounts for complex signal dynamics in structurally heterogeneous soils and enhances signal amplification to improve detection accuracy. Criteria for the neural network parameters that enable resonance effects were established. This optimization of the network architecture allowed for efficient processing of low-amplitude signals near the noise level. An algorithm for recognizing multispectral signals was designed, incorporating the intrinsic dynamics of oscillatory neurons. The completed tasks have resulted in the creation of an innovative approach to automated landmine detection, characterized by its accuracy, speed, and adaptability to various soil conditions.

Product Description

popup.authors

Vasyl Lytvyn
Peleshchak Ivan R.

popup.nrat_date

2025-11-24