The dissertation is devoted to solving the actual issue of investigating the experimental and theoretical aspects of radiation and propagation of transient electromagnetic waves in the media with complex spatial distribution. The work also covers the analysis of diffraction of these fields on objects located in these media, the application of methods for obtaining and processing reflected signals, and the development of algorithms for recognizing hidden objects and determining their position using artificial intelligence approaches.
The scientific novelty of the dissertation is in the following results:
1. Using the boundary conditions of classical electrodynamics, the relationship between the evolutionary coefficients from the Klein-Gordon equations, which describe the incident, transmitted, and reflected non-stationary waves in the first approximation, was found for the first time.
2. It was established for the first time that in the presence of Gaussian noise in the received signals, the final results of object position recognition by artificial neural networks have an advantage over the cross-correlation method at low noise levels. However, under significant noise presence, these two approaches do not show noticeable advantages over each other, except that artificial neural networks operate orders of magnitude faster, especially when implemented in specialized microchips.
3. For the first time, the approach of discrete tomography demonstrated the enhancement of the informational components of electromagnetic waves reflected from a hidden object, using the ray approximation of these waves while considering their temporal form, dielectric parameters of the soil, processes at the air-soil boundary, and the use of several distributed receiving antennas above the soil.
4. The optimal parameters of the system based on the discrete tomography method were determined for the first time: time window, number of receiving antennas, input data augmentation, and the proportion of pre-processed input signals, in the presence of high-level noise in the input tracts of the antenna receivers.
5. The quality of recognition by the artificial neural network method for detecting various anti-personnel mines, such as PMN-1, PMN-4, and PFM, in a heterogeneous environment with the presence of white noise in the received time dependencies was evaluated for the first time.
6. A new approach to determining the location of hidden objects in the soil using collective artificial intelligence, which simultaneously processes the same time dependencies obtained by ultra-wideband ground-penetrating radar, was proposed for the first time.
Practical significance of the results:
1. The relationship between the unknown coefficients from the Klein-Gordon equations is established, which describe the passage of transient waves into a medium, demonstrates the possibility of concentrating the energy of the electromagnetic wave in the soil, similar to the phenomenon of an “electromagnetic projectile,” to increase the energy of the wave reflected from the hidden object and, consequently, improve its recognition.
2. The analysis of recognizing hidden objects using artificial neural networks and the correlation approach allows a significant improvement in their detection when applied simultaneously.
3. The application of the discrete tomography approach to obtain the additional set of input data for the artificial neural network by utilizing the physical processes during the propagation of impulse waves in the soil reduces the necessary computational resources without losing the accuracy of recognizing hidden objects. This is beneficial for new radars capable of detecting hidden dangerous objects in real time.
4. The carried out research on artificial neural networks for detecting anti-personnel mines, including in a heterogeneous environment, allows the creation of new unique demining systems capable of detecting hidden explosive devices that practically do not contain metal parts.
5. The new proposed approach to determining the location of hidden objects in the soil using collective artificial intelligence with data obtained from an ultra-wideband ground-penetrating radar not only increases the accuracy of the object’s location, but also improves the quality of its recognition. This has practical significance for geological research, construction, and military applications.
