Home – Home – Jornals – Avtometriya 2026 number 1

An automatic method is proposed for counting maize seedlings under conditions of substantial weed infestation and partial occlusion using ultra-high-resolution (<0.5 cm/pixel) RGB imagery acquired from an unmanned aerial vehicle. The method is based on a combination of computer vision and machine learning algorithms. Experimental results demonstrate that the accuracy of estimating the number of maize seedlings at early growth stages averaged 97%.

A histogram scene-based algorithm is developed to compensate for gain and bias non-uniformity in infrared focal-plane arrays. The statistical scene-based calibration method is proposed. Experimental results illustrating the effectiveness of the method are presented.

Issues related to the variational approach to the problem of non-parametric a priori uncertainty are disclosed is a method of converting a system of a priori stochastic differential equations, which enables to reduce the initial problem of nonlinear filtering of a multidimensional Markov process to a problem of filtering a new Markov process, characterised by a zero drift vector. The main provisions of the method of independent first integrals are given, allowing to provide the required transformation of linear systems of a priori stochastic differential equations. Filtering algorithms were constructed, in which the method of transforming multidimensional densities is used to restore the initial a priori and a posteriori densities of the probability distribution. An example of filtering the phase of a narrowband random process using the proposed conversion method is given. It has been shown that the proposed method of transformation using numerical integration reduces the requirements for the selection of parameters of the posterior density of the probability distribution of the vector of filtered parameters

The article discusses algorithms and neural network models for semantic segmentation of the Amur River water surface areas obtained using aerial photography. A dataset has been prepared for training a neural network based on aerial photography materials of the Amur River water area. The article presents the results of research on the accuracy of prediction of the most popular models in the field of semantic segmentation, such as UNet++, DeepLabV++, FPNet, and SAM. The experiments used the IoU (Jaccard similarity measure) and Boundure IoU (object boundary segmentation accuracy assessment) metrics. Computational experiments were conducted to measure the accuracy of the trained models in order to select the optimal parameters. As a result, it was found that the UNet++ model has an advantage in terms of segmentation accuracy, with an average Boundure IoU score of > 0.9. The developed algorithms and trained neural network models can be used in river water surface monitoring systems based on orthophoto images to determine the boundaries of the coastal zone.

The superior efficiency of the piecewise linear approximation method based on Delaunay triangulation has been demonstrated. This method completely eliminates negative concentration values on the resulting surface and exhibits the most favorable gradient distribution, with only 3.33% of gradients exceeding 75 ppm across the entire point cloud surface. The significance of the obtained results lies in the confirmed ability of this approximation method to generate highly reliable and physically accurate regime maps. These maps are essential for predicting critical emission levels, such as carbon monoxide (CO) and nitrogen oxides (NOx), during liquid hydrocarbon fuel combustion, enabling the development of effective burner optimization strategies to minimize environmental impact and maximize combustion efficiency in practical power installations.

Modern trends in underwater robotics development are focused on creating fully autonomous unmanned underwater vehicles (UUVs) equipped with multi-link manipulators (MLMs) and capable of performing complex research, manipulation, and technological operations in dynamic and uncertain underwater environments without direct operator involvement. The key challenge in UUV implementation is the generation and adaptive real-time adjustment of control programs based on data about the current environmental conditions obtained from the UUV's onboard sensors and MLM sensors. This paper presents a comprehensive set of tools for an intelligent UUV mission planner, including its information and software components, as well as development methodology. The proposed planner provides decomposition of strategic tasks into elementary operations with dynamic selection of their execution sequence and parameters depending on actual data and external conditions. The architecture of the developed software is based on principles of modularity and scalability, allowing the system to be adapted for various classes of UUVs and mission types. The presented solution contributes to the advancement of autonomous UUV control methods and can be used in designing advanced robotic systems for marine research, monitoring, and underwater technological operations.

This work is devoted to the study of acoustic phonon scattering in aluminum nitride (AlN) thin films grown by magnetron sputtering with an emphasis on their potential application as piezoactive layers for microwave resonators. The main objective of the work is to establish a correlation between the data obtained by various analytical methods, including Mandelstam-Brillouin scattering (MBS), Raman scattering (RS), infrared (IR) and terahertz (THz) spectroscopy, as well as X-ray diffraction (XRD), to assess the quality and homogeneity of AlN films. The data analysis confirmed the high stability of the MBS peaks in the studied films. It was found that the stability of the MBS phonon peak reaches 87.6 ± 0.5 GHz, which corresponds to the literature data, and confirms the homogeneity of the films with an error of less than 1 %. The presence of a specific peak in the XRD patterns and the minimum width of the phonon peak obtained by Raman and FTIR spectroscopy confirm the possibility of using these methods to characterize such micron-thick films. Considering that the peak frequency of the phonon resonance did not change for films with a thickness of 4.5 μm, it is assumed that the analysis of the phonon line width may be sufficient to evaluate the grown films. The Raman method seems to be especially promising, which in the future can allow in situ analysis during the growth of AlN films. Thin AlN films with a thickness of less than 7 μm have the quality sufficient for the development of acoustic microwave resonators, which opens the way for the experimental creation of multilayer structures with different acoustic impedances based on the obtained data.

In this work, the interaction of terahertz radiation with thin ITO films (< 300 nm) deposited on fused silica substrates was studied. The complex refractive index of the films was measured in the range of 0.2-1.2 THz. It was found that the Drude model accurately describes the dielectric properties of the films in this region. The parameters of the model were determined by fitting the experimental data across a wide spectral range, from millimeter to visible wavelengths. A linear correlation between conductivity and thickness of the film was revealed. The result is valuable for the design of antireflective coatings for terahertz radiation applications. The studies on the effects of intense terahertz radiation with peak and average intensities of up to 2 MW/cm² and 1.2 kW/cm², respectively, showed the absence of laser-induced damage to the films and made it possible to outline the scope of ITO films' application as dichroic mirrors in transport channels for intense terahertz beams.

The paper presents the results of studies of coherent population trapping (CPT) resonances recorded in a cell with 87Rb vapors with the addition of a buffer gas under multifrequency excitation by radiation from an external cavity diode laser (ECDL). The laser diode injection current is modulated simultaneously by microwave (MW) and very high frequency (VHF) signals. It is found that for a certain range of VHF modulation frequencies, the CPT resonance amplitude is proportional to the product of the electric field strengths of the spectral components contributing to the resonance excitation. The proposed model for calculating the resonance amplitude is in good agreement with the experimentally obtained data.

The properties of the anodic oxidized Si layers implanted with CO⁺ ions at an energy of 90 keV to were studied as a function of ion dose ranging from 2.5x10¹⁵ to 3x10¹⁶ cm^-2. After the implantation, the samples were annealed at 700 ^oC for 1 hour in the N₂ ambient. Spectral ellipsometry, atomic force microscopy and Fourier transform infrared spectroscopy were employed to study the oxidized layers. The refractive index of oxidized layers was higher than that in the thermally oxidized SiO₂ and dropped from 1.58 to 1.52 as the ion dose increased from 2.5x10¹⁵ to 1x10¹⁶ cm^-2. As the ion dose increased, the absorption band corresponding to the optical absorption by CO₂ molecules grew. The refractive index of the anodic oxidized layers as a function of ion dose is discussed in the frame of the microcavity formation in the SiO₂ as result of CO₂ molecules evaporation.

Experimental data on the energy efficiency (EE) of input/output of radiation along its entire propagation path in the medium of photopolymer holographic gratings and a planar waveguide were obtained and analyzed. It was found that the main contribution to radiation power losses is made by absorption in the photopolymer medium due to incomplete bleaching of the dye during post-processing of holograms, as well as their relatively low diffraction efficiency (DE). Due to insufficient matching of the spectral characteristics of reflection holograms and LED displays, only part of their radiation power is used as a useful signal. Taking into account the obtained experimental data, the final estimate of the EE was ~ 10 % relative to the LED radiation power at the input of the entire "input/output gratings - waveguide" structure. To increase the efficiency, it is necessary to use recording media with a small thickness (less than 10 μm) and, at the same time, with a large amplitude of photoinduced change in the refractive index, which can provide the necessary DE and the degree of matching of the specified spectral characteristics, as well as reduce absorption in the photopolymer medium. In this case, the expected EE value can be ~ 40 %.

A study of the complex spatial flow of a freon jet flowing out of a long triangular cross-section channel in the laminar-turbulent transition mode was conducted. In order to obtain a detailed picture of the flow, a multi-angle measuring complex of Gilbert diagnostics was implemented based on the IAB-463M device. It was shown that the channel geometry has a weak effect on the laminar-turbulent transition scenario in the channel, with the formation of turbulent spots. It was found that when a turbulent spot hits a jet flowing out of a triangular cross-section channel, it causes turbulization and generates secondary disturbances in the mixing layer. The results may be of practical importance in CVD combustion and deposition processes.

A new method for improving the quality of images reconstructed from digital holograms is considered. It utilizes spatial subpixel shifts, i.e., shifts by a certain value smaller than the resolution provided by the recording devices. In digital holography, photodetector matrices are used as a recording medium for holograms. The resolution is determined by the type of matrices used. To obtain holograms, a camera with a photodetector matrix with a resolution of 9000 × 6752 and a single element size of 0.86 × 0.86 μm was used. The size of the matrix used was 7.7 × 5.8 mm, the spatial resolution of such devices is approximately 500-1000 lines/mm. This already makes it possible to obtain digital holograms. However, to improve the quality of reconstructed images, it is necessary to increase the hologram resolution during registration. To increase the resolution, a device for introducing a spatial subpixel shift of the photodetector matrix was used in this study. Insufficient matrix resolution was compensated for using subpixel spatial shifts. The shift was performed along the X and Y coordinates by 430 nm, which was half the size of one matrix element. The digital hologram size increased fourfold (18,000 x 13,504 pixels). Image reconstruction was performed using a fast Fourier transform. To speed up the process, a graphics accelerator was used to reconstruct images from holograms.