Home – Home – Jornals – Avtometriya 2024 number 2

When solving various applied problems, it becomes necessary to reconstruct a three-dimensional scene and determine distances to objects from images obtained by cameras from different shooting points. Determining the shooting parameters and camera characteristics (calibration of the recording system) directly from the received images is an important stage of reconstruction, which determines the reliability of the results obtained. The need to use cameras with a small field of view leads to significant errors in calibration and reconstruction. The paper discusses a way to increase the accuracy of calibration by expanding the angle of view by "stitching" a series of images obtained from one shooting point into a wide-angle panorama. In addition to objects of interest, the panorama must include other shooting points. An example of estimating the distances to objects in a real urban scene is given. It is shown that the proposed method provides accuracy comparable to measuring distances on a large-scale map.

Based on the mathematical apparatus of stochastic gradient adaptation, an algorithm for estimating the parameters of image registration according to the criterion of the maximum Shannon mutual information is synthesized. Expressions for the gradient of mutual information are derived. At the same time, the derivatives of the entropy of the images according to the estimated parameters are found using the Parzen window method. Optimization of the synthesized algorithm is carried out, aimed at improving its performance.

A method for visualizing complex objects representing a composition of volumes and functionally defined surfaces using graphics processors is proposed. The method uses hierarchical 3D data structures that do not require parameterization. Combinations of complex translucent volumes and surfaces based on perturbation functions are described.

Currently, ongoing tasks involve research in the field of spray cooling, droplet interaction with complex surfaces, and droplet evaporation from solid walls. Automated digital image processing methods are utilized to conduct these experiments, particularly when recording large amounts of video data. This paper presents methods for recognizing objects in digital images to obtain quantitative characteristics of evaporating droplets in low-light conditions. New algorithms have been developed to identify and close object boundaries under conditions of limited visibility, enabling measurement of the geometric parameters of evaporating droplets. A comparison is made with data obtained using the manual image processing method, and the conditions for the applicability of the proposed algorithms are demonstrated.

Specific features of calculating the group delay time of a signal based on digital measurements of the S-parameters of microwave devices are discussed. Methods for calculating the group delay time of a signal with a random noise component based on differentiation by the increment method and approximation are discussed. A method for calculating the group delay time based on the differentiation method using a sliding piecewise-linear approximation is proposed. This algorithm is more resistant to the noise component of the signal in question than the standard method for calculating the group delay time of a signal based on the one-way difference method and has a smaller computational complexity than the method considered in the article based on quadratic approximation. The results of testing the operation of the described methods are presented, which illustrate their suitability for use in calculating the group delay time.

In this work, the surface morphology of InAlAs layers deviating from the lattice-matched composition with an InP substrate is studied. A linear relationship is shown between the dimensions of the pits and the thickness of the grown layer. The possibility of forming an array of nanopits uniform in size and shape on the surface of an InAlAs layer is demonstrated.

Vertical nanowire shape modification during the annealing process due to the readsorption of the material evaporated from neighboring wires and from the substrate are numerically calculated. The evaporation model from a set of isotropic point sources and the Knudsen-Lambert model are applied. The effects of the aspect ratio of the nanowire length to the distance between them and the material evaporation rate from the nanowire sidewall and the substrate on the nanowire shape are analyzed.

The methods of passivation of photodetector mesastructures based on InSb/In_1-xAl_xSb nBn heterostructures grown by molecular beam epitaxy have been investigated. The mesastructures are formed by photolithography and liquid etching methods. For passivation of the mesastructure surface, two different dielectrics Al₂O₃ and Si₃N₄ are used. It is shown that the Si₃N₄ dielectric completely passivates the surface of mesastructures and the bulk current is prevailing, while the surface leakage current dominates in mesastructures passivated by Al₂O₃. At small reverse biases, the correlation between the current value and the contact area is violated due to the influence of defects. Modelling of the mesastructure current-voltage characteristics at 77 K shows that the main contribution to the dark current is made by the tunneling current of electrons through the In_1-xAl_xSb barrier.

The effect of operation of a frequency converter based on an electro-optical Mach-Zehnder type modulator on the frequency spectrum of the output signal in the down-conversion mode has been studied. It is shown that the modulator does not introduce significant changes in the distribution of the power spectral density of the converted signal, including phase noise, compared to the signal at the input of the modulator.

Materials with non-standard physical properties are widely used in various fields of science and technology, from microelectronics to heavy industrial installations. International scientific teams carried out mathematical calculations showing that the most promising compounds of B-C structures, which have strong chemical sp²- and sp³-bonds under certain external conditions and concentration ratios. Such materials are characterized by high hardness, resistance to aggressive media, and electrical conductivity. This study is aimed at synthesizing and investigating carbon-containing BC_x films. The paper presents the results of measuring a thin BC₅ film. The spectral characteristics of the film, data on the layer thickness, conductivity, and composition of the film are obtained using X-ray, Raman, and photoelectron spectroscopy.

The influence of the number of elements of the ring antenna array on the quality of focusing the wave inside the ring and the ability to control the focus location is investigated. Controlled focusing is supposed to be used to scan the internal structure of the patient with a controlled focus for tomography or delivery of ultrasonic energy to a given area of the patient's body. The research is carried out by mathematical modeling. The simulation results show that thinning of the number of elements on the annular antenna array leads to the formation of interference rings in the area of focus image reconstruction. At the same time, the size of the focal spot remains unchanged.

Magnetometric measurements are widely used to solve problems in geology, navigation, and a number of other areas. Currently, there is a tendency to miniaturize and reduce the cost of magnetometers. However, the readings of inexpensive magnetometers are distorted by errors, which requires a calibration procedure. The article describes mathematical error models that are used in most well-known scientific research and adopted in this work. Mathematical modeling of the measuring signals of the magnetometer is carried out, and the principle of calibration of the magnetometric sensor, based on the accurate measurement of changes in its spatial position in a constant magnetic field, is proposed and researched. Estimates are given of the accuracy of restoration of calibrated parameters and the signal restoration errors achievable when performing the calibration operations proposed in the study.

The paper studies the possibility of using a long short-term memory (LSTM) neural network to simulate the operation of a PID controller. A PID regulator has been implemented to control a heater with a temperature sensor by using the Arduino framework. An LSTM model trained on the controller data has been developed. It is shown that the neural network model quite accurately reproduces the regulator operation and can completely replace it. In addition, the applicability of this model as a detector of abnormal operation of the PID controller is shown.

The problem of traffic control is solved using a model-oriented method. To develop a mathematical model, an automaton formalism is proposed. In strict accordance with the problem conditions, the states of a deterministic finite automaton (DFA) and predicates of directed transitions are defined. To implement the DFA, a transition is made to the Harel diagram (statechart), which is a unified mathematical and software support for modern global modeling tools. The solution to the problem posed is proved constructively by a table of states with the corresponding dynamics and computational experiments of the selected scenario in advanced domestic and foreign instrumental environments for modeling the designated class of event systems. Computational experiments show complete agreement with constructive analysis.