Home – Home – Jornals – Avtometriya 2024 number 1

The article addresses current issues in the field of fiber-optic data transmission, related to the constant increase in demand for communication system bandwidth and nonlinear response. The main machine learning methods used to compensate for nonlinear signal distortions in long-haul coherent communication lines are presented, including neural networks of various architectures. The paper emphasizes the promising nature of machine learning-based solutions to enhance the performance of optical fiber communication systems, thanks to their ability to derive effective and adaptive signal recovery schemes with low computational complexity.

Lasers of various power classes from tens of µW to tens of W based on the stimulated Raman scattering (SRS) effect in waveguide elements made of tellurite glasses are studied. Due to the high SRS gain and large SRS frequency shift, tellurite glasses are promising materials for the development of laser sources at a wavelength of about 2.3 μm with a pump at a wavelength of about 2 μm, which was used in the work. A detailed numerical simulation of Raman lasers based on single-core and multicore special tellurite fibers is carried out. It is shown that the conversion efficiency of pump power to SRS wave power for optimal parameters can exceed 50%. Detailed modeling of low-power Raman lasers based on high-Q microresonators is performed; optimal parameters and factors limiting generation are found.

Resonance properties of the local field of metal fractal clusters are considered. The effects of surface-enhanced Raman scattering, photoluminescence enhancement, and nonlinear responses are studied. It is demonstrated that the model developed by S. G. Rautian, based on the quantum theory of nonlinearity, offers a more accurate description than previous theories. Among the scientific results of the theory proposed by Rautian and his team, the following points can be highlighted: metal nanoparticles with a diameter < 30 nm exhibit discrete energy levels, in contrast to semiconductor quantum dots; Rautian’s theory provides an accurate description of the linear part of the dielectric function, corresponding to the classical Drude model; the cubic nonlinear susceptibility according to Rautian’s model is consistent with experiments, indicating the contribution of conduction electrons.

The article provides a brief overview of modern achievements in quantum information science, problems and prospects for the development of quantum computing. The elementary mathematical model of quantum computing and the concept of quantum supremacy are discussed. The use of ultracold atoms for the implementation of quantum processors is considered.

The operation of our atmospheric quantum cryptography setup using the BB84 protocol and polarization coding has been investigated experimentally. The "sifted" quantum key distribution rate and the quantum bit error rate in the key remain constant for 1 hour and are equal to 7558±83 bit/s and 5.1±0.84%, respectively, at a distance of 20 m between the transmitter and receiver. The minimum possible transmittance of a quantum channel is calculated for these parameters of our setup.

The room-temperature visible photoluminescence of the In⁺ and As⁺ ion-implanted SiO₂ films is investigated as a function of As⁺ ion energy, annealing temperature, and excitation wavelength. As⁺ ions at the energy of 40, 80, or 135 keV and In⁺ ions at the energy of 50 keV, providing an average range ratio R_p^As/R_p^In of 1, 2 or 3, respectively, are used. The subsequent annealing is carried out at a temperature of 900 and 1100 °C. The photoluminescence spectra are excited with a laser wavelength l_ex = 442 and 473 nm. The photoluminescence peak near 550 nm is obtained at an excitation wavelength of 473 nm. The wavelength position of this peak is shifted to 520-530 nm at l_ex = 442 нм. As the R_p^As/R_p^In ratio increases, the photoluminescence intensity drops down, and the photoluminescence intensity ratio as a function of the annealing temperature is changed. The resultant effect is discussed in the frame of the electron and hole recombination within InAs nanocrystals.

The development of a multichannel application-specific integrated circuit (ASIC) for registration and processing of signals from microstrip sensors in synchrotron radiation coordinate detectors developed at the Budker Institute of Nuclear Physics of the Siberian Branch of the Russian Academy of Sciences to equip the experimental stations of the SKIF is presented. The ASIC contains 64 independent photon registration channels with four energy separation thresholds. The range of registered energies is from 3 to 60 keV. The structure and basic parameters of the electronics of the registration channel for direct photon counting are described in detail.

A comprehensive analysis of multispectral data obtained in different growing periods makes it possible to reconstruct the current state of the anthropogenically transformed natural environment and identify areas of historical nature management. Ambiguous displaying of secondary succession of forests requires comparison of changes in the spectral features of vegetation and features of the landscape objects texture in multi-season images. The methodological basis of such studies is an algorithm of multispectral data processing adapted to apply Haralick's textural features at the feature extraction stage. Joint segmentation of multi-seasonal multispectral images “smoothes out” local, less significant changes in the texture of landscape objects, which makes it possible to more clearly demonstrate the general trends in the distribution of vegetation. The area of interest included a fragment of the Pudemskiy ironworks dacha, where “timber and firewood forests” were cut down for the production needs in the 1st half of the 19th century. The segmented image demonstrates the fundamental differences between the vegetation cover of the western and eastern parts of the survey area: the site of historical cutting areas and the part that was not subjected to an intense anthropogenic impact.

The effect of multiplicative noise acting simultaneously with white noise on the accuracy of measuring the frequency of a coherent signal given as a sequence of coherent pulses is considered and analyzed. The condition for the absence of systematic errors at the output of the tracking meter is determined. The steepness of its discrimination characteristic, the spectral density of fluctuations at the discriminator output, and the coefficients that determine the normalizing action of automatic gain control are found. These values determine the amount of deterioration in the accuracy characteristics of the meter under the influence of multiplicative noise. It is shown that the influence of multiplicative noise increases with a decrease in the level of the undistorted part of the signal and with an increase in the ratio of the signal energy coherently accumulated in the linear circuits of the discriminator to the power spectral density of additive noise. It is shown that the effect of multiplicative noise is most pronounced in those cases where the width of the spectrum of the noise modulation function is commensurate with the bandwidth of the linear circuits of the discriminator.

The approach of applying machine learning methods for automatic detection of deer individuals in images is studied. The neural network technology is used to accurately count the number of deer from photographs. Deep learning methods for convolutional neural networks (Resnet50, DenseNet, CenterNet, InceptionV3, Xception) are used in conjunction with the “transfer learning” technique. Based on the Faster R-CNN Resnet50 network, a neural network is trained, which makes it possible to determine deer individuals from graphic images with an accuracy of 0.91 on a sample using the F1-score metric with a threshold value of 0.6.

The paper describes a technique for restoring images distorted by blurring from an unmanned aerial vehicle (UAV). Image reconstruction occurs using the point spread function (PSF), which is calculated based on a priori information about the basic flight and photographing parameters, as well as the physical properties of the optical system used. Well-known methods of image processing are used, such as restoration using the Wiener method, Lucy-Richardson method, Tikhonov method, and blind deconvolution method.

The paper deals with a new method of resolution enhancement in optical microscopy by the method of spatial subpixel shifts, i.e., shifts by some value smaller than the resolution provided by the objective lens. The resolution of optical microscopes is determined by the type of objective lens used. Professional microscopes have a set of microlenses with different magnifications, which are mounted on a turret containing several lenses. It is shown that a single objective lens can be used instead of a set of microlenses if subpixel shifts can be provided. An increase in spatial resolution is provided using the subpixel shift technique. In this case, the spectrum of the feature is augmented by a multiplier, whose type depends on the type of the lens aperture. To obtain high-resolution features, it is necessary to divide the Fourier spectrum of an image formed from several subpixel shifted images by a multiplier depending on the aperture type. This multiplier is called the aperture function. The aperture function is determined by the type of the lens used and can be its passport value. The paper illustrates an experimental method of obtaining the aperture function of a low-resolution (8×) lens calibration based on images obtained with high-resolution (40×) lenses. After determining the aperture function, one low-resolution lens can be used to acquire images with a resolution smaller than that of the selected high-resolution (40×) lens.

The accuracy of the construction of "depth maps" of space using an active-pulse television measurement system is evaluated on the basis of experimental data. Two multi-area methods of constructing "depth maps" are described. Results of estimation of the absolute and root-mean-square errors of range measurements for "depth maps" constructed from real and averaged frames obtained by the multi-area methods mentioned above in experimental studies are presented. The approximating functions of linear and polynomial types calculated by the least squares technique are taken as measuring functions used to construct the depth maps.