Optical Memory and Neural Networks最新文献

We investigate the topological properties of reflection zeros of three-layer structures consisting of a dielectric layer sandwiched between two metal layers, which can be used as optical differentiators. We show that the reflection zeros possess non-zero topological charges, which makes them topologically protected. With a small perturbation of the parameters of the structure (e.g., a change in one of the layer thicknesses), the reflection zero does not disappear, but shifts in the parameter space, i.e., appears at different wavelength and angle of incidence. We demonstrate that with a further parameter change, two zeros with opposite topological charges (+1 and –1) approach each other, merge, and then disappear. We believe that the obtained results give useful insight regarding the operation of layered metal-dielectric-metal structures possessing reflection zeros.

The paper considers beams in the form of geometric progression of optical vortices. Numerical modelling of the propagation of such optical fields in turbulent media is simulated using the Fresnel integral. The topological charges of the initial and resulting fields have been calculated. As expected, the analysis of the obtained results showed that superpositions with a smaller number of beams are more resistant to distortions by strongly turbulent media. However, in the case of a superposition in the form of a geometric progression with a parameter, the stability of beam propagation is affected not only by the medium parameters, but also by the parameters of the superposition.

Currently, wireless laser communication technologies demonstrate high throughput, but the implemented communication systems are not widely used. This feature is due to the low reliability of the communication channels being formed. Recent technological developments have shown successful results in the field of sealing and increasing the noise immunity of communication channels. Therefore, this paper presents a review of modern achievements in the field of multichannel atmospheric optical communication in the visible and near–infrared ranges. The advantages of using diffraction optical elements (DOE) in such systems, which form vortex beams of laser radiation with the required amplitude-phase structure for multiplexing tasks and increasing the noise immunity of information channels, are shown.

A metalens for detection an incident field with initially a fractional topological charge in the range from –2 to 0 is considered in this work. The metalens is constructed utilizing a spiral zone plate with a topological charge of –1.5. A change in the topological charge of the focused incident beam is shown by simulation to lead to a displacement of its focal spot from the center on the optical axis and to a change in the intensity maximum value, which results in the change in the intensity on the optical axis by 6.9, the change from –0.6 to –1.5 of the topological charge of the incident beam was considered. The intensity at the focus on the optical axis is also affected by the rotation of the beam with a fractional topological charge. This makes it possible using the metalens to measure the tilt angle of the incident beam in the range from 0° to 110°.

The effectiveness of using convolutional neural networks to optimize the parameters of a spatial-frequency ring filter that provides contrasting edge detection is investigated. To create a data set, arbitrary images in the form of test objects and their Fourier transform are used. It was found that, value regardless of the internal and external radius, the intensity maximum is detected in the test figure corners of a square and a triangle. However, these values affect the uniformity of energy distribution along the contour of the figures. The energy distribution along the contour of the test circle figure occurs in the same way, virtually size regardless of the internal and external annular diaphragm radius. As for the contour width, it increases in direct proportion to the inner radius size. A convolutional neural network with 8 layers was trained. The images were classified into two groups according to the required contrast in order to determine the optimal parameters of the bandpass filter for identifying edges in an arbitrary test image. The criterion for dividing the training set into two classes is the specified contrast threshold value. After 10 epochs of training the convolutional neural network, an accuracy rate of 0.836 was obtained for the “hook” test image.

This paper proposes an approach for detecting smoke in industrial production using computer vision. The task of detecting smoke and fire can be framed as a detection problem, making modern convolutional neural network models well-suited for this task. The main issues of detection in industrial production are considered, and solutions to these problems are proposed. In the study, the Faster R-CNN, MobileNet SSD v2, and YOLOv8 models were trained and tested in combination with various image preprocessing algorithms. The best result was achieved by the YOLOv8 model combined with the adaptive histogram equalization algorithm for image preprocessing, showing a precision value of 80.1%. As a result, it was demonstrated that deep convolutional networks are well-suited for the task of detecting smoke and fire. Additionally, the main problems and solutions for preparing data for training deep convolutional models were explored.

An adaptive method for wavefront aberrations compensating has been developed based on the use of a spatial light modulator, the phase function of which is matched to a set of Zernike functions. It is proposed to use the second central moment of intensity of the focal image as a functional. A study of the second central moment was carried out for both individual wavefront aberrations and their superposition. It is shown that achieving the reference value of the second moment can serve as a sign of sufficient compensation for aberration.

In this work, we investigated tight focusing characteristics of beams, which do not contain longitudinal component of intensity. The investigated beams have azimuthal or sector-azimuthal polarization and could contain vortex phase. It was numerically shown that beams with azimuthal and sector azimuthal polarization do not contain longitudinal component of intensity. Moreover, the helical phase added to the beams does not add longitudinal component to the electric field; however, it could be used for manipulation with longitudinal component of spin angular momentum in the tight focus. The possibility of generation of investigated beams was demonstrated using vector waveplates and spatial light modulator.

This article proposes an approach to the analysis of high-resolution hyperspectral images in the applied problem of analyzing the state of river waters. This method allows you to detect blooming or contamination of water by foreign substances. High-resolution hyperspectral images were obtained using a hyperspectrometer mounted on a small unmanned aerial vehicle. The difference between the spectra of river areas with different intensity of algal blooms is demonstrated. Samples of river water were taken, chemical analysis was carried out, which confirmed the different content of magnesium and calcium in all samples, corresponding to the intensity of algal blooms in the water. The effectiveness of using machine learning algorithms and the construction of index images for the classification of water areas with different intensity of algal blooms is shown.

In this paper, we consider a method for estimating camera motion parameters from images acquired from this camera, which is based on the use of vector coplanarity estimation. It has been previously shown that the proposed approach can be effectively applied to three-dimensional scenes invariant to their depth. However, due to the criterion used, it is difficult to utilize the RANSAC method to ensure the robustness of the developed method. In this paper, an approach based on the minimum covariance determinant estimation method is proposed. The proposed approach allows us to select the most consistent observations and make an estimation based on these observations. An experimental study of the proposed approach on synthetic data has been carried out. It is shown that the proposed algorithm can provide a significant increase in the reliability of motion parameters determination even in conditions of a small number of corresponding points