Home – Home – Jornals – Avtometriya 2024 number 5

For spaces of object representations with given distances, the lower bounds to the error probability of object classification subject to the fixed values of the amount of processed information are investigated. The bounds are defined by the strictly decreasing functions of the minimal average mutual information between the submitted objects and the estimations of their classes depending on the error probability. So, the inverse functions yield the minimal error probability for any fixed value of the average mutual information. For spaces of tree-structured and vector-based object representations, numerical realizations of the bounds are calculated. It is shown that the lower bound to the error probability is lower in the space of vector-based representations as against the similar bound in the space of tree-structured representations. Also, a possibility of decreasing the bound to the error probability by combining object representations with various distance functions is demonstrated.

This work proposes new algorithms for combining superpixels into segments using a “greedy” strategy and a combined quality measure that includes two components: a measure of information redundancy and variation of information. The “greedy” strategy was previously used by the author to speed up image segmentation from the condition of minimum information redundancy, and a two-component information quality measure was used in the problem of combining different segmentation maps. The joint use of the “greedy” strategy and the combined quality measure in new algorithms is aimed at accelerating the generation of segmentation maps and improving their quality through a compromise between the requirements of minimizing the number of informationally important segments and minimizing the information difference between the original images and the generated partitions. A computational experiment on test images shows that the proposed algorithms can speed up the segmentation process compared to the method based on minimizing information redundancy previously used by the author and improve the information characteristics of the resulting segmentation maps.

The paper proposes a method for protecting 3D printed products made using the FDM technology from unauthorized copying. This method is based on the controlled deviation of the trajectory of filament application on a flat section of the surface layer of the product. The amount of deviation of the line trajectory is determined by the built-in security mark, which plays the role of a watermark. Changes made in this way are difficult to detect visually and even more difficult to automatically read when scanning the shape of a 3D product and subsequently reproduce it when creating its copy. In this case, verification of the authenticity of a product by detecting a security mark can be performed by shooting the product with a digital camera and specialized processing based on spectral analysis and narrow-band filtering.

Experiments are conducted on the use of mathematical processing methods for an expanded comparative analysis of the composition of biomaterials and brefomatrices. The linear discriminant analysis and ROC analysis are used as the main mathematical methods for processing the Raman spectra of the biomaterials under study. It is found that demineralized biomaterials from juvenile dentin are less antigenic compared to brefomatrices.

The work is devoted to studying the possibility of increasing the accuracy of measuring the coordinates and brightness of small-sized moving objects in a sequence of images formed by a scanning multi-row photodetector. The measurement is based on approximating the object image with a function whose parameters are the object brightness and the shift of its projection center relative to the nodes of the array of photosensitive cells (PSC) of the receiver. The approximation quality criteria are, on the one hand, the root-mean-square difference between the object image and the approximating function, and on the other hand, the spread of its coordinates and brightness estimates depending on the position relative to the center of the photodetector cell in a series of images. Several of the most popular functions commonly used to approximate a discrete image of a small-sized object are compared on a series of images obtained by projecting an illuminated small-diameter diaphragm into the plane of the photosensitive layer of the receiver. It is shown that, even with a sufficiently high level of random noise, the use of an approximating model allows one to estimate the coordinates of an object with a root-mean-square error not exceeding 20% of the PSC lattice step.

The article proposes a technique to identify the location of the round-shaped object in the image and to calculate its diameter. The technique is based on calculating the cross-correlation values of the analyzed image and several specific pattern images and on the subsequent analysis which of the patterns and for what its position relative to the inspected image is the most suitable one. The technique allows one not only to calculate the object size and coordinates, but also to determine the actual presence or absence of the object in the field of view. The feasibility and practical applicability of the technique are shown using an example of inspecting the ball-type wire bonding to the surface of a die in integrated circuits. The algorithm can be used as a part of software for non-destructive testing systems intended to inspect semiconductor products during their manufacturing before casing them. In particular, the algorithm allows one to check if there is a ball-type wire bonding at the place where it should be and to determine if both the location of the ball and its diameter are acceptable.

This article explores the application of the transformer architecture for encoding states in reinforcement learning algorithms. A new approach is presented, which integrates transformers with existing methods, such as Soft Actor-Critic (SAC), to enhance their performance and generalization ability. The results of experimental studies show that the proposed approach can improve learning in complex tasks of acquiring movement skills in a three-dimensional space.

The work is devoted to the experimental study of the effectiveness of methods of classification of green vegetation and camouflage materials according to their spectral characteristics. The studies are performed using real spectral data obtained on a multichannel spectrograph based on a dispersing element and a multi-element CCD ruler in the visible and near-infrared wavelength range. A technique is proposed for processing a set of spectral data in order to form a representative system of features and subsequent application of adaptive threshold algorithms for detecting objects. The effectiveness of using vegetation indices calculated for spectral channels with different widths and at different positions of the central wavelength as signs for classifying types of materials is investigated. Special attention is paid to the detection of differences between fresh and anthropogenic vegetation. It is experimentally shown that both two-channel and three-channel indices can be used as informative features in the classification of objects.

The purpose of this work is to improve the accuracy of measurements of the optical rotation angle (ORA) by analyzing the effect of the adjustment error of the differential phase polarimeter analyzer. The paper presents a scheme of a differential phase polarimeter used for the study of optically active substances. To analyze the effect of the alignment error of the rotating analyzer on the ORA measurements results, mathematical modeling of the process of forming an informative signal in the presence of a deviation from the perpendicular plane of the analyzer relative to the axis of radiation propagation is applied. The results of the study of the influence of the adjustment error on the accuracy of restoring the phase difference of two harmonic signals are presented. The results obtained make it possible to develop requirements for the adjustment of the elements of the differential phase polarimeter to increase the accuracy of the ORA measurements.

The dynamics of effective and physical (geometric) shrinkage of a holographic photopolymer material is studied by measuring the evolution of transmission spectra of volume reflection holograms during their recording. It is shown experimentally for the first time that physical shrinkage occurs to a significant extent already at an early stage of hologram formation, when its diffraction efficiency is insignificant (less than 0.05). It is established that the effective shrinkage values at the end of the exposure for small and large grating periods differ significantly, while the physical shrinkage is approximately the same. For the grating samples studied, the average refractive indices before and after exposure are estimated. A method and device for measuring the shrinkage dynamics of reflection photopolymer holograms have been developed, ensuring the minimal errors in estimating the shrinkage parameters. The research results may be useful for deepening the knowledge about the nature of hologram formation in photopolymer materials, as well as for practical use in the development of holographic elements and devices based on them.

The method of calibration of scales with an irregular structure typical for absolute angle encoders based on pseudo-random codes has been developed and tested. The calibration results at a level of ±2.2ʺ with an uncertainty of ±0.5ʺ, characteristic for the measuring and diagnostic unit AE.1686 used, are experimentally obtained. The reliability of the results of calibration of irregular scales is confirmed by the identity of the curves obtained when comparing them with the calibration curve of a regular scale formed at another radius in a single technological cycle together with one of the controlled irregular scales.

A fundamentally new model of an optical switch with decentralized control is proposed. The advantages of the proposed switch are the absence of optoelectronic conversion of the information signal and the possibility of constructing scalable circuits of low complexity. The switch circuit includes the following main elements: electrically controlled switch elements based on lithium niobate, Bragg filters, optical valves, focusing systems, a control lens system, and a local system for generating control signals. The calculation of the optical devices of the switch shows the possibility of its implementation and operation in the range of 1310-1460 nm for transmitting control signals and 1550 nm for the information signal to ensure interaction between data center devices.

The development of acousto-optical tunable filter with a synthesized spectral transmission characteristic is presented. The wide-aperture (wide-angle, tangent) geometry of acousto-optical tunable filters widely used in modern hyperspectral systems is considered. It is shown that they can be described with a single-coordinate model of acousto-optic interaction. The relationships between the driving radiofrequency signal characteristics, the parameters of the acoustic wave forming the grating, and the spectral transmission function of the acousto-optical filter are determined. The fundamental limitations on the spectral transmission function characteristics are revealed. In a linear transmission approximation, the deduced relations describe the operation of an acousto-optical filter with an arbitrary law of amplitude and frequency modulation of the driving signal. The results are applicable for development of acousto-optical spectral systems with a controlled transmission function.

The problem of determining the optimal initial composition of the reaction mixture is formulated in general terms. A method for solving this problem based on the simulated annealing algorithm is described. Computational experiments are conducted for the Michaelis-Menten enzymatic reaction and the aminomethylation of thiols. Parameters for the efficient operation of the algorithm are selected. Using the proposed method, the optimal initial composition of the reaction mixture ensuring the maximum yield of the target product is identified. An analysis of the effectiveness of this algorithm is presented by comparing it with the results obtained from other optimization methods.