Search

We consider the problem of calculating a refractive element with two surfaces, forming a flat front and a given distribution of illumination. The supporting quadrics method is formulated for calculating a given optical element and it is shown that this method coincides with the gradient method for some functional related to the problem of the Monge-Kantorovich mass transfer problem. This enables adaptive selection of the step in the supporting quadric method. At the end of the article a design example is given.

The use of neural networks to detect differences in radiographic images of patients with pneumonia and COVID-19 is demonstrated For the optimal selection of resize and neural network architecture parameters, hyperparameters, and adaptive image brightness adjustment, precision, re-call, and f1-score metrics are used The high values of these metrics of classification quality (> 0 91) strongly indicate a reliable difference between radiographic images of patients with pneumonia and patients with COVID-19, which opens up the possibility of creating a model with good predictive ability without involving ready-to-use complex models and without pre-training on third-party data, which is promising for the development of sensitive and reliable COVID-19 ex-press-diagnostic methods © 2021, Institution of Russian Academy of Sciences All rights reserved

For non-contact 3D measurements in hydrodynamic tunnels by photogrammetry methods, it is necessary to refine the standard model of image formation in the camera by taking into account an effect of refraction of rays at the boundaries of optical media, namely, at an air-glass boundary and glass-working fluid boundary. The article presents a model of image formation for shooting in a working environment that includes various optical media and methods for calibrating an optical system for 3D measurements of the coordinates of scene objects, while taking into account the real boundaries of the optical media. Experimental results on calibrating the system of three-dimensional measurements when an object image is formed by rays passing through two optical boundaries are discussed.

Melanoma skin cancer is one of the most dangerous forms of skin cancer because it grows fast and causes most of the skin cancer deaths. Hence, early detection is a very important task to treat melanoma. In this article, we propose a skin lesion segmentation method for dermoscopic images based on the U-Net architecture with VGG-16 encoder and the semantic segmentation. Base on the segmented skin lesion, diagnostic imaging systems can evaluate skin lesion features to classify them. The proposed method requires fewer resources for training, and it is suitable for computing systems without powerful GPUs, but the training accuracy is still high enough (above 95 %). In the experiments, we train the model on the ISIC dataset – a common dermoscopic image dataset. To assess the performance of the proposed skin lesion segmentation method, we evaluate the Sorensen-Dice and the Jaccard scores and compare to other deep learning-based skin lesion segmentation methods. Experimental results showed that skin lesion segmentation quality of the proposed method are better than ones of the compared methods.

In this work, the splitting of resonance lines in a Fabry-Perot resonator formed by a section of a standard curved single-mode fiber with metal-coated ends is investigated numerically and experimentally. It is shown that this splitting is similar to the Autler-Townes splitting and results from a strong coupling between the fundamental mode of the core and the whispering gallery mode of the cladding of the curved fiber. The influence of all basic parameters of the curved resonator on the splitting of its resonance lines in the reflection and transmission spectra is considered. Prospects for the practical application of the effects studied for high-resolution optical refractometry, as well as the direction of further research are outlined.

Many recent studies have focused on developing image reconstruction algorithms in optical systems based on flat optics. These studies demonstrate the feasibility of applying a combination of flat optics and the reconstruction algorithms in real vision systems. However, additional causes of quality loss have been encountered in the development of such systems. This study investigates the influence on the reconstructed image quality of such factors as limitations of mass production technology for diffractive optics, lossy video stream compression artifacts, and specificities of a neural network approach to image reconstruction. The paper offers an end-to-end deep learning-based image reconstruction framework to compensate for the additional factors of quality losing. It provides the image reconstruction quality sufficient for applied vision systems.

Optical character recognition (OCR) in images captured from arbitrary angles requires preliminary normalization, i.e. a geometric transformation resulting in an image as if it was captured at an angle suitable for OCR. In most cases, a surface containing characters can be considered flat, and a pinhole model can be adopted for a camera. Thus, in theory, the normalization should be projective. Usually, the camera optical axis is approximately perpendicular to the document surface, so the projective normalization can be replaced with an affine one without a significant loss of accuracy. An affine image transformation is performed significantly faster than a projective normalization, which is important for OCR on mobile devices. In this work, we propose a fast approach for image normalization. It utilizes an affine normalization instead of a projective one if there is no significant loss of accuracy. The approach is based on a proposed criterion for the normalization accuracy: root mean square (RMS) coordinate discrepancies over the region of interest (ROI). The problem of optimal affine normalization according to this criterion is considered. We have established that this unconstrained optimization is quadratic and can be reduced to a problem of fractional quadratic functions integration over the ROI. The latter was solved analytically in the case of OCR where the ROI consists of rectangles. The proposed approach is generalized for various cases when instead of the affine transform its special cases are used: scaling, translation, shearing, and their superposition, allowing the image normalization procedure to be further accelerated.

The analysis of mobile technologies used to improve the quality of management in the banking sector, service and public administration is carried out. Based on the analysis of vulnerabilities of mobile stations (smartphones, tablets, smart devices, various peripheral devices of smart phones, etc.), technologies for providing Internet access for mobile systems (used in cellular networks, wireless access), as well as mobile services of information systems. A classification of information security threats is proposed and an analysis of the possibilities for implementing these threats is carried out. The aim of the work is to develop an up-to-date model of security threats to mobile technologies and to study the completeness and consistency of currently used mobile system protection tools, such as a mobile device manager, mobile application manager, trusted mobile application store, mobile application security gateway, etc. The paper discusses the sources of threats, vulnerabilities of technologies of mobile systems, the channels of exposure to threats, objects of exposure and the resulting damage from the implementation of threats that are characteristic of mobile systems and have different applications of threat implementations and vectors. The analysis of the context of the use of mobile medicine and the impact on the processes of providing medical services is carried out.

The paper is devoted to the actual problem of compression of processed and transmitted data, as well as of the data storage and protection. The analysis and features of data compression methods are presented. The method of universal coding by triples of binary sets (CT) is considered. The method is intended for lossless compression of binary data under conditions of unknown message source statistics. The method is based on dividing the original sequence of binary data into blocks of length n . Based on the analysis of the contents of the n–block, three parameters are assigned to each block: the number of units in the block, the sum of the unit position numbers, and the number of this particular combination in the corresponding prefix class. Approaches to reducing the complexity of procedures for calculating the parameters of CT are considered. A tabular algorithm for calculating coefficients and an algorithm for filling table rows using a recursive relation for elements of sets and an estimate of the amount of memory for storing tables as well as the ways to reduce the memory capacity are presented. The dependence of the bit width of the QT codewords on the bit width of n -blocks is considered. The approaches and features of the use of coding by triplets for information protection are analyzed.

We show that a 70-nm dielectric nanoparticle placed on the optical axis near the surface (at a distance less than 100 nm) of a high-NA gradient microlens made of silicon, which is illuminated by a laser beam of 1.55 μm wavelength, is attracted to the lens surface with a piconewton force. The profile of the lens refractive index is described by a hyperbolic secant function. If a cut-out is made in the lens output surface, then the nanoparticle will be pulled into this cut-out, producing a kind of 'optical magnet'. If a reverse energy flow is to be generated on the optical axis near the output surface of such a gradient lens, this will lead to an absorbing dielectric nanoparticle being pulled toward the surface with a greater force than a similar non-absorbing particle. In the absence of a reverse flow, both absorbing and non-absorbing particles will be attracted to the surface with an equal force. The electromagnetic fields involved are calculated using a finite difference time domain (FDTD) method and the acting forces are calculated using a Maxwell stress tensor.

We propose a new approach to non-contact recording of respiratory function based on the analysis of a sequence of Infrared-terahertz images of the human face, allowing the processes that occur during breathing to be visualized. To obtain quantitative estimates of the respiratory function, two methods are proposed. The first one utilizes a probe which implements the function of spatial differentiation and provides high sensitivity, but requires an increased accuracy of positioning the probe at the nasal opening and scaling in accordance with the camera angle. The other one is a histogram method for obtaining quantitative estimates of the external respiratory function, which is scale invariant and does not require precise positioning, but has a lower sensitivity compared to the first one. The methods proposed have made it possible to remotely evaluate the respiratory rate, which correlates well with the data obtained by a contact method of respiratory function registration.

The possibility of extending the patterns characteristic of the information process to the educational process is substantiated in this paper. The features of the manifestation of information security properties in the context of educational activities related to the training of specialists in the field of information security are considered. The disadvantages of evaluating this property by quantitative methods are analyzed. It substantiates the possibility of assessing the security of educational activity by qualitative methods. Classification grounds for systematization of security characteristics are determined. The structure of the characteristics of the security of educational activities for the training of specialists in the field of information security is given. Methodological provisions for the implementation of the synthesis procedure of this structure are formulated. Variants of characteristic tables and an example of a decision table for assessing the values of the security characteristics of educational activities in terms of a linguistic scale are given. The results obtained will allow us to assess the feasibility of carrying out measures practically to ensure the information security of educational activities. The developed method can be considered as a methodological support for solving the practical problem of assessing the security of educational activities.

The numerical simulation and selection of optimal parameters of the diffraction grating for a newly developed technology for the manufacture of plane waveguide systems are performed. In contrast to the use of ready-made silicon wafers on an insulator, the new technology has been developed for the manufacture of a fully autonomous radiation input system, a coupling element and the waveguide itself. A general description of the technology of the ‘radiation input – propagation – radiation output’ system is given. Concrete fabrication parameters of the lattice height, the substrate and coating layers are found. The coupling efficiency of radiation input into the waveguide is found to be 30%.

When laser coagulation of eye retina is carried out, the laser beam is directed to target retinal areas selected by an ophthalmologist. The exposure to laser light produces a photocoagulate. When using laser coagulation, the main problem is selecting both the laser exposure areas that would cover all pathological zones and the laser exposure parameters to prevent retina damage. Any patient has an individual fundus structure. The individual structure of pathological and anatomical elements must be taken into account to achieve the desired therapeutic effect. The formation of coagulates in all hemorrhage-affected areas results in the desired therapeutic effect. The vascular layer must be heated to a sufficient temperature to form a coagulate. Such an effect can be predicted using mathematical modeling of laser exposure. In this paper, we consider methods of mathematical modeling of laser exposure based on the solution of a heat equation. The methods are compared in terms of their computational complexity and solution stability. An analysis of the possibility of predicting the therapeutic effect using methods of mathematical modeling of laser exposure is carried out.

The paper discusses different approaches to image and video processing aiming to solve the problems of detecting, tracking and estimating the parameters of moving objects. The developed algorithms for solving these problems are described in relation to the field of transport analytics. When developing the algorithms, attention was given to solving the problems on an embedded platform of video surveillance cameras, which imposes restrictions on the computational complexity. The first (basic) algorithm for moving object detection and parameter estimation is based on processing two associated areas of an image. This algorithm includes a computationally efficient adaptive procedure for evaluating and updating the background component of an image. The procedure is based on the physics of the process of movement of the object of interest through a processing zone. The second algorithm performs object tracking based on an optical flow method initialized by feature points. The third algorithm is based on object segment tracking and is computationally efficient for the implementation on an embedded platform of intelligent cameras. Results of experimental studies of the proposed algorithms are presented, as well as a comparison with some well-known algorithms. It is shown that tracking algorithms can improve the accuracy of moving object parameter estimation. Tracking also reduces the number of classification errors compared to the basic approach to object detection and parameter estimation.

Using electronic beam lithography and reactive ion beam etching, a metalens is manufactured in a thin layer of amorphous silicon of a 130-nm depth, a 30-µm diameter, and a 633-nm focal length (equal to the illumination wavelength). The metalens is composed of 16 sectored subwavelength binary gratings with a 220-nm period. The uniqueness of this metalens is that when illuminated by left-handed circularly polarized light, it is capable of generating a left-handed circularly polarized vortex beam with a topological charge of 2, generating a second-order cylindrical vector beam when illuminated by linearly polarized light. Both for linear and circular incident polarization, an energy backflow is found to be generated in the vicinity of the tight focus. Transverse intensity distributions measured with a scanning near-field optical microscope near the focus of the metalens are in qualitative agreement with the intensity distributions calculated by the FDTD method. This confirms that a backward energy flow takes place at the focus of the metalens. A metalens generating an energy backflow near its focus is fabricated and characterized for the first time.

In this work, the effect of the irradiation intensity on the radiation hardness of CMOS integrated circuits (ICs) was studied. We investigated 3 types of ICs manufactured according to design rules from 0.8 to 3 microns, and used in various information transmission systems: 1586IN4, HEF4093BT and MC14504B. The U-31/33 accelerator of electrons has been used for the study in both gamma-ray mode (for moderate intensity) and pulse electron generation mode (for high intensity). As a result, it was revealed that an increase in the intensity of exposure by 4 orders of magnitude might result in a radiation hardness decrease up to factor of 30. This circumstance has to be taken into account when testing CMOS ICs for radiation hardness to high intensity effects. The obtained experimental data confirm theoretical understanding of the influence of increased irradiation intensity on the dose hardness of CMOS ICs.

The paper considers a problem of extracting user preferences based on their photo gallery. We propose a novel approach based on image captioning, i.e., automatic generation of textual descriptions of photos, and their classification. Known image captioning methods based on convolutional and recurrent (Long short-term memory) neural networks are analyzed. We train several models that combine the visual features of a photograph and the outputs of an Long short-term memory block by using Google's Conceptual Captions dataset. We examine application of natural language processing algorithms to transform obtained textual annotations into user preferences. Experimental studies are carried out using Microsoft COCO Captions, Flickr8k and a specially collected dataset reflecting the user’s interests. It is demonstrated that the best quality of preference prediction is achieved using keyword search methods and text summarization from Watson API, which are 8 % more accurate compared to traditional latent Dirichlet allocation. Moreover, descriptions generated by trained neural models are classified 1 – 7 % more accurately when compared to known image captioning models.

Technological features of the growth of single crystal silicon carbide inevitably create condi-tions for the formation of crystal structure defects in them. A method is proposed for recognizing and analyzing a dislocation structure of single crystal silicon carbide based on the use of optical microscopy and a direct distribution neural network. The method was tested on homoepitaxial lay-ers of 4H-polytype silicon carbide. Software has been developed that allows building maps of the dislocation structure distribution over the surface of single crystal silicon carbide. The software was tested on digital images of the surface of silicon carbide epitaxial layers. The accuracy of recognition of dislocation structure was 95%. The dislocation mapping is used in the development of process technologies for reducing their density during the growth of single crystals.

This paper presents the results of an analysis of the necessary propagation conditions in a step-index optical fiber with a Kerr nonlinearity of two modes, LP01 and LP11, during the transmission of high-power optical pulses. All results were obtained by solving a system of two nonlinear equations for these modes, obtained by the Gauss approximation method, and the subsequent use of a procedure for refining estimates using the mixed finite elements method. The necessary conditions are determined, estimates of the boundaries for the range of normalised frequencies for which they are fulfilled are obtained, and an approximate formula is proposed for estimating the upper limit of this range.

On the basis of diffraction transformations of an optical wave field a mathematical model for the formation of speckle modulated interference patterns and signals at the output of a speckle interferometer is developed, which allows us to identify their properties and quantitative parameters. Speckle interferometers based on a Michelson arrangement are considered, where objects with scattering surfaces are used instead of mirrors in the reference and object arms. Results of numerical simulation of speckle modulated interference patterns on the basis of diffraction transformations of wave fields in an interferometer are discussed. Simulated images obtained at the output of the interferometer when focusing laser beams on the scattering surfaces of the controlled and reference objects are considered. Experimental results of using a speckle interferometer with a digital matrix photodetector for measuring the temperature micro-displacements of an object with a scattering surface and a quantitative comparison of experimental data with the results obtained by a numerical experiment using a diffraction model of a speckle interferometer are presented.

The existing action recognition systems mainly focus on generalized methods to categorize human actions. However, the generalized systems cannot attain the same level of recognition performance for new users mainly due to the high variance in terms of human behavior and the way of performing actions, i.e. activity handling. The use of personalized models based on similarity was introduced to overcome the activity handling problem, but the improvement was found to be limited as the similarity was based on physiognomies rather than the behavior. Moreover, human interaction with contextual information has not been studied extensively in the domain of action recognition. Such interactions can provide an edge for both recognizing high-level activities and improving the personalization effect. In this paper, we propose the context-aware personalized human activity recognition (CAPHAR) framework which computes the class association rules between low-level actions/sensor activations and the contextual information to recognize high-level activities. The personalization in CAPHAR leverages the individual behavior process using a similarity metric to reduce the effect of the activity handling problem. The experimental results on the “daily lifelog” dataset show that CAPHAR can achieve at most 23.73% better accuracy for new users in comparison to the existing classification methods.

Cloud computing can provide users with basic hardware resources, and there are three instance types: reserved instances, on-demand instances and spot instances. The price of spot instance is lower than others on average, but it fluctuates according to market demand and supply. When a user requests a spot instance, he/she needs to give a bid. Only if the bid is not lower than the spot price, user can obtain the right to use this instance. Thus, it is very important and challenging to predict the price of spot instance. To this end, we take the most popular and representative Amazon EC2 as a testbed, and use the price history of its spot instance to predict future price by building a k-Nearest Neighbors (kNN) regression model, which is based on our mathematical description of spot instance price prediction problem. We compare our model with Linear Regression (LR), Support Vector Machine Regression (SVR), Random Forest (RF), Multi-layer Perception Regression (MLPR), gcForest, and the experiments show that our model outperforms the others.

Today, many people use dashcams, and videos recorded on dashcams are often used as evidence of accident fault. People can upload videos of dashcam recordings with specific accident clips and share the videos with others who request them, by providing the time or location of an accident. However, dashcam videos are erased when the dashcam memory is full, so periodic backup is necessary for video sharing. It is inconvenient for dashcam owners to search for and transmit a requested video clip from backup videos. In addition, anonymity is not ensured, which may reduce location privacy by exposing the video owner’s location. To solve this problem, we propose a video sharing scheme with accident detection using deep learning coupled with automatic transfer to the cloud; we also propose ensuring data and operational integrity along with location privacy by using blockchain smart contracts. Furthermore, our proposed system uses proxy re-encryption to enhance the confidentiality of a shared video. Our experiments show that our proposed automatic video sharing system is cost-effective enough to be acceptable for deployment.

The paper is devoted to the development of an effective semantic segmentation algorithm for automation of airport infrastructure labelling in RGB satellite images. This task is addressed using algorithms based on deep convolutional artificial neural networks, as they have proven themselves in a wide range of tasks, including the terrestrial imagery segmentation, where they show consistently high results. A new dataset was labelled for this particular task and a comparative analysis of different architectures and backbones was carried out. A conditional random field model (CRF) was used for postprocessing and accounting of contextual information and neighborhood of objects of different classes in order to eliminate outliers. Features of the solutions applied at all preparatory stages of the algorithm were described, including data preparation, neural network training and post-processing of the training results.

Parameterized computation theory has developed rapidly over the last two decades. In theoretical computer science, it has attracted considerable attention for its theoretical value and significant guidance in many practical applications. We give an overview on parameterized algorithms for some fundamental NP-hard problems, including MaxSAT, Maximum Internal Spanning Trees, Maximum Internal Out-Branching, Planar (Connected) Dominating Set, Feedback Vertex Set, Hyperplane Cover, Vertex Cover, Packing and Matching problems. All of these problems have been widely applied in various areas, such as Internet of Things, Wireless Sensor Networks, Artificial Intelligence, Bioinformatics, Big Data, and so on. In this paper, we are focused on the algorithms’ main idea and algorithmic techniques, and omit the details of them.

Cloud computing based on OpenStack is widely used as a distributed computing platform. OpenStack has progressed at a rapid pace, incorporating a variety of service modules; it is supported by many companies, has a community of active developers, and a diverse user base. OpenStack uses message queue to coordinate and exchange operation and status information between services. OpenStack supports various message queue services including RabbitMQ, Qpid, and ZeroMQ, whereas its distribution architecture uses RabbitMQ. As an OpenStack’s message queue service, RabbitMQ runs on a controller node as a centralized service. In case of the centralized service, increased usage may cause slowed response times and security vulnerability. This paper proposes a Hybrid decentralized Practical byzantine fault tolerance Blockchain Framework with two-step verification for OpenStack message queue service. When compared to existing OpenStack message queue service, OpenStack with the proposed framework demonstrates identical reliability a faster response time by approximately 46.75% with a two-step verification process and decentralization approach. Additionally, a reduction in the security vulnerability in the OpenStack message queue information with saving the message queue information into each node by blockchain-based decentralized data duplication approach.

Handwritten signature is one of the most common methods of biometric authentication, where static and dynamic signature characteristics are used to confirm the user's identity. The existing developments are based on various technologies, such as the neural network, the hidden Markov model, and machine learning algorithms. This topic is rapidly developing, new approaches and algorithms for solving the problem improve the accuracy of verification and learning speed. The purpose of this study is to analyze existing approaches to the signature verification. The most promising algorithm will be used as the basis for the developed authentication system based on a handwritten signature.

Crowdsourcing and crowd computing are a trend that is likely to be increasingly popular, and there remain a number of research and operational challenges that need to be addressed. The human-centric computational abstraction called situation may be used to cope with these difficulties. In this paper, we focus on one such challenge, which is how to assign crowd assessment tasks about security and privacy in online social networks to the most appropriate users efficiently, effectively and accurately. Specifically, here we propose a novel task assignment method to facilitate crowd assessment, which improves the security and trustworthiness of social networking platforms, as well as a task assignment algorithm based on SocialSitu, which is a social-domain-focused situational analytics. Findings from our crowd assessment experiments on a real world social network Shareteches show that the precision and recall of the proposed method and algorithm are 0.491 and 0.538 higher than those of a random algorithm’s, as well as 0.336 and 0.366 higher than users’ theme-aware algorithm’s, respectively. Moreover, these results further suggest that our experimental evaluation enhance the security and privacy of online social networks.

The article proposes a modification of the Blurred Image Matching (BIM) method, a key point selection method in images, thus solving a problem of their accurate comparison. A new approach for blobs selection and comparison is proposed. The use of these modifications allows us to achieve an increase in the proportion of correctly matched pairs of images by 30.2% compared to the basic method when working with noisy data.

The effect of diffraction on the directional light transmittance of grating optical filters designed for smart windows is studied. The filter has an angular selective light transmission due to two thin-film gratings formed by "non-transmissive" (absorptive, reflective or scattering) parallel strips on the surfaces of a single or double glazed window. Methods are considered for calculating the op-timal slope angle of two filter’s gratings, their relative position on the opposite window surfaces and the widths of the strips of both gratings to minimize light transmission on a specific date and time of day for a given latitude and longitude of the building and azimuth of the window. A meth-od has been developed for calculating the diffraction at the input and output gratings of the filter with regard for a complex movement of the Sun relative to the window and a corresponding change in the incidence angle of the solar beams on the gratings. To evaluate a decrease in the light transmittance of the filter due to diffraction, an additional factor is introduced into the calcu-lation equation. Geometric parameters of the filters and diffraction are calculated. The values of the diffraction factor are 0.9999992 and 0.9999998 for single or double glazed windows, respec-tively, i.e., the influence of diffraction on the light transmittance of the grating filter can be ne-glected.

The article discusses techniques for the precise positioning of a probe tip on the surface under study in scanning probe microscopy. The precise tip positioning is of high importance when studying the same sample with various microscopes, probes, in various environments and conditions. It is offered that the precise positioning should be done by applying on the surface equidistant reference marks with a nanoindentor along a straight line. The article describes a procedure for shifting the microscope’s field of view for detecting the reference marks and subsequently moving the probe along the marks line. A reference marks recognition algorithm is presented. The algorithm is based on the image correlation analysis using a master mark image and the subsequent extraction of special points (at local maxima of the correlation coefficient). A descriptor of mutual distances between the special points and the Pearson coefficient (аs a quantitative criterion of recognition) are used for recognition. Three types of reference marks are considered: a circle, a circle with gradient filling, and a circle with continuous filling. It is shown that the number of the allocated special points can be significantly larger than that of the reference marks. A method for filtering special points is offered. The method is based on the threshold filtering of special points using values of the correlation coefficient. The filtering consists in excluding those special points in whose neighborhood the difference of the maximum and minimum of the correlation coefficient does not exceed a preset threshold. Results confirming the effectiveness of the offered filtering method are presented. Recommendations for choosing the parameters of the recognition and filtering algorithms are given.

In this paper a new method for contour detection in grayscale images is proposed. The pro-posed method is based on the use of an image weight model, which allows one to estimate its pix-els from the point of view of their significance for perception. In this case, the most significant pixels are those that contain characteristic features of the image, including brightness differences at the boundaries of the regions. To assess the significance of pixels, we propose a procedure for analyzing the contribution of the corresponding wavelet coefficients at different scale levels to the total energy of the image. The described method of contour detection involves building an image weight model, determining the directions of linear segments along the edges on the weight image, analyzing the significance of pixels and linking significant pixels. The advantage of the method is the high operation speed (the corresponding loop detector works on average four times faster than the Canny edge detector). In addition, the paper describes a detector of significant image areas, which is also based on the weight model. The proposed approach can be used in various systems of information processing and control based on methods and tools of computer vision, including control and navigation systems of unmanned vehicles, remote sensing of the Earth, systems for pavement defect detection, biometric systems, etc.

Unhealthy behaviors regarding nutrition are a global risk for health. Therefore, the healthiness of an individual’s nutrition should be monitored in the medium and long term. A powerful tool for monitoring nutrition is a food diary; i.e., a daily list of food taken by the individual, together with portion information. Unfortunately, frail people such as the elderly have a hard time filling food diaries on a continuous basis due to forgetfulness or physical issues. Existing solutions based on mobile apps also require user’s effort and are rarely used in the long term, especially by elderly people. For these reasons, in this paper we propose a novel architecture to automatically recognize the preparation of food at home in a privacy-preserving and unobtrusive way, by means of air quality data acquired from a commercial sensor. In particular, we devised statistical features to represent the trend of several air parameters, and a deep neural network for recognizing cooking activities based on those data. We collected a large corpus of annotated sensor data gathered over a period of 8 months from different individuals in different homes, and performed extensive experiments. Moreover, we developed an initial prototype of an interactive system for acquiring food information from the user when a cooking activity is detected by the neural network. To the best of our knowledge, this is the first work that adopts air quality sensor data for cooking activity recognition.

The paper presents the results of the analysis of scientific and technical literature and open regulatory and administrative documents of the international, Federal and departmental levels devoted to the process of functioning of information protection systems (SPI) from unauthorized access (UA) in automated systems (AS) at the objects of Informatization of internal Affairs bodies (ATS). Using the example of a typical certified SPI from UA «Strazh NT 4.0» the functionality of SPI from UA operating in protected ATS is analyzed, shortcomings are identified and the key aspects of improving subsystems of access control systems based on the use of new information and telecommunication technologies related to improving real security as on the objects of Informatization Department are defined. As a promising technology additional biometric authentication based on realization of recognition of the access subject on individual dynamics of a keyboard set (keyboard handwriting) is offered.

The paper discusses a method of visual attention based on vertex ranking of graphs on the basis of image features. The aim of the research is to develop a method that allows high-precision detection of objects in images with low color contrast between the selected and background areas. The image is pre-segmented into regions to calculate the saliency map. The graph is based on regions. Each region is associated with related regions, as well as with areas adjacent to adjacent regions. The regions are vertices of the graph. The vertices of the graph are ranked according to the characteristics of the corresponding image areas. The scope is highlighted based on requests from background areas. The saliency map is determined based on background area queries. Regions adjacent to the edges of the image belong to the background areas. Color features of the image were used in the existing approach of visual attention based on the manifold ranking. Texture features and shape features are additionally used in the proposed method to improve accuracy. Gabor's energy function is used to calculate texture features. The distance between centers of the regions is calculated by analyzing the form. The proposed method has shown good results for detecting objects in images in which the background color and object color are in similar ranges. The experimental results are presented on test images. Precision-recall curves showing the advantage of the developed method are constructed.

Artificial intelligence is one of the key technologies for progression to the fourth industrial revolution. This technology also has a significant impact on software professionals who are continuously striving to achieve high-quality software development by fixing various types of software bugs. During the software development and maintenance stages, software bugs are the major factor that can affect the cost and time of software delivery. To efficiently fix a software bug, open bug repositories are used for identifying bug reports and for classifying and prioritizing the reports for assignment to the most appropriate software developers based on their level of interest and expertise. Owing to a lack of resources such as time and manpower, this bug report triage process is extremely important in software development. To improve the bug report triage performance, numerous studies have focused on a latent Dirichlet allocation (LDA) using the k-nearest neighbors or a support vector machine. Although the existing approaches have improved the accuracy of a bug triage, they often cause conflicts between the combined techniques and generate incorrect triage results. In this study, we propose a method for improving the bug report triage performance using multiple LDA-based topic sets by improving the LDA. The proposed method improves the existing topic sets of the LDA by building two adjunct topic sets. In our experiment, we collected bug reports from a popular bug tracking system, Bugzilla, as well as Android bug reports, to evaluate the proposed method and demonstrate the achievement of the following two goals: increase the bug report triage accuracy, and satisfy the compatibility with other state-of-the-art approaches.

With the wide application of Light Detection and Ranging (LiDAR) in the collection of high-precision environmental point cloud information, three-dimensional (3D) object classification from point clouds has become an important research topic. However, the characteristics of LiDAR point clouds, such as unstructured distribution, disordered arrangement, and large amounts of data, typically result in high computational complexity and make it very difficult to classify 3D objects. Thus, this paper proposes a Convolutional Neural Network (CNN)-based 3D object classification method using the Hough space of LiDAR point clouds to overcome these problems. First, object point clouds are transformed into Hough space using a Hough transform algorithm, and then the Hough space is rasterized into a series of uniformly sized grids. The accumulator count in each grid is then computed and input to a CNN model to classify 3D objects. In addition, a semi-automatic 3D object labeling tool is developed to build a LiDAR point clouds object labeling library for four types of objects (wall, bush, pedestrian, and tree). After initializing the CNN model, we apply a dataset from the above object labeling library to train the neural network model offline through a large number of iterations. Experimental results demonstrate that the proposed method achieves object classification accuracy of up to 93.3% on average.

The technology of Artificial Intelligence (AI) brings tremendous possibilities for autonomous vehicle applications. One of the essential tasks of autonomous vehicle is environment perception using machine learning algorithms. Since the cyclists are the vulnerable road users, cyclist detection and tracking are important perception sub-tasks for autonomous vehicles to avoid vehicle-cyclist collision. In this paper, a robust method for cyclist detection and tracking is presented based on multi-layer laser scanner, i.e., IBEO LUX 4L, which obtains four-layer point cloud from local environment. First, the laser points are partitioned into individual clusters using Density-Based Spatial Clustering of Applications with Noise (DBSCAN) method based on subarea. Then, 37-dimensional feature set is optimized by Relief algorithm and Principal Component Analysis (PCA) to produce two new feature sets. Support Vector Machine (SVM) and Decision Tree (DT) classifiers are further combined with three feature sets, respectively. Moreover, Multiple Hypothesis Tracking (MHT) algorithm and Kalman filter based on Current Statistical (CS) model are applied to track moving cyclists and estimate the motion state. The performance of the proposed cyclist detection and tracking method is validated in real road environment.

Navigation systems help users access unfamiliar environments. Current technological advancements enable users to encapsulate these systems in handheld devices, which effectively increases the popularity of navigation systems and the number of users. In indoor environments, lack of Global Positioning System (GPS) signals and line of sight with orbiting satellites makes navigation more challenging compared to outdoor environments. Radio frequency (RF) signals, computer vision, and sensor-based solutions are more suitable for tracking the users in indoor environments. This article provides a comprehensive summary of evolution in indoor navigation and indoor positioning technologies. In particular, the paper reviews different computer vision-based indoor navigation and positioning systems along with indoor scene recognition methods that can aid the indoor navigation. Navigation and positioning systems that utilize pedestrian dead reckoning (PDR) methods and various communication technologies, such as Wi-Fi, Radio Frequency Identification (RFID) visible light, Bluetooth and ultra-wide band (UWB), are detailed as well. Moreover, this article investigates and contrasts the different navigation systems in each category. Various evaluation criteria for indoor navigation systems are proposed in this work. The article concludes with a brief insight into future directions in indoor positioning and navigation systems.

The necessity to improve security in a multi-cloud environment has become very urgent in recent years. Although in this topic, many methods using the message authentication code had been realized but, the results of these methods are unsatisfactory and heavy to apply, which, is why the security problem remains unresolved in this environment. This article proposes a new model that provides authentication and data integrity in a distributed and interoperable environment. For that in this paper, the authors first analyze some security models used in a large and distributed environment, and then, we introduce a new model to solve security issues in this environment. Our approach consists of three steps, the first step, was to propose a private virtual network to secure the data in transit. Secondly, we used an authentication method based on data encryption, to protect the identity of the user and his data, and finally, we realize an algorithm to know the integrity of data distributed on the various clouds of the system. The model achieves both identity authentication and the ability to inter-operate between processes running on different cloud’s provider. A data integrity algorithm will be demonstrated. The results of this proposed model can efficiently and safely construct a reliable and stable system in the cross-cloud environment.

We compare approaches for the automatic detection of pathological changes in brain MRI images that are visible to the naked eye. We analyse multi-stage approaches based on deep learning and threshold processing. A convolutional neural network was formed, a classifier was built based on the use of an ensemble of decision trees, and an algorithm was created for multi-stage image processing. Because of experimental studies, it was found that the most effective method for recognizing images of magnetic resonance imaging is an approach based on an ensemble of decision trees. With its help, 95 % of the images from the test sample were classified correctly. At the same time, using the convolutional neural network, it was possible to classify correctly all images containing the area of pathological changes. The data obtained can be used in practice for the diagnosis of brain diseases, for automating the processing of a large number of studies of magnetic resonance imaging.

The performance of a CNN based medical-image classification network depends on the intensities of the trained images. Therefore, it is necessary to generalize medical images of various intensities against degradation of performance. For lesion classification, features of generalized images should be carefully maintained. To maintain the performance of the medical image classification network and minimize the loss of features, we propose a method using a generative adversarial network (GAN) as a generator to adapt the arbitrary intensity distribution to the specific intensity distribution of the training set. We also select CycleGAN and UNIT to train unpaired medical image data sets. The following was done to evaluate each method’s performance: the similarities between the generalized image and the original were measured via the structural similarity index (SSIM) and histogram, and the original domain data set was passed to a classifier that trained only the original domain images for accuracy comparisons. The results show that the performance evaluation of the generalized images is better than that of the originals, confirming that our proposed method is a simple but powerful solution to the performance degradation of a classification network.

Object detection methods aim to identify all target objects in the target image and determine the categories and position information in order to achieve machine vision understanding. Numerous approaches have been proposed to solve this problem, mainly inspired by methods of computer vision and deep learning. However, existing approaches always perform poorly for the detection of small, dense objects, and even fail to detect objects with random geometric transformations. In this study, we compare and analyse mainstream object detection algorithms and propose a multi-scaled deformable convolutional object detection network to deal with the challenges faced by current methods. Our analysis demonstrates a strong performance on par, or even better, than state of the art methods. We use deep convolutional networks to obtain multi-scaled features, and add deformable convolutional structures to overcome geometric transformations. We then fuse the multi-scaled features by up sampling, in order to implement the final object recognition and region regress. Experiments prove that our suggested framework improves the accuracy of detecting small target objects with geometric deformation, showing significant improvements in the trade-off between accuracy and speed.

A method of image fusion based on wavelet decomposition of the original images is considered. An integrated monochrome image is formed from images of the same scene obtained in different spectral ranges. A strategy for fusing detail coefficients by comparing the proportions of their magnitudes for all original images is proposed. The fusion procedure does not require any thresholds. This fusion procedure can be performed for any number of original images. The algorithm does not introduce additional distortions and collects all the necessary information from the original images. The quantitative and qualitative evaluation of the results was performed. The proposed algorithm can be used in optoelectronic systems for automatic image processing.

Melanoma is considered to be one of the deadliest skin cancer types, whose occurring frequency elevated in the last few years; its earlier diagnosis, however, significantly increases the chances of patients’ survival. In the quest for the same, a few computer based methods, capable of diagnosing the skin lesion at initial stages, have been recently proposed. Despite some success, however, margin exists, due to which the machine learning community still considers this an outstanding research challenge. In this work, we come up with a novel framework for skin lesion classification, which integrates deep features information to generate most discriminant feature vector, with an advantage of preserving the original feature space. We utilize recent deep models for feature extraction, and by taking advantage of transfer learning. Initially, the dermoscopic images are segmented, and the lesion region is extracted, which is later subjected to retrain the selected deep models to generate fused feature vectors. In the second phase, a framework for most discriminant feature selection and dimensionality reduction is proposed, entropy-controlled neighborhood component analysis (ECNCA). This hierarchical framework optimizes fused features by selecting the principle components and extricating the redundant and irrelevant data. The effectiveness of our design is validated on four benchmark dermoscopic datasets; PH2, ISIC MSK, ISIC UDA, and ISBI-2017. To authenticate the proposed method, a fair comparison with the existing techniques is also provided. The simulation results clearly show that the proposed design is accurate enough to categorize the skin lesion with 98.8%, 99.2% and 97.1% and 95.9% accuracy with the selected classifiers on all four datasets, and by utilizing less than 3% features.

Problems of hidden storage and transmission of information are successfully solved by using steganography. The most common steganographic maskers for transmitting and storing information at a given time are media files, such as photos, videos, and audio files. The use of such maskers imposes a number of restrictions, one of which is a limit on the size of the embedded information. Thus, increasing the amount of hidden data changes the entropy properties of the masker and allows you to detect the fact of steganography. One of the alternative ways to ensure the possibility of steganographic concealment of large amounts of information is to use a fundamentally different method of concealment. In this regard, it becomes relevant to develop a method for steganographic storage of information in the file system, as a masker, in which you can secretly store quite large amounts of data. This paper presents an approach to steganographic storage of information in the file system of the extracted data carrier, which is usually an electronic data carrier with the FAT32 file system. During the analysis of existing steganography tools designed to hide information in file systems, their weaknesses and strengths were identified, on the basis of which the requirements for the developed tool were put forward. The paper presents a detailed description of the algorithms used in the new approach. Its main advantages and the situations in which its application is justified are described, as well as its perfection by the Cachin method is proved.

The paper presents an analysis of existing approaches to estimation of single event rate (SER) in integrated circuits under effects of charged particles of space radiation environment. These issues are of significant importance in the light of the expansion of the scope of practical application of cyber-physical control systems for space objects, since it is mainly due to the SER that information is lost in the register elements and in the memory cells of the electronic blocks of spacecraft. It is shown that existing models based on energy deposition in fixed sensitive volume are not applicable for SER estimations in case of high threshold linear energy transfer (LET) values. An alternative approach is proposed. It is based on diffusion charge collection model, which can be used to estimate the SER cross-sections in isotropic particle field. A universal dependence for SER estimation in integral circuits (ICs) at geostationary orbit is proposed and used as a basis for establishing classification of devices based on hardness assurance levels. The obtained results provide the grounds for setting test requirements that has to be met during single event effects testing of ICs.

The paper is devoted to the recent scientific problem of privacy-preserving machine learning. The problem actuality is determined by the growing need to use machine learning for personal data, as well as for data that make up commercial, medical, financial and other types of information protected by law. The aim of the study is to systematize the security models of machine learning, to identify algorithmic tools that can be used to ensure the privacy of the learning process and application of models, as well as to analyze the privacy-preserving machine learning systems. The paper presents the major concepts and definitions related to machine learning, provides a systematization of machine learning problems and methods of their solution, and pays attention to the modern and promising areas of development of machine learning. Among the tasks of machine learning are those for which it is important to ensure the privacy of data from training, test and work samples. Special cryptographic methods and protocols are correlated to the problems solved. A brief description of the known privacy-preserving machine learning systems is given. The machine learning methods supported by these systems, as well as the type of adversary that the system can resist, and the cryptographic primitives used for the implementation are described in the paper. Unsolved problems in the field of privacy-preserving machine learning and prospects for the development of this scientific field are considered.

In this paper, a person tracking algorithm for indoor video surveillance is presented. The algorithm contains the following steps: person detection, person features formation, features similarity calculation for the detected objects, postprocessing, person indexing, and person visibility determination in the current frame. Convolutional Neural Network (CNN) YOLO v3 is used for person detection. Person features are formed based on H channel in HSV color space histograms and a modified CNN ResNet. The proposed architecture includes 29 convolutional and one fully connected layer. As the output, it forms a 128-feature vector for every input image. CNN model was trained to perform feature extraction. Experiments were conducted using MOT methodology on stable camera videos in indoor environment. Main characteristics of the presented algorithm are calculated and discussed, confirming its effectiveness in comparison with the current approaches for person tracking in an indoor environment. Our algorithm performs real time processing for object detection and tracking using CUDA technology and a graphics card NVIDIA GTX 1060.