This article presents the results of research into methods and algorithms for processing image sequences aimed at application in autonomous navigation systems. The methods are grouped in the following areas: pre-processing, generation of 2D and 3D scene models, object recognition. and determination of motion parameters from a sequence of images. The article presents methods and algorithms proposed and researched by the authors of the article over the past decade. Description of the methods and algorithms is accompanied by examples and results of experimental studies.

The article discusses the possibilities of the formal analysis of the fine-art painting composition on the basis of the classical definitions of beauty and computational aesthetics’ approaches of the second half of the 20th century he authors define the problem and consider solutions for the formalization of aesthetic perception in the context of aesthetic text, i.e., as part of the fine arts composition – a formal sequence of signs simply ordered in accordance with the syntactic rules’ system. The methodology of the research is defined by the general semiotics, distinguishing semantics, syntax, and pragmatics of a sign, by the aesthetic analysis’ methods, ranging according to the author’s message aesthetics, receptive aesthetics, and text aesthetics, as well as by the computational analysis methods connected with neural network means of defining the images’ symmetry. The article reveals preconditions for the emergence and also the current state of computational aesthetics as an interdisciplinary branch of knowledge. Analyzing the problem from the perspective of philosophy, aesthetics, semiotics, and technology, the authors draw attention to the need to improve the computational aesthetics methods. Firstly, the existing methods do not always enable to describe the fine-art object adequately. Secondly, there exists the so-called reduction of aesthetic assertion transforming it into the assertion concerning the object’s external characteristics. As a result, the authors assume that the increasing complexity of the current mathematical models and the experts’ subjective assessment support will allow to reach a compromise solution that enables the development of computational aesthetics as a branch of knowledge. Enhancement and development of the mathematical models, taking into account the rules of semiotics and subjectivism of the human perception, is the relevant objective of computational analysis of the aesthetic fine-arts text. The results of the present research supports the classic statement regarding the underivability of semantic and pragmatic propositions from syntax. The research concludes that relevant objectives are to find a correlation between, one the one hand, the axes and points of symmetry, deriving from the neural simulation, and, on the other hand, aesthetic effect, emerging from the perception of fine-art paintings.

Human poses and the behaviour estimation for different activities in (virtual reality/augmented reality) VR/AR could have numerous beneficial applications. Human fall monitoring is especially important for elderly people and for non-typical activities with VR/AR applications. There are a lot of different approaches to improving the fidelity of fall monitoring systems through the use of novel sensors and deep learning architectures; however, there is still a lack of detail and diverse datasets for training deep learning fall detectors using monocular images. The issues with synthetic data generation based on digital human simulation were implemented and examined using the Unreal Engine. The proposed pipeline provides automatic “playback” of various scenarios for digital human behaviour simulation, and the result of a proposed modular pipeline for synthetic data generation of digital human interaction with the 3D environments is demonstrated in this paper. We used the generated synthetic data to train the Mask R-CNN-based segmentation of the falling person interaction area. It is shown that, by training the model with simulation data, it is possible to recognize a falling person with an accuracy of 97.6% and classify the type of person’s interaction impact. The proposed approach also allows for covering a variety of scenarios that can have a positive effect at a deep learning training stage in other human action estimation tasks in an VR/AR environment.

The truly relevant dataset creation tasks are aimed at assessing human action. The approach to detect and recognize a person falling allows to promptly warn about dangerous incidents for further analysis of the consequences that can help in the tasks of healthcare industry. The problem can be solved using modeled data based on digital human simulation. The result of a proposed modular pipeline for synthetic data generation of digital human interaction with the 3D environment was demonstrated in this paper. The research includes the following contributions: the synthetic dataset based on procedural generation of realistic movements and fall which taking into account physics model of a digital human; registering basic rgb and segmentation rendering maps while simulating a digital human fall; in segmentation maps, we present hitting coordinate masks with the interaction of the human model and 3D scene. The pipeline is implemented using Unreal Engine that provides automatic “playback” of various scenarios for simulation. We used the generated synthetic data to train the Mask R-CNN framework. It is shown that a fallen person can be recognized with an accuracy of 97.6% and the type of person’s impact can be classified, including hitting the head when falling with training the model on simulation data. The proposed method also allows covering a variety of scenarios that can have a positive effect at a CNN training stage in the tasks of data creation for human action estimation.

In this paper there has been presented the thesis concerning the improvement of the computational aesthetics methods necessary for carrying out the analysis of the fine-art paintings referring to the classical pictorial art. It has been indicated that the fine-art object does not necessarily involve clear description required for computational analysis. It has also been described how the image expresses aesthetics via the fine-art object’s external characteristics. This paper gives an overview of the algorithm performance results for the purpose of defining the type of painting’s composition. Within the frame of the paper there have been proposed solutions to the problem of computational analysis as exemplified by the extension of computable attributes referring to the composition of fine-art painting.

The article presents studies of deep reinforcement learning method for the autonomous positioning problem of a small robot in a simulation environment. In our experiments, the open source game engine Unreal Engine is used to simulate a physically adequate 3D scene with obstacles. Images obtained by a virtual robot camera in the simulation environment are entered into a neural network to determine the required direction of the target and obstacles localization in 3D environment and then analyze the training of a real robot with reinforcement. In this study, we investigate the agent’s ability to learn free movement without interacting and colliding with other static or moving objects on the scene.

This article deals with the problem of quantitative research of the aesthetic content of the fine-art object. The paper states that a fine-art object is a conceptually formed sequence of signs, and its composition is a structural form, that can be measured using mathematical models. The main approach is based on the perception of the formal order as a determinant of the aesthetic category of beauty. The composition of the image is directly related to the formation of aesthetic sensations and values, since it performs the function of controlling the viewer’s perception of a work of art. The research is based on the studies of computational aesthetics by G. D. Birkhoff and M. Bense, as well as the studies of the receptive aesthetics of R. Ingarden, W. Iser, H. R. Jauss and Ya. Mukarzhovsky. The computational aesthetics methods, such as CNN-based object detectors, and gestalt-based symmetry analysis, are used to detect symmetry axes in fine-art images. Experimental analysis demonstrates that the applied computational approach is consistent with the philosophical analysis and the expert evaluations of the fine-art images, therefore it allows to obtain more detailed fine-art paintings description.

Modification and improvement of visual odometry algorithms are essential for the successful and stable functioning of autonomous systems and robots. Existing real datasets are not well scalable and cover a limited set of scenarios and motion models in comparison with real cases. The provision of a new large volume of annotated data that is solved by obtaining the synthetic data using a computer simulation is an urgent problem. Such synthetic datasets have the advantage of being better scalable. The paper presents a large-scale synthetic dataset of indoor and outdoor video sequences for ground autonomous systems and robot navigation tasks. The main characteristics of our dataset are a high degree of realism and variability, simulation of lighting changes, presence of moving objects in virtual scene, as well as providing different types of trajectories for the movement of a ground robot. As a result, the direct visual odometry algorithm was tested on the created synthetic dataset.

Currently, in virtual reality simulations and laboratories much attention is focused on the development of a user interaction controller for better immersion and student experience. Of course, the visual experience is a base for the laboratory study of any students, but direct contact with the visualized environment is also important. This paper presents a virtual scalpel technique that simulates the experience of medical discipline student education using virtual reality (VR) and augmented reality (AR) laboratory software. User's hands and a pencil with a tracker marker are the main tracking system components. Due to the high sensitivity of hand sign recognition, the Leap Motion camera was chosen as the base device for tracking the interaction of the user's virtual scalpel and hand models. We used Unreal Engine to create and visualize such virtual laboratory. During the recognition "hand-pencil" system, the 3D scalpel model is activated in our VR and scene, with a collision in the proposed virtual blade area. The user manipulates such “hand-scalpel” system in VR and AR simulation process, where the collision area of blade interacts with an imitating an organic 3D object. In this paper we presented the sensitivity and efficiency of “hand-pencil” system inside the virtual scene. In addition, the comparison and application of the methodology were highlighted for VR and AR prototypes of the laboratory scene.

In the study, a classification algorithm of plant crops in hyperspectral images is analysed. The algorithm uses the conjugation index with a subspace formed by samples of a given class. The purpose of the work is to show that this algorithm, with the data pre-processing (weighting of the feature vectors components and forming of the subclasses), provides a higher classification quality compared to the most popular reference vector method (SVM). The experiments were conducted with the implementation of the SVM method. The Indian Pines test of close types of vegetation, including 16 marked classes of plant crops, was used in the recognition experiments. The test was rather complicated, as class samples are highly correlated. The results show the possibility of a reliable recognition of plant crops.

The article aimed to provide a sort of new education process including virtual reality based application. At present, in accordance with the established ways of archaeological research, archaeologists are forced to transfer the found samples for long-term storage. In such notation, there is a challenging issue to create a virtual museum with deepening experience user interaction. The modern approaches of the virtual reality were implemented by applying technologies such as the Unreal Engine (UE) and Leap Motion (LM). In the paper, we give the scheme of the implemented development workflow. The ability of interaction with objects using the interface and hand gestures on LM on UE was given.

The article discusses the technology of autonomous navigation by optical observations. A scheme based on the recognition of three-dimensional landmarks is considered. To improve the accuracy and reliability of recognition, a survey model is constructed for each landmark. When calculating the parameters of this model, the most conformed data is selected. The problem of recognition is solved using the method of support subspaces. Support subspaces are formed from a set of vectors of each reference landmark. An important advantage of the technology is the ability to refine the landmark models based on the results of current observations during operation.

This article is devoted to the problems of radar sensing. Herein, we have considered the tasks of modeling and recognizing radar images. The modeling technology was based on the independent creation of terrain models and objects, which were then integrated into a three-dimensional (3D) scene. This approach enabled the operative creation of a number of image variants of different classes. Recognition methods and algorithms were based on the use of the so-called conjugacy index as a measure of proximity. At the same time, support subspaces of the minimum dimension were formed by vectors, components of which were samples of the radar image. Problems of higher accuracy of recognition due to a division of classes into subclasses and a combination of the support subspace method with the neural convolutional networks were considered.

In the work there is a modernization of the parallel algorithm for the radar images formation of 3D models with the synthesis of the antenna aperture. In the formation of the scene description, the various structures are used in which it is possible to use more efficient and derived calculations. In addition, it is the topical task to recognize objects on radar images. Thus, on the basis of the implemented parallel program for modelling, the high performance required for simulating multiple radar images can be achieved.

We researched studied the implementation of the parallel algorithm of synthetic aperture radar images modelling. We used CUDA for computing the trajectory signal of scattered field from object and ground on 3D scene. The important task of effective recognition algorithm studying is a formation of a large database of image samples. Thus, we reached the high performance using parallel implementation of modelling program, which necessary to model many synthetic aperture radar images of an object in different positions. We discussed results of objects recognition with real images using modelled on the training stage. We achieved the result 80.24% of correct recognition of two classes using such approach.

The paper proposes a method for processing personal data that allows them to be divided into many segments or classes. The customer database is used as the source data. We use the indicator of conjugacy that has already proved the effectiveness in both recognition and clustering of data problems.

In this work the problem of recognition is studied using SAR images. The algorithm of recognition is based on the computation of conjugation indices with vectors of class. The support subspaces for each class are constructed by exception of the most and the less correlated vectors in a class. In the study we examine the ability of a significant feature vector size reduce that leads to recognition time decrease. The images of targets form the feature vectors that are transformed using pre-trained convolutional neural network (CNN).

The goal of this work is to develop a technology that can reduce recognition computational complexity with the rise of recognition quality. We use an approach based on implementation of the conjugation indices of the vectors with the class feature spaces. We suggest a new criterion of class separability based on the conjugation index and use it to form so-called support subspaces from the training vectors. This procedure decreases computing complexity at training stage about 1000 times in comparison with previous algorithm implementation and improves recognition quality. The most significant decrease of the computational complexity of the proposed technology is achieved by implementing the fractal compression to radar images. The results prove that using this technology leads to an increase of the recognition quality.

In this study, we examine our recognition approach using modelled SAR images as a test set. The supposed approach is compared with the technology of training using real data from the MSTAR dataset. We conducted the recognition experiments using three different objects. The SAR images were modelled with approximate parameters, such as those used in SAR constructions of real targets. As a recognition method, one proposed in our previous articles, the support subspaces method, is used. The experiment results show the ability to achieve high recognition quality with construction of the support subspaces (training) using modelled SAR images.

This article offers a new object recognition approach that gives high quality using synthetic aperture radar images. The approach includes image preprocessing, clustering and recognition stages. At the image preprocessing stage, we compute the mass centre of object images for better image matching. A conjugation index of a recognition vector is used as a distance function at clustering and recognition stages. We suggest a construction of the so-called support subspaces, which provide high recognition quality with a significant dimension reduction. The results of the experiments demonstrate that the proposed method provides higher recognition quality (97.8%) than such methods as support vector machine (95.9%), deep learning based on multilayer auto-encoder (96.6%) and adaptive boosting (96.1%). The proposed method is stable for objects processed from different angles.