We discuss the problem of estimating the state of a dynamic object by using observed images generated by an optical system. The work aims to implement a novel approach that would ensure improved accuracy of dynamic object tracking using a sequence of images. We utilize a vector model that describes the object image as a limited number of vertexes (reference points). Upon imaging, the object of interest is assumed to be retained at the center of each frame, so that the motion parameters can be considered as projections onto the axes of a coordinate system matched with the camera's optical axis. The novelty of the approach is that the observed parameters (the distance along the optical axis and angular attitude) of the object are calculated using the coordinates of specified points in the object images. For estimating the object condition, a Kalman-Bucy filter is constructed on the assumption that the dynamic object motion is described by a set of equations for the translational motion of the center of mass along the optical axis and variations in the angular attitude relative to the image plane. The efficiency of the proposed method is illustrated by an example of estimating the object's angular attitude.

We are exploring technology for recognizing radar images of objects observed from different viewing angles. To do this, we use many reference images corresponding to different viewing angles. As a measure of proximity, we use a conjugacy criterion, defined as the squared cosine of the angle between the current image vector and a space formed by the vectors of reference images. Based on the properties of the orthogonal decomposition of the conjugacy criterion, we also construct a computationally simpler approximate proximity criterion. Using the conjugacy criteria, we build a recognition technology with the formation of subsets of reference images corresponding to the most informative angles of the object in the sense of conjugacy criteria to the angle of the current (recognized) object. Results of comparative experiments using the open database of MSTAR radar images are presented, illustrating the quality of recognition with various modifications of the criterion and various sizes of the reference images subsets. Sizes of the reference image subsets are established for which the quality of recognition using the original conjugacy criterion exceeds the recognition quality provided by the familiar methods, with a simpler modified criterion also providing good results.

Рассматривается задача оценивания состояния динамического объекта по наблюдаемым изображениям, сформированным оптической системой. Цель исследования состоит в реализации нового подхода, обеспечивающего повышение точности автономного слежения за динамическим объектом по последовательности изображений. Используется векторная модель изображения объекта в виде ограниченного количества вершин (базовых точек). Предполагается, что в процессе регистрации объект удерживается в центральной области каждого кадра, поэтому параметры движения могут описываться в виде проекций на оси системы координат, связанной с оптической осью камеры. Новизна подхода состоит в том, что наблюдаемые параметры (расстояние вдоль оптической оси и угловое положение) объекта вычисляются по координатам заданных точек на изображениях объекта. Для оценки состояний объекта строится фильтр Калмана-Бьюси в предположении, что движение динамического объекта описывается совокупностью уравнений поступательного движения центра масс вдоль оптической оси и изменений углового положения относительно плоскости изображения. Приведен пример оценивания углового положения объекта, иллюстрирующий работоспособность предложенного метода.

The article proposes a method for hyperspectral image recognition, in whichthe conjugacy with subspaces formed by training class vectors is used as a measure of proximity. The geometric interpretation of the proximity measure used is given in a space formed by eigenvectors. An algorithm for hyperspectral image recognition is built with sequential selection of the most informative subspaces according to the criterion of maximum conjugacy. The results of vegetation recognition experiments on the test hyperspectral image «Indian Pines» that had been obtained within the project AVIRIS (Airborne Visible/Infrared Imaging Spectrometer) are presented. The experiment showed the possibility of achieving higher recognition quality in comparison with known methods.

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.

In this paper, we present a hybrid refractive-diffractive lens that, when paired with a deep neural network-based image reconstruction, produces high-quality, real-world images with minimal artifacts, reaching a PSNR of 28 dB on the test set. Our diffractive element compensates for the off-axis aberrations of a single refractive element and has reduced chromatic aberrations across the visible light spectrum. We also describe our training set augmentation and novel quality criteria called "false edge level" (FEL), which validates that the neural network produces visually appealing images without artifacts under a wide range of ISO and exposure settings. Our quality criteria (FEL) enabled us to include real scene images without a corresponding ground truth in the training process.

We offer a computer technology for modeling a process of optical imaging with a diffractive imaging lens. The central idea of the technology is to evaluate the quality of the optical system by matching the input and output images against criteria adopted in image processing. For this purpose, same-resolution hyperspectral images are fed to the input and generated at the output. Thanks to the large number of spectral components, a fairly accurate reproduction of the effects associated with the dependence of the refractive index on the wavelength is ensured. To compare input and output images in terms of PSNR (peak signal-to-noise ratio), standard three-component RGB images are "assembled" using standard matching functions over the entire optical range. Results of the study of the dependence of the PSNR indicator on the main parameters of the optical system are given: focal length, linear aperture and the number of diffraction orders taken into account.

The paper considers the problem of visual odometry based on a sequence of video framesformed using a camera perpendicularly downward facing the reference surface. The problem issolved under the assumption that the shooting frequency is high, so that the interframe rotationand shift parameters are small. The technology is implemented in the form of a sequence of thefollowing steps: determining the shift and rotation with an accuracy of an integer number of pixelsusing the correlation method, clarifying the shift and rotation parameters using the optical flowmethod, and correcting estimation errors associated with uneven motion and fluctuations in thedistance of the camera to the reference surface by estimating deviations of local calibrationcharacteristics from their mean values. The results of experimental studies of the technology ontest trajectories obtained by simulating the motion of a vehicle along the reference surface arepresented.

The article discusses the technology of constructing a recursive filter on a non-uniform grid of samples with the parameters identification on test images. This filter is the IIR-filter that has a physical feasibility problem. To over-come it, a multi-step procedure is implemented. Unfortunately, identifying the best filter in terms of a given criterion does not guarantee that a recursive implementation of that filter will be stable. In the paper, for the considered iterative scheme, stability conditions are obtained. It has been experimentally confirmed that if these conditions are met, it is possible to achieve a high quality of correction. Based on the obtained criteria, a technology for correcting defocusing with control over the stability of estimates is proposed. The results of image correction showing the effectiveness of the technology are presented.

We offer computer technology for simulating the imaging process using diffraction optical harmonic lenses. The technology is built in a paraxial approximation using the laws of geometric optics. To simulate the refractive index versus wavelength, we use a multispectral image with a uniform wavelength distribution in the optical range. To assess the quality of the generated images, the components of the input and output multispectral images are integrated using standard colorimetric observer functions, specified on a discrete set of wavelengths. An illustrative example of computer simulating is given for the case when a color image at the input is presented as a plane perpendicular to the lens optical axis.

The article discusses the technology of images recognition, which uses as a measure of proximity of the conjugation criterion with the so-called reference subspaces formed by training vectors of classes. With use of orthogonal decomposition a geometric interpretation of the proximity measure under consideration is given. Based on the discovered properties, an images recognition algorithm is proposed with sequential selection of the most informative reference subspaces. The results of experiments on vegetation recognition on a hyperspectral image are presented. The experiment used images that had been obtained within of the project AVIRIS (Airborne Visible/Infrared Imaging Speсtrometer). The image shows the Indian Pines test field located in the north-west of Indiana, United States.

The image processing technology using IIR filters in order to eliminate defocusing distortions is proposed. Using the property of central symmetry of distortions, the two-dimensional filter is reduced to a one-dimensional filter with nonuniform sampling on a system of circles with different radii. The technique for constructing transfer functions and difference equations of an IIR filter on a nonuniform sampling system is considered. Sufficient stability conditions for an iterative scheme for implementing an IIR filter are obtained. An example of stability analysis and synthesis of an IIR filter with a given degree of stability and the results of restoring a test image are given.

We consider the problem of visual odometry from a sequence of video frames, which are formed using a camera directed perpendicularly downward. We propose an adaptive visual odometry technology based on sequential determination of interframe shifts and episodic correction of current coordinate estimates. The frame shifts between two consecutive frames are determined by the correlation method with pixel precision. Then, those shifts are refined with subpixel precision using optical flow method. To improve reliability, the selection of the most consistent optical flow estimates is carried out. We present the results of experiments using publicly available test data.

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.

We consider the problem of visual odometry for a sequence of video frames using a cameradirected perpendicularly downward. We propose an adaptive two-stage visual odometrytechnology based on sequential determination of interframe shifts and regular correction of currentcoordinate estimates. At the first stage, the shift between two consecutive frames is determined bythe correlation method, with the compared video frames being aligned using the found shiftparameters up to a pixel. At the second stage, the shifts are refined with subpixel precision usingthe optical flow method. To improve reliability, the most consistent estimates of the optical floware selected. We present the results of experimental studies on publicly available survey data,which confirm the high reliability and accuracy of the estimates.

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.

In this study, we developed a two-stage technology for improving the sharpness of images. In the first stage, the correction was performed using a linear square exponential (SE) filter with a centrally symmetric frequency response in the form of quadratic and exponential functions. This stage included setting the parameters of the SE filter and the actual processing. In the second stage, non-linear correction was carried out. The idea of the filter was to increase the impact of the central value, if it was at the edge of different intensity levels. We assumed that an increase in the absolute value of the weighted average of the differences in the point neighbourhood could be an indicator of such edges. The central point of the reference area belonged to the edge if its value was considerably greater or lesser than the significant number of values in this area. The first experiment confirmed the possibility for the improvement of the quantitative criteria of image restoration by non-linear correction. The second experiment illustrated the increase in the image sharpness obtained using a diffraction Fresnel lens. The proposed technology has opened up prospects for the use of cameras based on diffraction optic elements in mobile devices.

The task of image motion analysis is over forty years old, but it has not lost its relevance. Optical flow competitions regarding the robust vision challenge are still commonly held. The main issues arise because images usually contain noise or differ in brightness, spectral range and orientation scale, etc. A scene image motion analysis is formulated via the task of constructing an optical flow created by a sequence of images. Optical flow is a velocity field with two components for each image point. It can be restored by matching a sequence of image frames. This paper considers approaches to solving the problem of determining parameters of optical flow based on the ideas of the quality criterion functionalization. For examining our method, we used the Middlebury optical flow benchmark.

The article discusses the technology of correcting blurred type distortions on images recorded by mobile devices. The aim of the development is to provide high-quality distortion correction with minimal computational costs. We use two-stage technology. At the first stage, blind identification of linear filter parameters is performed. It is assumed that the frequency response of the filter has central symmetry and consists of segments of quadratic and exponential functions. At the second stage, parameters of nonlinear filter are determined. Both automatic and manual visual settings are provided. The program code for Android OS and the results of experiments showing the effectiveness of the developed application to eliminate distortions in images are presented.

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.