Home – Home – Jornals – Avtometriya 2019 number 3

Among other problems, systems of space monitoring of the near-Earth space include detection of moving low-contrast objects in images with a powerful spatially unstable background significantly exceeding random (in most cases, weakly correlated) noise. The most effective method of increasing the signal-to-noise ratio under the conditions of different velocities of the objects and background is interframe processing of a sequence of images, which ensures suppression of the background component in the current frame by means of subtracting its estimate obtained from the previous frames. The problem is the presence of a priori unknown motion of the background, leading to significant errors in estimate formation in the regions of sudden changes in the background velocity. The algorithm of interframe processing is studied, which allows one to estimate moderate local motions of the background and to compensate for them down to fractions of the sampling step. Results of full-scale modeling are presented, which demonstrate the possibility of background component suppression down to the noise level even in regions with drastic changes in the background.

Spatial-local algorithms for signal and image filtering are widely used in practice because they have one or several parameters whose values significantly determine a filtering error. The choice of these parameters from the filtering error minimum condition is a known problem that has not been acceptably solved yet. A statistical algorithm is proposed that makes it possible to quite accurately estimate the optimal parameters of two spatial-local filters. An efficient algorithm for estimating the measurement noise variance is presented

A nonparametric algorithm of automatic classification of large arrays of statistical data is considered. Its synthesis is based on decomposition of initial data. The results of decomposition form a set of centers of multidimensional intervals and the corresponding frequencies of occurrence of values of random variables. Based on information obtained, classes corresponding to single-mode fragments of the probability density of features of examined objects are detected. The spatial interpretation of automatic classification results is analyzed. The nonparametric algorithms developed in the study are important tools of processing of data obtained by remote sensing of natural resources.

The comparative estimation of the probability distribution density parameters determined using the maximum likelihood and moment methods is carried out according to their accuracy and complexity of estimation algorithms. Expressions determining the Nakagami distribution parameters by the maximum likelihood method are obtained. A method for estimating the Nakagami distribution parameters by the moment method in which the distribution moments are replaced by their estimates is described. It is noted that parameter estimation by the maximum likelihood method has a smaller variance and displacement as compared with estimation by the moment method, especially with a small number of samples. It is shown that, unlike energy parameter estimation, the real laws of Nakagami distribution is approximated using a large volume of statistical data that describe the signal.

The results of temperature signal analysis are given. A mathematical model of a temperature signal in a form of an amplitude and phase modulated oscillation is presented. The amplitude and phase are calculated on the basis of the theory of analytical signals. The amplitude and phase of the temperature signals are applied as criteria for detecting climactic clusters. A climactic clustering algorithm is described

Errors of laser interferometers using quadrature detectors depend not only on the interferometers themselves, but also on the electronic systems for processing quadrature signals. There are three sources of errors of quadrature detector signals that affect the displacement measurement error of: the inequality of the amplitudes of quadrature signals, the phase shift of these signals (not equal to 90^◦), and the zero offset in each signal. A simple method for identifying and correcting these errors based on experimental data on signals obtained using analog-to-digital converters is considered. The proposed method is applicable to any interferometer with sinusoidal quadrature outputs

A problem of age recognition by a human's face is developed with the popularization of convolutional neural networks. They make it possible to determine the specific features of faces, unseen by a human eye, and interpret them as age characteristics. Existing approaches to age recognition are analyzed. Data from existing sets for learning with subsequent correction for reducing the errors made in markings by acquisition algorithms are used. Neural networks are taught and tested using the resulting data. There is a problem with head rotation, whose solution is carried out using the images of faces rotated using the PRNet neural network

It is experimentally demonstrated that the classification of fragments of a hyperspectral images with preliminary transformation of the spectral features of the image into the principal components and with the use of the Hilbert - Huang spectral transformation is fairly effective in the case of vegetation types that are difficult-to-distinguish on the basis of hyperspectra. This classification is compared with traditional methods, where hyperspectral features transformed to the principal components without using spatial information are used. Neural networks RBF are used in all methods at the final stage of the classification.

Theoretical and experimental estimates of the energy efficiency of holographic gratings as passive solar trackers were obtained. It is found that in central Russia, their use can provide an about 20 % increase in the signal of photovoltaic panels mounted in smart windows. Increase in the signal to a 35 % level is possible only by improving the diffraction properties of gratings, i.e., by expanding the angular selectivity contour while maintaining high diffraction efficiency. This can be achieved either by developing new materials with a refractive index modulation of more than 0.1 with a thickness of the recording layer of about 3 мкм, or using well-known hybrid surface-relief volume diffractive structures with refractive index modulation.

An approximate theory of the formation of images of one-dimensional binary objects in a partially coherent projection diffraction-limited system with an arbitrary ratio of the angular sizes of the light source and aperture diaphragm is developed. The solution of the two-dimensional problem for such objects can be reduced to a one-dimensional problem, i.e., to calculating the integral in the form of convolution of the object with a generalized aperture response whose spectrum has a tapered profile determined by system parameters. The convolution integral is further subjected to a nonlinear transformation whose level is defined by the ratio of the angular sizes of the source and the aperture diaphragm. It is shown that the profile of images formed in partially coherent light is determined by the threshold characteristic of the system that ensures reliable detection of the binary object boundaries. An approximation of this characteristic of this characteristic is proposed. The theory admits a simple physical interpretation of phenomena in partially coherent projection systems.

Proposals on compensation for distortions of circular trajectories implemented in laser image generators with circular scanning for increasing the accuracy of synthesis of the topology of optical elements being formed are analyzed. The expected gain in the accuracy of the element formation topology is estimated with the use of a photoelectric method of detecting the beatings of the axis of rotation of the laser generator spindle rotor

The dependence of the piezo-optical strain sensor gauge factor on the geometric parameters of the photoelastic element is obtained by accurate numerical simulations. It is shown that the piezo-optical gauge sensitivity to the applied force is more dependent on the photoelastic element shape than the sensitivity to deformation. A comparative analysis of the gauge factors and other parameters of strain-resistive, piezoelectric, fiber optic, and piezo-optical strain gauges was carried out. Strain-resistive and fiber-optic sensor gauge factors are three orders of magnitude lower than those of piezo-optical sensors. Correctly calculated piezoelectric sensor gauge factors are also two-three orders of magnitude inferior to those of piezo-optical sensors. A comparison of the basic properties and parameters of modern industrial strain sensors based on different physical principles is presented.

A method has been developed and tested to control the fabrication accuracy of precision synthesized structures on circular laser writing systems using specialized microstructures formed in two orthogonal directions. These microstructures consist of two parts, one of which is formed before the fabrication of a precision structure, and the second during the fabrication of the structure. The shift between the first and second part of the microstructure allows one to determine the errors of drift, substrate shift, etc. on each of the coordinates. The method provides an increase in the accuracy and certification reliability in the fabrication of precision synthesized structures.

The effect of linear acceleration on the characteristics of a optoelectronic vibration ring angular velocity transducer is analyzed taking into account the rigidity of the supports of the ring resonator. A method for compensating the effect of linear acceleration on the characteristics of the angular velocity transducer is proposed.