Informative Feature Analysis in Data Processing for Aerospace Monitoring

Khimenko, V. I.; Okhtilev, Michael Yu.; Klucharev, A. A.; Матяш, В. А.

doi:10.15217/issn1684-8853.2017.2.2

Cited by 3 publications

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Если на N-м шаге в кодовой строке установлен один элемент, то его положение указывает номер выделенного класса. При упорядоченном рабочем словаре X р = {X 1 , …, X j , …, X Tр I(X 1 ) ≥ I(X 2 ) ≥ … … ≥ I(X j ) ≥ … ≥ I(X Tр )}, где T р = X р ; I(X j ) -дискриминирующая сила X j [22][23], решение находится без перебора значе- Визуализатор распознавания демонстрирует на каждом шаге инициированные сегменты диаграммы и соответствующее сужение в подмножестве классифицирующих признаков пространства поиска «подмножество классов» (рис. 4):…”

Section: когнитивная диаграмма-визуализатор классовunclassified

The cognitive visualization of classifying rules extracted from data based on binary solver matrix model

Pimenov¹,

Voronov²,

Pimenov³

2020

ICS

View full text Add to dashboard Cite

Introduction: Huge volumes of data are generated in cyberspace or from internal information of various organizations. In order to obtain a set of output data with a clear structure, divide it into significant parts and develop rules of classification, machine learning methods are used. Most inductive methods simulate intermediate and high-level abstract categories in multidimensional space which are difficult to interpret. Purpose: Developing a model of machine learning in the form of a “white box” which explains the chosen solution using conventional production rules, along with cognitive visualizers for characterizing classes of objects. Methods: Formation of a binary decision matrix containing information about a combination of the selected informative sign values which imply the specified classes. Results: A binary decision matrix is formed automatically according to the results of cluster and discriminant analyzes. The learning procedure is reduced to setting interval thresholds and matrix elements, which makes it easy to implement a semantic interpretation of a solving rule. The object is recognized by elementwise conjunction of the matrix cells to which the values of the attributes are pointing, and by selection of a single cell corresponding to the class code. To interpret a rule, a universal algorithm for processing a binary matrix has been developed, which applies user-entered attribute values. The dimension of the viewed space is specified by adjustment rings on the recognition visualizer. The azimuth of an initiated diagram cell with the greatest dimensionality indicates the belonging of an object with the set features to a target class. For the characterization of classes, visualizers have been developed, demonstrating both the distinctive properties of a class and properties that several classes share. In many cases, the object type recognition stops when the depth of the scanned features space is significantly less than with a full search. Practical relevance: The proposed methods of cognitive analysis and data visualization provide not only the classification of data, determination of the significance of features, their ranking and selection, but also the development of rules which reveal the cause-and-effect relationship between the combination of factors and the type of a made decision.

show abstract

Section: когнитивная диаграмма-визуализатор классовunclassified

The cognitive visualization of classifying rules extracted from data based on binary solver matrix model

Pimenov¹,

Voronov²,

Pimenov³

2020

ICS

View full text Add to dashboard Cite

show abstract

“…Задача моделирования работы гиперспектрометра с получением модельного гиперспектрального изображения актуальна на различных стадиях создания прибора. Например, на предпроектной и проектной стадиях (при формировании требований к аппаратуре) данное моделирование позволяет: оценить качество изображения, линейную и спектральную разрешающую способность, качество спектральных характеристик; оценить возможности аппаратуры по разделению спектральных пар «объект-фон» [1]; проанализировать и выбрать информативные спектральные каналы и признаки для решения задач обнаружения и распознавания объектов на гиперспектральных изображениях [2,3]; исследовать, сформировать и оптимизировать алгоритмы обработки гиперспектральных изображений [4], и т.д.…”

unclassified

Modeling of image formation with a space-borne Offner hyperspectrometer

2020

View full text Add to dashboard Cite

In this paper, we developed a mathematical model of image formation that allows a predictive hyperspectral image to be generated. The model takes into account the formation of an optical image using a matrix photodetector. The paper presents a numerical modeling of hyperspectral image formation and gives estimates of spatial and spectral resolution, as well as analyzing the adequacy of the results.

show abstract

Model of process of processing rapid telemetrated parameters in real tim

Птицын

Zaitsev

Pavlov

et al. 2020

Voprosy radioèlektroniki

View full text Add to dashboard Cite

The paper considers the generalized problem of processing telemetry data. The solution of this problem in real time is due to the requirement of the operational stage of processing telemetry information in the form of a report on the flight of rocket and space technology. At the operational stage, 10% of the total number of telemetry parameters is processed. The consequence is that the results obtained are insufficient for an operational and reliable analysis of the technical condition of on-Board systems of rocket and space technology. To eliminate the insufficiency, it is necessary to increase the completeness of the results of processing telemetry information. A conceptual formulation of the problem of processing rapidly changing telemetry parameters is formulated, taking into account the requirement to include rapidly changing parameters in the processing of telemetry information at the operational stage. A model of the processing rapidly changing parameters in real time is constructed, which is based on the method of discrete linear mathematical programming. Restrictions on processing time, processing nodes, and completeness of the processing result are defined. Taking into account the constraints, the Pareto optimal set of acceptable solutions is determined. A narrowing of the set of acceptable solutions based on the use of nonlinear partial quality indicators is described. Conclusions are made about the expected results of the decision, as well as about the course of further research.

show abstract

Informative Feature Analysis in Data Processing for Aerospace Monitoring

Cited by 3 publications

References 0 publications

The cognitive visualization of classifying rules extracted from data based on binary solver matrix model

The cognitive visualization of classifying rules extracted from data based on binary solver matrix model

Modeling of image formation with a space-borne Offner hyperspectrometer

Model of process of processing rapid telemetrated parameters in real tim

Contact Info

Product

Resources

About