Popov A. Recognition of small radar objects in the conditions of a priory uncertainty with the use of invariant polarization signatures

Object of the research - the process of receiving informative signatures of radar objects for their recognition by polarization characteristics; Aim of the research - enhancing quality of real-time recognition of small radar objects with given electro-physical properties against the background interference from an underlying surface; Methods of research - the methods of polarimetric signals theory applied to radar systems; the methods of probability theory and mathematical statistics for description of objects classes; the methods of statistical modeling for simulations; the methods of information theory for revealing informative signatures of recognition objects; the methods of statistical decision-making theory for radar objects classifier construction; results - the relevant applied scientific task of real-time recognition of small radar objects in the conditions of a priory uncertainty as to their characteristics by polarization parameters of backscattered signals is solved; novelty - for the first time a mathematical operator model of a radar object that gives separate description of transformations of amplitude, phase and polarization of the backscattered signal independent of sensing signal polarization parameters is suggested, which provided the possibility of revealing new informative polarization signatures for radar objects recognition; for the first time invariant polarization signatures of a radar object that depend only on geometric and electro-physical characteristics of the object and are mathematically independent are suggested, which allowed increasing probability of recognition of small objects against an underlying surface; for the first time a method for estimating parameters of the mathematical operator model that allows receiving estimates of polarization transfer functions from the results of the radar object polarization matrix measuring is developed, which allowed estimating distributions of new polarization signatures of real small radar objects and underlying surfaces experimentally; for the first time on the basis of multi-dimensional statistical models received in field experiments a comparative analysis of informativity of the polarization signatures of small radar objects was carried out, which allowed receiving quite informative vector polarization signatures for recognizing small objects with given shapes or of given materials against common underlying surfaces; the theory of multi-dimensional non-Gaussian random processes is further developed, namely on the basis of the known Johnson distributions a method for estimation parameters of the statistical model of a non-Gaussian multi-dimensional stationary process with account of correlation of its components and sample sequences by experimental data is developed, which allowed receiving numerical values of parameters of multi-dimensional statistical models of polarization signatures of real radar objects and underlying surfaces; the theory of signal parameter informativity is further developed, namely a method for estimating distinguishability of a set of object classes by the vector signature with non-Gaussian distribution that does not depend on the decision-making method and is applicable for analysis of informativity of signatures with non-Gaussian distributions even when one object class is fully masked by other classes, which allows revealing signatures that group objects into classes according to their electro-physical and geometrical properties and revealing combinations of signatures that ensure polarization contrast sufficient for recognition of small objects against a background of an underlying surface; the theory of decision-making in the conditions of a priory uncertainty is further developed, namely a method of polarization-probability filtering that does not require a priory information concerning underlying surface signatures distributions is developed, which ensures recognition of objects with given geometrical and electro-physical properties by polarization signatures against the background of the underlying surface in the conditions of a priori uncertainty as to its polarization signatures; the theory of automatic classification of remote sensing objects is further developed, namely a method for automatic classification of radar objects by their polarization signatures over the criterion of minimal sufficient quantity of information at a priory unknown number of object classes is developed; this method can be applied in real-time and allows indentifying objects by their polarization signatures; extent of adoption - O.Ya. Usikov Institute for Radiophysics and Electronics of National Academy of Sciences of Ukraine, Institute of Telecommunications and Global Information Space of National Academy of Sciences of Ukraine and State Enterprise Research and Design Institute 'Soyuz'; sector of implementation - telecommunications and radio engineering.