• PDF / 174,018 Bytes
• 4 Pages / 594 x 792 pts Page_size
• 52 Downloads / 232 Views

DOWNLOAD

REPORT

---

AN APPROACH TO CONSTRUCTING A MATHEMATICAL MODEL OF THE SIGNATURE TO DETERMINE OBJECTS OF MONITORING USING THE INFORMATIVITY OF ITS MONITORING FEATURES O. Iliashov,1† V. Êomarov,1‡ O. Yurchenko,2 and O. Deineha1††

UDC 621.391

Abstract. The authors consider the use of informativity of monitoring features of the signature of objects of monitoring for specific conditions of solving applied recognition problems. The approach to construction of the mathematical model of the signature of definition of objects of monitoring by varying the difference between the probabilities of recognition of the objects of monitoring is proposed. Keywords: monitoring feature, signature, informativity of the monitoring feature. INTRODUCTION Nowadays, of importance is developing the models that would allow the formation of monitoring features and signatures of the required source (object) of monitoring (SOM), as well as formalization of the problem of compliance of the selected SOM to one of the given images of the sampling of values of its information signature followed by the synthesis of decision-making algorithms. Formalization of signatures and monitoring features is complicated by their heterogeneous structure and the presence of intersecting image classes; therefore, the problems of formalizing a reference description of the SOM and their recognition by the relevant monitoring features (MF) are not solved in full [1–3]. PROBLEM STATEMENT The heterogeneous, fuzzy nature of the MF and the signatures based on them, as well as the lack of training sample, significantly complicate the development of a reference description (signature) using probabilistic-statistical models [3]. Application of the information approach to description of SOM signatures makes it possible to obtain MFs with certain errors (losses), which increases entropy, causes loss of feature informativity, and reduces the probability of correct SOM recognition. However, in some cases, the set of non-informative MF will provide the required level of SOM classification [4]. The proposed approach as to using only priority MFs [3] will allow, to some extent and in some cases, eliminating the above difficulties associated with the process of generating the signature with the necessary relative informativity of the MFs that compose it. 1 Central Scientific and Research Institute of the Armed Forces of Ukraine, Kyiv, Ukraine, [email protected]; ‡[email protected]; ††[email protected]. 2Igor Sikorsky National Technical University of Ukraine “Kyiv Polytechnic Institute,” [email protected]. Translated from Kibernetika i Sistemnyi Analiz, No. 6, November–December, 2020, pp. 115–119. Original article submitted May 26, 2020.

†

1060-0396/20/5606-0959 ©2020 Springer Science+Business Media, LLC

959

Since generally there is no functional dependence between the recognition error and the amount of information for making a decision about recognition of SOM [4, 5], there is a need to construct an approach to solving applied problems in specific conditions that allow r