Pridruživanje radarskih podataka manevrirajućem cilju uz smetnje uzrokovane odjekom od mora

doctoral thesis

• Title (english): Manoeuvring target radar data association in presence of sea clutter
• Author: Bojan Vondra; MBZ: -331061
• Mentor: Davor Bonefačić (mentor)
• Committee member: Davor Bonefačić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; (Department of Communication and Space Technologies); Zagreb
• Defense date and country: 2021, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing
• Universal decimal classification (UDC): 004 - Computer science and technology. Computing. Data processing

Abstract

Različitost Jadranskog mora u odnosu na otvoreno more, ocean, najviše se reflektira u specifi-čnoj gustoći valne energije, specifičnom odnosu brzine vjetra prema efektivnoj visini valova, malom privjetrištu, velikom broju otoka. U dostupnoj literaturi ne postoje zapisi o elektromag-netskom raspršenju od morske površine za specifične uvjete Jadrana, pa nije poznata razina točnosti klasičnih parametarskih modela, razvijenih na temelju mjerenja na otvorenom moru, oceanu. Alternativno se statistika odjeka smetnje (i cilja) može ekstrahirati iz estimirane razdi-obe, isključivo na temelju uzoraka (otisaka) u okolici praćenog cilja. Tretiranje mjerenja otisaka kao bešumnih, otvara prostor estimaciji razdiobe primjenom neuronskih mreža, pri čemu se iskorištava svojstvo njihove univerzalne aproksimativnosti. U estimaciji amplitudne razdiobe, radijalne mreže pokazuju najpovoljniji omjer točnosti i računalnog opterećenja, važnog za stvarnovremensku aplikaciju. U scenariju praćenja male gumene brodice maskirane valovima, tradicionalni parametarski Swerlingov model ne opisuje dovoljno točno odjek cilja kontamini-ran odjecima valova. Djelovanje neuronske mreže kao estimatora razdiobe u scenariju smetnje izrazito teškog repa, rezultira vjerojatnošću zadržavanja staze od 0,3, u odnosu na vjerojatnost 0,02 ostvarene primjenom parametarskog Swerlingovog modela. Minimalno 64 kvantizacijskih razina potrebno je za prijenos amplitudne informacije iz udaljenog radarskog senzora bez zna-čajnih gubitaka.

Abstract (english)

In comparison with open seas, properties of Adriatic Sea are distinct wave spectrum, specific ratio of wind velocity to effective wave height, limited fetch, large number of islands. Data of electromagnetic scattering from sea surface in Adriatic basin does not exist in open literature, so accuracy of employing traditional parametric models developed on the basis of measure-ments under condition of open and fully developed seas, is not known. Alternatively, statis-tics of clutter (and target) returns can be extracted from estimated probability density, exclu-sively on the basis of radar return samples (signatures), taken in the vicinity of tracked target. Treating measurements as noiseless opens up possibility of their probability density estimation by means of a neural network, employing property of universal approximation. Radial basis function network exhibits favorable ratio of estimation accuracy to computing time which is important for real-time application. In scenario of tracking a small rubber boat masked with waves, traditional Swerling model does not describe target echo, contaminated with echoes from waves, well. Neural network as density estimator employed in scenario of heavy-tailed clutter results in probability of tracking true target of 0.3, in comparison with 0.01 which re-sults from employment of parametric Swerling model. Minimum of 64 quantization levels are needed for amplitude information transfer from remote radar sensor without significant losses.

• Keywords: Jadransko more; smetnja mora; Viterbijev algoritam; primjetljivost; više-modelna estimacija; neuronska mreža; IPIX; K-razdioba; Paretova razdioba; nekoherentni ra-darski senzor
• Keywords (english): Adriatic Sea; sea clutter; Viterbi algorithm; perceivability; multimodel estima-tion; neural network; IPIX; K-distribution; Pareto distribution; non-coherent radar sensor
• Language: croatian
• URN:NBN: urn:nbn:hr:168:822654
• Study programme: Title: Computer Science; Study programme type: university; Study level: postgraduate; Academic / professional title: Doktor znanosti (Doktor znanosti)
• Catalog URL: http://lib.fer.hr/cgi-bin/koha/opac-detail.pl?biblionumber=52326
• Type of resource: Text
• Extent: iii, 455 str.
• File origin: Born digital
• Access conditions: Open access
• Created on: 2022-03-09 14:08:05