Application of Kernel Method to Air Radar Target Identification and Underwater Communication Positioning

  • 詹 勝智

Student thesis: Doctoral Thesis

Abstract

This dissertation focuses on two points: air radar target identification and underwater communication positioning Kernel method (i e kernel principal component analysis KPCA and kernel scatter-difference-based discriminant analysis KSDA) and maximum scatter difference (MSD) are used as the basis for determining success and accuracy rates In the research on air radar target identification simple ship models (fishing boat naval ship and container vessel) are chosen as the targets simulations were conducted on undulating sea level and commercial electromagnetic software Ansys HFSS is utilized to calculate angular-diversity and frequency-diversity radar cross section (RCS) as basis of identification RCS of electromagnetic wave is related to not only shape dimension structure and material of target but also frequency incidence angle and polarization mode of incident electromagnetic wave These data are usually enormous irregular and difficult to store analyze and recognize Therefore for the sake of simplifying RCS scattering data and thus benefiting radar target identification this dissertation applies KPCA KSDA and MSD to air radar target identification Principle of KPCA and KSDA is to project original data into high-dimensional space and then perform linear calculation (i e principal component analysis PCA and MSD) MSD an improved version of Fisher linear discriminant analysis (FLDA) reduces complexity of calculation and accelerates computing process Finally Euclidean distance is used to calculate type of the ship From numerical simulation results we know these three kinds of calculations can achieve a high air radar target identification rate To be closer to actual measurements we actively add random variables of Gaussian distribution as noises Results of numerical simulation show noise-caused disturbances are effectively decreased and verify feasibility of applying KPCA KSDA and MSD to theme of this study In the research on underwater communication positioning the concept of location fingerprinting is introduced so as to avoid influences of reflected signals on measuring results To demonstrate such idea is not affected by reflected signals or multipath communication signals experiments are done in bordered towing tank We propose a method of using frequency component to simulate sound projector of underwater communication to lower hardware cost in underwater environment The experiment consists of training and test state The collected underwater communication signals are subject to parametric training through KPCA Under test state maximum likelihood (ML) is used to estimate coordinate of reception location of underwater communication signal Finally Euclidean distance is used to calculate distance between actual location and estimated location as positioning error in underwater environment Experimental results show it’s possible to realize wireless underwater communication positioning by an application of probability model identification into feature space of KPCA This dissertation is divided into nine chapters The first chapter is Introduction The second chapter is an explanation of algorithmic theories including KPCA KSDA and MSD The third fourth and fifth chapters apply KPCA KSDA and MSD to angular-diversity RCS The sixth and seventh chapters apply KPCA and KSDA to frequency-diversity RCS The eighth chapter combines KPCA and ML and apply them to underwater communication positioning based on probability comparison The ninth chapter is conclusion of this dissertation and recommended direction of future research
Date of Award2014 Feb 18
Original languageEnglish
SupervisorKun-Chou Lee (Supervisor)

Cite this

Application of Kernel Method to Air Radar Target Identification and Underwater Communication Positioning
勝智, 詹. (Author). 2014 Feb 18

Student thesis: Doctoral Thesis