Modeling recognizing behavior of radar high resolution range profile using multi-agent system

  • Authors:

  • Jiansheng Fu;Kuo Liao;Daiying Zhou;Wanlin Yang

  • Affiliations:

  • College of Electronic Engineering, University of Electronic Science and Technology of China, Chengdu, China;College of Electronic Engineering, University of Electronic Science and Technology of China, Chengdu, China;College of Electronic Engineering, University of Electronic Science and Technology of China, Chengdu, China;College of Electronic Engineering, University of Electronic Science and Technology of China, Chengdu, China

  • Venue:
  • WSEAS Transactions on Information Science and Applications
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

In an Automatic Target Recognition (ATR) system, target recognition-makers need assistance to determine which class a new High Resolution Range Profile (HRRP) belongs to. Note that the HRRP data can be obtained from an Open Database (ODB) freely, we present a new Multi-Agent System (MAS) model in which specialized intelligent agents, namely Individual Target Analyzing (ITA) agents, are designed to perform recognizing behaviour on behalf of their corresponding target classes, and then show their identity information and claims that Public Recognition Arbitrating (PRA) agent may adopt for HRRP analyzing and judging. In order to describe the details, we apply Generalized Discriminant Analysis (GDA) in the model, and accordingly, two new GDA variations come forth, called Distributed-GDA (D-GDA) and Synthetic-GDA (S-GDA) respectively. Generally, the traditional application of GDA is to emphasize the Common-Discrimination Information (C-DI) among all targets while D-GDA prefers to the Individual-Discrimination Information (1-DI) against other targets one by one, so their syntheses S-GDA can obtain more useful discrimination information than both of them. Experimental results for measured and simulated data show that GDA and D-GDA are complementary in many facets and can be considered as a feature extraction method couple. Furthermore, compared with GDA and D-GDA, the proposed S-GDA not only achieves better and better recognition performance with the number of targets increasing, but also is more robust to many challenges, such as noise disturbance, aspect variation, Small Sample Size (SSS) problem and etc. All these experimental results confirm the effectiveness of the MAS model proposed in this paper.