The Issue of Robustness in the Acquisition of Relocatable Targets - An Overview

Abstract

Features are a means of statistical pattern recognition that ATR algorithms use to discriminate ground targets from the surrounding clutter background and, subsequently, to sort potential targets into one of several target classes (including the non-target case). Problems for ATR arise from the specular nature of radar imagery because small changes to the configuration of targets can result in significant changes to the resulting target signature [3][4]. This adds to the challenge of constructing a classifier that is both robust to changes in target configuration and target aspect, and which is capable of generalizing to previously unseen targets. ATR features have to provide at the same time good inter-class separability and good intra-class stability. The reference vectors usually are obtained from former measurements of the respective target either on a turntable or by means of SAR and are stored in look-up tables. The test vectors are obtained on-line while the seeker is passing over the target area. In order for the ATR to provide reliable results both the test vectors and the reference vectors have to show robustness against target modifications, preferably including camouflage, different target realizations or articulations, slight changes in depression angle, aspect angle changes that occur during the time-on-target, and many more.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
May 01, 2005
Accession Number
ADA471157

Entities

People

  • Hartmut Schimpf

Tags

Communities of Interest

  • Materials and Manufacturing Processes
  • Sensors

DTIC Thesaurus Topics

  • Acquisition
  • Aspect Angle
  • Classification
  • Cross Correlation
  • Data Science
  • Depression Angles
  • Factor Analysis
  • Frequency
  • Information Science
  • Measurement
  • Millimeter Waves
  • Pattern Recognition
  • Probability
  • Recognition
  • Statistics
  • Target Recognition
  • Target Signatures

Readers

  • Computer Vision.
  • Radar Systems Engineering.

Technology Areas

  • AI & ML