Monte Carlo Techniques for Automated Target Recognition

Motivated by the extraordinary performance of the human visual system, automated recognition of targets from their remotely sensed images has become an active area of research. In particular, Bayesian techniques based on statistical models of system components such as targets, sensors, and clutter have emerged. Since the probability models associated with such physical systems are complicated, general closed-form analytical solutions are ruled out, and computational approaches become vital. This chapter explores a family of computational solutions applied to finding the unknowns associated with Automated Target Recognition (ATR) problems. In general ATR, remote sensors (such as visual or infrared cameras, or microwave or laser radars) observe a scene containing a number of targets, either moving or stationary. These sensors produce measurements that are analyzed by computer algorithms to detect, track and recognise the targets of interest in that scene. (Dudgen and Lacoss 1993) and (Srivastava, Miller and Grenander 1999) offer more detailed introductions.

Editors and Affiliations

1. Department of Electrical and Electronic Engineering, The University of Melbourne, Victoria, 3010, Australia

   Arnaud Doucet

2. Computer Science Division, University of California, 387 Soda Hall, 94720-1776, Berkeley, CA, USA

   Nando de Freitas

3. Pattern and Information Processing, Defence Evaluation and Research Agency, St. Andrews Road, WR14 3PS, Malvern, Worcs, UK

   Neil Gordon

Reprints and permissions

© 2001 Springer Science+Business Media New York

Policies and ethics