 |
|
 |
| The Hunt for Red October: Forays into Bayesian Data Fusion | ||
|---|---|---|
| Tuesday 15 June 1999 | 17.00 | Brian Spratt Room |
| Dr Andrew Runnalls Computing Laboratory University of Kent |
||
Data fusion is the branch of systems engineering concerned with combining data from multiple sensor systems to provide as accurate a picture as possible of the 'real world' that the sensors are monitoring. Areas of application include air traffic control, monitoring of ship traffic in congested waterways, and integrated navigation systems, along with numerous defence applications.
Data fusion must take into account the fact that each sensor system will measure a different aspect of the objects under surveillance, and will be affected by various systematic errors, noise, and other foibles. It has long been recognised that Bayes' Theorem of probability theory can form the foundation stone for a unifying theoretical framework for data fusion, but until recently only a limited range of data fusion problems could be addressed in an explicitly Bayesian fashion. This is now beginning to change, partly because of theoretical advances in statistics, and partly because of the increasing affordability of fast computers.
This seminar will describe some of the problems that data fusion faces, explain some of the new algorithms that are available, and illustrate their operation with examples drawn loosely from the field of submarine tracking.
The seminar is based on two talks given at colloquia organised by the Computing and Control Division of the IEE.
