Return Home

About Tracking and Data Fusion

This two-day short course is based on the tutorial that was prepared with Dr Alfonso Farina (Scientific Director, SELEX, Italy). The course is a thorough overview of tracking and data fusion for surveillance systems with applications both to defense and civilian systems. Filters are the basic building block of radar defense systems that track targets, provide surveillance, avoid collisions, and guide flight. The most common type of filter is the Kalman filter (KF). For most applications the KF is reliable and efficient, but it is limited to a relatively restricted class of linear Gaussian problems. To solve problems beyond this restricted class, nonlinear filters (such as EKF, UKF, particle filter, IMM) are used. This course introduces the latest advances in nonlinear filters, data association techniques and tracking system designs and examines their application to ballistic missile tracking, GMTI tracking, tracking through the blind Doppler zone and others.

The course will help to develop more accurate and reliable filter designs and more precisely predict these designs' performance. It is intended for those who interested in tracking and radar system designs: engineers, scientists, technicians and others, of military or civil employment.

Course Topics

• State estimation for stochastic dynamic systems: optimal Bayesian estimator, linear filters, nonlinear filters, hybrid state estimation, IMM filter, case studies.
• Data association: gating, (global) nearest neighbour algorithm, (joint) probabilistic data association, multiple hypotheses tracking, multi-dimensional assignment, mixture reduction data association.
• Tracking system design issues: management of track life stages (initialisation, termination), tracking coordinate systems, performance prediction and evaluation and electronic counter-counter measures.
• Multiple sensor tracking and fusion: fusion architectures, information graphs, distributed fusion, track association and fusion, registration of multiple sensors.
• Sensor management: the role of sensor management, optimisation criteria, look-ahead optimisation techniques, case studies.
• Decision support systems: representation of uncertain attributes and knowledge, combination of evidence, evidential networks, local computation algorithms, applications to target ID and threat assessment.
• Selected applications: ballistic missile tracking, GMTI radar tracking, angle-only tracking, tracking using TDoA measurements, track-before-detect, bi-static radar tracking, video tracking, wireless sensor networks.

Guest Speaker: Dr Daniel Clark

This unit of the course will be about 1.5 to 2 hours long, and will be delivred after the afternoon break on the first day.

Topics:

• Bayesian tracking using Random finite sets;
• Introduction to the Bayesian framework for multi-target filtering based on the theory of random finite sets;
• PHD filter and the Cardinalised PHD (CPHD) filter; particle-PHD/CPHD filters; the Gaussian mixture PHD/CPHD filters.

Daniel Clark was awarded his PhD on Multiple Target Tracking at Heriot-Watt University, Scotland in 2006.  He worked as a researcher in Cambridge and Melbourne before returning to Heriot-Watt as a Lecturer in 2007. His research interests are in the theory and applications of random-set stochastic filtering techniques for multi-object tracking. Using this framework, he has developed algorithms for multi-target tracking, established their convergence properties and demonstrated them on a range of different sensor data, including sonar, millimetre wave images and audio data. He has deployed these techniques on an autonomous underwater vehicle for oil pipeline tracking which achieved a world record in autonomous navigation.

For further information please contact:
Anne-Marie Eliseo
Industry Education Manager
phone: +61-8-8302-3928
email: industryeducation@nicta.com.au