Home | Search | In Swedish

Luleň University of Technology

MSc Programmes in Engineering Electrical Engineering / 2003:199
TITLE
Target tracking and definition of lost track

AUTHOR
Saltin, Per-Johan

DATE
2003-06-11

DEPARTMENT
Computer Science and Electrical Engineering / Signal Processing

SUMMARY
In modern warfare, systems equipped with video cameras for target tracking are becoming more common. Cameras for visual and infrared light are those most commonly used. Much research is being conducted on tracking. This report will describe target tracking and will mainly determine when a tracker has lost track of its target.

A tracker generally assumes that it tracks its expected target. However, this is not always the case. For example when a tracked target passes a hill, tracking can shift to a house situated on the hill and thus the house becomes the expected target. This is not acceptable. The tracker must be capable of determining when the target is ``lost'', i.e. has disappeared, has shifted to a new object, etc. It must also be able to determine if a tracked target is regained, at least within a limited period of time.

Means of determining when a track has been lost vary with the different tracking principles. In this report a correlation tracker has been used.

Different value types were combined in algorithms and tested. The test was conducted on large numbers of sequences where an artificial target had been overlaid real background images. These were split so that the target could pass in front of or behind various obstacles.

The results show that is is possible to achieve 98 - 99 percent probability to decide when a target is lost by the tracker. Value types combined in the evaluation algorithm are correlation accuracy, position and size.

ISSN 1402-1617 / ISRN LTU-EX--03/199--SE / NR 2003:199

Home | Search | The University | The Library


To the Library
LULE┼ UNIVERSITY LIBRARY