Generative-Model-Based Tracking by Cluster Analysis of Image Differences

Robotics and Autonomous Systems, vol.39 (3-4), pp. 181-194, 2002.

Full text supplied by ScienceDirect

A postscript copy is available from the author on request.

E-mail: aecp@diku.dk

Abstract

The EM algorithm is used to track moving objects as clusters of pixels significantly different from the corresponding pixels in a reference image. The underlying cluster model is Gaussian in image space, but not in grey-level-difference distribution. The generative model is used to derive criteria for the elimination and merging of clusters, while simple heuristics are used for the initialisation and splitting of clusters. The system is competitive with other tracking algorithms based on image differencing.

Related publications

A.E.C. Pece (2002)
From Cluster Tracking to People Counting
Proceedings of the 3rd IEEE Int. Workshop on Performance Evaluation in Tracking and Surveillance:
PETS 2002, Copenhagen, 1 June 2002
A.E.C. Pece (2001)
Tracking of Non-Gaussian Clusters in the PETS2001 Image Sequences.
Proceedings of the 2nd IEEE Int. Workshop on Performance Evaluation in Tracking and Surveillance:
PETS2001, Kauai (Hawaii), 9 Dec. 2001.
A.E.C. Pece (2000)
Tracking by Cluster Analysis of Image Differences.
in: Proceedings of the 8th Int. Symposium on Intelligent Robotic Systems:
SIRS 2000, The University of Reading, Reading, UK, 18-20 July 2000.
(J.M. Ferryman and A.D. Worrall, eds.)