Abstract
In this paper a robust method for block-based motion analysis in monocular image sequences is considered. Due to the realized recognition of false measurements by a neural recognition system, gross errors in the motion trajectories are avoided. Assuming correspondences between regions in successive images, a model based recursive estimation technique is applied to estimate the motion of the observed image region. For this purpose a Kalman filter is used. The underlying kinematic model contains assumptions about the motion, especially constant velocity components are assumed. Particularly, the existence of problematic image situations (e.g. partial occlusion of objects) leads to gross errors in the measuring values and false motion parameters are estimated. In order to cope with this problem an extension of the filtering algorithm by a neural recognition system is proposed. This system recognizes typical problematic image situations and controls the adaptation of the filter. Selected results for real-world image sequences are described.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Musmann, H.G., Pirsch, P., Grallert, H.J.: Advances in Picture Coding. Proc. of the IEEE 73(4), 523–548 (1985)
Zhang, Z.: Token Tracking in a Cluttered Scene. Research Report No 2072, INRIA (October 1993) ISSN 0249-6399
Blake, A., et al.: Affine-Invariant Contour Tracking With Automatic Control of Spatiotemporal Scale. In: Proc. of 4. Int. Conf. on Computer Vision, Berlin, May 11th-14th. IEEE Computer Society Press, Los Alamitos (1993) ISBN 0-8186-3870-2
Maybeck, P.S.: Stochastic Models, Estimation and Control, vol. 1-3. Academic Press Inc., London (1982) ISBN 0-12-48070(1or2or3)-X
Sorenson, H.W.: Kalman Filtering: Theory and Application. IEEE Press, Los Alamitos (1985) ISBN 0-87942-191-6
Kalman, R.E.: A New Approach to Linear Filtering and Prediction Problems. In: [5] from this list, pp. 16–26 (1960)
Haykin, S.: Adaptive Filter Theorie, 3rd edn. Prentice-Hall Inc., Englewood Cliffs (1996)
Schnelting, O., Michaelis, B., Mecke, R.: Motion Estimation By Using Artificial Neural Networks. In: Proc. of IEEE-Workshop on Nonlinear Signal and Image Processing, Neos Marmaras, Greece, June 20th-22th, vol. 2, pp. 714–717 (1995)
Mecke, R.: Grauwertbasierte Bewegungsschätzung in monokularen Bildsequenzen unter besonderer Berücksichtigung bildspezifischer Störungen, PhD Thesis, Ottovon- Guericke-University of Magdeburg, Department Electrical Engineering (1999)
Rao, R.P.N.: Robust Kalman Filters for Prediction, Recognition, and Learning. Technical Report 645, University of Rochester, Computer Science Department, Rochester, New York (Dezember 1996)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2000 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Mecke, R., Michaelis, B. (2000). A Robust Method for Motion Estimation in Image Sequences. In: Nagel, HH., Perales López, F.J. (eds) Articulated Motion and Deformable Objects. AMDO 2000. Lecture Notes in Computer Science, vol 1899. Springer, Berlin, Heidelberg. https://doi.org/10.1007/10722604_10
Download citation
DOI: https://doi.org/10.1007/10722604_10
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-67912-7
Online ISBN: 978-3-540-44591-3
eBook Packages: Springer Book Archive