Konferenční příspěvek

Haindl M., Žid P., Holub R.

10th International Conference on Information Sciences, Signal Processing and their Applications

IEEE

Los Alamitos

2010

978-1-4244-7166-9

prof. Ing. Michal Haindl, DrSc. (ÚTIA - Oddělení rozpoznávání obrazů)

Range segmentation, Markov random fields

An unsupervised range video segmentation method based on a spatial probabilistic model for intended vehicle-based safety and warning system applications is introduced. Statistical range data discontinuities are represented by a wide-sense Markov model which guides the subsequent line-based region growing process. Single frame segmentations are mutually corrected using the continuity constraint. The resulting segmentation allows tracking moving objects and estimating their distance and velocity. The method is illustrated on synthetic range video data.