Mobility tracking by interactive graph-cut segmentation with Bi-elliptical shape prior

Abstract

Tracking of partially occluded or unevenly illuminated objects is the most challenging problem of the analysis of moving objects. Both problems can solve the graph cut segmentation with elliptical shape prior. Interactive control is necessary in the training phase of system as well as in a critical situation requiring the human intervention. Bi-elliptical shape prior model corresponds to the anatomy of animals and improves the precision of tracking. The system was tested in the analysis of animal's mobility which corresponds to the level of brain damage by a neurodegenerative disease.