ICCCI'12 Proceedings of the 4th international conference on Computational Collective Intelligence: technologies and applications - Volume Part I
Hi-index | 0.00 |
In this paper a system is presented for object tracking based on the novel connection of the one-class SVM classifier with the mean shift tracker. An object for tracking is defined by feature vectors composed of the components of the orthogonal color space, as well as local phase and coherence components of the structural tensor which convey information on texture. The binary output of the SVM is mapped into a membership field with a proposed transformation function. Tracking is performed with the continuously adaptive mean shift method operating in the membership field. The method shows high discriminative power and fast run-time properties.