This paper is aimed to construct a human-machine interacting system based on the visual tracking techniques. A monocular camera is set up in front of the interacting user, and the interaction is performed through a held object to express the user's intention. Since the object motion during interaction is arbitrary, the target modeling with multiple visual clues must be considered at the same time to reliably track the target in such challenged scenario. The target appearance is modeled as a linear combination of several target templates and trivial templates in various color spaces with sparse coefficient vectors. To achieve the real-time performance for human-machine interaction, the adaptive particle filtering algorithm is proposed to balance the tracking robustness and processing instantaneity. The dominant templates in the discriminative color channels will be selected to verify the tracking hypotheses. The sparse coefficient vectors of each hypothesis corresponding to the selected templates are then efficiently estimated by the particle swarm optimization. The selected templates and the estimated sparse coefficient vectors are dynamically changing over time. The overall performance has been validated in the experiments.
[1]
R. Collins,et al.
On-line selection of discriminative tracking features
,
2003,
Proceedings Ninth IEEE International Conference on Computer Vision.
[2]
Carlo Tomasi,et al.
Good features to track
,
1994,
1994 Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.
[3]
Neil J. Gordon,et al.
A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking
,
2002,
IEEE Trans. Signal Process..
[4]
Xiaoqin Zhang,et al.
Sequential particle swarm optimization for visual tracking
,
2008,
2008 IEEE Conference on Computer Vision and Pattern Recognition.
[5]
Allen Y. Yang,et al.
Robust Face Recognition via Sparse Representation
,
2009,
IEEE Transactions on Pattern Analysis and Machine Intelligence.
[6]
Yasushi Yagi,et al.
Integrating Color and Shape-Texture Features for Adaptive Real-Time Object Tracking
,
2008,
IEEE Transactions on Image Processing.