-
Articulated Object Tracking
- Abstract
- Tracking Results
- Figure Skater Sequence 1 [Video]
Top: Tracking results using the holistic histogram based algorithm. Bottom: Tracking results using our algorithm. Note the skater changes pose changes 360o a few times. - Figure Skater Sequence 2 [Video]
Top: Tracking results using the holistic histogram based algorithm. Bottom: Tracking results using our algorithm. Note that our
algorithm is able to track the skater across shots (e.g., frame 372 to 373 and onward).
- Dancer Sequence 1 [Video]
Top: Tracking results using the holistic histogram based algorithm. Bottom: Tracking results using our algorithm. Note the
target undergoes large shape and scale variation.
- Dancer Sequence 2 [Video]
Top: Tracking results using the holistic histogram based algorithm. Bottom: Tracking results using our algorithm. Note the
target undergoes large shape and appearance variation.
- Tom and Jerry [Video]
Tracking results using our algorithm. Note there is a large variation in shape, scale and appearance of the target. In addition, the target exhibits ballistic and non-rigid motions.
We propose an algorithm for accurate tracking of articulated objects using online update of appearance and shape. The challenge here is to model foreground appearance with histogram in a way that is both efficient and accurate. In this algorithm, constantly changing foreground shape is modeled as a small number of rectangular blocks, whose positions within the tracking window are adaptively determined. Under the general assumption of stationary foreground appearance, we show that robust object tracking is possible by adaptively adjusting the importance of these blocks. Implemented in MATLAB without substantial optimization, our tracker runs already at 3.7 frames per second on a 3GHz machine. Borne out with experiments, we demonstrate that our algorithm is able to efficiently track articulated objects undergoing large variation in appearance and shape.
We collected a set of image sequences from the web for experiments. These gray scale videos consist of figure skaters, dancers as well as cartoon figures whose shapes and appearances vary rapidly. Tracking result on these sequences are presented here: