Incremental Learning for Visual Tracking
Most existing tracking algorithms construct a representation of
a target object prior to the tracking task starts, and utilize
invariant features to handle appearance variation of the target caused
by lighting, pose, and view angle change.
In this paper, we present an efficient and effective online algorithm
that incrementally learns and adapts a low dimensional eigenspace
representation to reflect appearance changes of the target, thereby
facilitating the tracking task.
Furthermore, our incremental method correctly updates the sample mean
and the eigenbasis, whereas existing incremental subspace update methods
ignore the fact the sample mean varies over time.
The tracking problem is formulated as a state inference problem within
a Markov Chain Monte Carlo framework and a particle filter is incorporated
for propagating sample distributions over time.
Numerous experiments demonstrate the effectiveness of the proposed tracking
algorithm in indoor and outdoor environments where the target objects
undergo large pose and lighting changes.
For each panel, the number on the upper left corner denotes the frame number,
and the tracking results are shown in the first row, where the green boxes
represent particles with large confidence and the red box is the final
The images in the second row shows the current sample mean, tracked image
region, reconstructed image with the mean and eigenbases,
and the reconstruction error respectively.
The third and forth rows show 10 largest eigenbases.
A person moves from dark toward bright area with large lighting and pose
result: [ avi(16MB) ]
An animal doll moving with large pose, lighting variation in a cluttered
background. The first result shows two failure-and-recovery in severe pose
changes before it loses track after 900th frame. The second result with
forgetting factor, it gives more robust tracking result.
without forgetting factor [ avi(31MB) ]
with forgetting factor [ avi(32MB) ]
A person moves underneath a trellis with large illumination change and cast
shadows while changing his pose. The tracker works well over the severe
shadows and illumination changes. Without forgetting factor, it fails when
the person changes his pose suddenly while the lighting is also changing.
The second result with forgetting factor does not fail, but the drift in
scale and rotation is noticeable.
without forgetting factor [ avi(12MB) ]
with forgetting factor [ avi(17MB) ]
This sequence is originally used in Jepson et al.'s tracking paper.
We subsampled the original video of 30fps into 15fps to give faster motions,
and our tracker could successfully track it.
This sequence includes pose changes, facial expression and appearance (glasses)
without forgetting factor [ avi(34MB) ]