Unformatted text preview: on from the moment she enters the subway platform till she gets on
the train ( 3600 frames). The tracking performance is
remarkable, taking into account the low quality of the
processed sequence, due to the compression artifacts. A
thorough evaluation of the tracker, however, is subject
to our current work.
The minimum value of the distance (18) for each
frame is shown in Figure 6. The compression noise
determined the distance to increase from 0 (perfect
match) to a stationary value of about 0:3. Signi cant
deviations from this value correspond to occlusions generated by other persons or rotations in depth of the target. The large distance increase at the end signals the
complete occlusion of the target. Figure 4: Subway1 sequence: The frames 500, 529,
600, 633, and 686 are shown (leftright, topdown).
histogram of a face model onto the incoming frame.
However, the direct projection of the model histogram
onto the new frame can introduce a large bias in the
estimated location of the target, and the resulting measure is scale variant. Gradient based region tracking has
been formulated in 2] by minimizing the energy of the
deformable region, but no realtime claims were made. 6 Discussion By exploiting the spatial gradient of the statistical
measure (18) the new method achieves realtime tracking performance, while e ectively rejecting background
clutter and partial occlusions.
Note that the same technique can be employed
to derive the measurement vector for optimal prediction schemes such as the (Extended) Kalman lter 1,
p.56, 106], or multiple hypothesis tracking approaches
5, 9, 17, 18]. In return, the prediction can determine
the priors (de ning the presence of the target in a given
neighborhood) assumed equal in this paper. This connection is however beyond the scope of this paper. A
patent application has been led covering the tracking
algorithm together with the Kalman extension and various applications 29].
We nally observe that the idea of centroid computation is also employed in 22]. The mea...
View
Full Document
 Fall '10
 Staff
 Math, The Land, Nonparametric statistics, kernel, Kernel density estimation, Density estimation, Bhattacharyya coe cient

Click to edit the document details