Robust Dual-Kernel Tracking Using Both Foreground and Background - Advances in Image and Graphics Technologies - page 88

The subjective sense from Figure 1 and Figure 2 indicates that the proposed
tracker outperforms the traditional kernel-based tracker and its amelioration in most
situations.
We quantitatively evaluate the performance using the center location error metric.
It measures the Euclidean distance between the center locations between the tracking
result and the ground truth. A smaller error indicates a better performance. Figure 3
shows the error plots of the trackers. Numerically, the mean errors of KBT, BC-KBT
and DKT are 138.9, 137.0 and 5.4 for the first sequence, and 62.5, 59.1 and 17.9 for
the second sequence. The results demonstrate that the proposed tracker exceeds the
referenced trackers.
We refer to Table 1 for more details of the mean values of the center location er-
rors comparison.

300

KBT
BC-KBT
DKT

250

200

150

100

50

0

0 50 100 150 200 250 300 350 400 450

Frames

300

KBT
BC-KBT
DKT

250

200

150

100

50

0

0 50 100 150 200 250 300 350

Frames

Fig. 3. The center location error plots of KBT, BC-KBT and DKT for the sequence with
obvious background change (Top) and the sequence with complicated background clutters
(Bottom)