Unformatted text preview: rrent measurements and old radius. 5 Experiments The proposed method has been applied to the task
of tracking a football player marked by a hand-drawn
ellipsoidal region ( rst image of Figure 1). The sequence has 154 frames of 352 240 pixels each and
the initial normalization constants (determined from
the size of the target model) were (hx hy ) = (71 53).
The Epanechnikov pro le (4) has been used for histogram computation, therefore, the mean shift iterations were computed with the uniform pro le. The target histogram has been derived in the RGB space with
32 32 32 bins. The algorithm runs comfortably at
30 fps on a 600 MHz PC, Java implementation.
The tracking results are presented in Figure 1. The
mean shift based tracker proved to be robust to partial
occlusion, clutter, distractors (frame 140 in Figure 1), 2. Derive the weights fwi gi=1:::nh according to (25).
3. Based on the mean shift vector, derive the new
location of the target (14)
y1 = Pnh i=1 xi wi g Pnh i=1 wi g 2 ^
y 0 ;x i
h 2 : (26)
4 and camera motion. Since no motion model has been
assumed, the tracker adapted well to the nonstationary
character of the player's movements, which alternates
abruptly between slow and fast action. In addition,
the intense blurring present in some frames and due to
the camera motion, did not in uence the tracker performance (frame 150 in Figure 1). The same e ect,
however, can largely perturb contour based trackers.
16 Mean Shift Iterations 14
0 50 100 150 Frame Index Figure 2: The number of mean shift iterations function
of the frame index for the Football sequence. The mean
number of iterations is 4:19 per frame.
The number of mean shift iterations necessary for
each frame (one scale) in the Football sequence is shown
in Figure 2. One can identify two central peaks, corresponding to the movement of the player to the center
of the image and back to the left side. The last and
largest peak is due to the fast movement from the left
to the right...
View Full Document