mean shift

The last and largest peak is due to the fast movement

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: side. 1 Bhattacharyya Coefficient 0.9 0.8 0.7 0.6 0.5 0.4 0.3 Initial location Convergence point 0.2 40 20 0 −20 −40 Y 40 0 20 −20 −40 X Figure 3: Values of the Bhattacharyya coe cient corresponding to the marked region (81 81 pixels) in frame 105 from Figure 1. The surface is asymmetric, due to the player colors that are similar to the target. Four mean shift iterations were necessary for the algorithm to converge from the initial location (circle). To demonstrate the e ciency of our approach, Figure 3 presents the surface obtained by computing the Bhattacharyya coe cient for the rectangle marked in Figure 1, frame 105. The target model (the selected elliptical region in frame 30) has been compared with the target candidates obtained by sweeping the elliptical region in frame 105 inside the rectangle. While most of the tracking approaches based on regions 3, 14, 21] Figure 1: Football sequence: Tracking the player no. 75 with initial window of 71 53 pixels. The frames 30, 75, 105, 140, and 150 are shown. 5 must perform an exhaustive search in the rectangle to nd the maximum, our algorithm converged in four iterations as shown in Figure 3. Note that since the basin of attraction of the mode covers the entire window, the correct location of the target would have been reached also from farther initial points. An optimized computation of the exhaustive search of the mode 13] has a much larger arithmetic complexity, depending on the chosen search area. The new method has been applied to track people on subway platforms. The camera being xed, additional geometric constraints and also background subtraction can be exploited to improve the tracking process. The following sequences, however, have been processed with the algorithm unchanged. A rst example is shown in Figure 4, demonstrating the capability of the tracker to adapt to scale changes. The sequence has 187 frames of 320 240 pixels each and the initial normalization constants were (hx hy ) = (23 37). Figure 5 presents six frames from a 2 minute sequence showing the tracking of a pers...
View Full Document

Ask a homework question - tutors are online