Information Technology Reference
In-Depth Information
5 Conclusions
The Signature-based Tracking Across Cameras (STAC) algorithm as part of a
distributed tracking framework enables real-time multi-camera tracking without
a training phase. The kernel-based tracking algorithm covers the entire field of
view of each camera rather than only entry and exit points, and continuously
collects and updates tracking statistics. Reusing kernels enables the collection
of tracking statistics. Also, reusing kernels places a bound on memory usage,
allowing implementation in an embedded application. The novel location-based
kernel matching method uses tracking statistics to accommodate abrupt and
unpredictable changes in the visual characteristics of objects within and across
camera views. We showed that STAC's tracking accuracy and speed were im-
proved over seven test sets by the addition of location-based kernel matching.
References
1. Arsic, D., Lyutskanov, A., Rigoll, G., Kwolek, B.: Multi camera person tracking ap-
plying a graph-cuts based foreground segmentation in a homography framework. In:
12th IEEE Int Workshop on Performance Evaluation of Tracking and Surveillance
(PETS-Winter), pp. 1-8. IEEE, Los Alamitos (2009)
2. Khan, S., Shah, M.: Consistent labeling of tracked objects in multiple cameras with
overlapping fields of view. IEEE Trans. Pattern Anal. Mach. Intell. 25, 1355-1360
(2003)
3. Lee, L., Romano, R., Stein, G.: Monitoring activities from multiple video streams:
establishing a common coordinate frame. IEEE Trans. Pattern Anal. Mach. In-
tell. 22, 758-767 (2000)
4. Makris, D., Ellis, T., Black, J.: Bridging the gaps between cameras. In: Int Conf
Computer Vision Pattern Recognition (CVPR), pp. 205-210. IEEE, Los Alamitos
(2004)
5. Cichowski, A., Madden, C., Detmold, H., Dick, A., van den Hengel, A., Hill, R.:
Tracking hand-off in large surveillance networks. In: 24th Int Conf on Image and
Vision Computing New Zealand (IVCNZ), pp. 276-281. IEEE, Los Alamitos (2009)
6. Sheikh, Y., Li, X., Shah, M.: Trajectory Association across Non-overlapping Mov-
ing Cameras in Planar Scenes. In: Int Conf Computer Vision Pattern Recognition
(CVPR), pp. 1-7. IEEE, Los Alamitos (2007)
7. Javed, O., Rasheed, Z., Shafique, K., Shah, M.: Tracking across multiple cameras
with disjoint views. In: 9th IEEE Int Conf on Computer Vision (ICCV), pp. 952-957.
IEEE, Los Alamitos (2003)
8. Petrushin, V.A., Wei, G., Gershman, A.V.: Multiple-camera people localization in
an indoor environment. Knowl Inf. Syst. 10, 229-241 (2006)
Search WWH ::
Custom Search