![](bg8.png)
Table 1: Comparisons of tracking successful ratio among
M-PF, PF+Detector, FaceAPI, and M-PFDMA (proposed).
When we calculated the tracking successful ratio, the
frames which were not able to track due to occlusion were
excluded.
successful tracking ratio
Method video 1 video2
M-PF 37.5% 36.9%
PF+Detector 52.4% 62.2%
FaceAPI 64.2% 53.7%
M-PFDMA (proposed) 81.5% 90.1%
distribution prediction. The left figure shows the
output of M-PF, and the right figure shows that of
M-PFDMA. As shown in Fig.8, M-PF stored his-
tory covered only the top-right area, where tracking
started. This means that the tracker failed to redis-
cover target when the target moved to other areas af-
ter occlusions due to large changes in position/pose
while occlusions. On the contrary, the stored memory
of M-PFDMA covered the entire field of view. The
numbers of stored memories of this sequence by M-
PF and M-PFAP are 455 and 808, respectively. The
memory acquisition performance of M-PFDMA un-
der severe occlusion was confirmed.
5 CONCLUSIONS
A memory-based particle filter with detection-based
memory acquisition, in short M-PFDMA, was pro-
posed for vision-based object tracking. M-PFDMA
offers robust memory acquisition under severe oc-
clusion since it creates a synergistic combination of
detection-based memory acquisition and the memory-
based approach. M-PFDMA was shown to achieve
high accuracy and quick recovery in real-world sit-
uations. We verified its effectiveness in facial pose
tracking experiments.
Future works include memory management. M-
PFDMA stores the correctly estimated target state
in memory. The correctness is automatically judged
by using the maximum likelihood among particles.
Though it works well in most cases, the quality of
stored data is very important for memory-based prior
distribution prediction. Therefore, we will consider
more precise ways of judging tracking correctness.
Future works also include automatic determination of
the mixing weight α of detection-based and memory-
based prior distribution. Results on this paper em-
ployed static α. However, the ideal α varies by condi-
tions such as current tracking stability. We would like
to reveal factors affecting α, and then, would like to
tackle automatic setting of α according to the factors.
REFERENCES
Andriluka, M., Roth, S., and Schiele, B. (2010). Monocu-
lar 3D pose estimation and tracking by detection. In
CVPR 2010, pages 623–630.
Ba, S. and Odobez, J.-M. (2008). Probabilistic head pose
tracking evaluation in single and multiple camera.
Multimodal Technologies for Perception of Humans,
4625/2008:276–286.
Bradski, G. R. (1998). Computer vision face tracking for
use in a perceptual user interface. In Proc. IEEE Work-
shop Applications of Computer Vision, pages 214–
219.
Comaniciu, D., Ramesh, V., and Meer, P. (2003). Kernel-
based object tracking. IEEE Trans. PAMI, 25:564–
577.
Gordon, N., Salmond, D., and Smith, A. F. M.
(1993). Novel approach to non-linear and
non-gaussian bayesian state estimeation. lEE
Proc.F:Communications Rader and Signal Process-
ing, 140(2):107–113.
Isard, M. and Blake, A. (1998). Condensation - condi-
tional density propagation for visual tracking. IJCV,
29(1):5–28.
Kobayashi, Y., Sugimura, D., Sato, Y., Hirasawa, K.,
Suzuki, N., Kage, H., and Sugimoto, A. (2006). 3D
head tracking using the particle filter with cascaded
classifiers. In BMVC.
Lozano, O. M. and Otsuka, K. (2008). Real-time visual
tracker by stream processing. Journal of VLSI Signal
Processing Systems.
Mikami, D., Otsuka, K., and Yamato, J. (2009). Memory-
based particle filter for face pose tracking robust under
complex dynamics. In Proc. CVPR, pages 999–1006.
Murphy-Chutorian, E. and Trivedi, M. M. (2008). Head
pose estimation in computer vision: A survey. IEEE
Trans. PAMI.
Otsuka, K., Araki, S., Ishizuka, K., Fujimoto, M., Hein-
rich, M., and Yamato, J. (2008). A realtime multi-
modal system for analyzing group meeting by com-
bining face pose tracking and speaker diarization. In
Proc. ACM ICMI, pages 257–264.
Ozuysal, M., Lepetit, V., Fleuret, F., and Fua, P. (2006).
Feature harvesting for tracking-by-detection. In LNCS
3953(ECCV2006), pages 592–605.
Seeingmachines. FaceAPI. http://
www.seeingmachines.com/product/faceapi/.
Tian, L., Ando, S., Suzuki, A., and Koike, H. (2010). A
probabilistic approach for fast and robust multi-view
face detection using compact local patterns. In Proc.
the IIEEJ Image Electronics and Visual Computing
Workshop.
Tua, J., Taob, H., and Huang, T. (2007). Face as mouse
through visual face tracking. CVIU, 108(1-2):35–40.
ENHANCING MEMORY-BASED PARTICLE FILTER WITH DETECTION-BASED MEMORY ACQUISITION FOR
ROBUSTNESS UNDER SEVERE OCCLUSION
215