Multi-object Tracking Using Particle Swarm Optimization on Target Interactions - Advances in Heuristic Signal Processing and Applications - page 76

also demonstrated that the PSO algorithm is an effective tool for solving such non-
linear and nonconvex energy functions. Since the PSO does not rely on any gradient
information, smoothness, or continuity properties, it is possible to employ in the ob-
jective functions the terms that employ information, for instance, about the nearest
neighbors, identity switches, etc. The PSO-algorithm has also demonstrated great
usefulness in single object tracking where swarms consisting of 20 particles and
in 10 iterations are able to follow objects, even in case of considerable temporal
occlusions. The discussed algorithms were implemented in MATLAB/C.
4.6 Conclusions
We demonstrated that in multi-object tracking, considerable improvement of the
tracking accuracy can be obtained through the use of an optimization algorithm for
the refinement of the results obtained by individual trackers, even if they are built
on highly discriminative appearance models. In the presented algorithm, the joint
state is optimized in some moving temporal widow. The state vector consists of the
states determined by the individual trackers in the current frame and the states that
were progressively refined in the previous frames. We demonstrated that the particle
swarm optimization is an effective tool for solving such nonlinear and nonconvex
energy functions. Individual object tracking was considered as a numerical opti-
mization problem, where a particle swarm optimization was utilized in searching
for the best local mode of the similarity measure.
References
1. Andriyenko, A., Schindler, K.: Globally optimal multi-target tracking on a hexagonal lattice.
In: Proc. of the 11th European Conf. on Computer Vision: Part I, pp. 466-479 (2010)
2. Andriyenko, A., Schindler, K.: An analytical formulation of global occlusion reasoning for
multi-target tracking. In: IEEE Int. Workshop on Visual Surveillance, pp. 1839-1846 (2011)
3. Andriyenko, A., Schindler, K.: Multi-target tracking by continuous energy minimization. In:
IEEE Int. Conf. on CVPR, pp. 1265-1272 (2011)
4. Arsic, D., Lyutskanov, A., Rigoll, G., Kwolek, B.: Multi-camera person tracking applying a
graph-cuts based foreground segmentation in a homography framework. In: IEEE Int. Work-
shop on Performance Evaluation of Tracking and Surveillance, pp. 30-37 (2009)
5. Arsigny, V., Fillard, P., Pennec, X., Ayache, N.: Log-Euclidean metrics for fast and simple
calculus on diffusion tensors. Magn. Reson. Med. 56 , 411-421 (2006)
6. Berclaz, E.T.J., Fleuret, F., Fua, P.: Multiple object tracking using k -shortest paths optimiza-
tion. IEEE Trans. Pattern Anal. Mach. Intell. 33 (9), 1806-1819 (2011)
7. Boykov, Y., Veksler, O., Zabih, R.: Fast approximate energy minimization via graph cuts.
IEEE Trans. Pattern Anal. Mach. Intell. 23 (11), 1222-1239 (2001)
8. Breitenstein, M.D., Reichlin, F., Leibe, B., Koller-Meier, E., Van Gool, L.J.: Robust tracking-
by-detection using a detector confidence particle filter. In: ICCV'09, pp. 1515-1522 (2009)
9. Cai, Y., de Freitas, N., Little, J.J.: Robust visual tracking for multiple targets. In: ECCV,
vol. IV, pp. 107-118 (2006)
10. Cheng-Hao, K., Huang, C., Nevatia, R.: Multi-target tracking by on-line learned discriminative
appearance models. In: IEEE Int. Conf. on CVPR, pp. 685-692 (2010)