![](bg8.png)
5 CONCLUSIONS AND
OUTLOOK
In this work, we firstly had a review on the recent stud-
ies of multi-object tracking using a single camera or
multiple cameras and discussed the recent researches
for global optimization based on flow networks or
graphs. After the formulation of data association into
two MAP problems. we proposed a two-stage graph-
based multi-person multi-camera tracking approach.
Firstly, a hypothesis graph was constructed to extract
possible associated tracklets from reconstructed 3D
detections. While most of the papers considered global
features only, we incorporated local features such as
appearance, size into the computation of costs for the
edges in the hypothesis graph. Importantly, the hy-
potheses for tracking on the ground plane were arose
from the reconstructions of 2D detections from each
view at the same time step. Those reconstructed 3D de-
tections who were regarded to be the same object were
replaced by the one with the minimum back projection
error. Consequently, the task of the second graph was
to link tracklets into complete tracks. For this sake,
the cost function accordingly took temporal and spa-
tial distances into account. All in all, this framework
is general for multi-object tracking in multi-camera
systems.
From our experiments, we conclude that it is im-
portant to have the optimal outcome from the first
step of hypothesis generation for multi-object track-
ing in multi-camera systems. Due to the impact of
calibration and object detection errors, the precision
of recognizing of identical objects can be improved
by more restrict constraints. Hence, in the future, we
would like to focus on the modeling of calibration and
detection errors and incorporating them into the frame-
work. Incremental learning might be able to refine
the modeling as more and more frames are processed.
Additionally, the tracklet linking stage could consider
more information such as histogram of motion in the
cost function to reduce the false positive rate and the
number of Id switches.
REFERENCES
Andriyenko, A. and Schindler, K. (2011). Multi-target track-
ing by continuous energy minimization. In CVPR,
pages 1265–1272.
Berclaz, J., Fleuret, F., Turetken, E., and Fua, P. (2011).
Multiple object tracking using k-shortest paths opti-
mization. TPAMI, 33:1806–1819.
Bernardin, K. and Stiefelhagen, R. (2008). Evaluating multi-
ple object tracking performance: The clear mot metrics.
EJIVP, 246309.
Bredereck, M., Jiang, X., K
¨
orner, M., and Denzler, J. (2012).
Data association for multi-object tracking-by-detection
in multi-camera networks. In ICDSC.
Breitenstein, M. D., Reichlin, F., Leibe, B., Koller-Meier, E.,
and Gool, L. V. (2011). Online muti-person tracking-
by-detection from a single, uncalibrated camera. PAMI,
33:1820 – 1833.
Collins, R. T. (2012). Multitarget data association with
higher-order motion models. In CVPR.
Dijkstra, E. W. (1959). A note on two problems in connexion
with graphs. NUMERISCHE MATHEMATIK, 1:269–
271.
Felzenszwalb, P., Girshick, R., and McAllester, D. (2010).
Cascade object detection with deformable part models.
In CVPR.
Ferryman, J. and Shahrokni, A. (2009). Pets2009: Dataset
and challenge. In 2009 Twelfth IEEE International
Workshop on Performance Evaluation of Tracking and
Surveillance (PETS-Winter).
Fleuret, F., Berclaz, J., Lengagne, R., and Fua, P. (2008).
Multi-camera people tracking with a probabilistic oc-
cupancy map. TPAMI, 30:267–282.
Henriques, J. F., Caseiro, R., and Batista, J. (2011). Globally
optimal solution to multi-object tracking with merged
measurements. In ICCV.
Hofmann, M., Wolf, D., and Rigoll, G. (2013). Hypergraphs
for joint multi-view reconstruction and multi-object
tracking. In CVPR.
Huang, C., Wu, B., and Nevatia, R. (2008). Robust ob-
ject tracking by hierarchical association of detection
responses. In ECCV, pages 788–801.
J. Yang, Z. Shi, P. V. and Teizer, J. (2009). Probabilistic mul-
tiple people tracking through complex situations. In
IEEE Workshop Performance Evaluation of Tracking
and Surveillance.
Jiang, X., Haase, D., K
¨
orner, M., Bothe, W., and Denzler,
J. (2013). Accurate 3d multi-marker tracking in x-ray
cardiac sequences using a two-stage graph modeling
approach. In the 15th Conference on Computer Analy-
sis of Images and Patterns (CAIP).
Jiang, X., Rodner, E., and Denzler, J. (2012). Multi-
person tracking-by-detection based on calibrated multi-
camera systems. In ICCVG, pages 743–751.
Leal-Taix
´
e, L., Pons-Moll, G., and Rosenhahn, B. (2012).
Branch-and-price global optimization for multi-view
multi-target tracking. In CVPR, pages 1987–1994.
Satoh, Y., Okatani, T., and Deguchi, K. (2004). A color-
based tracking by kalman particle filter. In ICPR, pages
502–505.
Wu, Z., Kunz, T. H., and Betke, M. (2011). Efficient track
linking methods for track graphs using network-flow
and set-cover techniques. In CVPR, pages 1185–1192.
Wu, Z., Thangali, A., Sclaroff, S., and Betke, M. (2012).
Coupling detection and data association for multiple
object tracking. In CVPR.
Xing, J., Ai, H., and Lao, S. (2009). Multi-object track-
ing through occlusions by local tracklets filtering and
global tracklets association with detection responses.
In CVPR.
Zhang, L., Li, Y., and Nevatia, R. (2008). Global data
association for multi-object tracking using network
flows. In CVPR.
VISAPP2014-InternationalConferenceonComputerVisionTheoryandApplications
350