Object tracking is an important element of computer vision algorithms. This problem is difficult due to occlusion, illumination changes and shadows. We propose an appearance based occlusion-aware method for object tracking. Proposed method is based on particle filter tracking in a multi-camera environment. In this method, observations involving both position and appearance information are evaluated depending on whether the corresponding objects are involved in occlusion or not. Tracking is done using state vector in 3D coordinates and probabilities that objects are occluded is estimated to elevate tracking performance. Particles are graded according to their positions and appearances by taking occlusions into account. Weighting particles in terms of position information allows particles to imitate object position and motion. Appearance information help recognize objects after occlusion and track objects when position information is not available. In case of occlusion, particles are weighted according to occlusion probability in order not to make them affected by possibly false measurements. Appearance information is updated by time to account for appearance changes. Appearance is not updated if the object is involved in occlusion. Experiments with PETS and EPFL datasets revealed the success of proposed method and that our method can be applied to different camera configurations.
[1]
Larry S. Davis,et al.
Multi-camera Tracking and Segmentation of Occluded People on Ground Plane Using Search-Guided Particle Filtering
,
2006,
ECCV.
[2]
Neil J. Gordon,et al.
A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking
,
2002,
IEEE Trans. Signal Process..
[3]
Patrick Pérez,et al.
Color-Based Probabilistic Tracking
,
2002,
ECCV.
[4]
Bernhard P. Wrobel,et al.
Multiple View Geometry in Computer Vision
,
2001
.
[5]
Leonidas J. Guibas,et al.
Object tracking in the presence of occlusions using multiple cameras: A sensor network approach
,
2013,
TOSN.
[6]
Dar-Shyang Lee,et al.
Effective Gaussian mixture learning for video background subtraction
,
2005,
IEEE Transactions on Pattern Analysis and Machine Intelligence.
[7]
Luc Van Gool,et al.
An adaptive color-based particle filter
,
2003,
Image Vis. Comput..
[8]
Gang Hua,et al.
Tracking appearances with occlusions
,
2003,
2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings..
[9]
Kristine L. Bell,et al.
A Tutorial on Particle Filters for Online Nonlinear/NonGaussian Bayesian Tracking
,
2007
.
[10]
Andrew Zisserman,et al.
Multiple View Geometry in Computer Vision (2nd ed)
,
2003
.