Dealing with occlusions with multi targets tracking algorithms for the real road context

In this paper, we present a robust approach to occlusion problems for tracking vehicle and pedestrian on road context. Most multi-target tracking algorithms, like Multiple Hypothesis Tracker (MHT) or Cardinalized Probability Hypothesis Density (CPHD), are based on a sensor detection probability map. This paper proposes to solve the occlusion issue by modifying this detection probability map. We assume targets occlusion is provided by other targets and are treated as non detection event. The new detection probability map is computed by taking into account the width and the imprecision of the position of the targets that hide the others. Our system has been validated with simulated data and also with real measurements from a smart camera sensor embedded in a real car for road context.

[1]  F. Nashashibi,et al.  Laser-based vehicles tracking and classification using occlusion reasoning and confidence estimation , 2008, 2008 IEEE Intelligent Vehicles Symposium.

[2]  Lili Huang,et al.  Real-time multi-vehicle tracking based on feature detection and color probability model , 2010, 2010 IEEE Intelligent Vehicles Symposium.

[3]  R. Mahler,et al.  PHD filters of higher order in target number , 2006, IEEE Transactions on Aerospace and Electronic Systems.

[4]  G. Gate,et al.  Using targets appearance to improve pedestrian classification with a laser scanner , 2008, 2008 IEEE Intelligent Vehicles Symposium.

[5]  Francesc Serratosa,et al.  Dealing with occlusion in a probabilistic object tracking method , 2008, 2008 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops.

[6]  S.S. Blackman,et al.  Multiple hypothesis tracking for multiple target tracking , 2004, IEEE Aerospace and Electronic Systems Magazine.

[7]  Klaus C. J. Dietmayer,et al.  Pedestrian tracking using Random Finite Sets , 2011, 14th International Conference on Information Fusion.

[8]  Dietrich Fränken,et al.  Missed detection problems in the cardinalized probability hypothesis density filter , 2008, 2008 11th International Conference on Information Fusion.

[9]  Heungkyu Lee,et al.  Detection of occluded multiple objects using occlusion activity detection and object association , 2004, Proceedings of 2004 International Symposium on Intelligent Signal Processing and Communication Systems, 2004. ISPACS 2004..