Motion and Appearance Contexts for Tracking and Re-Acquiring Targets in Aerial Videos

In this paper, we use motion and appearance contexts for persistent tracking of objects in aerial imagery. The motion context in a given environment is a collection of trajectories of objects which are representative of the motion of the occluded or unobserved object. It is learned using a clustering scheme based on the Lyapunov characteristic exponent (LCE) which measures the mean exponential rate of divergence of the nearby trajectories. The learned motion context is then used in a regression framework to predict the location of the unobserved object. The appearance context of an occluded (target) object consists of appearance information of objects which are currently occluded or unobserved. It is incorporated by learning a distribution of interclass variation for each target-unobservable object pair. In addition, intra-class variation distribution is constructed for each occluded object using all of its previous observations. Qualitative and quantitative results are reported on challenging aerial sequences.

[1]  C. Froeschlé,et al.  The Lyapunov characteristic exponents-applications to celestial mechanics , 1984 .

[2]  A. Wolf,et al.  Determining Lyapunov exponents from a time series , 1985 .

[3]  Steve Mann,et al.  Video orbits of the projective group a simple approach to featureless estimation of parameters , 1997, IEEE Trans. Image Process..

[4]  Stuart J. Russell,et al.  Object Identification: A Bayesian Analysis with Application to Traffic Surveillance , 1998, Artif. Intell..

[5]  Ramin Zabih,et al.  Bayesian multi-camera surveillance , 1999, Proceedings. 1999 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Cat. No PR00149).

[6]  Leonid I. Piterbarg,et al.  Short-Term Prediction of Lagrangian Trajectories , 2001 .

[7]  Leonid I. Piterbarg,et al.  Predictability of Drifter Trajectories in the Tropical Pacific Ocean , 2001 .

[8]  Leonid I. Piterbarg,et al.  A Simple Prediction Algorithm for the Lagrangian Motion in Two-Dimensional Turbulent Flows , 2002, SIAM J. Appl. Math..

[9]  Mubarak Shah,et al.  Tracking across multiple cameras with disjoint views , 2003, Proceedings Ninth IEEE International Conference on Computer Vision.

[10]  Harpreet S. Sawhney,et al.  Vehicle identification between non-overlapping cameras without direct feature matching , 2005, Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1.

[11]  Ying Shan,et al.  Unsupervised learning of discriminative edge measures for vehicle matching between non-overlapping cameras , 2005, 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05).

[12]  Harpreet S. Sawhney,et al.  Vehicle fingerprinting for reacquisition & tracking in videos , 2005, 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05).

[13]  A. G. Amitha Perera,et al.  A unified framework for tracking through occlusions and across sensor gaps , 2005, 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05).

[14]  Andrew Gilbert,et al.  Tracking Objects Across Cameras by Incrementally Learning Inter-camera Colour Calibration and Patterns of Activity , 2006, ECCV.

[15]  Joseph L. Mundy,et al.  Augmenting Shape with Appearance in Vehicle Category Recognition , 2006, 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06).

[16]  A. G. Amitha Perera,et al.  Multi-Object Tracking Through Simultaneous Long Occlusions and Split-Merge Conditions , 2006, 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06).

[17]  Mubarak Shah,et al.  COCOA: tracking in aerial imagery , 2006, SPIE Defense + Commercial Sensing.

[18]  H. Sawhney,et al.  Unsupervised Learning of Discriminative Edge Measures for Vehicle Matching between Nonoverlapping Cameras , 2008, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[19]  Mubarak Shah,et al.  Tracking in Multiple Cameras with Disjoint Views , 2008 .