Conference Paper

Object Tracking in the Presence of Occlusions via a Camera Network

Stanford Univ. Stanford, Stanford
DOI: 10.1109/IPSN.2007.4379711 Conference: Information Processing in Sensor Networks, 2007. IPSN 2007. 6th International Symposium on
Source: IEEE Xplore

ABSTRACT This paper describes a sensor network approach to tracking a single object in the presence of static and moving occluders using a network of cameras. To conserve communication bandwidth and energy, each camera first performs simple local processing to reduce each frame to a scan line. This information is then sent to a cluster head to track a point object. We assume the locations of the static occluders to be known, but only prior statistics on the positions of the moving occluders are available. A noisy perspective camera measurement model is presented, where occlusions are captured through an occlusion indicator function. An auxiliary particle filter that incorporates the occluder information is used to track the object. Using simulations, we investigate (i) the dependency of the tracker performance on the accuracy of the moving occluder priors, (ii) the tradeoff between the number of cameras and the occluder prior accuracy required to achieve a prescribed tracker performance, and (iii) the importance of having occluder priors to the tracker performance as the number of occluders increases. We generally find that computing moving occluder priors may not be worthwhile, unless it can be obtained cheaply and to a reasonable accuracy. Preliminary experimental results are provided.

0 Bookmarks
 · 
80 Views
  • [Show abstract] [Hide abstract]
    ABSTRACT: Wireless Video Sensor Networks (WVSNs) provide opportunities to use large number of low-cost low-resolution wireless camera sensors for large-scale outdoor remote surveillance missions. Camera sensor deployment is crucial in achieving good coverage, accuracy and fault tolerance. In particular, with the decreased costs of wireless cameras, redundant camera deployment is attractive in order to get multiple disparate views of events for improved event identification. If the capturing of an event spans ${ 360^circ}$ , this is referred to as angular coverage. In this paper, we consider the problem of determining optimal camera placement to achieve angular coverage continuously over a given region. We develop a bi-level algorithm to find the minimum-cost camera placement. In the first level, we run a master problem that identifies the camera placement points to achieve angular coverage of a discrete set of points selected from the region of interest. Next, we use a sub-problem to identify points in the continuous region that are not covered by the cameras placed in the previous run of the master problem. We then add these uncovered points to discrete point set of the master problem and re-run the master problem. We continue running the master and sub-problems iteratively until the sub-problem becomes infeasible indicating that the entire region is covered. In the numerical experiments, we consider two cases 1) placement of homogeneous cameras with fixed resolutions; and 2) placement of heterogeneous cameras with different characteristics and resolutions. We also introduce varying resolution requirements for different parts of the region and place the cameras such that the required resolution is satisfied. The numerical results show the superiority of the bi-level approach respect to existing approaches.
    IEEE Transactions on Computers 07/2014; 63(7):1812-1825. DOI:10.1109/TC.2013.45 · 1.47 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Visual tracking is the problem of using visual sensor measurements to determine location and path of target object. One of big challenges for visual tracking is full occlusion. When full occlusions are present, image data alone can be unreliable, and is not sufficient to detect the target object. The developed tracking algorithm is based on bootstrap particle filter and using color feature target. Furthermore the algorithm is modified using nonretinotopic concept, inspired from the way of human visual cortex handles occlusion by constructing nonretinotopic layers. We interpreted the concept by using past tracking memory about motion dynamics rather than current measurement when quality level of tracking reliability below a threshold. Using experiments, we found (i) the performance of the object tracking algorithm in handling occlusion can be improved using nonretinotopic concept, (ii) dynamic model is crucial for object tracking, especially when the target object experienced occlusion and maneuver motions, (iii) the dependency of the tracker performance on the accuracy of tracking quality threshold when facing illumination challenge. Preliminary experimental results are provided.
  • [Show abstract] [Hide abstract]
    ABSTRACT: This article describes a sensor network approach to tracking a single object in the presence of static and moving occluders using a network of cameras. To conserve communication bandwidth and energy, we combine a task-driven approach with camera subset selection. In the task-driven approach, each camera first performs simple local processing to detect the horizontal position of the object in the image. This information is then sent to a cluster head to track the object. We assume the locations of the static occluders to be known, but only prior statistics on the positions of the moving occluders are available. A noisy perspective camera measurement model is introduced, where occlusions are captured through occlusion indicator functions. An auxiliary particle filter that incorporates the occluder information is used to track the object. The camera subset selection algorithm uses the minimum mean square error of the best linear estimate of the object position as a metric, and tracking is performed using only the selected subset of cameras. Using simulations and preselected subsets of cameras, we investigate (i) the dependency of the tracker performance on the accuracy of the moving occluder priors, (ii) the trade-off between the number of cameras and the occluder prior accuracy required to achieve a prescribed tracker performance, and (iii) the importance of having occluder priors to the tracker performance as the number of occluders increases. We find that computing moving occluder priors may not be worthwhile, unless it can be obtained cheaply and to high accuracy. We also investigate the effect of dynamically selecting the subset of camera nodes used in tracking on the tracking performance. We show through simulations that a greedy selection algorithm performs close to the brute-force method and outperforms other heuristics, and the performance achieved by greedily selecting a small fraction of the cameras is close to that of using all the cameras.
    ACM Transactions on Sensor Networks 03/2013; 9(2). DOI:10.1145/2422966.2422973 · 1.46 Impact Factor

Full-text (2 Sources)

Download
15 Downloads
Available from
Aug 18, 2014