System and method for object tracking and timing across multiple camera views

Patent metadata

Field	Value
Publication number	US-9641763-B2
Application number	US-201313973330-A
Country	US
Kind code	B2
Filing date	Aug 22, 2013
Priority date	Aug 29, 2012
Publication date	May 2, 2017
Grant date	May 2, 2017

Abstract

A system and method for object tracking and timing across multiple camera views includes local and global tracking modules for tracking the location of objects as they traverse particular regions of interest within an area of interest. A local timing module measures the time spent with each object within the area captured by a camera. A global timing module measures the time taken by the tracked object to traverse the entire area of interest or the length of the stay of the object within the area of interest.

First claim

What is claimed is:

1. A spatio-temporal object tracking system for identifying an object traversing a plurality of video sensor views, including:
a video acquisition system for sensing a plurality of local views of a global area of interest;
an object detection module for detecting an object entering one of the local views;
a local object tracking module for tracking a motion of the object through the one local view comprising:
assigning an indicator of an at least one invariant representative feature of the object through the global area of interest wherein the at least one invariant representative feature is selected from the group consisting of: license plates, vehicle color, vehicle make, vehicle model, Harris Corners, Scale Invariant Feature Transform (SIFT), Local Binary Patterns (LBP), and Speeded-up Robust Features (SURF);
extracting and maintaining each invariant representative feature in a full dataset of interest points per object;

a global object tracking module for tracking the object through the plurality of local views and the global area of interest;
a local timing module for measuring time data of the object in the one local view; and,
a global timing module for measuring time data of the object through the plurality of local views and a length of stay of the object through the global area of interest by synchronization of data across the plurality of video sensor views to achieve timing data normalization across the plurality of video sensor views.

2. The system of claim 1 , wherein the video acquisition system comprises a plurality of video cameras arrayed in a layout relative to the global area of interest.

3. The system of claim 1 , wherein the video acquisition system comprises a video camera with a Pan-Tilt-Zoom (PTZ) imaging system for acquiring the plurality of local views.

4. The system of claim 1 , wherein the local tracking module includes a processor for associating a local tracking indicator with the object in the one local view.

5. The system of claim 1 , wherein the global tracking module includes a processor for at least one of (1) re-identifying the object identified through local object tracking in overlapping multiple local views, and (2) object representative feature tracking of the at least one invariant representative feature through the plurality of local views.

6. The system of claim 1 , wherein the local timing module includes a processor for detecting time data of the object relative to a local region of interest in the one local view.

7. The system of claim 1 , wherein the global timing module includes a processor for detecting time data of the object relative to the global area of interest including local regions of interest therein.

8. The system of claim 7 , wherein the detecting time data by the processor includes receiving an output from the local timing and global tracking modules and normalizing time data across the plurality of local views based on a synchronization of the video acquisition system.

9. A method of identifying an object traversing a plurality of video sensor views, comprising:
acquiring video from a global area of interest from a plurality of local views;
detecting when a moving object enters the plurality of local views;
locally tracking the moving object across a first local view of the plurality of local views with a local tracking indicator comprising assigning an indicator of an at least one invariant representative feature of the object through the global area of interest wherein the at least one invariant representative feature is selected from the group consisting of: license plates, vehicle color, vehicle make, vehicle model, Scale Invariant Feature Transform (SIFT), Local Binary Patterns (LBP), and Speeded-up Robust Features (SURF);
extracting and maintaining each invariant representative feature in a full dataset of interest points per object; and
globally tracking the moving object across a plurality of local views by re-identifying the object identified through the local tracking through the plurality of local views including measuring time data of the object through the plurality of local views and a length of stay of the object through the global area of interest by synchronization of time data across the plurality of video sensor views to achieve timing data normalization across the plurality of video sensor views.

10. The method of claim 9 , wherein the globally tracking further includes tracking the at least one invariant representative feature through the plurality of local views.

11. The method of claim 10 , wherein the object invariant representative feature is invariant to the multiple viewing conditions across the video acquisition system.

12. The method of claim 9 , further including performing local time measurement of the moving object in the first local view of the plurality of local views, wherein the local time measurement includes obtaining time data with respect to a region of interest.

13. The method of claim 9 , wherein the global time measurement includes receiving local timing data and global tracking data to enable normalization of time data across the plurality of local views.

14. The method of claim 9 , further including obtaining sub-global time measurement data across portions of the area of interest by sharing data obtained from the video acquired across the plurality of local views.

15. The method of claim 12 , wherein the local time measurement is used to monitor the performance of a plurality of points of interest by measuring the time the moving object spends at each of the plurality of points of interest.

16. The method of claim 9 , wherein the locally tracking the moving object is performed according to at least one of a mean-shift tracking algorithm, a point feature-based tracking algorithm, a global feature-based tracking algorithm, a template matching algorithm, a silhouette/contour tracking algorithm, a Kalman Filter based tracking algorithm and a particle filter tracker.

Assignees

• Xerox Corp
• Conduent Business Services Llc

Inventors

• Bernal Edgar A
• Loce Robert P
• Wade Thomas F
• Paul Peter

Classifications

• H04N23/69 Primary

  Control of means for changing angle of the field of view, e.g. optical zoom objectives or electronic zooming · CPC title

• G08B13/19608

  Tracking movement of a target, e.g. by detecting an object predefined as a target, using target direction and or velocity to predict its new position · CPC title

• G08B13/19645 Primary

  Multiple cameras, each having view on one of a plurality of scenes, e.g. multiple cameras for multi-room surveillance or for tracking an object by view hand-over · CPC title

• H04N5/23296 Primary

  Electricity · mapped topic