Video object tracking in traffic monitoring

1. A method for object tracking in a video sequence comprising a plurality of video frames, the method comprising:

selecting a region of interest (ROI) in a current one of the video frames for tracking an object, the ROI including at least a location and a size associated with the object;

adjusting dynamically said ROI based on an estimated change determined as a result of tracking changes occurring to account for size changes of said object over a given time interval;

providing a booster tracker to learn about features of said object for tracking purposes;

determining a first new location for the ROI based on on-line updates of a plurality of features for tracking the object;

preventing tracking error carryover in subsequent ones of the plurality of video frames by establishing a permissible zone in each of the plurality of video frames calculated from a predicted velocity of the object such that a range of permissible locations for the object are constrained by the permissible zone;

predicting a second new location for the object by applying a filter to the ROI from the current video frame, if the determined first new location is outside the permissible zone;

recording the determined first new location, if the determined first new location is within said permissible zone; and

adjusting the size of the ROI based on an estimated change in a scale of the object, wherein the adjusted ROI is used for tracking the object in a given one of the subsequent video frames.

2. The method of claim 1 , wherein said filter is a Kalman filter.

3. The method of claim 1 , wherein the size of the ROI is dynamically adapted over a given time interval according to the estimated change in the scale of the object by looking at previous frames during subsequent frames and through user input or object detection during initial frames.

4. The method of claim 1 , wherein the estimated change in the scale of the object is determined from a history of tracking results previously obtained for the object.

5. The method of claim 1 , wherein the estimated change in the scale of the object is determined without any knowledge of camera parameters corresponding to the sequence.

6. The method of claim 1 , wherein the range of permissible locations is checked on a per-frame basis to prevent tracking error carryover in the subsequent ones of the video frames.

7. A method for tracking an object in a video sequence formed from a plurality of video frames, the method comprising:

selecting a region of interest (ROI) in a current video frame for tracking the object, the ROI including at least a location and a size associated with the object;

adjusting said ROI based on an estimated change determined as a result of tracking changes occurring to account for size changes of said object over a given time interval;

providing a booster tracker to learn about features of said object for tracking purposes;

determining a new location of the object based on on-line updates of object features;

preventing tracking error carryover in subsequent video frames by establishing a permissible zone in each of the plurality of video frames calculated from a predicted velocity of the object such that a range of permissible locations for the object are constrained by the permissible zone;

predicting a second new location for the object by applying a filter to the ROI from the current video frame, if the determined first new location is outside the permissible zone;

recording the determined first new location, if the determined first new location is within the range of permissible locations of said permissible zone calculated from the predicted velocity of the object; and

re-estimating the new location to obtain a re-estimated location of the object when the new location of the object is determined to be outside of the permissible zone,

wherein the new location is used when the new location of the object is determined to be within the range of permissible locations or the re-estimated location of the object is used when the new location of the object is determined to be outside of the range of permissible locations, to track the object in a subsequent one of the plurality of video frames.

8. The method of claim 7 , wherein a history of tracking results for the object is used to re-estimate the new location of the object.

9. The method of claim 7 , wherein the estimated change in a scale of the object is determined from a history of tracking results previously obtained for the object.

10. A system for object tracking in a video sequence comprising a plurality of video frames, comprising:

a first processor configured for

selecting a region of interest (ROI) in a current one of the plurality of video frames for tracking an object, the ROI including at least a location and a size associated with the object,

determining a first new location for the ROI based on on-line updates of a plurality of features for tracking the object, and

preventing tracking error carryover in subsequent video frames by establishing a permissible zone in each of the plurality of video frames calculated from a predicted velocity of the object such that a range of permissible locations for the object are constrained by the permissible zone;

predicting a second new location for the object by applying a filter to the ROI from the current video frame, if the determined first new location is outside the permissible zone;

a memory for recording the determined first new location, if the determined first new location is within any permissible zone in each video sequence calculated from the predicted velocity of the object;

the first processor further configured for adjusting the size of the ROI based on an estimated change in a scale of the object, wherein the adjusted ROI is used for tracking the object in a subsequent one of the plurality of video frames; and

a booster tracker for adjusting said ROI based on an estimated change determined as a result of tracking changes occurring to account for size changes of said object over a given time interval and for learning about features of said object for tracking purposes.

11. The system of claim 10 , further comprising a second processor for predicting a second new location for the video object, if the determined first new location is outside the range of permissible locations.

12. The system of claim 10 , wherein the first processor is configured for dynamically adapting the size of the ROI over a given time interval according to the estimated change in the scale of the object.