此应用程序的某些内容目前无法使用。
如果这种情况持续存在,请联系我们反馈与联系
1. (WO2019046077) PRIORITIZING OBJECTS FOR OBJECT RECOGNITION
注:相关文本通过自动光符识别流程生成。凡涉及法律问题,请以 PDF 版本为准

CLAIMS

WHAT IS CLAIMED IS:

1. An apparatus for prioritizing objects for object recognition in one or more video frames, comprising:

a memory configured to store video data associated with the one or more video frames; and

a processor configured to:

obtain a current video frame;

detect a plurality of objects in the current video frame;

determine state information associated with the plurality of objects; determine priorities for the plurality of objects, wherein a priority is determined for an object based on state information associated with the object; and

perform object recognition for one or more objects from the plurality of objects based on one or more priorities determined for the one or more objects, wherein object recognition is performed for objects having higher priorities before objects having lower priorities.

2. The apparatus of claim 1, wherein the processor is configured to:

obtain object identifiers, confidence scores, and history counter information for the plurality of objects, wherein history counter information for an object indicates at least a number of frames the object has had a current state; and

determine the state information for the plurality of objects based on the object identifiers, the confidence scores, and the history counter information for the plurality of objects.

3. The apparatus of claim 1, wherein the processor is configured to increase a priority for an object for the current video frame by a pre-defined amount in response to a state of the object being maintained in a current state for the current video frame.

4. The apparatus of claim 1, wherein the processor is configured to:

track the plurality of objects in the current video frame, wherein tracking includes matching an object in the current video frame with a previously detected object; and

update the state information associated with the plurality of objects based on results of the tracking.

5. The apparatus of claim 4, wherein an object detected in the current video frame that is not matched with any object detected in a previous video frame is assigned a new state, wherein objects assigned a new state are given a higher base priority than objects assigned to other states.

6. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a new state;

determine an object identifier for the object matches an object identifier of a previously detected obj ect that matches the obj ect in the current video frame;

determine a confidence score for the object is greater than a first confidence threshold and is less than a second confidence threshold, the first confidence threshold being less than the second confidence threshold;

determine a history counter for the object is less than a first counter threshold, the first counter threshold including a count of a number of video frames the confidence score for the object is greater than the first confidence threshold;

maintain a state of the object in the new state based on the confidence score for the object being greater than the first confidence threshold and less than the second confidence threshold and based on the history counter for the object being less than the first counter threshold; and increase a priority for the object by a pre-defined amount in response to the state of the object being maintained in the new state.

7. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a new state;

determine a confidence score for the object is greater than a second confidence threshold; change a state of the object from the new state to a confirmed state based on the confidence score for the object being greater than the second confidence threshold, the confirmed state indicating a confirmation that object recognition previously performed on the object was successful; and

change a priority for the object to a base priority associated with the confirmed state in response to changing the state of the object from the new state to the confirmed state, wherein the base priority associated with the confirmed state is less than a base priority associated with the new state.

8. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a new state;

determine an object identifier for the object matches an object identifier of a previously detected object that matches the object in the current video frame;

determine a confidence score for the object is greater than a first confidence threshold and is less than a second confidence threshold, the first confidence threshold being less than the second confidence threshold;

determine a history counter for the object is greater than a first counter threshold, the first counter threshold including a count of a number of video frames the confidence score for the object is greater than the first confidence threshold;

change a state of the object from the new state to a confirmed state based on the confidence score for the object being greater than the first confidence threshold and less than the second confidence threshold and based on the history counter for the object being greater than the first counter threshold; and

increase a priority for the object by a pre-defined amount in response to the state of the object being maintained in the new state.

9. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a new state;

determine an object identifier for the object does not match an object identifier of a previously detected object that matches the object in the current video frame;

change a state of the object from the new state to a non-confirmed state based on the object identifier for the object not matching the object identifier of the previously detected object; and

change a priority for the object to a base priority associated with the non-confirmed state in response to changing the state of the object from the new state to the non-confirmed state, wherein the base priority associated with the non-confirmed state is less than a base priority associated with the new state.

10. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a new state;

determine a confidence score for the object is less than a minimum confidence threshold, wherein the minimum confidence threshold is used to identify a stranger, and wherein a stranger is an object not registered in a database of known objects;

change a state of the object from the new state to a pre-stranger state based on the confidence score for the object being less than the minimum confidence threshold; and

change a priority for the object to a base priority associated with the pre-stranger state in response to changing the state of the object from the new state to the pre-stranger state, wherein the base priority associated with the pre-stranger state is less than a base priority associated with the new state.

11. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a new state;

determine a confidence score for the object is less than a minimum confidence threshold, wherein the minimum confidence threshold is used to identify a stranger, and wherein a stranger is an object not registered in a database of known objects;

determine a history counter for the object is greater than a minimum counter threshold, the minimum counter threshold including a count of a number of video frames the confidence score for the object is less than the minimum confidence threshold;

change a state of the object from the new state to a stranger state based on the confidence score for the object being less than the minimum confidence threshold and based on the history counter for the object being greater than the minimum counter threshold; and

change a priority for the object to a base priority associated with the stranger state in response to changing the state of the object from the new state to the stranger state, wherein the base priority associated with the stranger state is less than a base priority associated with the new state.

12. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a confirmed state, the confirmed state indicating a confirmation that object recognition previously performed on the obj ect was successful;

determine an object identifier for the object matches an object identifier of a previously detected object that matches the object in the current video frame;

maintain a state of the object in the confirmed state based on the object identifier for the object matching the object identifier of the previously detected object; and

increase a priority for the object by a pre-defined amount in response to the state of the object being maintained in the confirmed state.

13. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a confirmed state, the confirmed state indicating a confirmation that object recognition previously performed on the object was successful;

determine an object identifier for the object does not match an object identifier of a previously detected object that matches the object in the current video frame;

determine a confidence score for the object is greater than a minimum confidence threshold, wherein the minimum confidence threshold is used to identify a stranger, and wherein a stranger is an object not registered in a database of known objects;

change a state of the object from the confirmed state to a non-confirmed state based on the object identifier for the object not matching the object identifier of the previously detected object and based on the confidence score for the object being greater than the minimum confidence threshold; and

change a priority for the object to a base priority associated with the non-confirmed state in response to changing the state of the object from the confirmed state to the non-confirmed state, wherein the base priority associated with the non-confirmed state is higher than a base priority associated with the confirmed state.

14. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a confirmed state, the confirmed state indicating a confirmation that object recognition previously performed on the object was successful;

determine an object identifier for the object does not match an object identifier of a previously detected object that matches the object in the current video frame;

determine a confidence score for the object is less than a minimum confidence threshold, wherein the minimum confidence threshold is used to identify a stranger, and wherein a stranger is an object not registered in a database of known objects;

change a state of the object from the confirmed state to a pre-stranger state based on the object identifier for the object not matching the object identifier of the previously detected object and based on the confidence score for the object being less than the minimum confidence threshold; and

change a priority for the object to a base priority associated with the pre-stranger state in response to changing the state of the object from the confirmed state to the pre-stranger state, wherein the base priority associated with the pre-stranger state is greater than a base priority associated with the confirmed state.

15. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a non-confirmed state;

determine an object identifier for the object matches an object identifier of a previously detected object that matches the object in the current video frame;

determine a confidence score for the object is greater than a first confidence threshold and is less than a second confidence threshold, the first confidence threshold being less than the second confidence threshold;

change a state of the object from the non-confirmed state to a new state based on the object identifier for the object matching the object identifier of the previously detected object and based on the confidence score for the object being greater than the first confidence threshold and less than the second confidence threshold; and

change a priority for the object to a base priority associated with the new state in response to changing the state of the object from the non-confirmed state to the new state, wherein the base priority associated with the new state is greater than a base priority associated with the non-confirmed state.

16. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a non-confirmed state;

determine a confidence score for the object is greater than a second confidence threshold, wherein the second confidence threshold is used to confirm that object recognition previously performed on the object was successful;

change a state of the object from the non-confirmed state to a confirmed state based on the confidence score for the object being greater than the second confidence threshold; and

change a priority for the object to a base priority associated with the confirmed state in response to changing the state of the object from the non-confirmed state to the confirmed state, wherein the base priority associated with the confirmed state is less than a base priority associated with the non-confirmed state.

17. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a non-confirmed state;

determine an object identifier for the object does not match an object identifier of a previously detected object that matches the object in the current video frame;

maintain a state of the object in the non-confirmed state based on the object identifier for the object not matching the object identifier of the previously detected object; and

increase a priority for the object by a pre-defined amount in response to the state of the object being maintained in the non-confirmed state.

18. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a non-confirmed state;

determine a confidence score for the object is less than a minimum confidence threshold, wherein the minimum confidence threshold is used to identify a stranger, and wherein a stranger is an object not registered in a database of known objects;

change a state of the object from the non-confirmed state to a pre-stranger state based on the confidence score for the object being less than the minimum confidence threshold; and

change a priority for the object to a base priority associated with the pre-stranger state in response to changing the state of the object from the non-confirmed state to the pre-stranger state, wherein the base priority associated with the pre-stranger state is less than a base priority associated with the non-confirmed state.

19. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a pre-stranger state, wherein a stranger is an object not registered in a database of known objects;

determine an object identifier for the object matches an object identifier of a previously detected object that matches the object in the current video frame;

determine a confidence score for the object is greater than a first confidence threshold and is less than a second confidence threshold, the first confidence threshold being less than the second confidence threshold;

change a state of the object from the pre-stranger state to a new state based on the object identifier for the object matching the object identifier of the previously detected object and based on the confidence score for the object being greater than the first confidence threshold and less than the second confidence threshold; and

change a priority for the object to a base priority associated with the new state in response to changing the state of the object from the pre-stranger state to the new state, wherein the base priority associated with the new state is greater than a base priority associated with the pre-stranger state.

20. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a pre-stranger state, wherein a stranger is an object not registered in a database of known objects;

determine an object identifier for the object matches an object identifier of a previously detected object that matches the object in the current video frame;

determine a confidence score for the object is greater than a minimum confidence threshold and is less than a first confidence threshold, the minimum confidence threshold being less than the first confidence threshold;

change a state of the object from the pre-stranger state to a non-confirmed state based on the object identifier for the object matching the object identifier of the previously detected object and based on the confidence score for the object being greater than the minimum confidence threshold and less than the first confidence threshold; and

change a priority for the object to a base priority associated with the non-confirmed state in response to changing the state of the object from the pre-stranger state to the non-confirmed state, wherein the base priority associated with the non-confirmed state is greater than a base priority associated with the pre-stranger state.

21. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a pre-stranger state, wherein a stranger is an object not registered in a database of known objects;

determine an object identifier for the object matches an object identifier of a previously detected object that matches the object in the current video frame;

determine a confidence score for the object is less than a minimum confidence threshold, wherein the minimum confidence threshold is used to identify a stranger;

determine a history counter for the object is less than a minimum counter threshold, the minimum counter threshold including a count of a number of video frames the confidence score for the object is less than the minimum confidence threshold;

maintain a state of the object in the pre-stranger state based on the object identifier for the object matching the object identifier of the previously detected object, based on the confidence score for the object being less than the minimum confidence threshold, and based on the history counter for the object being less than the minimum counter threshold; and

increase a priority for the object by a pre-defined amount in response to the state of the object being maintained in the pre-stranger state.

22. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a pre-stranger state, wherein a stranger is an object not registered in a database of known objects;

determine an object identifier for the object matches an object identifier of a previously detected object that matches the object in the current video frame;

determine a confidence score for the object is less than a minimum confidence threshold, wherein the minimum confidence threshold is used to identify a stranger;

determine a history counter for the object is greater than a minimum counter threshold, the minimum counter threshold including a count of a number of video frames the confidence score for the object is less than the minimum confidence threshold;

change a state of the object from the pre-stranger state to a stranger state based on the object identifier for the object matching the object identifier of the previously detected object, based on the confidence score for the object being less than the minimum confidence threshold, and based on the history counter for the object being greater than the minimum counter threshold; and

change a priority for the object to a base priority associated with the stranger state in response to changing the state of the object from the pre-stranger state to the stranger state, wherein the base priority associated with the stranger state is less than a base priority associated with the pre-stranger state.

23. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a stranger state, wherein a stranger is an object not registered in a database of known objects;

determine an object identifier for the object matches an object identifier of a previously detected object that matches the object in the current video frame;

determine a confidence score for the object is greater than a first confidence threshold and is less than a second confidence threshold, the first confidence threshold being less than the second confidence threshold;

determine a history counter for the object is greater than a first counter threshold, the first counter threshold including a count of a number of video frames the confidence score for the object is greater than the first confidence threshold;

change a state of the object from the stranger state to a new state based on the object identifier for the object matching the object identifier of the previously detected object, based on the confidence score for the object being greater than the first confidence threshold and less than the second confidence threshold, and based on the history counter for the object being greater than the first counter threshold; and

change a priority for the object to a base priority associated with the new state in response to changing the state of the object from the stranger state to the new state, wherein the base priority associated with the new state is greater than a base priority associated with the stranger state.

24. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a stranger state, wherein a stranger is an object not registered in a database of known objects;

determine an object identifier for the object matches an object identifier of a previously detected object that matches the object in the current video frame;

determine a confidence score for the object is greater than a minimum confidence threshold and is less than a first confidence threshold, the minimum confidence threshold being less than the first confidence threshold;

change a state of the object from the stranger state to a pre-stranger state based on the object identifier for the object matching the object identifier of the previously detected object and based on the confidence score for the object being greater than the minimum confidence threshold and less than the first confidence threshold; and

change a priority for the object to a base priority associated with the pre-stranger state in response to changing the state of the object from the stranger state to the pre-stranger state, wherein the base priority associated with the pre-stranger state is greater than a base priority associated with the stranger state.

25. The apparatus of claim 1, wherein the processor is configured to:

determine an object detected in the current video frame has a stranger state, wherein a stranger is an object not registered in a database of known objects;

determine an object identifier for the object matches an object identifier of a previously detected object that matches the object in the current video frame;

determine a confidence score for the object is less than a minimum confidence threshold, wherein the minimum confidence threshold is used to identify a stranger; and

maintain a state of the object in the stranger state based on the object identifier for the object matching the object identifier of the previously detected object and based on the confidence score for the object being less than the minimum confidence threshold; and

increase a priority for the object by a pre-defined amount in response to the state of the object being maintained in the stranger state.

26. The apparatus of claim 1, wherein the plurality of objects include a plurality of faces, and wherein the object recognition includes face recognition.

27. The apparatus of claim 1, wherein the apparatus comprises a mobile device comprising one or more of a camera for capturing the one or more video frames and a display for displaying the one or more video frames.

28. A method of prioritizing objects for object recognition in one or more video frames, the method comprising:

obtaining a current video frame;

detecting a plurality of objects in the current video frame;

determining state information associated with the plurality of objects;

determining priorities for the plurality of objects, wherein a priority is determined for an object based on state information associated with the object; and

performing object recognition for one or more objects from the plurality of objects based on one or more priorities determined for the one or more objects, wherein object recognition is performed for objects having higher priorities before objects having lower priorities.

29. The method of claim 28, further comprising:

obtaining object identifiers, confidence scores, and history counter information for the plurality of objects, wherein history counter information for an object indicates at least a number of frames the object has had a current state; and

determining the state information for the plurality of objects based on the object identifiers, the confidence scores, and the history counter information for the plurality of objects.

30. The method of claim 28, further comprising increasing a priority for an object for the current video frame by a pre-defined amount in response to a state of the object being maintained in a current state for the current video frame.