REDUCING OBJECT DETECTION TIME BY UTILIZING SPACE LOCALIZATION OF FEATURES

• US 20140270344A1
• Filed: 03/12/2013
• Published: 09/18/2014
• Est. Priority Date: 03/12/2013
• Status: Active Grant

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: VRCELJ, Bojan, McGRATH, Gary G., KRISHNAMOORTHI, Raghuraman
• Granted Patent: US 9,202,129 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06V 10/40 Extraction of image or video features
• CPC Class Codes: G06V 10/757 Matching configurations of points or features
• CPC Class Codes: G06V 30/2504 Coarse or fine approaches, e.g. resolution of ambiguities or multiscale approaches

Abstract

In one example, a method for exiting an object detection pipeline includes determining, while in the object detection pipeline, a number of features within a first tile of an image, wherein the image consists of a plurality of tiles, performing a matching procedure using at least a subset of the features within the first tile if the number of features within the first tile meets a threshold value, exiting the object detection pipeline if a result of the matching procedure indicates an object is recognized in the image, and presenting the result of the matching procedure.

Claims

1. A method of exiting an object detection pipeline, the method comprising:

determining, while in the object detection pipeline, a number of features within a first tile of an image, wherein the image is divisible into a plurality of tiles;

performing a matching procedure using at least a subset of the features within the first tile if the number of features within the first tile meets a threshold value;

exiting the object detection pipeline if a result of the matching procedure indicates an object is recognized in the image; and

presenting the result of the matching procedure.

2. The method of claim 1 further comprising extracting features from a second tile within the image if the number of features in the first tile does not meet the threshold value.

3. The method of claim 2 wherein a location of a second tile within the image is based on a vertical raster pattern.

4. The method of claim 2 wherein a location of the second tile within the image is based on a predetermined search pattern.

5. The method of claim 2 wherein locations of the first and second tiles within the image are based on an image analysis performed on a downsampled version of the image.

6. The method of claim 2 further comprising:

extracting features from an N+1 tile within the image if a cumulative number of features in the N tiles does not meet a threshold value;

performing the matching procedure using the features extracted from the N+1 tile and the cumulative features extracted from the N tiles if the number of cumulative features from the N tiles and the N+1 tile meets a threshold value; and

exiting the object detection pipeline if a result of the matching procedure using the cumulative features from the N tiles and the N+1 tile indicates an object is recognized in the image.

7. The method of claim 1 further comprising extracting features from a second tile within the image if the results of the matching procedure do not indicate that an object is recognized.

8. The method of claim 1 further comprising determining the pose of a recognized object, and extracting features from a second tile within the image if the pose of the recognized object is not geometrically consistent with information in an objects database.

9. The method of claim 1 wherein the image is downsampled before the features are extracted from the first tile.

10. The method of claim 1, wherein performing the matching procedure comprises transmitting the subset of features to a remote device, and receiving the result from the remote device.

11. The method of claim 1 comprising:

determining location statistics based on the extracted features of the first tile;

detecting the extracted features on a lower octave version of the image based on the location statistics; and

calculating a pose of the object based on the features detected on the lower octave version of the image.

12. An apparatus configured to implement an object detection pipeline, the apparatus comprising:

a memory unit;

a processor coupled to the memory unit and configured to;

determine, while in the object detection pipeline, a number of features within a first tile of an image, wherein the image is divisible into a plurality of tiles;

perform a matching procedure using at least a subset of the features within the first tile if the number of features within the first tile meets a threshold value;

exit the object detection pipeline if a result of the matching procedure indicates an object is recognized in the image; and

store the result of the matching procedure in the memory unit.

13. The apparatus of claim 12 wherein the processor is further configured to extract features from a second tile within the image if the number of features in the first tile does not meet the threshold value.

14. The apparatus of claim 13 wherein the processor is configured to determine a location of a second tile within the image based on a vertical raster pattern.

15. The apparatus of claim 13 wherein the processor is configured to determine a location of the second tile within the image based on a predetermined search pattern.

16. The apparatus of claim 13 wherein the processor is configured to create a downsampled version of the image and determine locations of the first and second tiles within the image based on an image analysis performed on the downsampled version of the image.

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

extract features from an N+1 tile within the image if a cumulative number of features in the N tiles does not meet a threshold value;

perform the matching procedure using the features extracted from the N+1 tile and the cumulative features extracted from the N tiles if the number of cumulative features from the N tiles and the N+1 tile meets a threshold value; and

exit the object detection pipeline if a result of the matching procedure using the cumulative features from the N tiles and the N+1 tile indicates an object is recognized in the image.

18. The apparatus of claim 12 wherein the processor is configured to extract features from a second tile within the image if the results of the matching procedure do not indicate that an object is recognized.

19. The apparatus of claim 12 wherein the processor is configured to determine the pose of a recognized object, and extract features from a second tile within the image if the pose of the recognized object is not geometrically consistent with information in an objects database.

20. The apparatus of claim 12 wherein the processor is configured to downsample the image before the features are extracted from the first tile.

21. The apparatus of claim 12 wherein the processor is configured to perform the matching procedure by transmitting the subset of features to a remote device, and receiving the result from the remote device.

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

determine location statistics based on the extracted features of the first tile;

detect the extracted features on a lower octave version of the image based on the location statistics; and

calculate a pose of the object based on the features detected on the lower octave version of the image.

23. The apparatus of claim 12 further comprising a camera configured to capture the image and provide the image to the processor.

24. An apparatus for exiting an object detection pipeline, the apparatus comprising:

means for determining, while in the object detection pipeline, a number of features within a first tile of an image, wherein the image is divisible into a plurality of tiles;

means for performing a matching procedure using at least a subset of the features within the first tile if the number of features within the first tile meets a threshold value;

means for exiting the object detection pipeline if a result of the matching procedure indicates an object is recognized in the image; and

means for storing the result of the matching procedure.

25. The apparatus of claim 24 further comprising means for extracting features from a second tile within the image if the number of features in the first tile does not meet the threshold value.

26. The apparatus of claim 25 further comprising means for determining a location of a second tile within the image based on a vertical raster pattern.

27. The apparatus of claim 25 further comprising means for determining a location of the second tile within the image based on a predetermined search pattern.

28. The apparatus of claim 25 further comprising means for creating a downsampled version of the image and means for determining locations of the first and second tiles within the image based on an image analysis performed on the downsampled version of the image.

29. The apparatus of claim 25 further comprising:

means for extracting features from an N+1 tile within the image if a cumulative number of features in the N tiles does not meet a threshold value;

means for performing the matching procedure using the features extracted from the N+1 tile and the cumulative features extracted from the N tiles if the number of cumulative features from the N tiles and the N+1 tile meets a threshold value; and

means for exiting the object detection pipeline if a result of the matching procedure using the cumulative features from the N tiles and the N+1 tile indicates an object is recognized in the image.

30. The apparatus of claim 24 further comprising means for extracting features from a second tile within the image if the results of the matching procedure do not indicate that an object is recognized.

31. The apparatus of claim 24 further comprising means for determining the pose of a recognized object, and means for extracting features from a second tile within the image if the pose of the recognized object is not geometrically consistent with information in an objects database.

32. The apparatus of claim 24 comprising means for downsampling the image before the features are extracted from the first tile.

33. The apparatus of claim 24 wherein the means for performing the matching procedure comprises means for transmitting the subset of features to a remote device, and means for receiving the result from the remote device.

34. The apparatus of claim 24 comprising:

means for determining location statistics based on the extracted features of the first tile;

means for detecting the extracted features on a lower octave version of the image based on the location statistics; and

means for calculating a pose of the object based on the features detected on the lower octave version of the image.

35. The apparatus of claim 24 comprising means for capturing the image and providing the image to the processor.

36. A computer-readable medium having stored thereon instructions that, when executed, cause a processor to:

determine, while in an object detection pipeline, a number of features within a first tile of an image, wherein the image is divisible into a plurality of tiles;

perform a matching procedure using at least a subset of the features within the first tile if the number of features within the first tile meets a threshold value;

exit the object detection pipeline if a result of the matching procedure indicates an object is recognized in the image; and

store the result of the matching procedure.

37. The computer program product of claim 36 further comprising instructions to cause the processor to extract features from a second tile within the image if the number of features in the first tile does not meet the threshold value.

38. The computer program product of claim 36 further comprising instructions to cause the processor to extract features from a second tile within the image if the results of the matching procedure do not indicate that an object is recognized.