METHOD AND SYSTEM FOR AUTOMATING AN IMAGE REJECTION PROCESS

• US 20160148076A1
• Filed: 12/05/2014
• Published: 05/26/2016
• Est. Priority Date: 11/26/2014
• Status: Active Grant

• Current Assignee: Conduent Business Services, LLC (Conduent, Inc.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Kozitsky, Vladimir, Shreve, Matthew, Burry, Aaron M.
• Granted Patent: US 9,460,367 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 18/2413 based on distances to training or reference patterns
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/20 Ensemble learning
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69
• CPC Class Codes: G06V 20/62 Text, e.g. of license plates, overlay texts or captions on TV images
• CPC Class Codes: G06V 20/625 License plates
• CPC Class Codes: G08G 1/0175 by photographing vehicles, e.g. when violating traffic rules

Abstract

Systems and methods for automating an image rejection process. Features including texture, spatial structure, and image quality characteristics can be extracted from one or more images to train a classifier. Features can be calculated with respect to a test image for submission of the features to the classifier, given an operating point corresponding to a desired false positive rate. One or more inputs can be generated from the classifier as a confidence value corresponding to a likelihood of, for example: a license plate being absent in the image, the license plate being unreadable, or the license plate being obstructed. The confidence value can be compared against a threshold to determine if the image(s) should be removed from a human review pipeline, thereby reducing images requiring human review.

Claims

1. A method for automating an image rejection process, said method comprising:

extracting features from at least one image among a batch of images to train a classifier, said features comprising texture, spatial structure, and image quality characteristics;

calculating said features with respect to a test image for submission of said features to said classifier, given an operating point corresponding to a desired false positive rate;

generating at least one output from said classifier as a confidence value corresponding to a likelihood of at least one of the following;

a license plate being absent in said image, said license plate being unreadable, or said license plate being obstructed; and

comparing said confidence value against a threshold to determine if said at least one image should be removed from a human review pipeline, thereby reducing images requiring human review.

2. The method of claim 1 wherein said classifier comprises a random forest based classifier.

3. The method of claim 1 further comprising initially capturing said at least one image via an image-capturing unit.

4. The method of claim 2 further comprising processing said at least one image using an LPR engine after capturing said at least one image via said image-capturing unit.

5. The method of claim 1 further comprising;

sweeping an image-based classifier across said at least one image to identify a center of highest confidence ROI with respect to said image; and

computing a brightness, a contrast, and a character frequency for an ROI in said image.

6. The method of claim 5 wherein said image-based classifier comprises a SNoW classifier.

7. The method of claim 1 further wherein extracting features from at least one image among a batch of images to train a classifier, said features comprising texture, spatial structure, and image quality characteristics, further comprises:

splitting said image into an m×

n grid of blocks; and

computing LBP features for block among said m×

n grid of blocks to concatenate said LBP features.

8. The method of claim 2 further comprising;

sweeping an image-based classifier across said at least one image to identify a center of highest confidence ROI with respect to said image; and

computing a brightness, a contrast, and a character frequency for an ROI in said image.

9. The method of claim 2 further wherein extracting features from at least one image among a batch of images to train a classifier, said features comprising texture, spatial structure, and image quality characteristics, further comprises:

splitting said image into an m×

n grid of blocks; and

computing LBP features for block among said m×

n grid of blocks to concatenate said LBP features.

10. A system for automating an image rejection process, said system comprising:

at least one processor; and

memory comprising instructions stored therein, which when executed by said at least one processor, cause said at least one processor to perform operations comprising;

extracting features from at least one image among a batch of images, said features to train a classifier, said features comprising texture, spatial structure, and image quality characteristics;

calculating said features with respect to a test image for submission of said features to said classifier, given an operating point corresponding to a desired false positive rate;

generating at least one output from said classifier as a confidence value corresponding to a likelihood of at least one of the following;

a license plate being absent in said at least one image, said license plate being unreadable, or said license plate being obstructed; and

comparing said confidence value against a threshold to determine if said at least one image should be removed from a human review pipeline, thereby reducing images requiring human review.

11. The system of claim 10 wherein said classifier comprises a random forest based classifier.

12. The system of claim 10 further comprising an image-capturing unit that communicates electronically with said at least one processor and said memory, wherein said operations further comprise initially capturing said at least one image via said image-capturing unit.

13. The system of claim 12 wherein said at least one processor processes said at least one image using an LPR engine after capturing said at least one image via said image-capturing unit.

14. The system of claim 10 further comprising:

an image-based classifier that sweeps across said at least one image to identify a center of highest confidence ROI with respect to said at least one image; and

wherein said operations further comprise computing a brightness, a contrast, and a character frequency for an ROI in said at least one image.

15. The system of claim 14 wherein said image-based classifier comprises a SNoW classifier.

16. The system of claim 10 wherein extracting said features comprising texture, spatial structure, and Image quality characteristics from said at least one image, further comprises:

splitting said at least one image into an m×

n grid of blocks; and

computing LBP features for block among said m×

n grid of blocks to concatenate said LBP features.

20. The system of claim 14 wherein said image-based classifier comprises a SNoW classifier and wherein extracting said features comprising texture, spatial structure, and image quality characteristics from said at least one image, further comprises:

splitting said at least one image into an m×

n grid of blocks; and

computing LBP features for block among said m×

n grid of blocks to concatenate said LBP features.

17. A system for automating an image rejection process, said system comprising:

at least one image-capturing unit;

at least one processor that communicates electronically with said at least one image-capturing unit; and

memory comprising instructions stored therein, which when executed by said at least one processor, cause said at least one processor to perform operations comprising;

extracting features from at least one image among a batch of images to train a classifier, said features comprising texture, spatial structure, and image quality characteristics, wherein said at least one image is captured by said at least one image-capturing unit;

calculating said features with respect to a test image for submission of said features to said classifier, given an operating point corresponding to a desired false positive rate;

generating at least one output from said classifier as a confidence value corresponding to a likelihood of at least one of the following;

a license plate being absent in said at least one image, said license plate being unreadable, or said license plate being obstructed; and

comparing said confidence value against a threshold to determine if said at least one image should be removed from a human review pipeline, thereby reducing images requiring human review.

18. The system of claim 17 wherein said classifier comprises a random forest based classifier and wherein said at least one processor processes said at least one image using an LPR engine after capturing said at least one image via said image-capturing unit.

19. The system of claim 17 further comprising:

an image-based classifier that sweeps across said at least one image to identify a center of highest confidence ROI with respect to said at least one image; and

wherein said operations further comprise computing a brightness, a contrast, and a character frequency for an ROI in said at least one image.