Method and system for OCR-free vehicle identification number localization

• US 9,965,677 B2
• Filed: 12/09/2014
• Issued: 05/08/2018
• Est. Priority Date: 12/09/2014
• Status: Active Grant

• Current Assignee: Conduent Business Services, LLC (Conduent, Inc.)
• Original Assignee: Conduent Business Services, LLC (Conduent, Inc.)
• Inventors: Bulan, Orhan, Mizes, Howard, Kozitsky, Vladimir, Burry, Aaron M.
• Primary Examiner(s): Le, Vu
• Assistant Examiner(s): Mangialaschi, Tracy
• Application Number:
  US14/564,347
• Publication Number:
  US 20160162761A1
• Time in Patent Office: 1,246 Days
• Field of Search: None
• 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: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables

Abstract

Methods and systems for localizing numbers and characters in captured images. A side image of a vehicle captured by one or more cameras can be preprocessed to determine a region of interest. A confidence value of series of windows within regions of interest of different sizes and aspect ratios containing a structure of interest can be calculated. Highest confidence candidate regions can then be identified with respect to the regions of interest and at least one region adjacent to the highest confidence candidate regions. An OCR operation can then be performed in the adjacent region. An identifier can then be returned from the adjacent region in order to localize numbers and characters in the side image of the vehicle.

Claims

1. A method for localizing numbers and characters in captured images, said method comprising:

training a machine learning classifier in an offline training phase;

automatically preprocessing a side image of a vehicle digitally captured by at least one camera to determine at least one region of interest in said side image, said at least one region of interest among regions of interest, said preprocessing comprising computer vision including median filtering to remove salt and pepper noise from said side image;

determining by said machine learning classifier a confidence of a series of windows within said regions of interest of said side image, said regions of interest comprising regions of interest of different sizes and aspect ratios, said regions of interest containing a structure of interest in said side image, said confidence comprising a measure of certainty and said structure of interest comprising text including numbers and/or an image of a physical structure of at least one object contained in said side image;

identifying highest confidence candidate regions among said regions of interest that have said structure of interest and identifying at least one region adjacent to said highest confidence candidate regions;

performing an optical character recognition in said at least one adjacent region; and

returning an identifier from said at least one adjacent region in order to localize numbers and characters in said side image of said vehicle.

2. The method of claim 1 wherein said highest confidence candidate regions are identified with nonmaximal suppression.

3. The method of claim 1 wherein a window size of said at least one adjacent region is determined by a window size of at least one candidate region among said highest confidence candidate regions.

4. The method of claim 1 wherein said confidence is automatically determined with said classifier, said classifier comprising a KNN (k-nearest neighbor) classifier, and wherein said classifier is trained in said offline training phase based on extracted features that include a Fisher Vector.

5. The method of claim 4 wherein a size and aspect ratio of said series of windows spans an expected size and aspect ratio of said structure of interest.

6. The method of claim 4 wherein said identifier comprises a tag number.

7. The method of claim 4 wherein said returned identifier is calculated using said optical character recognition.

8. The method of claim 7 wherein a confidence of said returned identifier exceeds a threshold.

9. A system for localizing numbers and characters in captured images, said system comprising:

at least one camera;

a processor that communicates with said at least one camera; and

a non-transitory computer-usable medium embodying computer program code, wherein said computer-usable medium communicates with the processor, said computer program code comprising instructions executable by said processor and configured for;

training a machine learning classifier in an offline training phase;

automatically preprocessing a side image of a vehicle digitally captured by said at least one camera to determine at least one region of interest in said side image, said at least one region of interest among regions of interest, said preprocessing comprising computer vision including median filtering to remove salt and pepper noise from said side image;

determining by said machine learning classifier a confidence of a series of windows within said regions of interest of said side image, said regions of interest comprising regions of interest of different sizes and aspect ratios, said regions of interest containing a structure of interest in said side image, said confidence comprising a measure of certainty and said structure of interest comprising text including numbers and/or an image of a physical structure of at least one object contained in said side image;

identifying highest confidence candidate regions among said regions of interest that have said structure of interest and identifying at least one region adjacent to said highest confidence candidate regions;

performing an optical character recognition in said at least one adjacent region; and

returning an identifier from said at least one adjacent region so as to localize numbers and characters in said side image of said vehicle.

10. The system of claim 9 wherein said highest confidence candidate regions are identified with nonmaximal suppression.

11. The system of claim 9 wherein a window size of said at least one adjacent region is determined by a window size of at least one candidate region among said highest confidence candidate regions.

12. The system of claim 9 wherein said confidence is automatically determined with said machine learning classifier, said machine learning classifier comprising a KNN (k-nearest neighbor) classifier and wherein said classifier is trained in said offline training phase based on extracted features that include a Fisher Vector.

13. The system of claim 12 wherein a size and aspect ratio of said series of windows spans an expected size and aspect ratio of said structure of interest.

14. The system of claim 12 wherein said identifier comprises a tag number.

15. The system of claim 12 wherein said returned identifier is calculated using said optical character recognition.

16. The system of claim 15 wherein a confidence of said returned identifier exceeds a threshold.

17. A non-transitory processor-readable medium storing code representing instructions to cause a computer executable process for localizing numbers and characters in captured images to:

train a machine learning classifier in an offline training phase;

automatically preprocess a side image of a vehicle digitally captured by at least one camera to determine at least one region of interest in said side image, said at least one region of interest among regions of interest, said preprocessing comprising computer vision including median filtering to remove salt and pepper noise from said side image;

determine by said machine learning classifier a confidence of a series of windows within said regions of interest of said side image, said regions of interest comprising regions of interest of different sizes and aspect ratios, said regions of interest containing a structure of interest in said side image, said confidence comprising a measure of certainty and said structure of interest comprising text including numbers and/or an image of a physical structure of at least one object contained in said side image;

identify highest confidence candidate regions among said regions of interest that have said structure of interest and identifying at least one region adjacent to said highest confidence candidate regions;

perform an optical character recognition in said at least one adjacent region; and

return an identifier from said at least one adjacent region in order to localize numbers and characters in said side image of said vehicle.

18. The processor-readable medium of claim 17 wherein said highest confidence candidate regions are identified with nonmaximal suppression and wherein a window size of said at least one adjacent region is determined by a window size of at least one candidate region among said highest confidence candidate regions.

19. The processor-readable medium of claim 17 wherein said confidence is automatically determined with said machine learning classifier, said machine learning classifier comprising a KNN (k-nearest neighbor) classifier and wherein said classifier is trained in said offline training phase based on extracted features that include a Fisher Vector.

20. The processor-readable medium of claim 19 wherein a size and aspect ratio of said series of windows spans an expected size and aspect ratio of said structure of interest and wherein said identifier comprises a tag number.