Systems and methods for improving image recognition

• US 8,731,244 B2
• Filed: 07/28/2011
• Issued: 05/20/2014
• Est. Priority Date: 07/28/2011
• Status: Active Grant

• Current Assignee: Conduent Business Services, LLC (Conduent, Inc.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Wu, Wencheng
• Primary Examiner(s): Abdi, Amara
• Application Number:
  US13/193,287
• Publication Number:
  US 20130028481A1
• Time in Patent Office: 1,027 Days
• Field of Search: 382/105, 382/173, 382/299, 340/933, 345/427
• US Class Current: 382/105
• CPC Class Codes: G06T 2207/30236 Traffic on road, railway or crossing
• CPC Class Codes: G06T 5/80 Geometric correction
• CPC Class Codes: G06V 20/625 License plates
• CPC Class Codes: G06V 20/63 Scene text, e.g. street names

Abstract

An embodiment generally relates to systems and methods for processing license plate image data. A processing module can receive an image comprising a license plate and, from the image, localize a region of the license plate itself. The license plate region can be extracted and license plate distortion parameters can be inferred. The image can be geometrically corrected using the inferred license plate distortion parameters, and the license plate image can be iteratively sharpened to optimize the quality of the license plate image. Further, processing parameters can be modified and applied to the image data in a feedback-based loop for further optimization.

Claims

1. A method of processing data, the method comprising:

detecting image data comprising an image of a license plate;

localizing a region of the license plate within the image;

processing, by a processor, the region of the license plate to determine a set of distortion parameters for the license plate, wherein the processing comprises;

determining an optimal threshold, wherein determining the optimal threshold comprises determining a threshold that generates a binary image with a maximized convexity;

binarizing the image data based on the optimal threshold to generate a first binarized image; and

processing the first binarized image to determine a rotation angle; and

applying the set of distortion parameters to the image data to enhance the image of the license plate, wherein applying the set of distortion parameters comprises;

rotating the image based on the rotation angle to generate a rotated image; and

binarizing the rotated image to generate a second binarized image.

2. The method of claim 1, further comprising:

modifying the set of distortion parameters based on a confidence measure; and

applying the set of distortion parameters that was modified to the image data.

3. The method of claim 2, wherein the modifying and the applying are repeated a set number of times or until a threshold result is reached.

4. The method of claim 1, further comprising:

extracting license plate data from the image of the license plate that was enhanced; and

providing the license plate data to a user.

5. The method of claim 1, wherein applying the set of distortion parameters to the image data to enhance the image of the license plate further comprises:

performing a geometric correction on the image of the license plate to generate a corrected image;

resizing the corrected image to a set range; and

processing the corrected image that was resized to enhance a quality of the corrected image that was resized.

6. The method of claim 1, wherein processing the region of the license plate to determine the set of distortion parameters further comprises:

prior to binarizing the image data based on the optimal threshold, rotating the image of the license plate by a first rotation angle;

extracting an outline of the license plate from the second binarized image; and

smoothing the outline of the license plate using Fourier descriptors.

7. The method of claim 6, wherein processing the region of the license plate to determine the set of distortion parameters further comprises fitting the outline of the license plate that was smoothed with a quadrilateral.

8. The method of claim 6, wherein the rotation angle is derived from the image of the license plate.

9. The method of claim 1, further comprising processing the region of the license plate with a histogram equalization.

10. A system for processing data, the system comprising:

a processor coupled to memory and configured to perform actions comprising;

detecting image data comprising an image of a license plate;

localizing a region of the license plate within the image;

processing, by a processor, the region of the license plate to determine a set of distortion parameters for the license plate, wherein the processing comprises;

determining an optimal threshold, wherein determining the optimal threshold comprises determining a threshold that generates a binary image with a maximized convexity;

binarizing the image data based on the optimal threshold to generate a first binarized image; and

processing the first binarized image to determine a rotation angle; and

applying the set of distortion parameters to the image data to enhance the image of the license plate, wherein applying the set of distortion parameters comprises;

rotating the image based on the rotation angle to generate a rotated image; and

binarizing the rotated image to generate a second binarized image.

11. The system of claim 10, wherein the processor is further configured to perform actions comprising:

modifying the set of distortion parameters based on a confidence measure; and

applying the set of distortion parameters that was modified to the image data.

12. The system of claim 11, wherein the modifying and the applying are repeated a set number of times or until a threshold result is reached.

13. The system of claim 10, wherein the processor is further configured to perform actions comprising:

extracting license plate data from the image of the license plate that was enhanced; and

providing the license plate data to a user.

14. The system of claim 10, wherein applying the set of distortion parameters to the image data to enhance the image of the license plate further comprises:

performing a geometric correction on the image of the license plate to generate a corrected image;

resizing the corrected image to a set range; and

processing the corrected image that was resized to enhance a quality of the corrected image that was resized.

15. The system of claim 10, wherein processing the region of the license plate to determine the set of distortion parameters further comprises:

prior to binarizing the image data based on the optimal threshold, rotating the image of the license plate by a first rotation angle;

extracting an outline of the license plate from the second binarized image; and

smoothing the outline of the license plate using Fourier descriptors.

16. The system of claim 15, wherein processing the region of the license plate to determine the set of distortion parameters further comprises fitting the outline of the license plate that was smoothed with a quadrilateral.

17. The system of claim 15, wherein the rotation angle is derived from the image of the license plate.

18. The system of claim 10, wherein the processor is further configured to perform actions comprising:

processing the region of the license plate with a histogram equalization.