Label Detection

• US 20070058863A1
• Filed: 09/09/2005
• Published: 03/15/2007
• Est. Priority Date: 09/09/2005
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Eswara, Lalitha, Boregowda, Lokesh, Dornadula, Krishnamoorthy, Rajagopal, Anupama
• Granted Patent: US 7,613,328 B2
• Time in Patent Office: Days
• US Class Current: 382/170
• CPC Class Codes: G06V 20/63 Scene text, e.g. street names

Abstract

A system and a method to detect a label are described herein. The system and method are to extract a horizontal region of interest based on the image; to extract a vertical region of interest based on the image; and to identify presence of the primary label in the image from the horizontal region of interest and from the vertical region of interest

Claims

1. A method to detect an identification label comprising:

determining an input image having an identification label;

processing the input image;

determining a column difference image;

determining a row projection histogram of the column difference image associated with the identification label;

processing a row projected image to extract a horizontal region associated with the identification label; and

extracting a vertical region of interest associated with the identification label, wherein the extracting the horizontal region of interest is selected from a group including;

determining a column projection histogram of a row difference image associated with the identification label and processing a column projected image; and

using wavelet coefficient clustering of the row difference image.

2. The method of claim 1 further comprising:

identifying the identification label from the vertical region and from the horizontal region; and

performing character recognition on the identification label.

3. The method of claim 1 wherein using the wavelet coefficient clustering includes preprocessing vertical detailed coefficients and clustering the vertical detailed coefficients.

4. The method of claim 1 wherein an index of a largest cluster present is used to extract the vertical region.

5. The method of claim 1 wherein processing the input image is selected from a group including image enhancement, brightness correction, and contrast correction.

6. A method to identify presence of a primary label in an image comprising:

extracting a vertical region of interest based on the image using projection histogram analysis;

extracting a horizontal region of interest based on the image using projection histogram analysis; and

identifying presence of the primary label in the image from the horizontal region of interest and from the vertical region of interest.

7. The method of claim 6 wherein extracting the horizontal region of interest based on the image includes column difference image analysis and row projection histogram analysis.

8. The method of claim 6 wherein extracting the vertical region of interest based on the image includes row difference image analysis and column projection histogram analysis.

9. The method of claim 6 wherein extracting the vertical region of interest based on the image includes wavelet coefficient clustering.

10. The method of claim 6 further comprising pre-processing the image to enhance the primary label before extracting the vertical region of interest.

11. The method of claim 10 wherein the pre-processing is selected from a group including brightness correction and contrast correction.

12. The method of claim 10 wherein the pre-processing includes:

subjecting the image to zero meaning; and

stretching the histogram.

13. The method of claim 6 wherein the primary label includes an orientation of up to about 20 degrees from a horizontal.

14. A machine-readable medium storing a sequence of instructions that, when executed by a computer, cause the computer to perform a method of identifying presence of a primary label in an image, the method comprising:

extracting a vertical region of interest based on an image using projection histogram analysis;

extracting a horizontal region of interest based on the image using projection histogram analysis; and

identifying presence of a primary label in the image from the horizontal region of interest and from the vertical region of interest.

15. A method comprising:

detecting a vehicle in an image;

detecting a license plate of the vehicle using projection histogram analysis on a gradient image;

segmenting characters of the plate; and

recognizing the characters.

16. The method of claim 15 further comprising locating presence of characters in the primary label.

17. The method of claim 16 wherein the image is selected from a group including a gray scale image version and a Y channel image version.

18. A system to identify a label in an image comprising:

means for extracting a vertical region of interest based on the image using projection histogram analysis;

means for extracting a horizontal region of interest based on the image using projection histogram analysis; and

means for identifying presence of the label in the image from the horizontal region of interest and from the vertical region of interest.

19. The system of claim 18 wherein the image includes an object having the label, wherein the object at least one of slows down and stops when the image is taken.

20. The system of claim 18 wherein means for extracting the vertical region includes removing one-fourth of a row in a top part of the image, removing one-eighth of a row at a bottom part of the image;

and removing one-eighth of columns at both sides of the image.

21. The system of claim 18 wherein means for extracting the horizontal region includes an adaptive threshold to determine substantially exact row location where the label is present.

22. The system of claim 18 wherein means for extracting the vertical region uses row differencing and includes an adaptive threshold to prune the image to substantially exact columns.