Method and system for image identification and identification result output

• US 8,391,559 B2
• Filed: 07/30/2009
• Issued: 03/05/2013
• Est. Priority Date: 01/23/2009
• Status: Active Grant

• Current Assignee: Industrial Technology Research Institute
• Original Assignee: Industrial Technology Research Institute
• Inventors: Wu, Tung-Chuan, Tsai, Ya-Hui, Huang, Kuo-Tang, Lin, Yu-Ting, Chang, Chun-Lung
• Primary Examiner(s): Newman, Michael A
• Application Number:
  US12/512,575
• Publication Number:
  US 20100189364A1
• Time in Patent Office: 1,314 Days
• Field of Search: 382/105, 382/104, 382/176, 382/177, 382/181, 382/190, 382/209, 382/218, 382/220
• US Class Current: 382/105
• CPC Class Codes: G06V 20/625 License plates
• CPC Class Codes: G06V 20/63 Scene text, e.g. street names

Abstract

The present invention provides a method and system for image identification and identification result output, wherein a feature image under identification acquired from an image is compared with a plurality of sample images respectively stored in a database so as to obtain a plurality of similarity indexes associated with the plurality of sample images respectively. Each similarity index represents similarity between the feature image and the corresponding sample image. Thereafter, the plurality of similarity indexes are sorted and then a least one of comparison results is output. The present invention is further capable of being used for identifying identification marks with respect to a carrier. By sorting the similarity index with respect to each feature forming the identification marks, it is capable of outputting many sets of combinations corresponding to the identification marks so as to improve speed for targeting suspected carrier and enhance the identification efficiency.

Claims

1. A method for image identification and identification result output, comprising steps of:

providing an image;

acquiring a feature image from the image;

providing a plurality of sample images, each having respectively a standard image region and at least a non-standard image region, the standard image region having pixels corresponding to a first feature value respectively, and the non-standard image region having pixels corresponding to a second feature value respectively;

performing a calculation on a third feature value of each pixel in the feature image and the first feature value or the second feature value corresponding to each pixel in the plurality of sample images to obtain a similarity index of the feature image corresponding to the plurality of sample images respectively;

collecting a plurality of similarity indexes with respect to the feature image compared with the plurality of sample images; and

sorting the plurality of similarity indexes and outputting at least one of comparison results,

wherein the calculation is based on normalized correlation matching,

wherein the first feature value and the second feature value are respectively a combination of a weight value and a greyscale value, and the third feature value is a greyscale value, and

wherein normalized correlation matching is performed by dividing the product of the weight value corresponding to each pixel in the sample image, the difference between the greyscale value of each pixel and the average greyscale value in the sample image, and the difference between the greyscale value of each pixel and the average greyscale value in the feature image by the product of the standard deviation of the greyscale value of the sample image and the standard deviation of the greyscale value of the feature image.

2. The method for image identification and identification result output as recited in claim 1, further comprising a step of normalizing the feature image to adjust the size and the angle of the feature image so that the size of the feature image is identical to the size of the sample image after acquiring the feature image from the image.

3. The method for image identification and identification result output as recited in claim 1, wherein each sample image corresponds to an image of a digit or a character.

4. The method for image identification and identification result output as recited in claim 1, wherein the step of providing the plurality of sample images comprises the steps of:

determining the size of the sample images;

providing in each of the sample images the standard image region corresponding to a piece of information, and determining the first feature value corresponding to the pixels;

providing in each of the sample images the non-standard image region corresponding to another piece of information that the piece of information is taken for, and determining the second feature value corresponding to the pixels; and

repeating the aforegoing steps for a plurality of times to provide the plurality of sample images corresponding to different characters.

5. The method for image identification and identification result output as recited in claim 1, wherein the content in the standard image region is a piece of information corresponding to the sample image, and the content in the non-standard image region and the standard image region is another piece of information that the piece of information is taken for corresponding to the sample image.

6. A method for image identification and identification result output, comprising steps of:

providing an image of a carrier with an identification mark thereon;

acquiring from the image a plurality of feature images with respect to the identification mark;

providing a plurality of sample images, each having respectively a standard image region and at least a non-standard image region, the standard image region having pixels corresponding to a first feature value respectively, and the non-standard image region having pixels corresponding to a second feature value respectively;

performing a calculation on a third feature value of each pixel in the plurality of feature images and the first feature value or the second feature value corresponding to each pixel in the plurality of sample images to obtain a similarity index of each feature image corresponding to the plurality of sample images respectively;

collecting a plurality of similarity indexes with respect to each feature image compared with the plurality of sample images; and

sorting the plurality of similarity indexes corresponding to the identification mark and outputting at least one of comparison results,

wherein the calculation is based on normalized correlation matching,

wherein the first feature value and the second feature value are respectively a combination of a weight value and a greyscale value, and the third feature value is a greyscale value, and

wherein normalized correlation matching is performed by dividing the product of the weight value corresponding to each pixel in the sample image, the difference between the greyscale value of each pixel and the average greyscale value in the sample image, and the difference between the greyscale value of each pixel and the average greyscale value in the feature image by the product of the standard deviation of the greyscale value of the sample image and the standard deviation of the greyscale value of the feature image.

7. The method for image identification and identification result output as recited in claim 6, further comprising a step of normalizing the feature image to adjust the size and the angle of the feature image so that the size of the feature image is identical to the size of the sample image after acquiring the feature image from the image.

8. The method for image identification and identification result output as recited in claim 6, wherein the identification mark is a license plate number.

9. The method for image identification and identification result output as recited in claim 6, wherein each sample image corresponds to an image of a digit or a character.

10. The method for image identification and identification result output as recited in claim 6, wherein each sample image is further divided into feature images acquired in front, feature images acquired slantedly and feature images acquired at least one distances.

11. The method for image identification and identification result output as recited in claim 6, further comprising a step of excluding images of impossible characters or digits according to a relative location of the feature image of the identification mark when each feature image is compared with the plurality of sample images stored in the database.

12. The method for image identification and identification result output as recited in claim 6, wherein the step of providing the plurality of sample images comprises the steps of:

determining the size of the sample images;

providing in each of the sample images the standard image region corresponding to a piece of information, and determining the first feature value corresponding to the pixels;

providing in each of the sample images the standard image region corresponding to another piece of information that the piece of information is taken for, and determining the second feature value corresponding to the pixels; and

repeating the aforegoing steps for a plurality of times to provide the plurality of sample images corresponding to different characters.

13. The method for image identification and identification result output as recited in claim 6, wherein the content in the standard image region is a piece of information corresponding to the sample image, and the content in the non-standard image region and the standard image region is another piece of information that the piece of information is taken for corresponding to the sample image.

14. A system for image identification and identification result output, comprising:

a database capable of providing a plurality of sample images, each having respectively a standard image region and at least a non-standard image region, the standard image region having pixels corresponding to a first feature value respectively, and the non-standard image region having pixels corresponding to a second feature value respectively;

an image acquiring unit capable of acquiring an image of an object;

a feature acquiring unit capable of acquiring a feature image from the image;

an operation and processing unit capable of performing a calculation on a third feature value of each pixel in the feature image and the first feature value or the second feature value corresponding to each pixel in the plurality of sample images to obtain a similarity index of the feature image corresponding to the plurality of sample images respectively, and sorting a plurality of similarity indexes and outputting at least one of comparison results; and

an identification and output unit being connected to the operation and processing unit to output the comparison result identified by the operation and processing unit,

wherein the calculation is based on normalized correlation matching,

wherein the first feature value and the second feature value are respectively a combination of a weight value and a greyscale value, and the third feature value is a greyscale value, and

wherein normalized correlation matching is performed by dividing the product of the weight value corresponding to each pixel in the sample image, the difference between the greyscale value of each pixel and the average greyscale value in the sample image, and the difference between the greyscale value of each pixel and the average greyscale value in the feature image by the product of the standard deviation of the greyscale value of the sample image and the standard deviation of the greyscale value of the feature image.

15. The system for image identification and identification result output as recited in claim 14, wherein the object is a carrier, the image of the object is an image of an identification mark on the carrier, and the identification mark is a license plate number.

16. The system for image identification and identification result output as recited in claim 14, wherein each sample image corresponds to an image of a word, a digit or a character.

17. The system for image identification and identification result output as recited in claim 14, wherein the operation and processing unit excludes images of impossible characters or digits according to a relative location of the feature image in an identification mark when each feature image is compared with the plurality of sample images stored in the database by the operation and processing unit.

18. The system for image identification and identification result output as recited in claim 14, wherein the image acquiring unit normalizes the feature image to adjust the size and the angle of the feature image so that the size of the feature image is identical to the size of the sample image after acquiring the feature image from the image.

19. The system for image identification and identification result output as recited in claim 14, wherein the operation and processing unit comprises an enhancing unit and an identification and comparison unit, wherein the enhancing unit improves and normalizes the feature image so that the size of the feature image is identical to the size of the sample image and the identification and comparison unit compares the improved and normalized feature image with the sample image to obtain the plurality of similarity indexes corresponding thereto.

20. The system for image identification and identification result output as recited in claim 14, wherein the content in the standard image region is a piece of information corresponding to the sample image, and the content in the non-standard image region and the standard image region is another piece of information that the piece of information is taken for corresponding to the sample image.