METHODS AND SYSTEMS FOR VEHICLE TAG NUMBER RECOGNITION

• US 20160171328A1
• Filed: 12/11/2014
• Published: 06/16/2016
• Est. Priority Date: 12/11/2014
• Status: Active Grant

• Current Assignee: Conduent Business Services, LLC (Conduent, Inc.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Bulan, Orhan, Mizes, Howard, Kozitsky, Vladimir, Burry, Aaron M.
• Granted Patent: US 9,400,936 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06V 20/625 License plates
• CPC Class Codes: G06V 20/63 Scene text, e.g. street names
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/1473 Recognising objects as potential recognition candidates based on visual cues, e.g. shapes
• CPC Class Codes: G06V 30/268 Lexical context

Abstract

Methods and systems for tag recognition in captured images. A candidate region can be localized from regions of interest with respect to a tag and a tag number shown in the regions of interest within a side image of a vehicle. A number of confidence levels can then be calculated with respect to each digit recognized as a result of an optical character recognition operation performed with respect to the tag number. Optimal candidates within the candidate region can be determined for the tag number based on individual character confidence levels among the confidence levels. Optimal candidates from a pool of valid tag numbers can then be validated using prior appearance probabilities and data returned, which is indicative of the most probable tag to be detected to improve image recognition accuracy.

Claims

1. An image-processing method for tag recognition in captured images, said method comprising:

capturing a side image of a vehicle utilizing at least one camera of an automated vision-based recognition system comprising an image processor;

localizing with said image processor a candidate region from regions of interest with respect to a physical object comprising a tag and a tag number shown in regions of interest within said side image of said vehicle by classifying said side image using a classifier trained as a trained classifier in an offline phase;

calculating with said image processor a plurality of confidence levels with respect to each digit recognized as a result of an optical character recognition operation performed with respect to said tag number;

determining with said image processor optimal candidates within said candidate region for said tag number based on individual character confidence levels among said plurality of confidence levels; and

validating with said image processor said optimal candidates from a pool of valid tag numbers using prior appearance probabilities and returning data indicative of a most probable tag to be detected to improve image recognition accuracy with respect to data representing said physical object comprising said tag and tag number.

2. The method of claim 1 wherein said at least one camera comprises an NIR camera coupled with at least one IR illuminator for night vision capturing of said side image of said vehicle.

3. The method of claim 2 further comprising performing said optical character recognition operation with respect to said tag number.

5. The method of claim 3 further comprising training said classifier in said offline phase using a training set containing a variety of fonts and sizes, said training set comprising a positive sample set that spans said variety of fonts and sizes and a negative sample set that is generated utilizing randomly cropped patches that contain a same amount of structure as said positive sample set.

6. The method of claim 5 further comprising using said trained classifier in an online phase.

7. The method of claim 6 wherein said validating said optimal candidates from said pool of valid tag numbers, further comprises retrieving said pool of valid tag numbers from a database of tags.

8. The method of claim 7 wherein said database further comprises data comprising location of a test site and a time of tag detection.

9. The method of claim 7 wherein a prior probability of detection at a given location increases based on a number of prior detections at said given location.

10. The method of claim 7 wherein a prior probability of detection at a given location increases based on an existence of a detection of a same tag at a nearby facility within a time window determined by a travel time between a present facility and said nearby facility.

12. The system of claim 10 wherein said at least one camera comprises an NIR camera coupled with at least one IR illuminator for night vision capturing said side image of said vehicle.

13. The system of claim 12 further comprising an OCR engine for performing said optical character recognition operation with respect to said tag number.

4. (canceled)

11. An image-processing system for tag recognition in captured images, said system comprising:

at least one camera of an automated vision-based recognition system,

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

a computer-usable medium embodying computer program code, said computer-usable medium capable of communicating with the processor, said computer program code comprising instructions executable by said processor and configured for;

capturing a side image of a vehicle utilizing at least one camera of an automated vision-based recognition system;

localizing a candidate region from regions of interest with respect to a physical object comprising a tag and a tag number shown in regions of interest within said side image of said vehicle captured by said at least one camera by classifying said side image using a classifier trained as a trained classifier in an offline phase;

calculating a plurality of confidence levels with respect to each digit recognized as a result of an optical character recognition operation performed with respect to said tag number;

determining optimal candidates within said candidate region for said tag number based on individual character confidence levels among said plurality of confidence levels; and

validating said optimal candidates from a pool of valid tag numbers using prior appearance probabilities and returning data indicative of a most probable tag to be detected to improve image recognition accuracy with respect to data representing said physical object comprising said tag and tag number.

15. The system of claim 11 wherein said instructions are further configured for training said classifier in said offline phase using a training set containing a variety of fonts and sizes, said training set comprising a positive sample set that spans said variety of fonts and sizes and a negative sample set that is generated utilizing randomly cropped patches that contain a same amount of structure as said positive sample set.

16. The system of claim 15 wherein said instructions are further configured for retrieving said pool of valid tag numbers from a database of tags and wherein said database of tags further comprises data comprising a location of a test site and a time of tag detection.

17. The system of claim 16 wherein a prior probability of detection at a given location increases based on a number of prior detections at said given location.

18. The system of claim 16 wherein a prior probability of detection at a given location increases based on an existence of a detection of a same tag at a nearby facility within a time window determined by a distance between a present facility and said nearby facility.

14. (canceled)

19. A non-transitory processor-readable medium storing code representing instructions to cause a process for tag recognition in captured images, said code comprising code to:

capture a side image of a vehicle utilizing at least one camera of an automated vision-based recognition system comprising an image processor;

localize a candidate region from regions of interest with respect to a physical object comprising a tag and a tag number shown in regions of interest within said side image of said vehicle by classifying said side image using a classifier trained as a trained classifier in an offline phase;

calculate a plurality of confidence levels with respect to each digit recognized as a result of an optical character recognition operation performed with respect to said tag number;

determine optimal candidates within said candidate region for said tag number based on individual character confidence levels among said plurality of confidence levels; and

validate said optimal candidates from a pool of valid tag numbers using prior appearance probabilities and returning data indicative of a most probable tag to be detected to improve image recognition accuracy with respect to data representing said physical object comprising said tag and tag number.

20. The non-transitory processor-readable medium of claim 19 wherein said code further comprises code to train said classifier in said offline phase using a training set containing a variety of fonts and sizes, said training set comprising a positive sample set that spans said variety of fonts and sizes and a negative sample set that is generated utilizing randomly cropped patches that contain a same amount of structure as said positive sample set.

21. The non-transitory processor-readable medium of claim 20 wherein said code further comprises code to use said trained classifier in an online phase.

22. The non-transitory processor-readable medium of claim 20 wherein said code further comprises code to:

perform said optical character recognition operation with respect to said tag number and wherein said at least one camera comprises an NIR camera coupled with at least one IR illuminator for night vision capturing of said side image of said vehicle.