IDENTIFICATION AND VERIFICATION OF AN UNKNOWN DOCUMENT ACCORDING TO AN EIGEN IMAGE PROCESS

• US 20090154778A1
• Filed: 12/12/2007
• Published: 06/18/2009
• Est. Priority Date: 12/12/2007
• Status: Active Grant

• Current Assignee: Thales DIS France SA
• Original Assignee: 3M Innovative Properties Company (3M Company)
• Inventors: Lei, Yiwu, MacLean, James E.
• Granted Patent: US 8,194,933 B2
• Time in Patent Office: Days
• US Class Current: 382/112
• CPC Class Codes: G06F 18/2135 based on approximation criteria, e.g. principal component analysis
• CPC Class Codes: G06V 10/19 by sensing codes defining pattern positions
• CPC Class Codes: G06V 10/7715 Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables

Abstract

Techniques are described for identifying and validating security documents according to an Eigen image process method. For example, a security document authentication device selects one or more reference documents of different document types, calculates, from the reference documents, one or more Eigen images and Eigen values for the plurality of different document types, and calculates a reference weight coefficient vector of each of the plurality of document types. Upon receiving at least one captured image of an unknown document, the device calculates a weight coefficient vector of the captured image, compares the weight coefficient vector of the captured image and each of the reference weight coefficient vectors of the document types to calculate a plurality of distances, and, based on the plurality of distances, identifies the unknown document as one of the plurality of document types.

Claims

1. A method for verifying an unknown document as one of a plurality of possible document types, the method comprising:

selecting one or more reference documents of different document types;

calculating, from the reference documents, one or more Eigen images and Eigen values for the plurality of different document types;

calculating a reference weight coefficient vector for each of the plurality of document types based on the Eigen images and Eigen values calculated from the reference documents;

receiving at least one captured image of an unknown document;

calculating a weight coefficient vector for the captured image;

comparing the weight coefficient vector of the captured image and each of the weight coefficient vectors of the document types to calculate a plurality of distances; and

identifying the unknown document as one of the plurality of document types based on the plurality of calculated distances.

2. The method of claim 1,

wherein the Eigen images are othonormal and define a multi-dimensional space, and

wherein the plurality of distances represent distances of the captured image from the different document types in multi-dimensional space defined by the othonormal Eigen images.

3. The method of claim 2, wherein the each of the distances is expressed in the form of a certainty value for each possible document type that represents the distance on a unified scale.

4. The method of claim 1, wherein calculating a reference weight coefficient vector of the captured image comprises:

constructing a data matrix in which each column represents an Eigen image; and

multiplying the data matrix to produce the weight coefficient vector for the captured image, wherein each coefficient of the weight coefficient vector represents an expression of the captured image in multiple dimensional space formed by the othonormal Eigen images.

5. The method of claim 1, wherein calculating one or more Eigen images and Eigen values for a plurality of reference document types comprises:

collecting image templates for the possible document types;

normalizing the image data for these image templates as zero-mean;

forming a data matrix in which each column stores image data for a different one of the image templates;

calculating a set of Eigen vectors by computing the covariance matrix of the data matrix and sorting by respective Eigen values for each of the Eigen vectors;

applying a threshold to select only those Eigen vectors having Eigen values that exceed a threshold;

mapping the selected Eigen vectors back onto the Eigen images; and

storing the selected Eigen images.

6. The method of claim 1, wherein receiving at least one captured image includes receiving one or more of an infra-red (IR) image, a visible-light spectrum image, an ultraviolet (UV) image, and a retro-reflective image.

7. The method of claim 1, wherein receiving at least one captured image of the unknown document includes receiving a captured image of one of a passport, a driver'"'"'s license, a birth certificate, a financial document, a commercial paper, an identification card, and a social security card.

8. The method of claim 1, further comprising:

storing, within a data structure, a plurality of document type objects that represent the possible reference document types,

wherein the plurality of document type objects reference a plurality of recursive software processes for extracting attributes from the captured image to categorize and verify the unknown document as one of the possible document types represented by one of the document type objects.

9. The method of claim 1, wherein one or more of the document type objects reference an Eigenimage identification process that selects the reference document types, calculates the reference weight coefficient vector of each of the plurality of possible reference documents, calculates the weight coefficient vector of the captured image, compares the weight coefficient vector of the captured image and each of the weight coefficient vectors of the possible reference documents to calculate a plurality of distances, and, based on the plurality of distances, identifies the unknown document as one of the plurality of possible document types.

10. The method of claim 8, further comprising traversing the document type objects of the data structure in a variable order based on the attributes extracted by application of the plurality of the recursive processes to the captured image

11. The method of claim 9,

wherein storing the plurality of document type objects includes storing a classifier object that references one or more of the plurality of processes, and

wherein traversing the data structure includes traversing the classifier object to:

invoke the one or more processes referenced by the classifier object to determine a set of possible reference document type objects; and

traverse the set to identify the captured image as one of the plurality of possible reference document type objects.

12. The method of claim 9,

wherein storing the plurality of document type objects includes storing a minimum certainty value and a verifier object that references one or more of the plurality of processes, and

wherein traversing the set includes traversing the verifier object to:

invoke the one or more of the plurality of processes referenced by the verifier object;

calculate a certainty value from one or more return values received from invoking the one or more of the plurality of processes referenced by the verifier object to process attributes extracted from the unknown document by the classifier object in comparison to the reference document type objects; and

selectively discard one or more of the possible reference document type objects from the set based on the comparison between the certainty value and minimum certainty value.

13. The method of claim 1, further comprising:

invoking one or more of;

a first method to evaluate a possibility of a printing method for the unknown document is from intaglio or engrave printing technology;

a second method to discriminate a screening method used to print image and text;

or

a third method to evaluate whether a printing material for the unknown document includes a micro structure; and

based on the possibility, determine whether the identified document is authentic or not.

14. The method of claim 1, further comprising identifying and verifying the unknown document as a security document that combines embedded electronic information includes one or more of radio frequency identification (RFID) data, electronic passport data, smartcard data, or magnetic strip data.

15. The method of claim 1, further comprising presenting results of the identification and verification process to a user in visual or sound format.

19. The security document authentication device of claim 14, wherein the document processing engine calculates and stores the one or more Eigen images and Eigen values for the plurality of reference document types by:

collecting image templates for the possible document types;

normalizing the image data for these image templates as zero-mean;

forming a data matrix in which each column stores image data for a different one of the image templates;

calculating a set of Eigen vectors by computing the covariance matrix of the data matrix and sorting by respective Eigen values for each of the Eigen vectors;

applying a threshold to select only those Eigen vectors having Eigen values that exceed a threshold; and

mapping the selected Eigen vectors back onto Eigen images;

storing the selected Eigen images in the memory.

16. A security document authentication device comprising:

a memory storing one or more Eigen images and Eigen values for a plurality of reference document types;

an image capture interface that captures at least one image from an unknown document;

a document processing engine that, in response to the captured image, select one or more of the reference document types as possible reference document types, calculates a reference weight coefficient vector of each of the plurality of possible reference documents, calculates a weight coefficient vector of the captured image, compare the weight coefficient vector of the captured image and each of the weight coefficient vectors of the possible reference documents to calculate a plurality of distances, and, based on the plurality of distances, identifying the unknown document as one of the plurality of possible document types.

17. The security document authentication device of claim 16, wherein the document processing engine calculates the weight coefficient vector of the captured image by constructing a data matrix in which each column represents an Eigen image, and multiplying the data matrix to produce the weight coefficient vector for the captured image.

18. The security document authentication device of claim 16, wherein the document processing engine calculates and stores the one or more Eigen images and Eigen values for the plurality of reference document types.

20. The security document authentication device of claim 16,

wherein the memory stores a plurality of document type objects within a data structure according to a dynamic document identification framework, wherein the plurality of document type objects reference a plurality of recursive processes for extracting attributes from the captured image,

wherein the document processing engine traverses the document type objects of the data structure in a variable order based on the attributes extracted by application of the plurality of the recursive processes to the captured image, and

wherein the document processing engine identifies the unknown document as one of the plurality of document type objects upon traversing the data structure.

21. The security document authentication device of claim 16, wherein the image capture interface receives the captured image by receiving one or more of an infra-red (IR) image, a visible-light spectrum image, an ultraviolet (UV) image, and a retro-reflective image.

22. The security document authentication device of claim 16, wherein the image capture interface receives the captured image of the unknown document by receiving captured image of one of a passport, a driver'"'"'s license, a birth certificate, a financial document, a commercial paper, an identification card, or a social security card.

23. A computer-readable medium comprising instructions for causing a programmable processor to:

calculating one or more Eigen images and Eigen values for a plurality of reference document types,

receiving at least one captured image of an unknown document;

select one or more of the reference document types as possible reference document types,

calculate a reference weight coefficient vector of each of the plurality of possible reference documents;

calculate a weight coefficient vector of the captured image;

compare the weight coefficient vector of the captured image and each of the weight coefficient vectors of the possible reference documents to calculate a plurality of distances; and

based on the plurality of distances, identifying the unknown document as one of the plurality of possible document types.

24. A method for the identification or validation of an unknown document by a layout matching process comprising:

receiving a captured image of an unknown document;

segmenting and identifying a plurality of connected regions of the captured image;

constructing a graph to represent the captured image by classifying the plurality of connected regions and establishing a set of relationships between the plurality of connected regions;

comparing the plurality of connected regions to template data associated with one of the plurality of document type objects, wherein the template data specifies a pre-defined plurality of connected image regions and relationships between the plurality of pre-defined connected image regions; and

based on the comparison, determining whether the captured image belong to the one of the plurality of document type objects currently under comparison.