Fingerprint check to reduce check fraud

1. A method comprising:

receiving, at a computing system, a digital image of a surface of a check, wherein a check scanning device captures the digital image as the check moves through the check scanning device, wherein a fingerprint is impressed on the surface of the check;

using, at the computing system, the digital image to determine a match score, the match score representing an approximate probability that the fingerprint on the check matches a reference fingerprint, matching being performed by:

generating a target fingerprint template that represents the fingerprint;

storing a plurality of reference fingerprint templates, each of the reference fingerprint templates representing a fingerprint of a different person, the plurality of reference fingerprint templates including a reference fingerprint template representing the reference fingerprint; and

calculating the match score by comparing the target fingerprint template and the reference fingerprint template;

presenting an alert based on the match score; and

performing, at the computing system, a learning process that seeks to reduce a rate of false matches and false non-matches by monitoring at least one occurrence of a false match and/or a false non-match and adjusting at least one reference fingerprint template.

2. The method of claim 1 , wherein performing the learning process comprises:

when the computing system determines that the fingerprint template matches the reference fingerprint template, receiving input indicating that the fingerprint template does not match the reference fingerprint template; and

in response to receiving the input, adjusting the reference fingerprint template to improve a likelihood that the computing system will correctly determine that fingerprint templates match the reference fingerprint template.

3. The method of claim 1 , wherein using the digital image comprises:

applying one or more image processing operations to a fingerprint image, the fingerprint image being a digital image containing the fingerprint; and

using the fingerprint image to determine the match score.

4. The method of claim 3 , wherein using the digital image comprises:

after applying the one or more image processing operations to the fingerprint image, calculating a fingerprint image quality metric of the fingerprint image, wherein the fingerprint image quality metric of the fingerprint image is a measure of an expected accuracy of the fingerprint image in representing the fingerprint; and

adjusting the match score based on the fingerprint image quality metric of the fingerprint image.

5. The method of claim 3 , wherein performing the learning process comprises: evaluating whether the one or more image processing operations increase or decrease a likelihood of false matches or false non-matches.

6. The method of claim 3 , wherein applying the one or more image processing operations to the fingerprint image comprises applying one or more image processing operations to reduce artifacts caused by movement of the check through the check scanning device.

7. The method of claim 3 ,

wherein the method further comprises: detecting a fiducial mark on the check; and

wherein applying the one or more image processing operations comprises: using the fiducial mark during performance of at least one of the image processing operations.

8. The method of claim 1 , wherein using the digital image further comprises:

creating a list of likely suspects, the likely suspects being individuals who are most likely to commit fraud at a branch at which an individual presented the check;

identifying a prioritized set of reference fingerprint templates, each reference fingerprint template in the prioritized set of reference fingerprint templates being in the plurality of reference fingerprint templates and associated with one of the likely suspects;

calculating match scores for reference fingerprint templates in the prioritized set of reference fingerprint templates in an order based on priorities of the reference fingerprint templates; and

sending a search response when a given one of the match scores exceeds a search-ending threshold or when a time limit has expired, the search response indicating the given one of the match scores.

9. The method of claim 1 , wherein the digital image has a resolution less than 500 dpi.

10. A computing device comprising:

a processing unit; and

a computer-readable medium comprising instructions that, when executed by the processing unit, cause the computing device to:

determine a match score by comparing a target fingerprint template and a reference fingerprint template, the target fingerprint template representing a fingerprint impressed on a check, the target fingerprint template generated from an image of the check, the image of the check captured by a check scanning device as the check moved through the check scanning device, the reference fingerprint template representing a reference fingerprint, the match score representing an approximate probability that the fingerprint on the check matches the reference fingerprint;

cause a workstation device to present an alert when the match score exceeds an alert threshold; and

cause the computing device to perform a learning process that seeks to reduce a rate of false matches and false non-matches by monitoring at least one occurrence of a false match and/or a false non-match and adjusting at least one reference fingerprint template.

11. The computing device of claim 10 , wherein the instructions, when executed by the processing unit, further cause the computing device to:

cause the workstation device to instruct a cashier to complete a transaction involving the check when the match score does not exceed the alert threshold.

12. The computing device of claim 10 , wherein the instructions, when executed by the processing unit, cause the computing device to adjust the match score based on a fingerprint image quality metric, the fingerprint image quality metric being a measure of the expected accuracy of the fingerprint image in representing the fingerprint.