Document verification system, device, and method using a classification model

1. A computer-implemented document verification method, comprising:

obtaining an image of a to-be-verified document;

providing the image as input to a trained classification model;

obtaining an output of the trained classification model, wherein the output represents a probability at which the image passes an anti-counterfeiting identifier verification;

verifying whether an anti-counterfeiting identifier exists in the image based on the output of the trained classification model;

performing first verification on the image to determine a probability at which the image is from a screen print;

performing second verification on the image to determine a probability at which the image is from a copy;

determining a probability at which the image is from a physical document based on the probability at which the image is from a screen print and the probability at which the image is from a copy;

verifying a source of the image based on the probability at which the image is from a physical document; and

obtaining a verification result for the to-be-verified document based on the verifying of the anti-counterfeiting identifier and the verifying of the source of the image.

2. The method according to claim 1 , wherein verifying whether an anti-counterfeiting identifier exists in the image comprises:

determining a type of anti-counterfeiting identifier corresponding to a type of the to-be-verified document; and

determining a probability at which an anti-counterfeiting identifier corresponding to the determined type of anti-counterfeiting identifier exists in the image.

3. The method according to claim 2 , wherein the type of anti-counterfeiting identifier comprises at least one of watermark text, micro-text, shading lines, gravure printing, color cast, anti-counterfeiting secret marks, or fonts.

4. The method according to claim 1 , wherein verifying a source of the image comprises:

analyzing data of the image to determine a probability at which the image is from a physical document.

5. The method according to claim 4 , wherein analyzing data of the image to determine a probability at which the image is from a physical document comprises:

performing image recognition processing on data of an individual frame in the image.

6. The method according to claim 4 , wherein analyzing data of the image to determine a probability at which the image is from a physical document comprises:

obtaining at least two types of images of the to-be-verified document;

performing differencing processing on data of the at least two types of images to obtain image difference; and

using the image difference as an input to a pre-established classification model, the pre-established classification model being trained to output a verification result of the source of the image based on the input image difference.

7. The method according to claim 4 , wherein verifying an anti-counterfeiting identifier in the image comprises:

determining a type of anti-counterfeiting identifier corresponding to a type of the to-be-verified document;

determining a probability at which an anti-counterfeiting identifier corresponding to the determined type of anti-counterfeiting identifier exists in the image; and

wherein obtaining a verification result for the to-be-verified document based on the verifying of the anti-counterfeiting identifier and the verifying of the source of the image comprises:

determining the verification result based on the probability at which an anti-counterfeiting identifier corresponding to the determined type of anti-counterfeiting identifier exists in the image and the probability at which the image is from a physical document.

8. The method according to claim 1 , wherein verifying a source of the image comprises:

collecting a background image before or after obtaining the image of the to-be-verified document;

verifying the image of the to-be-verified document based on the background image to determine whether the image of the to-be-verified document and the background image are from a same photographing scene.

9. The method according to claim 8 , wherein verifying the image of the to-be-verified document based on the background image to determine whether the image of the to-be-verified document and the background image are from a same photographing scene comprises:

comparing a background region in the image of the to-be-verified document with the background image to determine an overlapping degree between the image of the to-be-verified document and the background image, the background region in the image of the to-be-verified document being a region outside a region of the to-be-verified document in the image; and

determining whether the image of the to-be-verified document and the background image are from the same photographing scene based on the overlapping degree.

10. The method according to claim 9 , wherein comparing a background region in the image of the to-be-verified document with the background image to determine an overlapping degree between the image of the to-be-verified document and the background image comprises:

comparing the background region in the image of the to-be-verified document with the background image based on first position information of a gyroscope of a device when the device collects the image of the to-be-verified document and second position information of the gyroscope of the device when the device collects the background image.

11. A document verification device, comprising:

a hardware processor; and

a non-transitory machine-readable storage medium encoded with instructions executable by the hardware processor to perform operations comprising:

obtaining an image of a to-be-verified document;

providing the image as input to a trained classification model;

obtaining an output of the trained classification model, wherein the output represents a probability at which the image passes an anti-counterfeiting identifier verification;

verifying whether an anti-counterfeiting identifier exists in the image based on the output of the trained classification model;

performing first verification on the image to determine a probability at which the image is from a screen print;

performing second verification on the image to determine a probability at which the image is from a copy;

determining a probability at which the image is from a physical document based on the probability at which the image is from a screen print and the probability at which the image is from a copy;

verifying a source of the image based on the probability at which the image is from a physical document; and

obtaining a verification result for the to-be-verified document based on the verifying of the anti-counterfeiting identifier and the verifying of the source of the image.

12. The document verification device according to claim 11 , wherein the verifying whether an anti-counterfeiting identifier exists in the image comprises:

determining a type of anti-counterfeiting identifier corresponding to a type of the to-be-verified document; and

determining a probability at which an anti-counterfeiting identifier corresponding to the determined type of anti-counterfeiting identifier exists in the image.

13. The document verification device according to claim 12 , wherein the type of anti-counterfeiting identifier comprises at least one of watermark text, micro-text, shading lines, gravure printing, color cast, anti-counterfeiting secret marks, or fonts.

14. The document verification device according to claim 11 , wherein the verifying a source of the image comprises:

analyzing data of the image to determine a probability at which the image is from a physical document.

15. The document verification device according to claim 14 , wherein the analyzing data of the image to determine a probability at which the image is from a physical document comprises:

performing image recognition processing on data of an individual frame in the image.

16. A non-transitory machine-readable storage medium encoded with instructions executable by a hardware processor of a document verification device, the machine-readable storage medium comprising instructions to cause the hardware processor to perform operations comprising:

obtaining an image of a to-be-verified document;

providing the image as input to a trained classification model;

obtaining an output of the trained classification model, wherein the output represents a probability at which the image passes an anti-counterfeiting identifier verification;

verifying whether an anti-counterfeiting identifier exists in the image based on the output of the trained classification model;

performing first verification on the image to determine a probability at which the image is from a screen print;

performing second verification on the image to determine a probability at which the image is from a copy;

determining a probability at which the image is from a physical document based on the probability at which the image is from a screen print and the probability at which the image is from a copy;

verifying a source of the image based on the probability at which the image is from a physical document; and

obtaining a verification result for the to-be-verified document based on the verifying of the anti-counterfeiting identifier and the verifying of the source of the image.

17. The non-transitory machine-readable storage medium according to claim 16 , wherein the verifying whether an anti-counterfeiting identifier exists in the image comprises:

determining a type of anti-counterfeiting identifier corresponding to a type of the to-be-verified document; and

determining a probability at which an anti-counterfeiting identifier corresponding to the determined type of anti-counterfeiting identifier exists in the image.

18. The non-transitory machine-readable storage medium according to claim 17 , wherein the type of anti-counterfeiting identifier comprises at least one of watermark text, micro-text, shading lines, gravure printing, color cast, anti-counterfeiting secret marks, or fonts.

19. The non-transitory machine-readable storage medium according to claim 16 , wherein the verifying a source of the image comprises:

analyzing data of the image to determine a probability at which the image is from a physical document.

20. The non-transitory machine-readable storage medium according to claim 19 , wherein the analyzing data of the image to determine a probability at which the image is from a physical document comprises:

performing image recognition processing on data of an individual frame in the image.