Document evaluation based on bounding boxes
The disclosure includes a system and method for obtaining at least one image of a document under test; determining a classification of the document under test; obtaining a document assembly object associated with the determined classification; performing object detection on the document under test; and evaluating the objects detected in the document under test against the document assembly object; and determining whether the document under test is a valid or abused document.
1 . A method comprising:
obtaining, using one or more processors, at least one image of a document under test;
determining, using the one or more processors, a classification of the document under test;
obtaining, using the one or more processors, a document assembly object associated with the determined classification, wherein the document assembly object was previously generated for a previously unsupported document thereby adding support for the previously unsupported document, wherein the document assembly object describes valid document instances, the document assembly object including a first check based on computer-vision generated bounding boxes derived from one or more detected objects common to a plurality of images of valid samples of the previously unsupported document, wherein the document under test is an instance of the previously unsupported, but now supported using the generated document assembly object, document;
performing, using the one or more processors, object detection on the at least one image of the document under test; and
evaluating, using the one or more processors, objects detected in the at least one image of the document under test against the document assembly object; and
determining, using the one or more processors, whether the document under test is a valid or abused document.
2 . The method of claim 1 , wherein the document assembly object includes one or more of:
a reconstructed background describing at least a portion of a background in valid document instances; and
a reconstructed microprinting describing at least a portion of microprinting in valid document instances.
3 . The method of claim 2 , wherein evaluating the objects detected in the document under test includes one or more of:
analyzing a non-occluded portion of the background within a bounding box associated with the document under test against a corresponding area in the reconstructed background included in the document assembly object; and
analyzing a non-occluded portion of the microprinting within the bounding box associated with the document under test against the corresponding area in the reconstructed microprinting included in the document assembly object.
4 . The method of claim 3 , wherein the evaluation is one or more of based on an average color in the document under test as compared to the average color in the reconstructed background, based on a frequency domain of color in the document under test as compared to a frequency domain of color in the reconstructed background, based on boundary characteristics between text and the background or microprint.
5 . The method of claim 1 , wherein the at least one image of the document under test includes a composite image based on multiple images and describing an optically dynamic security feature.
6 . The method of claim 1 , wherein the at least one image of the document under test includes a result of an edge detection, the result of the edge detection representing a presence or absence of a UV watermark.
7 . The method of claim 1 , wherein object detection on the document under test includes one or more of optical character recognition for text objects and object recognition of at least one non-text object, wherein the non-text object is associated with a security feature.
8 . The method of claim 1 , wherein the document assembly object further includes a direct check derived from issuer information.
9 . The method of claim 1 , wherein the evaluating the objects detected in the at least one image of the document under test includes one or more of evaluating a presence or absence of an object, an inter-bounding box evaluation, and an intra-bounding box evaluation.
10 . A system comprising:
a processor; and
a memory, the memory storing instructions that, when executed by the processor, cause the system to:
obtain at least one image of a document under test;
determine a classification of the document under test;
obtain a document assembly object associated with the determined classification, wherein the document assembly object was previously generated for a previously unsupported document thereby adding support for the previously unsupported document, wherein the document assembly object describes valid document instances, the document assembly object including a first check based on computer-vision generated bounding boxes derived from one or more detected objects common to a plurality of images of valid samples of the previously unsupported document, wherein the document under test is an instance of the previously unsupported, but now supported using the generated document assembly object, document;
perform object detection on the at least one image of the document under test; and
evaluate objects detected in the at least one image of the document under test against the document assembly object; and
determine whether the document under test is a valid or abused document.
11 . The system of claim 10 , wherein the document assembly object includes one or more of:
a reconstructed background describing at least a portion of a background in valid document instances; and
a reconstructed microprinting describing at least a portion of microprinting in valid document instances.
12 . The system of claim 11 , wherein evaluating the objects detected in the document under test includes one or more of:
analyzing a non-occluded portion of the background within a bounding box associated with the document under test against a corresponding area in the reconstructed background included in the document assembly object; and
analyzing a non-occluded portion of the microprinting within the bounding box associated with the document under test against the corresponding area in the reconstructed microprinting included in the document assembly object.
13 . The system of claim 12 , wherein the evaluation is one or more of based on an average color in the document under test as compared to the average color in the reconstructed background, based on a frequency domain of color in the document under test as compared to a frequency domain of color in the reconstructed background, based on boundary characteristics between text and the background or microprint.
14 . The system of claim 10 , wherein the image of the document under test includes a composite image based on multiple images and describing an optically dynamic security feature.
15 . The system of claim 10 , wherein the at least one image of the document under test includes a result of an edge detection, the result of the edge detection representing a presence or absence of a UV watermark.
16 . The system of claim 10 , wherein object detection on the document under test includes one or more of optical character recognition for text objects and object recognition of at least one non-text object, wherein the non-text object is associated with a security feature.
17 . The system of claim 10 , wherein the document assembly object further includes a direct check derived from issuer information.
18 . The system of claim 10 , wherein the evaluating the objects detected in the at least one image of the document under test includes one or more of evaluating a presence or absence of an object, an inter-bounding box evaluation, and an intra-bounding box evaluation.