Systems and methods for form recognition using visual signatures
Systems and methods for form recognition using visual signatures are disclosed. According to one embodiment, a first form of a first document defined by a first plurality of grids and a second form of a second document defined by a second plurality of grids are received. A determination is made as to whether the first form is similar to the second form by computing a pairwise difference between a first grid of the first plurality of grids with a second grid of the second plurality of grids.
1 . A computer-implemented method, comprising:
receiving a first form of a first document defined by a first plurality of grids and a second form of a second document defined by a second plurality of grids, wherein each grid of the first plurality of grids comprises a plurality of rows of pixels and a plurality of columns of pixels, and wherein each grid of the second plurality of grids comprises a plurality of rows of pixels and a plurality of columns of pixels, wherein the pixels of the first and second pluralities of grids comprise grayscale pixels varying in value between black and white;
for each grid of the first plurality of grids, summing grayscale values of all of the pixels in the plurality of rows and plurality of columns of the grid to generate a respective feature for each such grid of the first plurality of grids;
for each grid of the second plurality of grids, summing grayscale values of all of the pixels in the plurality of rows and plurality of columns of the grid to generate a respective feature for each such grid of the second plurality of grids; and
determining if the first form is similar to the second form by computing a Euclidean distance between the first plurality of grids and the second plurality of grids, wherein computing the Euclidean distance comprises calculating a pairwise difference comprising a square root of sum of squares between the respective feature of each grid of the first plurality of grids and the respective feature of a corresponding grid of the second plurality of grids.
2 . The computer-implemented method of claim 1 , wherein determining if the first form is similar to the second form comprises determining the second form is similar to the first form if the pairwise difference is less than a threshold.
3 . The computer-implemented method of claim 1 , wherein determining if the first form is similar to the second form comprises determining the second form is different from the first form if the pairwise difference is greater than a threshold.
4 . The computer-implemented method of claim 3 , further comprising clustering a plurality of forms using a silhouette score into a first cluster of forms and a second cluster of forms.
5 . The computer-implemented method of claim 4 , further comprising using a first data extraction process specific to a first structure of the first cluster of forms and using a second data extraction process specific to a second structure of the second cluster of forms.
6 . The computer-implemented method of claim 5 , further comprising assigning a unique identifier to each form of the first and second cluster of forms.
7 . The computer-implemented method of claim 6 , wherein the unique identifier comprises a document signature.
8 . The computer-implemented method of claim 7 , wherein the document signature comprises a document signature selected from the group consisting of a visa and immigration signature, a driving license signature and an invoice signature.
9 . A system, comprising:
a plurality of document databases containing forms, wherein the plurality of document databases comprises a first form of a first document database and a second form of a second document database; and
a document analysis server that is configured to perform the steps of:
receiving the first form of the first document database defined by a first plurality of grids and the second form of the second document database defined by a second plurality of grids, wherein each grid of the first plurality of grids comprises a plurality of rows of pixels and a plurality of columns of pixels, and wherein each grid of the second plurality of grids comprises a plurality of rows of pixels and a plurality of columns of pixels, wherein the pixels of the first and second pluralities of grids comprise grayscale pixels varying in value between black and white;
for each grid of the first plurality of grids, summing grayscale values of all of the pixels in the plurality of rows and plurality of columns of the grid to generate a respective feature for each such grid of the first plurality of grids;
for each grid of the second plurality of grids, summing grayscale values of all of the pixels in the plurality of rows and plurality of columns of the grid to generate a respective feature for each such grid of the second plurality of grids; and
determining if the first form is similar to the second form by computing a Euclidean distance between the first plurality of grids and the second plurality of grids, wherein computing the Euclidean distance comprises calculating a pairwise difference comprising a square root of sum of squares between the respective feature of each grid of the first plurality of grids and the respective feature of a corresponding grid of the second plurality of grids.
10 . The system of claim 9 , further comprising a network interconnecting the plurality of document databases.
11 . The system of claim 9 , wherein the document analysis server determines if the first form is similar to the second form if the pairwise difference is less than a threshold.
12 . The system of claim 9 , wherein the document analysis server determines if the first form is different from the second form if the pairwise difference is greater than a threshold.
13 . The system of claim 12 , wherein the document analysis server clusters a plurality of forms using a silhouette score into a first cluster of forms and a second cluster of forms.
14 . The system of claim 13 , wherein the document analysis server uses a first data extraction process specific to a first structure of the first cluster of forms and uses a second data extraction process specific to a second structure of the second cluster of forms.
15 . The system of claim 14 , wherein the document analysis server assigns a unique identifier to each form of the first and second cluster of forms.
16 . The system of claim 15 , wherein the unique identifier comprises a document signature.
17 . The system claim 16 , wherein the document signature comprises a document signature selected from the group consisting of a visa and immigration signature, a driving license signature and an invoice signature.