System and method for extracting target data from labels

1. A computing device for extracting target data from a source document, the computing device comprising:

a memory storing target data extraction rules;

a processor connected with the memory, the processor configured to:

obtain text recognition data returned from a text recognition process applied to an image of the source document, the text recognition data indicating locations of text structures on the source document;

define text lines based on the text recognition data by a smallest bounding box surrounding all words satisfying a same-line condition, wherein the same-line condition includes: (a) character heights of the words are within a predetermined percentage, and (b) the words are horizontally aligned;

identify a reference string from the text recognition data;

select a subset of the text lines based on a location of the reference string and the target data extraction rules; and

output the subset of the text lines as the target data.

2. The computing device of claim 1 , wherein the processor is configured to:

obtain the image of the source document; and

apply an optical character recognition process to the image to obtain the text recognition data.

3. The computing device of claim 2 , wherein the processor is further configured, prior to applying the optical character recognition process to the image, to:

detect superfluous features of the source document; and

remove the superfluous features of the source document.

4. The computing device of claim 1 , wherein the processor is configured, to define the text lines, to:

select, from the text recognition data, a leading word;

determine if any additional words defined in the text recognition data satisfy the same-line condition; and

define the smallest bounding box for the text line, the smallest bounding box surrounding all the additional words satisfying the same-line condition.

5. The computing device of claim 4 , wherein the same-line condition is further based on one or more of: a distance between words and a word orientation.

6. The computing device of claim 1 , wherein the processor is further configured to:

obtain barcode data representing a location of a barcode;

based on the barcode data, identify an approximate location of the reference string; and

select a searching subset of text lines within a threshold distance of the approximate location of the reference string; wherein the reference string is identified in one of the text lines in the searching subset.

7. The computing device of claim 1 , wherein the processor is further configured to:

obtain barcode data representing a location of a barcode; and

based on the barcode data, verify the target data based on a relative spatial relationship between the barcode and the subset of the text lines.

8. The computing device of claim 1 , wherein the processor is configured, to identify the reference string, to:

identify a word in the text recognition data matching a predefined regular expression as a potential reference string; and

verify the potential reference string against a predetermined list of valid reference strings.

9. The computing device of claim 1 , wherein the reference string is a ZIP code, and wherein the target data is a postal address.

10. The computing device of claim 9 , wherein, to select the subset of the text lines representing the postal address, the processor is configured to:

select, as part of the subset, the text line containing the ZIP code;

select, as part of the subset, text lines in a block having at least one text line within a threshold distance of the text line containing the ZIP code;

discard text lines failing font homogeneity and alignment conditions; and

discard text lines failing to match a regular expression.

11. A method for extracting target data from a source document, the method comprising:

storing target data extraction rules;

obtaining text recognition data returned from a text recognition process applied to an image of the source document, the text recognition data indicating locations of text structures on the source document;

defining text lines based on the text recognition data by a smallest bounding box surrounding all words satisfying a same-line condition, wherein the same-line condition includes: (a) character heights of the words are within a predetermined percentage, and (b) the words are horizontally aligned;

identifying a reference string from the text recognition data;

selecting a subset of the text lines based on a location of the reference string and the target data extraction rules; and

outputting the subset of the text lines as the target data.

12. The method of claim 11 , further comprising:

obtaining the image of the source document; and

applying an optical character recognition process to the image to obtain the text recognition data.

13. The method of claim 12 , further comprising, prior to applying the optical character recognition process to the image:

detecting superfluous features of the source document; and

removing the superfluous features of the source document.

14. The method of claim 11 , wherein defining the text lines comprises:

selecting, from the text recognition data, a leading word;

determining if any additional words defined in the text recognition data satisfy the same-line condition; and

defining the smallest bounding box for the text line, the smallest bounding box surrounding all the additional words satisfying the same-line condition.

15. The method of claim 14 , wherein the same-line condition is based on one or more of: a distance between words and a word orientation.

16. The method of claim 11 , further comprising:

obtaining barcode data representing a location of a barcode;

based on the barcode data, identifying an approximate location of the reference string; and

selecting a searching subset of text lines within a threshold distance of the approximate location of the reference string; wherein the reference string is identified in one of the text lines in the searching subset.

17. The method of claim 11 , further comprising:

obtaining barcode data representing a location of a barcode; and

based on the barcode data, verifying the target data based on a relative spatial relationship between the barcode and the subset of the text lines.

18. The method of claim 11 , wherein identifying the reference string comprises:

identifying a word in the text recognition data matching a predefined regular expression as a potential reference string; and

verifying the potential reference string against a predetermined list of valid reference strings.

19. The method of claim 11 , wherein the reference string is a ZIP code, and wherein the target data is a postal address.

20. The method of claim 19 , wherein selecting the subset of the text lines representing the postal address comprises:

selecting, as part of the subset, the text line containing the ZIP code;

selecting, as part of the subset, text lines in a block having at least one text line within a threshold distance of the text line containing the ZIP code;

discarding text lines failing font homogeneity and alignment conditions; and

discarding text lines failing to match a regular expression.