System and method for OCR based object registration
Some implementations of the present disclosure are directed to a computer-implemented method that includes: detecting a plurality of characters from a photo of an object; generating a set of character-based features based on the plurality of characters; matching the set of character-based features with a template of feature sets obtained from known objects; based on a matching set of character-based features, establishing a matching transformation between the object in the photo and the template of feature sets; and projecting the matching transformation to the photo such that the object is segmented from the photo.
1. A computer-implemented method, comprising:
generating a set of character-based features extracted from an object in a photo;
based on, at least in part, the set of character-based features, identifying a matching transformation capable of being projected to the photo containing the object such that the object is segmented from the photo;
in response to identifying the matching transformation, providing, to a template of feature sets, the set of character-based features extracted from the object in the photo;
matching the template of feature sets with a later set of character-based features extracted from an object in a later received photo;
based on a matching set of character-based features, establishing a later matching transformation between the object in the later received photo and the template of feature sets; and
projecting the later matching transformation to the later received photo such that the object on the later received photo is segmented from the later received photo.
2. The computer-implemented method of claim 1 , wherein the object comprises: an identification document.
3. The computer-implemented method of claim 1 , further comprising:
providing, to the template of feature sets, the later set of character-based features extracted from the object in the later received photo.
4. The computer-implemented method of claim 1 , wherein the set of characters comprise: a triplet of characters.
5. The computer-implemented method of claim 1 , wherein generating a set of character-based features comprises:
estimating spatial coordinates for each character from the set of characters.
6. The computer-implemented method of claim 1 , wherein identifying a matching transformation comprises identifying a homography capable of projecting a rotation of the photo, a translation of the photo, and a scaling of the photo.