Verification of user photo IDs
View Patent ↗The presence of a target object within a query image may be identified based on the spatial consistency of feature points matched between the query image and a template image describing the target object. A query image and a template image describing a target object to be sought in a query image are received. Feature points are extracted from the query image and the template image using a SIFT method. Query feature points are each matched to the nearest neighbor template feature point. A matched pair of query points is assigned a confidence indicator based on the distance between each parameter of the query feature point and template feature point. The confidence indicators are mapped within a binned four-dimensional space. If the number of confidence indicators mapped to any bin is greater than a threshold value, then the target object has been detected within the query image.
1. A computer-implemented method, comprising:
receiving, by a processor, a query image;
extracting, by a processor, a plurality of query feature points from the query image, wherein each query feature point is a vector comprising an appearance-based descriptor, two spatial coordinates, a scale factor, and an orientation factor;
matching, by a processor, the query feature points to one of a plurality of template feature points in a template image;
calculating, by a processor, a confidence indicator for each query feature point match;
mapping, by a processor, the confidence indicators within a four dimensional space divided into a plurality of bins, a first dimension of the four dimensional space corresponding to a first of the two spatial coordinates, a second dimension of the four dimensional space corresponding to a second of the two spatial coordinates, a third dimension of the four dimensional space corresponding to the scale factor, and a fourth dimension of the four dimensional space corresponding to the orientation factor; and
determining, by a processor, that the query image contains a match to the template image when a number of confidence indicators in at least one of the plurality of bins is greater than a threshold value.
2. The computer-implemented method of claim 1 , further comprising:
receiving, by a processor, a template image; and
extracting, by a processor, a plurality of template feature points from a template image.
3. The computer-implemented method of claim 1 , wherein matching the query feature points to the template feature points comprises, for each query feature point:
calculating, by a processor, a distance from a query feature point descriptor to each descriptor corresponding to the template feature points; and
identifying, by a processor, the template feature point having the closest descriptor as the matching template feature point.
4. The computer-implemented method of claim 1 , wherein calculating the confidence indicator comprises calculating a difference between each respective coordinate of a matched query feature point and template feature point.
5. The computer-implemented method of claim 1 , wherein the template image is an image of a government issued identification.
6. The computer-implemented method of claim 1 , wherein the query image contains an image of a government issued identification.
7. The computer-implemented method of claim 1 , further comprising:
locking, by a processor, out a user from a user account; and
receiving, by a processor, a request from the user to access the user account.
8. A non-transitory computer-readable medium storing instructions, which when executed by a processing device, cause the processing device to perform method comprising:
receiving a query image;
extracting a plurality of query feature points from the query image wherein each query feature point is a vector comprising an appearance-based descriptor, two spatial coordinates, a scaling factor, and an orientation factor;
matching the query feature points to one of a plurality of template feature points in a template image;
calculating a confidence indicator for each query feature point match;
mapping the confidence indicators within a four dimensional space divided into a plurality of bins, a first dimension of the four dimensional space corresponding to a first of the two spatial coordinates, a second dimension of the four dimensional space corresponding to a second of the two spatial coordinates, a third dimension of the four dimensional space corresponding to the scale factor, and a fourth dimension of the four dimensional space corresponding to the orientation factor; and
determining that the query image contains a match to the template image when a number of confidence indicators in at least one of the plurality of bins is greater than a threshold value.
9. The non-transitory computer-readable medium of claim 8 , the method further comprising:
receiving a template image; and
extracting a plurality of template feature points from a template image.
10. The non-transitory computer-readable medium of claim 8 , wherein matching the query feature points to the template feature points comprises, for each query feature point:
calculating the distance from a query feature point descriptor to each descriptor corresponding to the template feature points; and
identifying the template feature point having the closest descriptor as the matching template feature point.
11. The non-transitory computer-readable medium of claim 8 , wherein calculating the confidence indicator comprises calculating a difference between each respective coordinate of a matched query feature point and template feature point.
12. The non-transitory computer-readable medium of claim 8 , wherein the template image is an image of a government issued identification.
13. The non-transitory computer-readable medium of claim 8 , wherein the query image contains an image of a government issued identification.
14. An apparatus comprising:
a storage device to store instructions; and
a processing device, wherein the processing device executes instructions that cause the apparatus to:
receive a query image;
extract a plurality of query feature points from the query image wherein each query feature point is a vector comprising an appearance-based descriptor, two spatial coordinates, a scaling factor, and an orientation factor;
match the query feature points to one of a plurality of template feature points in a template image;
calculate a confidence indicator for each query feature point match;
map the confidence indicators within a four dimensional space divided into a plurality of bins, a first dimension of the four dimensional space corresponding to a first of the two spatial coordinates, a second dimension of the four dimensional space corresponding to a second of the two spatial coordinates, a third dimension of the four dimensional space corresponding to the scale factor, and a fourth dimension of the four dimensional space corresponding to the orientation factor; and
determine that the query image contains a match to the template image when a number of confidence indicators in at least one of the plurality of bins is greater than a threshold value.
15. The apparatus of claim 14 , wherein the instructions cause the processor to match the query feature points to the template feature points by, for each query feature point:
calculating the distance from a query feature point descriptor to each descriptor corresponding to the template feature points; and
identifying the template feature point having the closest descriptor as the matching template feature point.
16. The apparatus of claim 14 , wherein the instructions cause the processor to calculate the confidence indicator by calculating a difference between each respective coordinate of a matched query feature point and template feature point.