Patient identification systems and methods
Disclosed techniques relate to identifying subjects in digital images. In some embodiments, intake digital images ( 404 ) are acquired ( 1002 ) that capture a first subject. A subset of the intake digital images is selected ( 1004 ) that depict multiple different views of the first subject's face. Based on the selected subset of intake digital images, first subject reference templates are generated and stored in a subject database ( 412 ). Later, a second subject is selected ( 1008 ) for identification within an area. Associated second subject reference templates are retrieved ( 1010 ) from the subject reference database. Digital image(s) ( 420 ) that depict the area are acquired ( 1012 ). Portion(s) of the digital image(s) that depict faces of subject(s) in the area are detected ( 1014 ) as detected face image(s). A given detected face image is compared ( 1016 ) to the second subject reference templates to identify the second subject ( 1018 ) in the digital image(s) that capture the area.
1. A method implemented by one or more processors, the method comprising:
Acquiring with at least one camera a plurality of intake digital images that capture at least a first subject in an area being monitored by the at least one camera;
selecting, from the plurality of intake digital images, a subset of intake digital images that depict multiple different views of a face of the first subject;
generating, based on the selected subset of intake digital images, first subject reference templates, wherein the first subject reference templates are stored in a subject reference database in association with information related to the first subject, the subject reference database capable of storing subject reference templates for a plurality of subjects;
selecting a second subject to identify within the area;
acquiring a plurality of intake digital images of the second subject;
selecting, from the plurality of intake digital images, a subset of intake digital images that depict multiple different views of a face of the second subject;
generating, based on the selected subset of intake digital images of the second subject, second subject reference templates, which are stored in the subject reference database in association with information related to the second subject;
retrieving second subject reference templates from the subject reference database;
acquiring one or more digital images that depict the area;
detecting, as one or more detected face images, one or more portions of the one or more digital images that depict faces of one or more subjects in the area;
comparing a given detected face image of the detected one or more detected face images to the second subject reference templates; and
identifying, based on the comparing, if the second subject is in the one or more digital images that capture the area.
2. The method of claim 1 , wherein the area comprises a waiting room, the intake images are acquired using a first camera that is configured to capture a registration or triage area, and the digital images that depict the waiting room are acquired using a second camera that is configured to capture the waiting room.
3. The method of claim 1 , wherein the comparing comprises applying the given detected face image as input across a trained machine learning model to generate output that indicates a measure of similarity between the given detected face image and the second subject, wherein the machine learning model is trained based at least in part on the second subject reference templates.
4. The method of claim 3 , wherein the trained machine learning model comprises a linear discriminant analysis model.
5. The method of claim 4 , further comprising retraining the machine learning model in response to a new subject being added to the subject reference database or an existing subject being removed from the subject reference database.
6. The method of claim 3 , wherein the trained machine learning model is trained based on the subject reference templates related to the plurality of subjects.
7. The method of claim 1 , wherein one or more of the subset of intake digital images are selected based on being sufficiently dissimilar to one or more other intake digital images.
8. The method of claim 1 , further comprising normalizing the one or more face images so that each detected face image depicts a frontal view of a face.
9. The method of claim 8 , wherein the normalizing includes geometric warping.
10. A system comprising one or more processors and memory operably coupled with the one or more processors, wherein the memory stores instructions that, in response to execution of the instructions by one or more processors, cause the one or more processors to:
Acquire from at least one camera a plurality of intake digital images that capture at least a first subject in an area being monitored by the at least one camera;
select, from the plurality of intake digital images, a subset of intake digital images that depict multiple different views of a face of the first subject;
generate, from the selected subset of intake digital images, first subject reference templates, wherein the first subject reference templates are stored in a subject reference database in association with information related to the first subject, the subject reference database capable of storing subject reference templates for a plurality of subjects;
select a second subject to identify within the area;
acquire a plurality of intake digital images of the second subject;
select, from the plurality of intake digital images, a subset of intake digital images that depict multiple different views of a face of the second subject;
generate, based on the selected subset of intake digital images of the second subject, second subject reference templates, which are stored in the subject reference database in association with information related to the second subject;
retrieve second subject reference templates related to the second subject from the subject reference database;
acquire one or more digital images that depict the area;
detect, as one or more detected face images, one or more portions of the one or more digital images that depict faces of one or more subjects in the area;
compare a given detected face image of the detected one or more detected face images to the second subject reference templates; and
identify, based on the comparing, if the second subject is in the one or more digital images that capture the area.
11. The system of claim 10 , wherein the area comprises a waiting room, the intake images are acquired using a first camera that is configured to capture a registration or triage area, and the digital images that depict the waiting room are acquired using a second camera that is configured to capture the waiting room.
12. The system of claim 10 , further comprising instructions to apply the given detected face image as input across a trained machine learning model to generate output that indicates a measure of similarity between the given detected face image and the second subject, wherein the machine learning model is trained based at least in part on the second subject reference templates.
13. The system of claim 12 , wherein the trained machine learning model comprises a linear discriminant analysis model.
14. At least one non-transitory computer-readable medium comprising instructions that, in response to execution of the instructions by one or more processors, cause the one or more processors to perform the following operations:
acquiring with at least one camera a plurality of intake digital images that capture at least a first subject in an area being monitored by the at least one camera;
selecting, from the plurality of intake digital images, a subset of intake digital images that depict multiple different views of a face of the first subject;
generating, based on the selected subset of intake digital images, first subject reference templates, wherein the first subject reference templates are stored in a subject reference database in association with information related to the first subject, the subject reference database capable of storing subject reference templates for a plurality of subjects;
selecting a second subject to identify within the area;
acquiring a plurality of intake digital images of the second subject;
selecting, from the plurality of intake digital images, a subset of intake digital images that depict multiple different views of a face of the second subject;
generating, based on the selected subset of intake digital images of the second subject, second subject reference templates, which are stored in the subject reference database in association with information related to the second subject;
retrieving second subject reference templates from the subject reference database;
acquiring one or more digital images that depict the area;
detecting, as one or more detected face images, one or more portions of the one or more digital images that depict faces of one or more subjects in the area;
comparing a given detected face image of the detected one or more detected face images to the second subject reference templates; and
identifying, based on the comparing, if the second subject is in the one or more digital images that capture the area.
15. The at least one non-transitory computer-readable medium of claim 14 , wherein the area comprises a waiting room, the intake images are acquired using a first camera that is configured to capture a registration or triage area, and the digital images that depict the waiting room are acquired using a second camera that is configured to capture the waiting room.