Reference image enrollment and evolution for security systems
Person or object authentication can be performed using artificial intelligence-enabled systems. Reference information, such as for use in comparisons or assessments for authentication, can be updated over time to accommodate changes in an individual's appearance, voice, or behavior. In an example, reference information can be updated automatically with test data, or reference information can be updated conditionally, based on instructions from a system administrator. Various types of media can be used for authentication, including image information, audio information, or biometric information. In an example, authentication can be performed wholly or partially at an edge device such as a security panel in an installed security system.
1 . A method for authenticating, the method comprising:
receiving reference image information about a face of a first person, wherein the reference image information about the face of the first person includes information about the face of the first person when the face of the first person is fully visible without the first person wearing a face mask;
receiving a test image including information about a face of an unknown person when the face of the unknown person is partially obscured by the unknown person wearing a partial face mask; and
determining a relatedness metric indicating at least a numerical value of similarity between the first person and the unknown person by comparing a plurality of features of the face of the first person and the face of the unknown person, and based on the relatedness metric at least one of:
generating updated reference image information using the test image; and
providing a validation request to an administrator, wherein the validation request includes the relatedness metric.
2 . The method of claim 1 , wherein the partial face mask covers less than all of the face of the unknown person.
3 . The method of claim 1 , wherein determining the relatedness metric includes applying the information about the test image to a neural network-based classifier to determine a likelihood that the unknown person in the test image is the first person.
4 . The method of claim 1 , wherein the reference image information about the face of the first person is received at a first time when the first person has first facial characteristics, and wherein the test image includes information about the face of the first person at a second time when the first person has different second facial characteristics.
5 . The method of claim 1 , comprising receiving reference image information about a first vehicle from a reference image of the first vehicle, wherein the test image includes information about the first vehicle or a different vehicle.
6 . The method of claim 5 , wherein receiving the reference image information about the first vehicle includes receiving information about one or more reference attributes of the first vehicle, wherein the one or more reference attributes of the first vehicle are determined using the reference image of the first vehicle, and the method further comprising:
identifying a second vehicle in the test image; and
identifying candidate attributes of the second vehicle using the test image;
wherein determining the relatedness metric includes comparing the one or more reference attributes of the first vehicle and the candidate attributes of the second vehicle.
7 . The method of claim 1 , further comprising:
receiving a validation result from the administrator; and
updating the reference image information using the test image based on the validation result, wherein updating the reference image information includes providing the test image to a machine learning system.
8 . The method of claim 7 , wherein the machine learning system is configured to recognize an evolutionary characteristic between the test image and the reference image information and to apply the evolutionary characteristic to enable recognition of other candidate objects in other test images.
9 . A method for authenticating, the method comprising:
receiving reference image information about a first object;
receiving a test image including information about a second object;
determining an elapsed time between receiving the reference image information and receiving the test image; and
responsive to the elapsed time exceeding a specified threshold elapsed time limit,
providing a validation request to an administrator, wherein the validation request includes one or more of a relatedness metric, information about the first object, or information about the second object, wherein the relatedness metric indicates a magnitude of relatedness of the first object to the second object.
10 . The method of claim 9 , comprising determining the relatedness metric using a neural network-based classifier to determine a likelihood that the second object is the same object as the first object.
11 . The method of claim 9 , wherein the information about the second object includes information from a sensor, wherein the sensor includes one or more of an image sensor, an audio sensor, a biometric sensor, or an environmental condition sensor.
12 . The method of claim 11 , comprising determining the relatedness metric including transmitting at least one of the information about the second object or the information from the sensor to a neural network-based classifier to determine a likelihood that the second object in the test image is the first object.
13 . The method of claim 9 , further comprising:
receiving a validation result from the administrator; and
updating the reference image information using the test image based on the validation result, wherein updating the reference image information includes providing the test image to a machine learning system, and wherein the machine learning system is configured to recognize an evolutionary characteristic between the test image and the reference image information and to apply the evolutionary characteristic to enable recognition of other candidate objects in other test images.
14 . A system for remote authentication, the system comprising:
an authorization device to maintain reference information about one or more objects; and
at least one edge device, remote from and communicatively coupled to the authorization device, wherein the at least one edge device includes processing circuitry to:
receive reference information about a target object from at least one of a local database or the authorization device, wherein the reference information includes information about a face of a first person when the face of the first person is fully visible without the first person wearing a face mask;
receive a test image including information about a candidate object, wherein the test image includes information about the face of the first person when the face of the first person is partially obscured by the unknown person wearing a partial face mask;
determine an elapsed time between acquisition of the reference information and the test image; and
determine a relatedness metric indicating at least a numerical value of similarity between the target object and the candidate object by comparing a plurality of features of the face of the first person and the face of the unknown person, and based on the relatedness metric, at least one of:
generate updated reference information using the test image; and
provide a validation request to the authorization device, wherein the validation request includes the relatedness metric, and is provided to the authorization device when the elapsed time exceeds a specified threshold elapsed time limit.
15 . The system of claim 14 , wherein the authorization device is configured to receive the validation request from the at least one edge device and update the reference information based on the information about the candidate object.
16 . The system of claim 15 , wherein the authorization device is further configured to provide an access control instruction to the at least one edge device in response to receiving the validation request, and wherein the at least one edge device is configured to use the access control instruction to control operation of a physical barrier or generate an alert.
17 . The system of claim 14 , wherein the processing circuitry is further to:
generate updated reference information using information about the candidate object when the relatedness metric as-determined, meets or exceeds a specified relatedness threshold.
18 . The system of claim 14 , wherein the authorization device is configured to update the reference information about the target object using data received from at least one of a system administrator or one or more sensors communicatively couple to the authorization device.
19 . A method for authenticating, the method comprising:
receiving reference image information about a face of a first person, wherein the reference image information about the face of the first person includes information about the face of the first person when the face of the first person is fully visible without the first person wearing a face mask;
receiving a test image including information about a face of an unknown person when the face of the unknown person is partially obscured by the unknown person wearing a partial face mask; and
determining a relatedness metric indicating a numerical value of similarity of the faces of the first person and the unknown person by comparing a plurality of features of the face of the first person and the face of the unknown person, and based on the relatedness metric at least one of:
generating updated reference image information using the test image; and
providing a validation request to an administrator, wherein the validation request includes one or more of the relatedness metric, information about the first person, or information about the unknown person,
wherein the information about the face of the first person when the face of the first person is fully visible includes information about the face of the first person when the first person is without a face mask, and wherein the information about the face of the first person when the face of the first person is partially obscured includes information about a face of the first person when the first person is wearing a partial face mask.