US 11,989,961 B2
Authentication machine learning from multiple digital presentations
Frederic Jordan, Vevey (CH); Nicolas Rudaz, Vevey (CH); Yves Delacretaz, Vevey (CH); and Martin Kutter, Vevey (CH)
Assigned to ALPVISION S.A., Vevey (CH)
Filed by ALPVISION S.A., Vevey (CH)
Filed on Sep. 13, 2022, as Appl. No. 17/931,578.
Application 17/931,578 is a continuation of application No. 16/956,028, granted, now 11,461,582, previously published as PCT/EP2018/086442, filed on Dec. 20, 2018.
Claims priority of provisional application 62/608,352, filed on Dec. 20, 2017.
Prior Publication US 2023/0080164 A1, Mar. 16, 2023
Int. Cl. G06K 9/62 (2022.01); G06F 18/20 (2023.01); G06F 18/2111 (2023.01); G06F 18/214 (2023.01); G06F 18/24 (2023.01); G06T 7/70 (2017.01); G06V 10/70 (2022.01); G06V 10/771 (2022.01); G06V 20/00 (2022.01); G06V 20/10 (2022.01)
CPC G06V 20/95 (2022.01) [G06F 18/2111 (2023.01); G06F 18/214 (2023.01); G06F 18/24 (2023.01); G06F 18/285 (2023.01); G06T 7/70 (2017.01); G06V 10/771 (2022.01); G06V 10/87 (2022.01); G06V 20/10 (2022.01); G06T 2207/20081 (2013.01)] 4 Claims
OG exemplary drawing
 
1. A computer-implemented method for determining if an object is similar to a set of genuine objects, the method comprising:
a) obtaining a digital representation of the object to be authenticated;
b) inputting the digital representation of the object to be authenticated to a classifier algorithm to obtain a decision value indicating if the digital representation of the object to be authenticated is of sufficient quality to enable its classification as a fake or as a genuine object, wherein the classifier algorithm has been trained using only digital representations of a set of genuine objects, wherein the classifier algorithm has been trained using a plurality of digital representations for each of a plurality of genuine objects, each digital representation of each genuine object being acquired using a least one sensor, and wherein the plurality of digital signal representations for each genuine object includes digital representations obtained at a plurality of predetermined genuine object positions and orientations relative to the sensor position and orientation;
c) determining if the digital representation of the object to be authenticated is of a good enough quality to enable the object classification as a fake or as a genuine object according to the decision value;
d) in response to determining that the digital representation of the object to be authenticated is of a good enough quality to enable the object classification as a fake or as a genuine object, calculate a cross-correlation signal SNR based on the digital representation of the object to be authenticated and a reference pattern representative of the set of genuine objects and check that the cross-correlation signal SNR is above a predetermined threshold value to identify from the feature vectors that the object is similar to the set of genuine objects; and
e) in response to determining that the digital representation of the object to be authenticated is not of a good enough quality to enable the object classification as a fake or as a genuine object, identify that the object is not similar to the set of genuine objects but also not similar to any possible set of fake objects.