Anti-spoofing detection using multi-light point irradiation
AssigneeHuawei Cloud Computing Technologies Co., Ltd.
InventorsZiwen Hu, Anping Li, Xiaoyuan Yu, Pu Chen
Life of this patent
Estimating the term…
Abstract
A front-end device is configured to capture a light point image of an authenticatee and send the light point image to the back-end device, where the light point image is an image captured from the authenticatee under irradiation of multi-light points, and the light point image includes a face of the authenticatee. The back-end device of the authentication system is configured to perform face anti-spoofing detection on the authenticatee based on the received light point image to obtain an authentication result.
Claims22
An apparatus, comprising:
- an image sensor configured to capture a light point image of an authenticatee under irradiation of multi-light points, wherein the light point image comprises a plurality of central bright spots from reflected light, a plurality of diffusion spots from diffused light, and a non-light point region, wherein each of the central bright spots is surrounded by one of the diffusion spots, and wherein the diffusion spots are surrounded by the non-light point region;
- and a communication interface coupled to the image sensor and configured to:
- send, to a remote device, the light point image;
- and receive, from the remote device, an authentication result that is based on face anti-spoofing detection on the light point image.
An apparatus, comprising:
- a memory configured to store instructions;
- and one or more processors coupled to the memory and configured to execute the instructions to:
- receive a light point image of an authenticatee, wherein the light point image is from the authenticatee under irradiation of first multi-light points, wherein the light point image comprises a plurality of central bright spots from reflected light, a plurality of diffusion spots from diffused light, and a non-light point region, wherein each of the central bright spots is surrounded by one of the diffusion spots, and wherein the diffusion spots are surrounded by the non-light point region;
- and a face of the authenticatee;
- and perform, based on a face anti-spoofing detection model and the light point image, face anti-spoofing detection on the authenticatee to obtain an authentication result, wherein the face anti-spoofing detection model is a neural network model trained based on images in a light point image training set.
A method, comprising:
- obtaining a light point image of an authenticatee, wherein the light point image is from the authenticatee under irradiation of first multi-light points, wherein the light point image comprises a plurality of central bright spots from reflected light, a plurality of diffusion spots from diffused light, and a non-light point region, wherein each of the central bright spots is surrounded by one of the diffusion spots, and wherein the diffusion spots are surrounded by the non-light point region;
- and performing, based on the light point image, face anti-spoofing detection on the authenticatee to obtain an authentication result.
Dependencies are inferred from the claim text.