Face liveness detection method, system, apparatus, computer device, and storage medium

Patent metadata

Field	Value
Publication number	US-12347239-B2
Application number	US-202217964688-A
Country	US
Kind code	B2
Filing date	Oct 12, 2022
Priority date	Aug 14, 2020
Publication date	Jul 1, 2025
Grant date	Jul 1, 2025

Abstract

A face liveness detection method is provided, and includes: receiving an image transmitted by a terminal, the image including a face of an object; performing data augmentation on the image, to obtain an extended image corresponding to the image, a number of extended images corresponding to the image being more than one; performing liveness detection on the extended images corresponding to the image, to obtain intermediate detection results of the extended images, a liveness detection model used in liveness detection being obtained by performing model training on an initial neural network model according to a sample image and extended sample images corresponding to the sample image; and obtaining a liveness detection result of the object in the image after fusing the intermediate detection results of the extended images.

First claim

What is claimed is:

1. A face liveness detection method performed by a computer device, the method comprising:
receiving an image from a terminal, the image comprising a face of an object captured by a camera of the terminal;
performing data augmentation on the image, to obtain a plurality of extended images corresponding to the image, wherein each extended image is a variation of the image;
performing liveness detection on the plurality of extended images independently, to obtain a plurality of intermediate detection results of the plurality of the extended images, wherein each intermediate detection result is obtained from the liveness detection on a corresponding extended image, a liveness detection model used in liveness detection being obtained by performing model training on an initial neural network model according to a sample image and extended sample images corresponding to the sample image;
counting a first number of intermediate detection results indicating that the object is a liveness;
counting a second number of intermediate detection results indicating that the object is not a liveness;
obtaining, when the first number is greater than the second number, a final liveness detection result indicating that the object is a liveness; and
performing a predefined operation in accordance with an identity of the object and the final liveness detection result.

2. The method according to claim 1 , wherein the performing data augmentation on the image comprises:
obtaining a first transformation parameter corresponding to a preset geometric transformation manner; and
performing geometric transformation on the image according to the first transformation parameter and the preset geometric transformation manner corresponding to the first transformation parameter.

3. The method according to claim 2 , wherein the preset geometric transformation manner comprises at least one of image flipping, image cropping, image rotation, and image translation.

4. The method according to claim 2 , wherein the performing geometric transformation on the image according to the first transformation parameter and the preset geometric transformation manner corresponding to the first transformation parameter comprises at least one of the following:
flipping the image along a preset direction according to a random flipping parameter;
cropping the image according to a random cropping parameter;
rotating the image according to a random rotation parameter; and
translating the image along a preset direction according to a random translation parameter.

5. The method according to claim 1 , wherein the performing data augmentation on the image comprises:
obtaining a second transformation parameter corresponding to a preset image attribute adjustment manner; and
performing image attribute adjustment on the image according to the second transformation parameter and the preset image attribute adjustment manner corresponding to the second transformation parameter.

6. The method according to claim 5 , wherein the preset image attribute adjustment manner comprises at least one of image random occlusion processing, grayscale processing, image brightness adjustment, and image contrast adjustment.

7. The method according to claim 5 , wherein the performing image attribute adjustment on the image according to the second transformation parameter and the preset image attribute adjustment manner corresponding to the second transformation parameter comprises at least one of the following:
determining a random occluded area in the image according to a random number, and replacing a pixel value of each pixel in the random occluded area with a preset value;
performing grayscale processing on the image, to obtain a corresponding grayscale image;
adjusting brightness of the image according to a random brightness adjustment parameter; and
adjusting contrast of the image according to a random contrast adjustment parameter.

8. The method according to claim 1 , wherein the performing liveness detection on the plurality of extended images, to obtain a plurality of intermediate detection results of the plurality of the extended images comprises:
performing model training on the initial neural network model according to the sample image, the extended sample images corresponding to the sample image, and real liveness categories respectively corresponding to the sample image and the extended sample images, to obtain the liveness detection model; and
performing, by the liveness detection model, liveness detection respectively on the plurality of extended images corresponding to the image, to obtain the plurality of intermediate detection results of the plurality of extended images.

9. A computer device, comprising a memory and one or more processors, the memory storing a plurality of computer-readable instructions that, when executed by the one or more processors, cause the one or more processors to implement a face liveness detection method including:
receiving an image from a terminal, the image comprising a face of an object captured by a camera of the terminal;
performing data augmentation on the image, to obtain a plurality of extended images corresponding to the image, wherein each extended image is a variation of the image;
performing liveness detection on the plurality of extended images independently, to obtain a plurality of intermediate detection results of the plurality of the extended images, wherein each intermediate detection result is obtained from the liveness detection on a corresponding extended image, a liveness detection model used in liveness detection being obtained by performing model training on an initial neural network model according to a sample image and extended sample images corresponding to the sample image;
counting a first number of intermediate detection results indicating that the object is a liveness;
counting a second number of intermediate detection results indicating that the object is not a liveness;
obtaining, when the first number is greater than the second number, a final liveness detection result indicating that the object is a liveness; and
performing a predefined operation in accordance with an identity of the object and the final liveness detection result.

10. The computer device according to claim 9 , wherein the performing data augmentation on the image comprises:
obtaining a first transformation parameter corresponding to a preset geometric transformation manner; and
performing geometric transformation on the image according to the first transformation parameter and the preset geometric transformation manner corresponding to the first transformation parameter.

11. The computer device according to claim 10 , wherein the preset geometric transformation manner comprises at least one of image flipping, image cropping, image rotation, and image translation.

12. The computer device according to claim 10 , wherein the performing geometric transformation on the image according to the first transformation parameter and the preset geometric transformation manner corresponding to the first transformation parameter comprises at least one of the following:
flipping the image along a preset direction according to a random flipping parameter;
cropping the image according to a random cropping parameter;
rotating the image according to a random rotation parameter; and
translating the image along a preset direction according to a random translation parameter.

13. The computer device according to claim 9 , wherein the performing data augmentation on the image comprises:
obtaining a second transformation parameter corresponding to a preset ima

Assignees

• Tencent Tech Shenzhen Co Ltd

Inventors

• Luo Zhao
• Bai Kun

Classifications

• G06V10/82

  using neural networks · CPC title

• G06V40/172

  Classification, e.g. identification · CPC title

• G06V10/774

  Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06V10/24

  Aligning, centring, orientation detection or correction of the image · CPC title

• G06F21/32

  using biometric data, e.g. fingerprints, iris scans or voiceprints · CPC title