A multi-model-based face anti-counterfeiting method, storage medium and detection device

Abstract

The invention discloses a multi-model-based face anti-counterfeiting method, which includes: constructing a face detection model, a frame detection model and a two-classification anti-counterfeiting model respectively; completing face position recognition in an image to be detected through the face detection model, and sending the image containing face information that has passed the detection to the frame detection model; completing the detection of whether the image has a frame through the frame monitoring model, and when the image does not have a frame, it is sent to the two-classification anti-counterfeiting model; when there is a frame in the image and the face is within the frame range, the frame type is identified, so that the image that has passed the recognition is sent to the two-classification anti-counterfeiting model, and the image that has not passed the recognition is screened out; the two-classification anti-counterfeiting model completes the face authenticity judgment based on the face reflectivity and / or texture features in the obtained image. It can adapt to face anti-counterfeiting within different distances, does not require action coordination, and installs additional hardware, which better solves the face anti-counterfeiting problem.

Description

Technical Field

[0001] The present invention belongs to the field of face recognition, and in particular relates to a multi-model-based face anti-counterfeiting method, storage medium and detection equipment. Background Art

[0002] As the most widely used non-invasive biometric technology, the security of face recognition technology is becoming increasingly important. The first line of defense for face recognition security is face anti-counterfeiting technology, which is to distinguish between real faces and forged faces. Face anti-counterfeiting is an effective means to prevent illegal users from passing the recognition system with forged faces. The technical difficulty lies in how to ensure a high enough pass rate for real people while increasing the rejection rate for forged faces.

[0003] Current facial anti-counterfeiting mainly adopts 1. active action coordination, such as turning the head, raising the head, opening the mouth, and blinking; 2. acquiring multimodal information such as infrared images and depth information by adding infrared cameras and depth cameras; 3. using border information and texture features to distinguish forged faces, such as photo borders, photo reflections, screen textures, etc.

[0004] Among them, the anti-counterfeiting effect of motion coordination is better, but the experience is poor and the recognition speed is slow. Infrared or structured light methods require the addition of additional hardware and the cost is high. Single-frame silent live experience is better, but the recognition is more difficult and the scene adaptability is poor. In general, a single feature model is difficult to fully cover the endless counterfeiting methods and complex and changing application scenarios. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the prior art and provide a multi-model based face anti-counterfeiting method. Through the coordinated use of multiple models, it can adapt to face anti-counterfeiting within different distance ranges, and does not require action coordination or the installation of additional hardware, which can better solve the face anti-counterfeiting problem.

[0006] On the one hand, the purpose of the present invention is achieved by the following technical solutions:

[0007] A multi-model-based face anti-counterfeiting method comprises: respectively constructing a face detection model, a frame detection model and a binary classification anti-counterfeiting model; completing face position recognition in an image to be detected by the face detection model, and sending the image containing face information that has passed the detection to the frame detection model; completing detection of whether an image has a frame by the frame detection model, and when the image does not have a frame, the image is sent to the binary classification anti-counterfeiting model; when a frame exists in the image and the face is within the range of the frame, the frame type is identified, so that the image that has passed the recognition is sent to the binary classification anti-counterfeiting model, and the image that has not passed the recognition is screened out; the binary classification anti-counterfeiting model completes the face authenticity judgment based on the face reflectivity and / or texture features in the obtained image.

[0008] According to a preferred embodiment, the face detection model is configured to detect the rectangular frame position of the face in the picture.

[0009] According to a preferred embodiment, the frame detection model completes the recognition of whether the frame is a background object based on the motion state of the frame. When the frame to be recognized is recognized as a background object, the image is verified by the frame detection model.

[0010] According to a preferred embodiment, the motion state of the frame is achieved by collecting a number of frames of continuous motion images to complete the position comparison of the frames of each image type. If the frame is in a relatively static state, it is determined to be a background object.

[0011] According to a preferred embodiment, the frame types include background objects and non-background objects.

[0012] According to a preferred embodiment, the background objects at least include door and window frames, wall frames and billboard frames.

[0013] According to a preferred embodiment, in the binary classification anti-counterfeiting model, when the reflectivity and / or texture features of the face in the image exceed a preset value, it is determined to be a fake face image.

[0014] According to a preferred embodiment, in the binary classification anti-counterfeiting model, when the reflectivity and / or texture features of the face in the image exceed the feature interval of real person samples collected by big data, it is judged as a fake face image.

[0015] On the other hand, the present invention also discloses:

[0016] A detection device with an external storage medium, the device comprising: an image acquisition unit, used to realize data acquisition of an image to be identified; a processor, used to realize the steps of the aforementioned face anti-counterfeiting method when executing a computer program.

[0017] On the other hand, the present invention also discloses:

[0018] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the aforementioned face anti-counterfeiting method.

[0019] The aforementioned main scheme of the present invention and its further options can be freely combined to form multiple schemes, all of which are schemes that can be adopted and claimed for protection by the present invention. After understanding the scheme of the present invention, those skilled in the art can understand that there are multiple combinations based on the prior art and common knowledge, all of which are technical schemes to be protected by the present invention, and they are not exhaustively listed here.

[0020] Beneficial effects of the present invention: The multi-model-based face anti-counterfeiting method of the present invention distinguishes fake faces through the inherent characteristic information of the outer frame of paper or screen, eliminating most cases with obvious borders; forged faces without obvious borders due to close distance or other special circumstances can be further distinguished through a two-classification anti-counterfeiting model; and, for faces in different scenes, the scene data can be updated to further train the classification model. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of the face anti-counterfeiting method based on multiple models of the present invention. DETAILED DESCRIPTION

[0022] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0023] It should be noted that in order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments.

[0024] Embodiment 1:

[0025] refer to Figure 1 As shown, the present invention discloses a multi-model-based face anti-counterfeiting method, which includes: separately constructing a face detection model, a frame detection model and a binary classification anti-counterfeiting model.

[0026] The face detection model is configured to detect faces. The border detection model is configured to identify the border of a printed photo or the border of an electronic device screen. The binary classification model is configured to distinguish between real and fake faces.

[0027] Preferably, the face detection model is configured to detect the rectangular frame position of the face in the picture.

[0028] Specifically, the face position recognition in the image to be detected is completed through the face detection model, and the image containing face information that has passed the detection is sent to the border detection model.

[0029] Most printed photos or electronic screens will have a rectangular border within the field of view of the camera at medium and long distances. If the border can be detected and the face is within the border, it can be preliminarily determined to be a fake face.

[0030] Preferably, the border detection model is used to detect a quadrilateral border around a face.

[0031] Preferably, the border detection model is used to detect whether the image has a border, and when the image does not have a border, the image passes the detection and is sent to the binary classification anti-counterfeiting model;

[0032] Furthermore, when there is a border in the image and the face is within the border, the type of the border is identified, so that the image that passes the identification is sent to the binary classification anti-counterfeiting model, and the image that fails the identification is screened out.

[0033] Preferably, the frame detection model completes the recognition of whether the frame is a background object based on the motion state of the frame. When the frame to be recognized is recognized as a background object, the image is verified by the frame detection model.

[0034] Furthermore, the motion state of the frame is achieved by collecting a number of frames of continuous motion images to complete the position comparison of the frames of each image type. If the frame is in a relatively static state, it is determined to be a background object.

[0035] Specifically, after multiple frame comparisons, if the frame is relatively static for a long time, it can be determined that the frame is an inherent background object and cannot be used as anti-counterfeiting evidence. Background objects include door and window frames, wall frames, and billboard frames.

[0036] Preferably, the binary classification anti-counterfeiting model completes the face authenticity judgment based on the face reflectivity and / or texture features in the obtained image.

[0037] In the binary classification anti-counterfeiting model, when the reflectivity and / or texture features of the face in the image exceed a preset value, it is determined to be a fake face image.

[0038] In the binary classification anti-counterfeiting model, when the reflectivity and / or texture features of the face in the image exceed the feature interval of real person samples collected by big data, it is judged as a fake face image.

[0039] Specifically, the two-classification anti-counterfeiting model inputs the face image in the rectangular frame in the previous step to determine whether it is a forged face. The main features of this model are high-frequency texture detail features. For frameless faces, the distance is generally close, and the reflection of the paper and the texture of the screen are clear, so the classification effect is relatively better than that of medium and long distances. After testing, it can basically achieve an anti-counterfeiting accuracy of more than 99%.

[0040] The multi-model-based face anti-counterfeiting method of the present invention distinguishes fake faces through the inherent characteristic information of the outer frame of paper or screen, excluding most cases with obvious borders; forged faces without obvious borders due to close distance or other special circumstances can be further distinguished through a two-classification anti-counterfeiting model; and, for faces in different scenes, the scene data can be updated to further train the classification model.

[0041] Example 2

[0042] The present invention also discloses: a detection device for an external storage medium. The device comprises: an image acquisition unit for acquiring data of an image to be identified; and a processor for implementing the steps of the face anti-counterfeiting method as described in Example 1 when executing a computer program.

[0043] Example 3

[0044] The present invention also discloses: a computer-readable storage medium having a computer program stored thereon, and the computer program, when executed by a processor, implements the steps of the face anti-counterfeiting method as described in Example 1.

[0045] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-model based face anti-counterfeiting method, characterized in that: The face anti-counterfeiting method comprises: constructing a face detection model, a frame detection model and a binary classification anti-counterfeiting model respectively; The face detection model is used to identify the face position in the image to be detected, and the image containing the face information that has passed the detection is sent to the border detection model; The frame detection model detects whether the image has a frame, and when the image does not have a frame, it is sent to the two-classification anti-counterfeiting model; when there is a frame in the image and the face is within the frame, the frame type is identified, so that the image that passes the identification is sent to the two-classification anti-counterfeiting model, and the image that fails the identification is screened out; The binary classification anti-counterfeiting model determines the authenticity of a face based on the reflectivity and / or texture features of the face in the obtained image; The frame detection model completes the recognition of whether the frame is a background object based on the motion state of the frame. When the frame to be recognized is recognized as a background object, the image is verified by the frame detection model. The motion state of the frame is achieved by collecting a number of frames of continuous motion images to complete the position comparison of the frames of each image type. If the frame is in a relatively static state, it is determined to be a background object.

2. The face anti-counterfeiting method according to claim 1, characterized in that: The face detection model is configured to detect the rectangular frame position of the face in the picture.

3. The face anti-counterfeiting method according to any one of claims 1 to 2, characterized in that: The frame types include background objects and non-background objects.

4. The face anti-counterfeiting method according to claim 3, characterized in that: The background objects at least include door and window frames, wall frames and billboard frames.

5. The face anti-counterfeiting method according to claim 1, characterized in that: In the binary classification anti-counterfeiting model, when the reflectivity and / or texture features of the face in the image exceed a preset value, it is determined to be a fake face image.

6. The face anti-counterfeiting method according to claim 1, characterized in that: In the binary classification anti-counterfeiting model, when the reflectivity and / or texture features of the face in the image exceed the feature interval of real person samples collected by big data, it is judged as a fake face image.

7. A detection device for an external storage medium, characterized in that: The device comprises: An image acquisition unit, used to acquire data of an image to be recognized; A processor, used to implement the steps of the face anti-counterfeiting method as described in any one of claims 1 to 6 when executing a computer program.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the face anti-counterfeiting method according to any one of claims 1 to 6 are implemented.

Images