Face recognition method, system and device based on silent living body detection algorithm

Abstract

The invention relates to the technical field of face recognition, and discloses a face recognition method, system and device based on a silent living body detection algorithm, and the key point of the technical scheme of the method is that the method comprises the following steps: S1, responding to a face recognition demand, starting an image collection device, and collecting a video image of a target region; s2, performing face position detection on the target area video image, and marking a face position in the target area video image to obtain a face image; s3, performing feature extraction on the face image to obtain a face feature vector; s4, performing living body recognition on the face feature vector through a silent living body detection algorithm; and S5, after the fact that the real face exists in the target area is obtained, face storage or face recognition detection is performed according to a face recognition requirement, and a recognition result is obtained.

Description

Technical Field

[0001] The present invention relates to the field of face recognition technology, and more specifically, to a face recognition method, system and device based on a silent living body detection algorithm. Background Art

[0002] In recent years, face recognition technology has gradually penetrated into many aspects of our lives, such as finance, security and anti-terrorism, education, social entertainment, equipment, access control / attendance, transportation, smart business, etc. Face recognition methods such as face payment, face entry, face pickup, face access, etc. have brought convenient and intelligent experience to our lives.

[0003] However, in most of the existing face recognition processes, people are required to make corresponding actions in order to better identify the face. However, this recognition method requires the cooperation of people, which is inconvenient in actual use. Therefore, silent liveness detection came into being.

[0004] Silent liveness detection does not require the user to do anything, just face the camera naturally for 3 or 4 seconds. Since real faces are not absolutely static, there are micro-expressions, such as the movement of eyelids and eyeballs, blinking, and the expansion and contraction of lips and cheeks, etc., which can be used to prevent deception.

[0005] The silent liveness detection algorithm is a technology based on machine learning algorithms. It captures facial images through an RGB camera and can quickly determine whether it is a real face without the user's cooperation in making specific actions, such as opening the mouth, turning the head, blinking, etc.

[0006] The core of silent liveness detection technology is to be able to identify and analyze whether the object in front of the camera is a real face, rather than distinguishing between real faces and fake face attacks by analyzing details such as moiré patterns that appear in screen shots and reflections on paper photos. This technology covers a wide range of application scenarios, including but not limited to scenarios that require real-name authentication, such as mobile phone unlocking, bank transactions, and social media registration. By analyzing the macro environment in which a face exists, such as lighting, facial accessories, gender, hairstyle, and mask material, silent liveness detection technology can effectively exclude scenes where it is obviously impossible for a face to exist, thereby avoiding serious losses such as personal information leakage caused by fake face attacks.

[0007] In terms of technical implementation, the silent liveness detection algorithm improves its accuracy and efficiency through a variety of methods. For example, by training models in different fields under different data sets, these models are used to extract features, train feature generators and discriminators in the field, until the features output by the generator can successfully deceive the discriminators in various fields, thereby learning a generalized feature space expression. In addition, by designing the triplet loss between domains, that is, for each subject, it is hoped that its positive distance across domains is smaller than the negative distance across domains, and increasing the task of predicting depth to enhance discriminability, these methods are all aimed at improving the performance and accuracy of the silent liveness detection algorithm.

[0008] When using silent liveness detection technology, attention should be paid to protecting user privacy and ensuring that user personal information is collected, used and stored.

[0009] In summary, the silent liveness detection algorithm is an important security technology. It provides users with a safer and more convenient way of identity authentication through advanced machine learning algorithms and continuous algorithm updates. Therefore, it is necessary to continuously upgrade and optimize the existing silent liveness detection algorithm to achieve higher accuracy and convenience in the actual face recognition process. Summary of the invention

[0010] The purpose of the present invention is to provide a face recognition method, system and device based on a silent liveness detection algorithm, which can effectively identify real faces by extracting and analyzing multiple facial features and using liveness detection technology to prevent financial fraud and identity theft and improve the security and reliability of the financial system.

[0011] The above technical objectives of the present invention are achieved through the following technical solutions: A face recognition method based on a silent liveness detection algorithm comprises the following steps:

[0012] S1. In response to a face recognition requirement, start an image acquisition device to acquire a video image of a target area;

[0013] S2, performing face position detection on the target area video image, marking the face position in the target area video image, and obtaining a face image;

[0014] S3, extracting features from the face image to obtain a face feature vector;

[0015] S4, performing liveness recognition on the facial feature vector through a silent liveness detection algorithm;

[0016] S5. After it is determined that there is a real face in the target area, face preservation or face recognition detection is performed according to face recognition requirements to obtain a recognition result.

[0017] As a preferred technical solution of the present invention, the image acquisition device is a high-resolution camera; the high-resolution camera has an adaptive dimming function and an automatic focusing function.

[0018] As a preferred technical solution of the present invention, the duration of the target area video image is at least t1, and the target area video image includes at least n image frames, where n>2.

[0019] As a preferred technical solution of the present invention, in S2, before performing face detection, the target area video image is optimized and preprocessed, and the optimization preprocessing includes: adjusting the target area video image to a preset image size, and graying the target area video image.

[0020] As a preferred technical solution of the present invention, in S2, a face detection algorithm is used to perform face recognition on the target area image, determine the face position, and separately extract the face image of the face area.

[0021] As a preferred technical solution of the present invention, in S3, extracting a facial feature vector from a facial image includes: a facial static feature vector, a facial dynamic feature vector and a facial depth feature vector.

[0022] As a preferred technical solution of the present invention, in S4, living body recognition includes:

[0023] Perform optical reflection analysis and skin texture analysis based on the static feature vector of the face;

[0024] Micro-expression detection is performed based on the dynamic feature vector of the face;

[0025] Perform facial depth structure analysis based on facial depth feature vector;

[0026] When the above three vector analyses are consistent with the real face information data, the face in the current face image is considered to be a real face.

[0027] As a preferred technical solution of the present invention, in S5, the face recognition requirements include collection and verification categories;

[0028] If it is a collection type, the face feature vector is saved in the face feature database, indicating that the face collection is successful;

[0029] If it is a verification class, the facial feature vector is normalized to obtain a standardized facial feature vector;

[0030] The standardized facial feature vector is input into the intelligent classifier to obtain the judgment result.

[0031] The intelligent classifier is a classifier trained and tested with labeled real face and photo datasets.

[0032] A face recognition system based on silent liveness detection algorithm, comprising:

[0033] An image acquisition module is used to start an image acquisition device to acquire a video image of a target area in response to a face recognition requirement;

[0034] A face positioning module is used to detect the face position of the target area video image, mark the face position in the target area video image, and obtain the face image;

[0035] A feature extraction module is used to extract features from a face image and obtain a face feature vector;

[0036] A liveness recognition module is used to perform liveness recognition on facial feature vectors through a silent liveness detection algorithm;

[0037] The face recognition module is used to save the face or perform face recognition detection to obtain the recognition result according to the face recognition requirements after it is concluded that there is a real face in the target area.

[0038] A face recognition device based on a silent living body detection algorithm comprises: a processor and a memory, wherein the memory stores a computer program executable by the processor, and the processor implements the above method when executing the computer program.

[0039] In summary, the present invention has the following beneficial effects: by extracting and analyzing multiple facial features and using liveness detection technology, it is possible to accurately and efficiently identify real faces to prevent financial fraud and identity theft, and improve the security and reliability of the financial system. And unlike traditional user interactive verification, the face recognition method of the present invention does not require active cooperation from the user, and liveness detection can be achieved by analyzing the images captured by the camera, which is more convenient for users to use and can improve the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0041] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0042] like Figure 1 As shown, the present invention provides a face recognition method and system based on a silent liveness detection algorithm, wherein the method is executed by various modules in the system, and includes the following steps:

[0043] S1, image acquisition module, in response to the face recognition requirement, starts the image acquisition device to acquire the video image of the target area;

[0044] The image acquisition device is a high-resolution camera, and the images it acquires can have higher definition, so it can capture the user's facial details more clearly;

[0045] The high-resolution camera has adaptive dimming and autofocus functions. The adaptive dimming function can ensure that the user's facial image is clearly captured in different lighting conditions, avoiding more negative interference from ambient light factors on user imaging, thereby expanding the actual application scenarios. The autofocus function can automatically find the user's face, so that the focal length and the user's position are adapted, making the final image clearer and obtaining more details of the user's face.

[0046] The duration of the target area video image is at least t1, and generally t1 is greater than two seconds. During this period, the user's effective micro-movements can be captured. The target area video image includes at least n image frames, n>2, so that there are multiple image frames for mutual comparison, thereby obtaining changes in the user's micro-expressions, which is convenient for actual face recognition.

[0047] S2, face positioning module, performs face position detection on the target area video image, marks the face position in the target area video image, and obtains the face image;

[0048] In S2, before performing face detection, the target area video image is optimized and preprocessed, and the optimization and preprocessing includes: adjusting the target area video image to a preset image size, and graying the target area video image.

[0049] In S2, a face detection algorithm is used to perform face recognition on the target area image, determine the face position, and extract the face image of the face area separately. The face detection algorithm can use a Haar feature cascade classifier or a deep learning model (such as MTCNN).

[0050] S3, feature extraction module, extracts features from the face image to obtain a face feature vector;

[0051] In S3, facial feature vectors extracted from facial images include: facial static feature vectors, facial dynamic feature vectors, and facial depth feature vectors. Among them, facial static features include facial geometry, texture features, etc.; dynamic features include various facial micro-expressions, such as eye blinking, mouth opening and closing, facial muscle movements, etc.; depth features are the depth of the facial structure, expressing the three-dimensional structural data of the face.

[0052] S4, liveness recognition module, performs liveness recognition on facial feature vectors through silent liveness detection algorithm;

[0053] In S4, liveness detection includes:

[0054] Based on the static feature vector of the face, optical reflection analysis and skin texture analysis are performed to determine whether it is a real face, a picture or a video;

[0055] Micro-expression detection is performed based on the dynamic feature vector of the face. Since a real face will not remain static, but will have certain micro-expressions, it can be used to determine whether it is a real face.

[0056] Based on the face depth feature vector, the face depth structure analysis is performed. Because the face is a three-dimensional structure, through depth analysis, it can be determined whether the face is a three-dimensional structure. If not, it means it is not a real face. This can be achieved through the function of the image acquisition device itself, or the camera can be optimized. Using a depth camera or an infrared camera can obtain depth information faster.

[0057] When the above three vector analyses are consistent with the real face information data, the face in the current face image is considered to be a real face.

[0058] S5, the face recognition module, after determining that there is a real face in the target area, performs face preservation or face recognition detection according to face recognition requirements to obtain a recognition result.

[0059] In S5, face recognition requirements include collection and verification;

[0060] If it is a collection type, the face feature vector is saved in the face feature database, indicating that the face collection is successful;

[0061] If it is a verification class, the facial feature vector is normalized to obtain a standardized facial feature vector;

[0062] The standardized facial feature vector is input into the intelligent classifier to obtain the judgment result.

[0063] The intelligent classifier is a classifier that has been trained and tested on a dataset of labeled real faces and photos. In actual use, you only need to input the extracted adjustment vector into the intelligent classifier in real time to automatically get the judgment result.

[0064] As an embodiment of the present invention, before the method of the present invention is applied, the accuracy and robustness of the algorithm are verified by experiments, and its performance under different light, angles and backgrounds is evaluated. Specifically, it includes: Dataset construction: Collect a large number of real face and photo samples to build training and test data sets. Performance indicators: accuracy, recall rate, F1 value, false positive rate and false negative rate, etc. Experimental result analysis: Compare the performance of different feature extraction methods and classifiers, and optimize the algorithm parameters. Thereby, the method of the present invention has a more reliable application effect.

[0065] Corresponding to the above method, a face recognition device based on a silent liveness detection algorithm includes: a processor and a memory, the memory stores a computer program executable by the processor, and the processor implements the above methods S1-S5 when executing the computer program.

[0066] The advantages of the face recognition method, system and device based on the silent liveness detection algorithm of the present invention are: by extracting and analyzing multiple face features and using liveness detection technology, real faces can be effectively identified to prevent financial fraud and identity theft, thereby improving the security and reliability of the financial system. In addition, unlike traditional user interactive verification, the face recognition method of the present invention does not require active cooperation from the user, and liveness detection can be achieved by analyzing the images captured by the camera, which is more convenient for users to use and can improve the user experience.

[0067] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A face recognition method based on a silent liveness detection algorithm, characterized in that: The steps include: S1. In response to a face recognition requirement, start an image acquisition device to acquire a video image of a target area; S2, performing face position detection on the target area video image, marking the face position in the target area video image, and obtaining a face image; S3, extracting features from the face image to obtain a face feature vector; S4, performing liveness recognition on the facial feature vector through a silent liveness detection algorithm; S5. After it is determined that there is a real face in the target area, face preservation or face recognition detection is performed according to face recognition requirements to obtain a recognition result.

2. The face recognition method based on silent living body detection algorithm according to claim 1 is characterized in that: The image acquisition device is a high-resolution camera; the high-resolution camera has an adaptive dimming function and an automatic focusing function.

3. The face recognition method based on silent living body detection algorithm according to claim 2 is characterized in that: The duration of the target area video image is at least t1, and the target area video image includes at least n image frames, where n>2.

4. The face recognition method based on silent liveness detection algorithm according to claim 3 is characterized in that: S2 In the method, before performing face detection, the target area video image is optimized and preprocessed, and the optimization and preprocessing includes: adjusting the target area video image to a preset image size, and graying the target area video image.

5. According to claim 4, a face recognition method based on a silent liveness detection algorithm is characterized in that: in S2, a face detection algorithm is used to perform face recognition on the target area image, determine the face position, and separately extract the face image of the face area.

6. The face recognition method based on silent living body detection algorithm according to claim 5 is characterized in that: In S3, facial feature vectors are extracted from the facial image, including: a facial static feature vector, a facial dynamic feature vector, and a facial depth feature vector.

7. The face recognition method based on silent living body detection algorithm according to claim 6, Its characteristics are: In S4, living body recognition includes: Perform optical reflection analysis and skin texture analysis based on the static feature vector of the face; Micro-expression detection is performed based on the dynamic feature vector of the face; Perform facial depth structure analysis based on facial depth feature vector; When the above three vector analyses are consistent with the real face information data, the face in the current face image is considered to be a real face.

8. The face recognition method based on silent living body detection algorithm according to claim 7, Its characteristics are: in S5, face recognition requirements include collection and verification categories; If it is a collection type, the face feature vector is saved in the face feature database, indicating that the face collection is successful; If it is a verification class, the facial feature vector is normalized to obtain a standardized facial feature vector; The standardized facial feature vector is input into the intelligent classifier to obtain the judgment result. The intelligent classifier is a classifier trained and tested with labeled real face and photo datasets.

9. A face recognition system based on a silent liveness detection algorithm, characterized by: include: An image acquisition module is used to start an image acquisition device to acquire a video image of a target area in response to a face recognition requirement; A face positioning module is used to detect the face position of the target area video image, mark the face position in the target area video image, and obtain the face image; A feature extraction module is used to extract features from a face image and obtain a face feature vector; A liveness recognition module is used to perform liveness recognition on facial feature vectors through a silent liveness detection algorithm; The face recognition module is used to save the face or perform face recognition detection to obtain the recognition result according to the face recognition requirements after it is concluded that there is a real face in the target area.

10. A face recognition device based on a silent liveness detection algorithm, characterized in that: include: A processor and a memory, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

Images