Face verification method and device, equipment and storage medium
By performing multi-dimensional feature extraction and risk probability assessment on ordinary image acquisition devices, the high cost of generative model deception static detection and multimodal fusion detection is solved, and high-accuracy face verification is achieved.
Patent Information
- Application Number
- CN202511342729.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-19
- Publication Date
- 2026-01-13
AI Technical Summary
In existing technologies, high-quality single-frame images generated by generative models can deceive static detection models, resulting in poor accuracy of face verification. Furthermore, multimodal fusion detection requires special sensors, leading to high hardware costs and hindering its widespread adoption.
The verification video is acquired using a common image acquisition device, and multi-dimensional features are extracted, including texture, identity, motion, and edge features. The risk score of each feature is calculated and weighted and fused to determine the risk probability. The verification result is determined by combining the risk score with the business probability threshold.
Without adding hardware, highly accurate face verification was achieved, improving the universality and accuracy of verification and reducing equipment costs.
Smart Images
Figure CN121330732A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present invention relate to the field of artificial intelligence technology, and in particular to a face verification method, apparatus, device and storage medium. Background Technology
[0002] With the development of computer technology, facial recognition technology has been widely used in recent years. For example, facial verification systems using facial recognition technology are widely deployed on financial trading platforms, allowing users to quickly complete financial transactions through the facial verification system.
[0003] In existing technologies, static detection models are typically used to analyze a single face image to verify whether the face is a real face.
[0004] However, the latest generative models can generate extremely high-quality single-frame images, enough to fool most static detection models, resulting in poor accuracy of face verification. Summary of the Invention
[0005] This invention provides a face verification method, apparatus, device, and storage medium to improve the accuracy of face verification.
[0006] In a first aspect, embodiments of the present invention provide a face verification method, including:
[0007] After confirming that facial verification is enabled for the current business, obtain the verification video;
[0008] Multi-dimensional feature extraction is performed on the verification video to obtain the multi-dimensional video features corresponding to the verification video;
[0009] Calculate the risk score corresponding to each video feature in the multi-dimensional video features, and fuse the risk scores to determine the risk probability;
[0010] The face verification result is determined based on the risk probability and the probability threshold corresponding to the current service.
[0011] The technical solution of this invention provides a face verification method, comprising: after determining that face verification for the current business is enabled, acquiring a verification video; performing multi-dimensional feature extraction on the verification video to obtain multi-dimensional video features corresponding to the verification video; calculating the risk score corresponding to each video feature in the multi-dimensional video features, and fusing the risk scores to determine the risk probability; and determining the face verification result based on the risk probability and the probability threshold corresponding to the current business. The above technical solution, after determining to enable face verification for the current business, can acquire a verification video for face verification of the current business. This verification video is directly acquired through the image acquisition device of the terminal equipment, eliminating the need for additional hardware and improving the convenience of acquiring the verification video. Furthermore, features can be extracted from the verification video in multiple dimensions to obtain a multi-dimensional feature stream corresponding to the verification video, achieving multi-dimensional feature extraction. Then, the risk score corresponding to each video feature in the multi-dimensional features is determined, and the risk scores corresponding to each video feature are fused to determine the risk probability of the verification video, achieving accurate risk probability assessment of the verification video. By comparing the risk probability corresponding to the current business with the probability threshold, the face verification result is determined as either verification passed or verification rejected, achieving accurate face verification and improving the accuracy of face verification.
[0012] Furthermore, after obtaining the verification video, it also includes:
[0013] The target video is obtained by performing face detection, face alignment, and face region cropping on the verification video.
[0014] Furthermore, multi-dimensional feature extraction is performed on the verification video to obtain multi-dimensional video features corresponding to the verification video, including:
[0015] Multi-dimensional feature extraction is performed on the verification video to obtain the texture feature stream, identity feature stream, motion feature stream, and edge feature stream corresponding to the verification video.
[0016] Further, the risk score corresponding to each video feature in the multi-dimensional video features is calculated, including:
[0017] The texture jitter risk score is calculated based on the texture feature stream, the identity mutation risk score is calculated based on the identity feature stream, the motion violation risk score is calculated based on the motion feature stream, and the edge flicker risk score is calculated based on the edge feature stream.
[0018] Furthermore, the risk probability is determined by integrating the aforementioned risk scores, including:
[0019] The risk probability is obtained by weighting and fusing the texture jitter risk score, the identity mutation risk score, the motion violation risk score, and the edge flicker risk score.
[0020] Further, based on the risk probability and the probability threshold corresponding to the current service, the face verification result is determined, including:
[0021] If the risk probability is greater than the probability threshold corresponding to the current service, the face verification result is determined to be authentication rejection; otherwise, the face verification result is determined to be authentication success.
[0022] Secondly, embodiments of the present invention also provide a face verification device, comprising:
[0023] The acquisition module is used to acquire the verification video after determining that face verification for the current business is enabled;
[0024] The extraction module is used to perform multi-dimensional feature extraction on the verification video to obtain the texture feature stream, identity feature stream, motion feature stream and edge feature stream corresponding to the verification video;
[0025] The calculation module is used to calculate a texture jitter score based on the texture feature stream, an identity mutation score based on the identity feature stream, a motion violation score based on the motion feature stream, and an edge flicker score based on the edge feature stream, and to fuse the texture jitter score, identity mutation score, motion violation score, and edge flicker score to obtain a risk probability;
[0026] The verification module is used to determine the face verification result based on the risk probability and the probability threshold corresponding to the current business.
[0027] Thirdly, embodiments of the present invention also provide a computer device, the computer device comprising:
[0028] At least one processor; and a memory communicatively connected to said at least one processor;
[0029] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the face verification method as described in any of the first aspects.
[0030] Fourthly, embodiments of the present invention also provide a storage medium containing computer-executable instructions, wherein the computer-executable instructions, when executed by a computer processor, are used to perform the face verification method as described in any of the first aspects.
[0031] Fifthly, this application provides a computer program product including computer instructions that, when executed on a computer, cause the computer to perform the face verification method as provided in the first aspect.
[0032] It should be noted that the aforementioned computer instructions may be stored, in whole or in part, on a computer-readable storage medium. This computer-readable storage medium may be packaged together with the processor of the face verification device, or it may be packaged separately from the processor of the face verification device; this application does not impose any limitations on this.
[0033] The descriptions of the second, third, fourth, and fifth aspects in this application can be referenced to the detailed description of the first aspect; and the beneficial effects of the descriptions of the second, third, fourth, and fifth aspects can be referenced to the analysis of the beneficial effects of the first aspect, which will not be repeated here.
[0034] In this application, the name of the aforementioned face verification device does not limit the device or functional module itself. In actual implementation, these devices or functional modules may appear under other names. As long as the function of each device or functional module is similar to that of this application, it falls within the scope of the claims of this application and its equivalents.
[0035] These or other aspects of this application will become more readily apparent in the following description. Attached Figure Description
[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0037] Figure 1 A flowchart of a face verification method provided in an embodiment of the present invention;
[0038] Figure 2 A flowchart of another face verification method provided in an embodiment of the present invention;
[0039] Figure 3 This is a schematic diagram of the structure of a face verification device provided in an embodiment of the present invention;
[0040] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation
[0041] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.
[0042] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.
[0043] The terms "first" and "second," etc., used in the specification and drawings of this application are used to distinguish different objects or to distinguish different treatments of the same object, rather than to describe a specific order of objects.
[0044] Furthermore, the terms "comprising" and "having," and any variations thereof, used in the description of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus.
[0045] Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but may also have additional steps not included in the figures. The process can correspond to a method, function, procedure, subroutine, subroutine, etc. Moreover, embodiments and features in the embodiments of the present invention can be combined with each other without conflict.
[0046] It should be noted that in the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.
[0047] In the description of this application, unless otherwise stated, "a plurality of" means two or more.
[0048] In existing technologies, face verification methods can include static detection, liveness detection, and multimodal fusion detection. Static detection can be understood as static single-frame detection, typically using a static detection model to analyze a single face image, looking for subtle artifacts left by the generation model, such as unnatural textures, asymmetrical pupils, and distorted decorations, to verify whether the face is real. Liveness detection can be understood as simple action command liveness detection, usually requiring the user to perform randomly specified actions (such as blinking, opening the mouth, or shaking the head), and determining whether the person is alive by tracking the movement of key facial points, thus verifying whether the face is real. Multimodal fusion detection typically uses dedicated hardware (such as infrared cameras or 3D structured light sensors) to collect multimodal data, and analyzes physical features such as infrared reflection characteristics and 3D depth information to distinguish real faces from forgeries such as screens or masks.
[0049] However, the current generative models generate extremely high-quality single-frame images, enough to fool most static detection models, making accurate face verification difficult in static detection. Furthermore, current generative models can generate coherent motion videos based on instructions, or simulate simple actions like nodding or shaking their heads by driving 3D face models, making accurate face verification difficult with simple motion-based liveness detection. While multimodal fusion detection offers strong security, it requires specialized sensors in the terminal device, resulting in high hardware costs and preventing its widespread adoption in terminal devices equipped only with ordinary image acquisition devices.
[0050] Therefore, this application proposes a face verification method that achieves highly accurate face verification with only a common image acquisition device.
[0051] The face verification method proposed in this application will be described in detail below with reference to the illustrations and embodiments.
[0052] Figure 1 This is a flowchart illustrating a face verification method provided in an embodiment of the present invention. This embodiment is applicable to situations requiring improved face verification accuracy. The method can be executed by a face verification device, such as... Figure 1 As shown, the specific steps include the following:
[0053] Step 110: After confirming that face verification is enabled for the current business, obtain the verification video.
[0054] The current business can be understood as the business that requires facial verification. The operation of these businesses needs to be secure, therefore, facial verification is required.
[0055] Specifically, facial verification can be enabled before or during the current business operation. Facial verification requires verifying whether the face of the current user matches the face associated with the current account. Specifically, after enabling facial verification for the current business, a verification video can be obtained. This can be achieved by acquiring the facial video of the current user using the image acquisition device installed on the terminal device.
[0056] It should be noted that the verification video is a three-color image obtained through a common three-color image acquisition device (an image acquisition device that uses red, green, and blue color channels to capture images).
[0057] In this embodiment of the invention, the acquisition of verification videos that require face verification is achieved without the need for additional hardware, thereby improving the universality of face verification.
[0058] Step 120: Extract multi-dimensional features from the verification video to obtain the multi-dimensional video features corresponding to the verification video.
[0059] Multidimensional video features can include at least texture features, identity features, motion features, and edge features.
[0060] Specifically, multi-dimensional feature extraction can be performed on the verification video. For example, texture features of the verification video can be extracted to obtain the texture feature stream corresponding to the verification video. Identity features of the verification video can also be extracted to obtain the identity feature stream corresponding to the verification video. Motion features of the verification video can also be extracted to obtain the motion feature stream corresponding to the verification video. Edge features of the verification video can also be extracted to obtain the edge feature stream corresponding to the verification video.
[0061] In this embodiment of the invention, feature extraction of the verification video is performed on the verification video in multiple dimensions to obtain a feature stream corresponding to the verification video in multiple dimensions, thereby realizing multi-dimensional feature extraction of the verification video.
[0062] Step 130: Calculate the risk score corresponding to each video feature in the multi-dimensional video features, and fuse the risk scores to determine the risk probability.
[0063] Among them, the first risk score corresponding to texture features can be determined by the stability of the texture feature stream, the second risk score corresponding to identity features can be determined by the similarity of adjacent identity features in the identity feature stream, the third risk score corresponding to motion features can be determined by whether the motion direction and amplitude of the face and background in the motion feature stream conform to motion dynamics, and the fourth risk score corresponding to edge features can be determined based on the stability of edge features.
[0064] Specifically, the first risk score corresponding to a texture feature can be determined by calculating the stability between each texture feature in the texture feature stream; the second risk score corresponding to an identity feature can be determined by calculating the similarity between adjacent identity features in the identity feature stream; the third risk score corresponding to a motion feature can be determined by determining whether the motion direction and amplitude of the face and background in the motion feature stream conform to motion dynamics; and the fourth risk score corresponding to an edge feature can be determined by determining the stability of each edge feature in the edge feature stream.
[0065] After determining the risk score corresponding to each video feature, the risk probability can be determined by fusing the risk scores corresponding to each video feature. Specifically, the risk scores corresponding to the first risk score of the texture feature, the second risk score of the identity feature, the third risk score of the motion feature, and the fourth risk score of the edge feature can be weighted and fused based on the first weight corresponding to the texture feature, the second weight corresponding to the identity feature, the third weight corresponding to the motion feature, and the fourth weight corresponding to the edge feature to obtain the risk probability corresponding to the verification video.
[0066] In this embodiment of the invention, after determining the risk score corresponding to each video feature in the multi-dimensional features of the verification video, the risk probability of the verification video is determined by fusing the risk scores corresponding to each video feature, thereby achieving an accurate risk probability assessment of the verification video.
[0067] Step 140: Determine the face verification result based on the risk probability and the probability threshold corresponding to the current service.
[0068] The probability threshold corresponding to the current business can be understood as the maximum risk probability of the current business operating safely.
[0069] Specifically, firstly, the probability threshold corresponding to the current business can be determined. Secondly, the risk probability corresponding to the current business can be compared with the probability threshold. If the risk probability is not greater than the probability threshold, the face verification result can be determined as authentication passed. If the risk probability is greater than the probability threshold, the face verification result can be determined as authentication rejected.
[0070] In this embodiment of the invention, by comparing the risk probability and probability threshold corresponding to the current business, the face verification result is determined to be either verification passed or verification rejected, thereby achieving accurate face verification and improving the accuracy of face verification.
[0071] The face verification method provided in this embodiment of the invention includes: after determining that face verification for the current business is enabled, obtaining a verification video; performing multi-dimensional feature extraction on the verification video to obtain multi-dimensional video features corresponding to the verification video; calculating the risk score corresponding to each video feature in the multi-dimensional video features, and fusing the risk scores to determine the risk probability; and determining the face verification result based on the risk probability and the probability threshold corresponding to the current business. The above technical solution, after determining to enable face verification for the current business, can acquire a verification video for face verification of the current business. This verification video is directly acquired through the image acquisition device of the terminal equipment, eliminating the need for additional hardware and improving the convenience of acquiring the verification video. Furthermore, features can be extracted from the verification video in multiple dimensions to obtain a multi-dimensional feature stream corresponding to the verification video, achieving multi-dimensional feature extraction. Then, the risk score corresponding to each video feature in the multi-dimensional features is determined, and the risk scores corresponding to each video feature are fused to determine the risk probability of the verification video, achieving accurate risk probability assessment of the verification video. By comparing the risk probability corresponding to the current business with the probability threshold, the face verification result is determined as either verification passed or verification rejected, achieving accurate face verification and improving the accuracy of face verification.
[0072] Figure 2 This is a flowchart illustrating another face verification method provided in an embodiment of the present invention. This embodiment is a specific modification based on the above embodiments. Figure 2 As shown, in this embodiment, the method may further include:
[0073] Step 210: After confirming that face verification is enabled for the current business, obtain the verification video.
[0074] Currently, it can also handle different risk levels of services such as regular login, password modification, remote account opening, large-amount transfers, and financial transactions.
[0075] Specifically, after determining that facial verification for the current business is enabled, the facial video of the current user can be obtained based on the image acquisition device installed on the terminal device. For example, a 3-5 second image of the three primary colors can be obtained to achieve the acquisition of the verification video.
[0076] Furthermore, during the process of obtaining the verification video, users do not need to perform any specific actions or remain absolutely still, resulting in a more natural and smooth user experience, which helps to improve the user conversion rate of the current business.
[0077] In this embodiment of the invention, the acquisition of verification videos that require face verification is achieved without the need for additional hardware, so that face verification can be seamlessly deployed on existing terminal devices equipped with ordinary three-primary-color image acquisition devices, thereby improving the universality of face verification.
[0078] Step 220: Obtain the target video by performing face detection, face alignment, and face region cropping on the verification video.
[0079] Specifically, the verification video acquired by the image acquisition device may contain environmental content unrelated to the face. In order to improve the accuracy of face verification, the verification video can be processed after it is acquired. Specifically, firstly, face detection can be performed on the verification video to determine the faces contained in each video frame. Secondly, the faces in each video frame can be visually aligned. Then, face region cropping can be performed on each video frame, that is, the face region in each video frame can be retained, and the region unrelated to the face can be cropped out to obtain the target video containing only the face region.
[0080] In this embodiment of the invention, the verification video is processed by performing face detection, face alignment, and face region cropping to obtain a target video containing only face regions.
[0081] Step 230: Extract multi-dimensional features from the verification video to obtain the multi-dimensional video features corresponding to the verification video.
[0082] In one implementation, step 230 may specifically include:
[0083] Multi-dimensional feature extraction is performed on the verification video to obtain the texture feature stream, identity feature stream, motion feature stream, and edge feature stream corresponding to the verification video.
[0084] Specifically, texture features can be extracted from the verification video to obtain the texture feature stream corresponding to the verification video. Specifically, firstly, a preset facial region can be selected for each video frame in the verification video. For example, the cheek or forehead can be selected. The selection of the specific facial region can be preset according to actual needs and is not specifically limited here. Secondly, texture features can be extracted from the selected preset facial region to obtain the texture features corresponding to each video frame. For example, the local binary pattern features of each video frame can be calculated and determined as texture features. Then, the texture feature stream can be determined based on the texture features corresponding to each video frame in the verification video.
[0085] Alternatively, identity features can be extracted from the verification video to obtain the identity feature stream corresponding to the verification video. Specifically, the verification video can be input into a pre-trained face recognition model so that the face recognition model can determine the high-dimensional identity features corresponding to each video frame. The high-dimensional identity features corresponding to each video frame in the verification video can then constitute the identity feature stream.
[0086] It should be noted that the face recognition model can be obtained by pre-training an initial recognition model using historical verification videos and their corresponding identity feature streams.
[0087] Motion features can also be extracted from the verification video to obtain the motion feature stream corresponding to the verification video. Specifically, the motion vectors of each pixel between consecutive video frames can be calculated based on the optical flow method to obtain the motion field sequence describing the motion of the face and background, thus obtaining the motion feature stream.
[0088] Edge features can also be extracted from the verification video to obtain the edge feature stream corresponding to the verification video. Specifically, firstly, the face contour can be extracted from each video frame to obtain the face contour corresponding to each video frame. Secondly, the gradient intensity of the edge region of the face contour can be calculated to obtain the edge intensity corresponding to each video frame. The edge intensity corresponding to each video frame in the verification video can constitute the edge feature stream.
[0089] While a perfect single-frame video is easy to forge, the absolute consistency of all physical and biological features over a long time span is usually difficult to forge. Therefore, by taking a temporal approach, we can determine the temporally consistent texture feature stream, identity feature stream, motion feature stream, and edge feature stream corresponding to the verification video. Based on the feature stream in the temporal dimension, we can extract features from the verification video to obtain multi-dimensional video features that combine temporal information.
[0090] In this embodiment of the invention, feature extraction is performed on the verification video in four dimensions: texture features, identity features, motion features, and edge features, to obtain the texture feature stream, identity feature stream, motion feature stream, and edge feature stream corresponding to the verification video, thereby realizing multi-dimensional feature extraction of the verification video.
[0091] Step 240: Calculate the risk score corresponding to each video feature in the multi-dimensional video features.
[0092] In one implementation, step 240 may specifically include:
[0093] The texture jitter risk score is calculated based on the texture feature stream, the identity mutation risk score is calculated based on the identity feature stream, the motion violation risk score is calculated based on the motion feature stream, and the edge flicker risk score is calculated based on the edge feature stream.
[0094] Specifically, real face videos are typically relatively stable, while fake face videos usually exhibit high-frequency jitter. That is, the variance of the texture feature stream corresponding to a real face video is stable, while the variance of the texture feature stream corresponding to a fake face video shows abnormal variance peaks. Therefore, the texture jitter risk score can be determined by identifying the variance of the texture feature stream on the time axis; in other words, the variance of the texture feature stream on the time axis can be used to determine the texture jitter risk score.
[0095] In real face videos, adjacent frames are typically quite similar, while in fake face videos, abrupt changes often occur. That is, in real face videos, the similarity of adjacent identity features in the corresponding feature stream is high, while in fake face videos, the similarity of adjacent identity features in the corresponding feature stream exhibits abrupt drops. Therefore, the risk score for identity mutation can be determined by assessing the similarity of adjacent identity features in the texture feature stream; specifically, the average similarity of adjacent identity features in the texture feature stream can be used as the risk score for identity mutation.
[0096] In real face videos, the motion of the foreground and background typically conforms to foreground-background motion dynamics, while the motion of the foreground and background in fake face videos is often disjointed or violates physical laws. Therefore, the face can be considered as the foreground region, and a motion violation risk score can be determined by analyzing whether the motion direction and amplitude of the foreground and background regions conform to foreground-background motion dynamics. Specifically, the motion feature stream can be input into a pre-trained motion risk assessment model, which then determines the corresponding motion violation risk score based on the motion feature stream. The motion risk assessment model is obtained by training an initial model based on historical motion feature streams and their corresponding risk scores.
[0097] In genuine face videos, the fusion point between the face and background is typically stable, indicating edge stability. In contrast, in fake face videos, periodic and irregular gradient flickering occurs at this fusion point. Therefore, motion violation risk scores can be determined by analyzing the edge gradient stability of the face and background regions. For example, edge feature streams can be input into a pre-trained edge risk assessment model, which then determines the corresponding edge flickering risk score based on the edge feature streams. This edge risk assessment model is obtained by training a second initial model based on historical edge feature streams and their corresponding risk scores.
[0098] In this embodiment of the invention, by analyzing the texture feature stream, identity feature stream, motion feature stream, and edge feature stream corresponding to the verification video, the texture jitter risk score corresponding to the texture feature, the identity mutation risk score corresponding to the identity feature, the motion violation risk score corresponding to the motion feature, and the edge flicker risk score corresponding to the edge feature are determined. This enables the determination of the risk score corresponding to each video feature in the multi-dimensional features of the verification video, thereby achieving a multi-dimensional risk assessment of the verification video.
[0099] Step 250: Combine the risk scores to determine the risk probability.
[0100] In one implementation, step 250 may specifically include:
[0101] The risk probability is obtained by weighting and fusing the texture jitter risk score, the identity mutation risk score, the motion violation risk score, and the edge flicker risk score.
[0102] Specifically, the risk probability can be determined by fusing the risk scores corresponding to each video feature. Specifically, the risk scores of texture jitter corresponding to texture features, identity mutation corresponding to identity features, motion violation corresponding to motion features, and edge flicker corresponding to edge features can be input into a weighted fusion classifier, which can then output the corresponding risk probabilities.
[0103] In this embodiment of the invention, the risk probability of the verification video is determined by fusing the risk scores corresponding to each video feature, thereby achieving an accurate risk probability assessment of the verification video.
[0104] Step 260: Determine the face verification result based on the risk probability and the probability threshold corresponding to the current business.
[0105] In one implementation, step 260 may specifically include:
[0106] If the risk probability is greater than the probability threshold corresponding to the current service, the face verification result is determined to be authentication rejection; otherwise, the face verification result is determined to be authentication success.
[0107] The probability thresholds for each business can be preset and adjusted flexibly. The probability thresholds for each business are stored in the database in the form of business name-probability threshold.
[0108] Specifically, firstly, the probability threshold corresponding to the current business can be determined. This can be done by searching the database based on the business name of the current business. Secondly, the risk probability corresponding to the current business can be compared with the probability threshold. If the risk probability is not greater than the probability threshold, the face verification result can be determined as authentication passed. If the risk probability is greater than the probability threshold, the face verification result can be determined as authentication rejected.
[0109] In practical applications, once the face verification result is determined to be authentication rejection, a secondary verification of the verification video can be triggered. Specifically, the secondary verification can be achieved through authentication methods such as manual review and SMS verification.
[0110] In this embodiment of the invention, by comparing the risk probability and probability threshold corresponding to the current business, the face verification result is determined to be either verification passed or verification rejected, thereby achieving accurate face verification and improving the accuracy of face verification.
[0111] The face verification method provided in this invention includes: after determining that face verification for the current service is enabled, acquiring a verification video; obtaining a target video by performing face detection, face alignment, and face region cropping on the verification video; extracting multi-dimensional features from the verification video to obtain multi-dimensional video features corresponding to the verification video; calculating a risk score corresponding to each video feature in the multi-dimensional video features; fusing the risk scores to determine a risk probability; and determining a face verification result based on the risk probability and a probability threshold corresponding to the current service. The above technical solution, after determining to enable face verification for the current business, can acquire the verification video used for face verification of the current business. It achieves the acquisition of the verification video requiring face verification without adding hardware, enabling seamless deployment of face verification on existing terminal devices equipped with ordinary three-primary-color image acquisition devices. This improves the convenience of acquiring verification video and the universality of face verification. Furthermore, it can perform face detection, face alignment, and face region cropping on the verification video. Through preprocessing, a target video containing only face regions is obtained. Then, features can be extracted from the target video in four dimensions: texture features, identity features, motion features, and edge features, yielding the corresponding texture feature stream, identity feature stream, motion feature stream, and edge feature stream. Feature streams enable multi-dimensional feature extraction from target videos. By analyzing texture feature streams, identity feature streams, motion feature streams, and edge feature streams, risk scores are determined for texture features (texture jitter risk score), identity features (identity mutation risk score), motion features (motion violation risk score), and edge features (edge flicker risk score). This allows for the determination of risk scores for each video feature within the multi-dimensional features, enabling multi-dimensional risk assessment of the verification video. By fusing the risk scores of each video feature, the risk probability of the verification video is determined, achieving accurate risk probability assessment. By comparing the risk probability corresponding to the current business with the probability threshold, the face verification result is determined as either verification passed or rejected, achieving precise face verification and improving its accuracy.
[0112] Furthermore, the temporal domains of the texture feature stream, identity feature stream, motion feature stream, and edge feature stream used to determine the risk probability are consistent. Moreover, it is usually difficult for forged videos to achieve multi-dimensional temporal domain consistency. Therefore, by fusing multi-dimensional features to determine the risk probability, an accurate risk probability assessment for the verification video can be achieved, thereby increasing the attack cost of face verification.
[0113] Figure 3 This is a schematic diagram of a face verification device provided in an embodiment of the present invention. This device can be applied to situations requiring improved face verification accuracy. The device can be implemented through software and / or hardware and is generally integrated into electronic devices, such as computer devices.
[0114] like Figure 3 As shown, the device includes:
[0115] The acquisition module 310 is used to acquire the verification video after determining that face verification for the current business is enabled;
[0116] Extraction module 320 is used to perform multi-dimensional feature extraction on the verification video to obtain the texture feature stream, identity feature stream, motion feature stream and edge feature stream corresponding to the verification video;
[0117] The calculation module 330 is used to calculate a texture jitter score based on the texture feature stream, an identity mutation score based on the identity feature stream, a motion violation score based on the motion feature stream, and an edge flicker score based on the edge feature stream, and to fuse the texture jitter score, identity mutation score, motion violation score, and edge flicker score to obtain a risk probability;
[0118] The verification module 340 is used to determine the face verification result based on the risk probability and the probability threshold corresponding to the current business.
[0119] The face verification device provided in this embodiment, after determining that face verification for the current service is to be enabled, acquires a verification video; performs multi-dimensional feature extraction on the verification video to obtain multi-dimensional video features corresponding to the verification video; calculates the risk score corresponding to each video feature in the multi-dimensional video features, and fuses the risk scores to determine the risk probability; and determines the face verification result based on the risk probability and the probability threshold corresponding to the current service. The above technical solution, after determining that face verification for the current service is to be enabled, can acquire a verification video for face verification of the current service. Furthermore, this verification video is directly acquired through the image acquisition device of the terminal device, without the need for additional hardware, improving the convenience of acquiring the verification video. This allows for feature extraction from the verification video in multiple dimensions, obtaining a multi-dimensional feature stream corresponding to the verification video, achieving multi-dimensional feature extraction of the verification video. Then, the risk score corresponding to each video feature in the multi-dimensional features is determined, and the risk scores corresponding to each video feature are fused to determine the risk probability of the verification video, achieving accurate risk probability assessment of the verification video. By comparing the risk probability corresponding to the current service with the probability threshold, the face verification result is determined as verification passed or verification rejected, achieving accurate face verification and improving the accuracy of face verification.
[0120] Based on the above embodiments, the device further includes:
[0121] The processing module is used to obtain the target video by performing face detection, face alignment, and face region cropping on the verification video after obtaining the verification video.
[0122] Based on the above embodiments, the extraction module 320 is specifically used for:
[0123] Multi-dimensional feature extraction is performed on the verification video to obtain the texture feature stream, identity feature stream, motion feature stream, and edge feature stream corresponding to the verification video.
[0124] Based on the above embodiments, the calculation module 330 is specifically used for:
[0125] The texture jitter risk score is calculated based on the texture feature stream, the identity mutation risk score is calculated based on the identity feature stream, the motion violation risk score is calculated based on the motion feature stream, and the edge flicker risk score is calculated based on the edge feature stream; the risk probability is obtained by weighted fusion of the texture jitter risk score, the identity mutation risk score, the motion violation risk score, and the edge flicker risk score.
[0126] Based on the above embodiments, the verification module 340 is specifically used for:
[0127] If the risk probability is greater than the probability threshold corresponding to the current service, the face verification result is determined to be authentication rejection; otherwise, the face verification result is determined to be authentication success.
[0128] The face verification device provided in this embodiment of the invention can execute the face verification method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the face verification method.
[0129] It is worth noting that in the above-described embodiments of the face verification device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.
[0130] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Figure 4 A block diagram of an exemplary computer device 4 suitable for implementing embodiments of the present invention is shown. Figure 4 The computer device 4 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.
[0131] like Figure 4 As shown, the computer device 4 is represented in the form of a general-purpose computing electronic device. The components of the computer device 4 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).
[0132] Bus 18 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. Examples of these architectures include, but are not limited to, industry-standard architecture buses, microchannel architecture buses, enhanced buses, Video Electronics Standards Association (VESA) local buses, and peripheral component interconnect buses.
[0133] Computer device 4 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 4, including volatile and non-volatile media, removable and non-removable media.
[0134] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory 30 and / or cache memory 32. Computer device 4 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 4 Not shown; usually referred to as a "hard drive"). Although Figure 4 As not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.
[0135] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of the present invention.
[0136] Computer device 4 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with computer device 4, and / or with any device that enables computer device 4 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed through input / output interface 22. Furthermore, computer device 4 can also communicate with one or more networks (e.g., local area network, wide area network, and / or public network, such as the Internet) through network adapter 20. Figure 4 As shown, network adapter 20 communicates with other modules of computer device 4 via bus 18. It should be understood that, although... Figure 4 Not shown, it can be used in conjunction with computer device 4 with other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, independent redundant disk arrays, tape drives, and data backup storage systems.
[0137] Processing unit 16 executes various functional applications and page displays by running programs stored in system memory 28, such as implementing the face verification method provided in this embodiment of the invention, which includes:
[0138] After confirming that facial verification is enabled for the current business, obtain the verification video;
[0139] Multi-dimensional feature extraction is performed on the verification video to obtain the multi-dimensional video features corresponding to the verification video;
[0140] Calculate the risk score corresponding to each video feature in the multi-dimensional video features, and fuse the risk scores to determine the risk probability;
[0141] The face verification result is determined based on the risk probability and the probability threshold corresponding to the current service.
[0142] Of course, those skilled in the art will understand that the processor can also implement the technical solution of the face verification method provided in any embodiment of the present invention.
[0143] This invention provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements, for example, the face verification method provided in this invention, which includes:
[0144] After confirming that facial verification is enabled for the current business, obtain the verification video;
[0145] Multi-dimensional feature extraction is performed on the verification video to obtain the multi-dimensional video features corresponding to the verification video;
[0146] Calculate the risk score corresponding to each video feature in the multi-dimensional video features, and fuse the risk scores to determine the risk probability;
[0147] The face verification result is determined based on the risk probability and the probability threshold corresponding to the current service.
[0148] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: electrical connections having one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory (erasable programmable read-only memory or flash memory), optical fibers, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.
[0149] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.
[0150] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, electromagnetic waves, etc., or any suitable combination thereof.
[0151] Computer program code for performing the operations of this invention can be written in one or more programming languages or a combination thereof. Programming languages include object-oriented programming languages such as Java, Smalltalk, and C++, as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0152] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computing device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.
[0153] Furthermore, the acquisition, storage, use, and processing of data in the technical solution of this invention all comply with relevant laws and regulations.
[0154] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.
Claims
1. A face verification method, characterized in that, include: After confirming that facial verification is enabled for the current business, obtain the verification video; Multi-dimensional feature extraction is performed on the verification video to obtain the multi-dimensional video features corresponding to the verification video; Calculate the risk score corresponding to each video feature in the multi-dimensional video features, and fuse the risk scores to determine the risk probability; The face verification result is determined based on the risk probability and the probability threshold corresponding to the current service.
2. The face verification method according to claim 1, characterized in that, After obtaining the verification video, the following is also included: The target video is obtained by performing face detection, face alignment, and face region cropping on the verification video.
3. The face verification method according to claim 1, characterized in that, Multi-dimensional feature extraction is performed on the verification video to obtain the multi-dimensional video features corresponding to the verification video, including: Multi-dimensional feature extraction is performed on the verification video to obtain the texture feature stream, identity feature stream, motion feature stream, and edge feature stream corresponding to the verification video.
4. The face verification method according to claim 3, characterized in that, Calculating the risk score corresponding to each video feature in the multi-dimensional video features includes: The texture jitter risk score is calculated based on the texture feature stream, the identity mutation risk score is calculated based on the identity feature stream, the motion violation risk score is calculated based on the motion feature stream, and the edge flicker risk score is calculated based on the edge feature stream.
5. The face verification method according to claim 4, characterized in that, Determining the risk probability by combining the aforementioned risk scores includes: The risk probability is obtained by weighting and fusing the texture jitter risk score, the identity mutation risk score, the motion violation risk score, and the edge flicker risk score.
6. The face verification method according to claim 1, characterized in that, Based on the risk probability and the probability threshold corresponding to the current service, the face verification result is determined, including: If the risk probability is greater than the probability threshold corresponding to the current service, the face verification result is determined to be authentication rejection; otherwise, the face verification result is determined to be authentication success.
7. A face verification device, characterized in that, include: The acquisition module is used to acquire the verification video after determining that face verification for the current business is enabled; The extraction module is used to perform multi-dimensional feature extraction on the verification video to obtain the texture feature stream, identity feature stream, motion feature stream and edge feature stream corresponding to the verification video; The calculation module is used to calculate a texture jitter score based on the texture feature stream, an identity mutation score based on the identity feature stream, a motion violation score based on the motion feature stream, and an edge flicker score based on the edge feature stream, and to fuse the texture jitter score, identity mutation score, motion violation score, and edge flicker score to obtain a risk probability; The verification module is used to determine the face verification result based on the risk probability and the probability threshold corresponding to the current business.
8. A computer device, characterized in that, The computer device includes: At least one processor; and a memory communicatively connected to said at least one processor; The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the face verification method as described in any one of claims 1-6.
9. A storage medium containing computer-executable instructions, characterized in that, The computer-executable instructions, when executed by a computer processor, are used to perform the face verification method as described in any one of claims 1-6.
10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the face verification method according to any one of claims 1-6.