Prosthetic attack detection methods, systems, devices, and storage media

Abstract

The application provides a kind of prosthetic attack detection method, system, equipment and storage medium, comprising: obtaining the infrared image of target face, the infrared image is collected by infrared camera;The face pixel region of pre-set size is intercepted on the infrared image;The face pixel region is input into the pre-trained prosthetic attack detection model, whether the face pixel region is live body face image is judged by the prosthetic attack detection model, and live body detection result is output.The infrared image of input target face is detected by live body in the application through prosthetic attack detection model, and prosthetic attack detection model includes input layer, multiple convolution processing modules, maximum pooling layer, first 2D convolution layer and output layer, which significantly reduces the complexity of the model, facilitating the popularization and application of the model.

Description

TECHNICAL FIELD

[0001] The present application relates to living body detection, in particular to a prosthesis attack detection method, system, device and storage medium. BACKGROUND

[0002] Most of the anti-attack algorithms on the market are based on convolutional networks, and a fully connected layer is added after the multi-layer convolutional network. Many algorithms preprocess the input data, which includes taking a number of small face samples in the input face picture, then inputting these samples into the learning network respectively to obtain the output for the samples, and finally analyzing the results.

[0003] However, considering that convolution itself is a module taking process, therefore, this preprocessing process of taking multiple modules may be redundant. Therefore, a simple face recognition 2D paper attack algorithm based on a full convolution model is provided. SUMMARY

[0004] In view of the defects in the prior art, the purpose of the present application is to provide a prosthesis attack detection method, system, device and storage medium.

[0005] According to the prosthesis attack detection method provided by the present application, the following steps are included:

[0006] Step S1: acquiring an infrared image of a target face, the infrared image being collected by an infrared camera;

[0007] Step S2: cutting a face pixel region of a preset size on the infrared image;

[0008] Step S3: inputting the face pixel region into a pre-trained prosthesis attack detection model, determining whether the face pixel region is a living face image through the prosthesis attack detection model, and outputting a living body detection result.

[0009] Preferably, the prosthesis attack detection model includes an input layer, a plurality of convolution processing modules, a maximum pooling layer, a first 2D convolution layer and an output layer;

[0010] The input layer, the plurality of convolution processing modules, the maximum pooling layer, the first 2D convolution layer and the output layer are connected in sequence.

[0011] The convolution processing module includes a second 2D convolution layer, a batch normalization layer and a ReLU activation layer; the second 2D convolution layer, the batch normalization layer and the ReLU activation layer are connected in sequence.

[0012] Preferably, the output of the output layer is an N*N matrix, when the attack detection result is generated according to the N*N matrix,

[0013] a weight value is set for each output eigenvalue in the N×N matrix, wherein the weight value of the output eigenvalue located in the center of the matrix is greater than the weight value of the output eigenvalue located in the periphery of the matrix.

[0014] Preferably, the output of the output layer is the output eigenvalue in the N×N matrix, and the number of output eigenvalues greater than the first threshold value is generated according to the pre-set first threshold value;

[0015] According to the number of output eigenvalues, the proportion value of the number of output eigenvalues in the total number of output eigenvalues in the N×N matrix is generated;

[0016] According to the comparison of the proportion value and the pre-set second threshold value, it is determined whether the target face corresponding to the image is a live body.

[0017] Preferably, the prosthesis attack detection model comprises an input layer, a plurality of convolution processing modules, a first 2D convolution layer, a first batch normalization layer, a second 2D convolution layer, a sigmoid function layer and an output layer;

[0018] The input layer, a plurality of convolution processing modules, the first 2D convolution layer, the first batch normalization layer, the second 2D convolution layer, the sigmoid function layer and the output layer are sequentially connected.

[0019] The convolution processing module comprises a third 2D convolution layer, a second batch normalization layer and a ReLU activation layer; the second 2D convolution layer, the batch normalization layer and the ReLU activation layer are sequentially connected.

[0020] Preferably, the training process of the prosthesis attack detection model comprises the following steps:

[0021] Step M1: Collecting infrared images of real faces, and generating a positive sample training set after labeling the infrared images of the real faces;

[0022] Step M2: Collecting infrared images of attack photos, and generating a negative sample training set after labeling the infrared images of the attack photos;

[0023] Step M3: Training a pre-set convolutional neural network model according to the negative sample training set and the positive sample training set to generate a prosthesis attack detection model.

[0024] Preferably, the step S3 comprises the following steps:

[0025] Step S301: Inputting the face pixel region into a pre-trained prosthesis attack detection model;

[0026] Step S302: determining whether the face pixel region is a live face image through the prosthesis attack detection model, and generating a live detection result;

[0027] Step S303: when the probability of the live face image in the live detection result is greater than the probability of the prosthesis attack image, it is determined that the face pixel region is a live face image, otherwise it is determined that the face pixel region is a prosthesis face image.

[0028] According to the prosthesis attack detection system provided by the application, the following modules are included:

[0029] An image acquisition module acquires an infrared image of a target face, and the infrared image is collected by an infrared camera;

[0030] A face region acquisition module is used to intercept a face pixel region of a preset size on the infrared image;

[0031] An attack detection module is used to input the face pixel region into a pre-trained prosthesis attack detection model, determine whether the face pixel region is a live face image through the prosthesis attack detection model, and output a live detection result.

[0032] According to the prosthesis attack detection device provided by the application, the following are included:

[0033] A processor;

[0034] A memory having executable instructions of the processor stored therein;

[0035] The processor is configured to execute the steps of the prosthesis attack detection method by executing the executable instructions.

[0036] According to the computer readable storage medium provided by the application, a program is stored, and the program is executed to realize the steps of the prosthesis attack detection method.

[0037] Compared with the prior art, the application has the following beneficial effects:

[0038] In the application, the infrared image of the input target face is live detected by the prosthesis attack detection model, the prosthesis attack detection model includes an input layer, a plurality of convolution processing modules, a maximum pooling layer, a first 2D convolution layer and an output layer, which significantly reduces the complexity of the model and facilitates the popularization and application of the model;

[0039] In the application, a weight value is set for each output feature value in the N*N matrix output by the output layer, and the weight value of the output feature value located in the center of the matrix is greater than the weight value of the output feature value located in the periphery of the matrix, thereby providing the accuracy of the prosthesis attack detection judgment;

[0040] The application generates a proportion value of the output characteristic value quantity in the total amount of output characteristic values in the N*N matrix according to the output characteristic value quantity, and determines whether the target face corresponding to the image is a living body according to the comparison of the proportion value with a preset second threshold value. The application can not only cope with infrared image input of different resolutions, but also be applicable to non-full face attacks under full false attack training. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of the provided drawings. Other features, objects and advantages of the present application will become more apparent through reading the following detailed description of the non-limiting embodiments with reference to the accompanying drawings:

[0042] Figure 1 The step flow chart of the prosthesis attack detection method in the embodiment of the present application;

[0043] Figure 2 The step flow chart of the prosthesis detection by the prosthesis attack detection model in the embodiment of the present application;

[0044] Figure 3 The step flow chart of the prosthesis attack detection model training in the embodiment of the present application;

[0045] Figure 4 The module schematic diagram of the prosthesis attack detection model in the embodiment of the present application;

[0046] Figure 5 The module schematic diagram of the prosthesis attack detection model in the embodiment of the present application;

[0047] Figure 6 The loss change schematic diagram of the prosthesis attack detection model in the embodiment of the present application;

[0048] Figure 7 The change process schematic diagram of the loss value of each feature in the input image in the embodiment of the present application;

[0049] Figure 8 The change process schematic diagram of the EER before taking the fixed weight matrix in the embodiment of the present application;

[0050] Figure 9 The change process schematic diagram of the EER after taking the fixed weight matrix in the embodiment of the present application;

[0051] Figure 10The FAR and FRR change process schematic diagram of P1 taking 0.5 in the embodiment of the present application is shown in the figure;

[0052] Figure 11 The FAR and FRR change process schematic diagram of the prosthetic attack detection model in the modified example of the present application is shown in the figure;

[0053] Figure 12 The threshold value P1, P2 and EER change process schematic diagram in the embodiment of the present application is shown in the figure;

[0054] Figure 13 The FAR and FRR change process schematic diagram under the optimal threshold value P1, P2 in the embodiment of the present application is shown in the figure;

[0055] Figure 14 The application example schematic diagram of the prosthetic attack detection model in the embodiment of the present application is shown in the figure;

[0056] Figure 15 The module schematic diagram of the prosthetic attack detection system in the embodiment of the present application is shown in the figure;

[0057] Figure 16 The structure schematic diagram of the prosthetic attack detection device in the embodiment of the present application is shown in the figure;

[0058] Figure 17 The structure schematic diagram of the computer readable storage medium in the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0059] The present application will be described in detail below with specific embodiments. The following embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any form. It should be pointed out that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made. These are within the scope of the present application.

[0060] The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-mentioned figures are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein, for example, can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0061] The technical solutions of the present application are described in detail below with specific examples. The following specific examples can be combined with each other, and the same or similar concepts or processes may not be described in detail in some examples.

[0062] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems are described in detail below with specific examples. The following specific examples can be combined with each other, and the same or similar concepts or processes may not be described in detail in some examples. The embodiments of the present application will be described below with reference to the accompanying drawings.

[0063] Figure 1 For the step flow chart of the prosthesis attack detection method in the embodiments of the present application, as shown in Figure 1 The prosthesis attack detection method provided by the present application comprises the following steps:

[0064] Step S1: Obtain an infrared image of a target face, wherein the infrared image is collected by an infrared camera;

[0065] Step S2: Crop a face pixel region of a preset size from the infrared image;

[0066] Step S3: Input the face pixel region into a pre-trained prosthesis attack detection model, determine whether the face pixel region is a live face image through the prosthesis attack detection model, and output a live detection result.

[0067] Figure 2 For the step flow chart of the prosthesis attack detection model in the embodiments of the present application, as shown in Figure 2 The step S3 comprises the following steps:

[0068] Step S301: Input the face pixel region into a pre-trained prosthesis attack detection model;

[0069] Step S302: Determine whether the face pixel region is a live face image through the prosthesis attack detection model, and generate a live detection result;

[0070] Step S303: When the probability of a live face image in the live detection result is greater than the probability of a prosthesis attack image, determine that the face pixel region is a live face image, otherwise determine that the face pixel region is a prosthesis face image.

[0071] Figure 4 For the module schematic diagram of the prosthesis attack detection model in the embodiments of the present application, as shown in Figure 4 The prosthesis attack detection model comprises an input layer, a plurality of convolution processing modules, a maximum pooling layer, a first 2D convolution layer, and an output layer;

[0072] The input layer, the plurality of convolution processing modules, the max pooling layer, the first 2D convolution layer and the output layer are sequentially connected.

[0073] The convolution processing module comprises a second 2D convolution layer, a batch normalization layer and a ReLU activation layer; the second 2D convolution layer, the batch normalization layer and the ReLU activation layer are sequentially connected.

[0074] In the embodiment of the present application, after obtaining the 6*6 feature map, global average is used to determine whether an input infrared image is a live face or a fake face. The training target is set as a 6*6 all-1 matrix for a live face and a 6*6 all-0 matrix for a fake face, and MSE (Mean Squared Error) is used as the loss function.

[0075] Figure 6 The loss change diagram of the fake attack detection model in the embodiment of the present application is shown in FIG. 2. Figure 6 As shown in FIG. 2, with the increase of the training time, the accuracy of the fake attack detection model gradually increases. Figure 7 The loss value change process diagram of each feature in the input image in the embodiment of the present application is shown in FIG. 3. Figure 7 As shown in FIG. 3, it can be observed that the loss values of the features around the circle, especially the four corners, are relatively higher and change more evenly. Since it is mentioned above that each feature in the feature map output by the fully convolutional network corresponds to the feature of the corresponding region of the original image, the information is extracted by the learning network, therefore, the features around the circle collect more background environment interference rather than the actual face.

[0076] Another observation is that the loss of the four features in the middle of the lower side is also relatively low, which should be corresponding to the neck part in the face picture. The left and right sides are sometimes the background, and the upper side has a high probability to be the hair. Therefore, this analysis why the loss of the features in the middle of the lower side and the 4*4 features in the center basically falls below 0.05, but the loss of the features in other side regions is higher.

[0077] Therefore, in the embodiment of the present application, the output of the output layer is an N*N matrix, when the attack detection result is generated according to the N*N matrix,

[0078] A weight value is set for each output feature value in the N*N matrix, wherein the weight value of the output feature value located in the center of the matrix is greater than the weight value of the output feature value located in the periphery of the matrix.

[0079] Therefore, according to the rule that the weight of the features in the four corners is less than the weight of the features in the four sides, and the weight of the features in the four sides is less than the weight of the features in the center, a fixed weight matrix can be taken when calculating the output result, as shown below.

[0080]

[0081] Figure 8 Fig. 1 is a schematic diagram of the EER change process before taking the fixed weight matrix in an embodiment of the present application, Figure 9 Fig. 2 is a schematic diagram of the EER change process after taking the fixed weight matrix in an embodiment of the present application, as shown in Figure 8 , Figure 9 The method of taking the fixed weight matrix indeed achieves obvious effect, and the EER (Equal Error Rate) of the prosthesis attack detection model is reduced by about 1.2% after adding the weight matrix.

[0082] In an embodiment of the present application, the output of the output layer is an N*N matrix, and the output feature value of each output feature value in the matrix is determined according to a pre-set first threshold P1 to generate the number of output feature values greater than the first threshold;

[0083] According to the number of output feature values, a proportion value of the number of output feature values in the total amount of output feature values in the N*N matrix is generated;

[0084] According to the proportion value and a pre-set second threshold P2, it is determined whether the target face corresponding to the image is a live body.

[0085] The method can cope with infrared image inputs of different resolutions. Since the convolution process does not require fixed input size, for larger resolution infrared images, the same trained network can be used to output larger feature maps, and then the same judgment method is used to obtain the result. The full convolution network has another advantage that it can perform autonomous segmentation on the original image. For example, each feature output by the 2D convolution is the reading and judgment of the maximum 31*31 region of the original image, so the full fake attack can be used for training, and then different judgment methods can be used to cope with non-full face attacks. For example, reduce the value of P2, so that if part of the output feature map is false, the corresponding part of the original image is also likely to be false, so as to cope with non-full face attacks.

[0086] Figure 10 Fig. 3 is a schematic diagram of the FAR (False Acceptance Rate) and FRR (False Rejection Rate) change process when P1 is 0.5 in an embodiment of the present application, as shown in Figure 10 P2 is no longer average, and the high P2 value indicates that the fitting score of the live body face is very close to 1, while the prosthesis is relatively chaotic.

[0087] Figure 5 Fig. 4 is a schematic diagram of the module of the prosthesis attack detection model in a deformation example of the present application, as shown in Figure 5As shown in the figure, the prosthetic attack detection model comprises an input layer, a plurality of convolution processing modules, a first 2D convolution layer, a first batch normalization layer, a second 2D convolution layer, a sigmoid function layer and an output layer;

[0088] The input layer, the plurality of convolution processing modules, the first 2D convolution layer, the first batch normalization layer, the second 2D convolution layer, the sigmoid function layer and the output layer are sequentially connected.

[0089] The convolution processing module comprises a third 2D convolution layer, a second batch normalization layer and a ReLU activation layer; the second 2D convolution layer, the batch normalization layer and the ReLU activation layer are sequentially connected.

[0090] Figure 11 The figure shows the change process of FAR and FRR in the prosthetic attack detection model in the deformation example of the present application, as Figure 11 The network is expanded to a maximum of 64 channels, the last ReLU activation is removed, and the stride of all convolutions except the last layer is changed to 2. The output of the model is still a 6x6 feature map, and the total output feature value has increased to about 20,000. Using the same training set and test set, the EER can be reduced to about 1% at P1=0.5, and then the EER can be reduced to less than 1% on a random test set of 4000.

[0091] Figure 12 The figure shows the change process of threshold P1, P2 and EER in the embodiment of the present application, Figure 13 The figure shows the change process of FAR and FRR under the optimal threshold P1, P2 in the embodiment of the present application, as Figure 12 、 Figure 13 As shown in the figure, when the threshold P1 is near 0.95, the EER can take the lowest value, although this lowest value is not necessarily the global minimum value, but from the figure we can infer that the global minimum value is almost only possible to appear here. Therefore, by adjusting the value range of the P1 axis, gradually narrowing the interval, finally P1=0.960, P2=0.306 can take EER=0.07% when P1=0.960, P2=0.306. In fact, the FRR and FAR at this time are 0.05% and 0.10%, respectively.

[0092] Figure 3 The figure shows the step flow chart of the training of the prosthetic attack detection model in the embodiment of the present application, as Figure 3 The training process of the prosthetic attack detection model comprises the following steps:

[0093] Step M1: Collecting infrared images of real faces, generating a positive sample training set after labeling the infrared images of the real faces;

[0094] Step M2: collect the infrared image of the attack photo, and generate a negative sample training set after labeling the infrared image of the attack photo;

[0095] Step M3: training a pre-set convolutional neural network model according to the negative sample training set and the positive sample training set to generate a prosthesis attack detection model.

[0096] Figure 14 For the application example of the prosthesis attack detection model in the embodiment of the application, as shown in Figure 14 It can be noted that in the printed picture in the lower left corner, although some regions of the feature map are white (close to 1), the total output score is still 0. This is because the threshold P1=0.96 is very high, so these points are still judged as false points.

[0097] Figure 15 For the module schematic diagram of the prosthesis attack detection system in the embodiment of the application, as shown in Figure 15 The prosthesis attack detection system provided by the application includes the following modules:

[0098] An image acquisition module acquires an infrared image of a target face, and the infrared image is collected by an infrared camera;

[0099] A face region acquisition module is used to crop a face pixel region of a pre-set size on the infrared image;

[0100] An attack detection module is used to input the face pixel region into a pre-trained prosthesis attack detection model, judge whether the face pixel region is a live face image through the prosthesis attack detection model, and output a live detection result.

[0101] The embodiment of the application further provides a prosthesis attack detection device, which includes a processor and a memory having executable instructions of the processor stored therein. The processor is configured to perform the steps of the prosthesis attack detection method via execution of the executable instructions.

[0102] As described above, in the embodiment of the application, the infrared image of the input target face is detected by the prosthesis attack detection model, the prosthesis attack detection model includes an input layer, a plurality of convolution processing modules, a maximum pooling layer, a first 2D convolution layer and an output layer, which significantly reduces the complexity of the model and facilitates the popularization and application of the model.

[0103] Those skilled in the art can understand that various aspects of the present application can be implemented as a system, a method or a program product. Therefore, various aspects of the present application can be embodied in a form of entirely hardware, entirely software (including firmware, microcode, etc.), or a combination of hardware and software, which can be collectively referred to as "circuitry", "module" or "platform" herein.

[0104] Figure 16 is a structural schematic diagram of the prosthetic attack detection device in the embodiment of the present application. The electronic device 600 according to this embodiment of the present application will be described below with reference to Figure 16 Figure 16 The displayed electronic device 600 is merely an example and should not bring any limitation to the function and use range of the embodiment of the present application.

[0105] As shown in Figure 16 , the electronic device 600 is in the form of a general computing device. The components of the electronic device 600 can include, but are not limited to, at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.

[0106] The storage unit stores program codes, which can be executed by the processing unit 610, so that the processing unit 610 performs the steps according to various exemplary embodiments of the present application described in the prosthetic attack detection method part of the present specification. For example, the processing unit 610 can perform the steps as shown in Figure 1

[0107] The storage unit 620 can include a readable medium in the form of a volatile storage unit, such as a random access memory (RAM) 6201 and / or a cache memory unit 6202, and can further include a read-only memory (ROM) 6203.

[0108] The storage unit 620 can further include a program / utility 6204 having a set of (at least one) program modules 6205, which include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or a combination thereof can include implementation of a network environment.

[0109] The bus 630 can represent one or more of several types of bus structures, including a storage unit bus or storage unit controller, a peripheral bus, a graphics acceleration port, a processing unit bus, or a local bus using any of a variety of bus architectures.

[0110] ​​The electronic device 600 can also communicate with one or more external devices 700 such as a keyboard or pointing device, a Bluetooth device, or a database, and / or one or more devices that enable a user to interact with the electronic device 600 and / or one or more devices (e.g., a router, a modem, a server, etc.) that enable the electronic device 600 to communicate with one or more other computing devices. Such communication can occur via an input / output (I / O) interface 650. Still yet, the electronic device 600 can communicate with one or more networks, such as a local area network (LAN), a wide area network (WAN), and / or the Internet, through a network adapter 660. The network adapter 660 can be any of a plurality of different types of adapters to enable the electronic device 600 to communicate with such networks and / or devices through a peripheral interface bus, such as a Universal Serial Bus (USB) bus, a Bluetooth ® connection, a 1394 bus, and / or the like. It should be understood that, although not shown, other hardware and / or software components could be used in conjunction with the electronic device 600. These components, such as the operating system, the platform software, microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data archival storage systems, etc., can also be utilized in connection with the embodiments of the present application. Figure 16

[0111] The embodiment of the present application also provides a computer readable storage medium for storing a program, the program being executed to implement the steps of the prosthesis attack detection method. In some possible implementation manners, various aspects of the present application can also be implemented in the form of a program product, which includes program codes for causing a terminal device to perform the steps of various exemplary embodiments of the present application described in the above-mentioned prosthesis attack detection method part of the specification when the program product is run on the terminal device.

[0112] As shown above, the program of the computer readable storage medium of the embodiment, when executed, performs living body detection on the input target human face infrared image through the prosthesis attack detection model, the prosthesis attack detection model includes an input layer, a plurality of convolution processing modules, a maximum pooling layer, a first 2D convolution layer and an output layer, which significantly reduces the complexity of the model and facilitates the popularization and application of the model.

[0113] Figure 17 is a structural schematic diagram of the computer readable storage medium in the embodiment of the present application. As shown in Figure 17 The program product 800 for implementing the above-mentioned method according to the embodiment of the present application can be in the form of a portable compact disc read-only memory (CD-ROM) and includes program codes, and can be run on a terminal device, such as a personal computer. However, the program product of the present application is not limited to this, and in this document, the readable storage medium can be any tangible medium containing or storing a program, which can be used or combined with an instruction execution system, device or apparatus.

[0114] ​A program product can take any combination of one or more computer-readable media. The computer-readable media can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium include an electrical connection having one or more wires, a portable disc, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0115] A computer readable storage medium can include a data signal traveling in or on a baseband medium or a carrier wave by itself or in combination with another data signal traveling in or on the baseband medium, wherein the data signal(s) can be electronically readable computer readable program code. The data signal can be a propagated data signal, for example, a propagated signal, for example, an electromagnetic signal, an optical signal, or any suitable combination of the foregoing. A computer readable storage medium can also be any tangible storage medium which can electronically store instructions of the programming code coupled to it or in which the instructions of the programming code can be stored, including a storage device. The instructions of the programming code can be electronically transmittable to or electronically transferable to the computer readable storage medium.

[0116] The program code can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++, or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computing device, partly on the user's computing device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device or entirely on the remote computing device or server. In the latter scenario, the remote computing device can be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computing device, for example, through the Internet using an Internet Service Provider (ISP). The program code can be embodied in any combination of hardware and software.

[0117] In the embodiment of the present application, the infrared image of the input target face is subjected to live body detection through a prosthesis attack detection model, the prosthesis attack detection model comprises an input layer, a plurality of convolution processing modules, a maximum pooling layer, a first 2D convolution layer and an output layer, which significantly reduces the complexity of the model and facilitates the popularization and application of the model; a weight value is set for each output characteristic value in the N*N matrix output by the output layer, wherein the weight value of the output characteristic value located in the center of the matrix is greater than the weight value of the output characteristic value located in the periphery of the matrix, thereby providing the accuracy of the prosthesis attack detection judgment; the proportion value of the number of output characteristic values in the total amount of output characteristic values in the N*N matrix is generated according to the number of output characteristic values, and whether the target face corresponding to the image is a live body is determined according to the comparison between the proportion value and a second threshold value, which not only can cope with infrared images of different resolutions, but also can be applied to non-full face attacks under the training of full false attacks.

[0118] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0119] The specific embodiments of the present application are described above. It should be understood that the present application is not limited to the above specific embodiments, and various modifications or changes can be made by those skilled in the art within the scope of the claims, which does not affect the essential content of the present application.

Claims

1. A method of prosthetic attack detection, the method comprising: The method comprises the following steps: Step S1: acquiring an infrared image of a target face, the infrared image being collected by an infrared camera; Step S2: intercepting a face pixel region of a preset size on the infrared image; Step S3: inputting the face pixel region into a pre-trained prosthesis attack detection model, judging whether the face pixel region is a live face image through the prosthesis attack detection model, and outputting a live detection result; The prosthesis attack detection model comprises an input layer, a plurality of convolution processing modules, a maximum pooling layer, a first 2D convolution layer, and an output layer; The input layer, the plurality of convolution processing modules, the maximum pooling layer, the first 2D convolution layer, and the output layer are sequentially connected; The convolution processing module comprises a second 2D convolution layer, a batch normalization layer, and a ReLU activation layer; the second 2D convolution layer, the batch normalization layer, and the ReLU activation layer are sequentially connected; The output of the output layer is an N×N matrix, when the attack detection result is generated according to the N×N matrix, A weight value is set for each output feature value in the N×N matrix, wherein the weight value of the output feature value located in the center of the matrix is greater than the weight value of the output feature value located in the periphery of the matrix.

2. The prosthetic attack detection method of claim 1, wherein, Each output feature value in the N×N matrix output by the output layer is subjected to live judgment according to a pre-set first threshold value, and the number of output feature values greater than the first threshold value is generated; According to the output feature value number, a proportion value of the output feature value number to the total amount of output feature values in the N×N matrix is generated; According to the comparison between the proportion value and a pre-set second threshold value, it is determined whether the target face corresponding to the image is a live body.

3. The prosthetic attack detection method of claim 1, wherein, The prosthesis attack detection model comprises an input layer, a plurality of convolution processing modules, a first 2D convolution layer, a first batch normalization layer, a second 2D convolution layer, a sigmoid function layer, and an output layer; The input layer, the plurality of convolution processing modules, the first 2D convolution layer, the first batch normalization layer, the second 2D convolution layer, the sigmoid function layer, and the output layer are sequentially connected; The convolution processing module comprises a third 2D convolution layer, a second batch normalization layer, and a ReLU activation layer; the second 2D convolution layer, the batch normalization layer, and the ReLU activation layer are sequentially connected.

4. The prosthetic attack detection method of claim 1, wherein, The training process of the prosthesis attack detection model comprises the following steps: Step M1: collecting infrared images of real faces, and generating a positive sample training set after labeling the infrared images of the real faces; Step M2: collecting infrared images of attack photos, and generating a negative sample training set after labeling the infrared images of the attack photos; Step M3: training a pre-set convolutional neural network model to generate a prosthesis attack detection model according to the negative sample training set and the positive sample training set.

5. The prosthetic attack detection method of claim 1, wherein, The step S3 comprises the following steps: Step S301: inputting the face pixel region into a pre-trained prosthesis attack detection model; Step S302: determining whether the face pixel region is a live face image by the prosthesis attack detection model, and generating a live detection result; Step S303: when the probability of the live face image in the live detection result is greater than the probability of the prosthesis attack image, determining that the face pixel region is a live face image, otherwise, determining that the face pixel region is a prosthesis face image.

6. A prosthetic attack detection system, comprising: Comprise the following modules: An image acquisition module acquires an infrared image of a target face, wherein the infrared image is collected by an infrared camera; A face region acquisition module is configured to crop a face pixel region of a preset size from the infrared image; An attack detection module is configured to input the face pixel region into a pre-trained prosthesis attack detection model, determine whether the face pixel region is a live face image by the prosthesis attack detection model, and output a live detection result; The prosthesis attack detection model comprises an input layer, a plurality of convolution processing modules, a maximum pooling layer, a first 2D convolution layer, and an output layer; The input layer, the plurality of convolution processing modules, the maximum pooling layer, the first 2D convolution layer, and the output layer are sequentially connected; The convolution processing module comprises a second 2D convolution layer, a batch normalization layer, and a ReLU activation layer; the second 2D convolution layer, the batch normalization layer, and the ReLU activation layer are sequentially connected; The output of the output layer is an N×N matrix, when the attack detection result is generated according to the N×N matrix, Each output feature value in the N×N matrix is assigned a weight value, wherein the weight value of the output feature value located in the center of the matrix is greater than the weight value of the output feature value located in the periphery of the matrix.

7. A prosthetic attack detection device, characterized by, Comprise: A processor; A memory having executable instructions of the processor stored therein; The processor is configured to execute the steps of the prosthesis attack detection method of any one of claims 1 to 5 by executing the executable instructions.

8. A computer readable storage medium for storing a program, characterized in that, The program is executed to realize the steps of the prosthesis attack detection method of any one of claims 1 to 5.

Images