WO2021072870A1

WO2021072870A1 - Adversarial network-based fingerprint model generation method and related apparatus

Info

Publication number: WO2021072870A1
Application number: PCT/CN2019/118092
Authority: WO
Inventors: 王义文; 王健宗
Original assignee: 平安科技（深圳）有限公司
Priority date: 2019-10-15
Filing date: 2019-11-13
Publication date: 2021-04-22
Also published as: CN110929564B; CN110929564A

Abstract

An adversarial network-based fingerprint model generation method and apparatus, a computer device, and a storage medium. The adversarial network-based fingerprint model generation method comprises: inputting a fingerprint sample image into a first machine learning sub-model, and outputting, by the first machine learning sub-model, a fingerprint simulated image (S121); inputting the fingerprint simulated image into a second machine learning sub-model, and outputting, by the second machine learning sub-model, a determining result concerning whether it is a fingerprint simulated image, such that the first machine learning sub-model adjusts parameters according to the determination result and then the fingerprint simulated image output by the first machine learning sub-model is more similar to the finger sample image (S122); and if the recognition rate of the second machine learning sub-model reaches a predetermined recognition threshold, outputting the fingerprint simulated image as a fingerprint model (S124). A fingerprint model having high similarity to real fingerprints can be synthesized in the case of lacking a fingerprint database.

Description

Fingerprint model generation method and related device based on confrontation network

This application claims the priority of the Chinese patent application 201910979602.9 filed on October 25, 2019 with the application name "A fingerprint model generation method and related device based on a confrontation network", the entire content of which is incorporated herein by reference.

Technical field

This application relates to the field of machine learning technology, in particular to a fingerprint model generation method, device, computer equipment and storage medium based on a confrontation network.

Background technique

With the development of technology, fingerprints, a biological feature, are more and more widely used in scenarios such as forensics, airports, and smart phones. Therefore, related technologies for fingerprint research are becoming more and more popular, but there is a lack of a large number of fingerprint samples exclusively for research, and it is particularly difficult to collect a large number of fingerprint samples exclusively for research. To generate a fingerprint by synthesis, since each person's fingerprint has unique characteristics and patterns, it is difficult to generate a synthetic and realistic fingerprint. Due to the lack of fingerprint database support, the current synthetic fingerprints are not very similar to real fingerprints and are not suitable for research and other purposes.

Summary of the invention

Based on this, in order to solve the technical problem of low similarity between synthetic fingerprints and real fingerprints in related technologies, this application provides a fingerprint model generation method, device, computer equipment and storage medium based on a confrontation network.

In one aspect, a fingerprint model generation method based on a confrontation network includes: acquiring a fingerprint sample image; inputting the fingerprint sample image into a machine learning model, and the machine learning model outputs the generated fingerprint model; wherein, the machine learning model It includes a first machine learning sub-model and a second machine learning sub-model, said inputting the fingerprint sample image into a machine learning model, and outputting the generated fingerprint model from the machine learning model includes: inputting the fingerprint sample image into the first A machine learning sub-model, the first machine learning sub-model outputs a fingerprint simulation image; the fingerprint simulation image is input into the second machine learning sub-model, whether the second machine learning sub-model output is a fingerprint simulation image The judgment result, so that the first machine learning sub-model adjusts parameters according to the judgment result, so that the fingerprint simulation image output by the first machine learning sub-model is more similar to the fingerprint sample image; and the second machine learning sub-model is calculated The recognition rate of the model, wherein the recognition rate includes the judgment result of the second machine learning sub-model that the output of the positive sample is a fingerprint simulation image and the judgment result that the output of the negative sample is not a fingerprint simulation image accounts for the second machine learning The proportion of all judgment results output by the sub-model; if the recognition rate of the second machine learning sub-model reaches the predetermined recognition threshold, the fingerprint simulation image is output as the generated fingerprint model.

On the other hand, a fingerprint model generation device based on a confrontation network includes: a sample image acquisition unit for acquiring a fingerprint sample image; a machine learning output unit for inputting the fingerprint sample image into a machine learning model, and the machine The learning model outputs the generated fingerprint model; wherein, the machine learning model includes a first machine learning sub-model and a second machine learning sub-model, and the machine learning output unit includes: a simulated image output unit for converting the fingerprint sample An image is input to the first machine learning sub-model, and the first machine learning sub-model outputs a fingerprint simulation image; the judgment result output unit is configured to input the fingerprint simulation image into the second machine learning sub-model, and the first machine learning sub-model 2. Whether the output of the machine learning sub-model is the judgment result of a fingerprint simulation image; a recognition rate calculation unit for calculating the recognition rate of the second machine learning sub-model, wherein the recognition rate includes the second machine learning sub-model The ratio of the judgment result that the output of the positive sample is a simulated fingerprint image and the judgment result that the output of the negative sample is not the simulated fingerprint image accounts for all the judgment results output by the second machine learning sub-model; the fingerprint model output unit is used if the When the recognition rate of the second machine learning sub-model reaches the predetermined recognition threshold, the fingerprint simulation image is output as the generated fingerprint model.

On the other hand, a fingerprint model generation device based on a confrontation network includes a processor and a memory, and computer-readable instructions are stored on the memory. When the computer-readable instructions are executed by the processor, the above-mentioned Fingerprint model generation method based on confrontation network.

On the other hand, a computer-readable storage medium has a computer program stored thereon, and when the computer program is executed by a processor, the method for generating a fingerprint model based on a confrontation network as described above is realized.

The fingerprint model generation method, device, computer equipment, and storage medium based on the confrontation network described above input the fingerprint sample image into a machine learning model, and the machine learning model simulates and outputs a fingerprint model based on the fingerprint sample image. The machine learning model includes a first machine learning sub-model and a second machine learning sub-model. When the machine learning model simulates and outputs a fingerprint model based on the fingerprint sample image, it first inputs the fingerprint sample image into the first machine learning sub-model. Model, making the first machine learning sub-model output a fingerprint simulation image, and then input the fingerprint simulation image into the second machine learning sub-model, the second machine learning sub-model output is a fingerprint simulation image judgment As a result, the first machine learning sub-model adjusts parameters according to the judgment result, so that the output fingerprint simulation image of the first machine learning sub-model is more similar to the fingerprint sample image, so that the first machine learning sub-model and The output result of the second machine learning sub-model mutually influences the accuracy of the output result of the other party, forming an adversarial learning, so that the fingerprint model output by the machine learning model becomes closer and closer to the real fingerprint image. If the recognition rate of the second machine learning sub-model reaches the predetermined recognition threshold, it is proved that the machine learning model has been trained well enough, and the fingerprint simulation image is output as a fingerprint model. In this way, it is possible to synthesize a fingerprint model with higher similarity to the real fingerprint when there is no fingerprint database.

It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the application.

Description of the drawings

The drawings here are incorporated into the specification and constitute a part of the specification, show embodiments that conform to the application, and are used together with the specification to explain the principle of the application.

Fig. 1 is a flow chart showing a method for generating a fingerprint model based on a confrontation network according to an exemplary embodiment.

Fig. 2 is a specific implementation flow chart of step S120 in the method for generating a fingerprint model based on a confrontation network according to the corresponding embodiment of Fig. 1.

FIG. 3 is a specific implementation flow chart of the implementation of machine learning model training in the fingerprint model generation method based on the confrontation network according to the embodiment corresponding to FIG. 1.

Fig. 4 is a block diagram showing a fingerprint model generation device based on a confrontation network according to an exemplary embodiment.

Fig. 5 schematically shows an exemplary block diagram of a computer device for implementing the above-mentioned method for generating a fingerprint model based on a confrontation network.

Fig. 6 schematically shows a computer-readable storage medium for implementing the aforementioned method for generating a fingerprint model based on a confrontation network.

Fig. 7 is an implementation environment diagram of a user shunt method provided in an embodiment.

Through the above drawings, the specific embodiments of the present application have been shown, and there will be more detailed descriptions in the following. These drawings and text descriptions are not intended to limit the scope of the concept of the present application in any way, but by referring to specific embodiments. The concept of this application is explained to those skilled in the art.

Detailed ways

Here, an exemplary embodiment will be described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the application as detailed in the appended claims.

FIG. 7 is an implementation environment diagram of a fingerprint model generation method based on a confrontation network provided in an embodiment. As shown in FIG. 7, the implementation environment includes a computer device 100. The computer device 100 includes a fingerprint generator 101 and a fingerprint distinguisher 102.

The computer device 100 is a fingerprint generation system device, for example, a computer or a server used by maintenance personnel of the fingerprint generation system. The fingerprint generator 101 and the fingerprint distinguisher 102 are internal sub-modules. When the computer device 100 obtains the fingerprint sample image, it is input to the fingerprint generator 101, and then the fingerprint generator 101 generates a fingerprint simulation image, and inputs the fingerprint simulation image to the fingerprint distinguisher 102. Whether the output of the fingerprint distinguisher 102 is According to the judgment result of the fingerprint simulation image, the fingerprint generator 101 adjusts the parameters according to the judgment result so that the output fingerprint simulation image is more similar to the fingerprint sample image. Finally, the recognition rate of the fingerprint distinguisher 102 is calculated, where the recognition rate includes the judgment result that the fingerprint distinguisher 102 outputs a fingerprint simulation image for positive samples and the judgment result that the output of negative samples is not a fingerprint simulation image accounts for the output of the fingerprint distinguisher 102. If the recognition rate reaches the predetermined recognition threshold, the fingerprint simulation image is output as the generated fingerprint model.

It should be noted that the computer device 100 can be a smart phone, a tablet computer, a notebook computer, a desktop computer, etc., but is not limited to this. The computer device 100 may be connected via Bluetooth, USB (Universal Serial Bus, Universal Serial Bus) or other communication connection methods, and the present invention is not limited herein.

As shown in Figure 1, in one embodiment, a fingerprint model generation method based on a confrontation network is proposed. The fingerprint model generation method based on a confrontation network can be applied to a fingerprint generation device, and specifically may include the following steps: Step S110 , Obtain a fingerprint sample image; in one of the embodiments, the fingerprint sample image includes a fingerprint image and a fingerprint sketch. Step S110 may include: obtaining a complete fingerprint image or an incomplete fingerprint image; obtaining all closed curves Fingerprint sketches or fingerprint sketches with closed and unclosed curves. The true fingerprint image is a real fingerprint sample, and the fingerprint sketch is a preliminary simulated fingerprint image. Inputting the preliminary simulated fingerprint image into the machine learning model, making the machine learning model based on the preliminary simulated fingerprint image will generate a fingerprint model better and faster than generating a fingerprint model directly from a blank screen. Among them, when obtaining the fingerprint sample image, you can first obtain the complete fingerprint image or the incomplete fingerprint image, and then obtain the fingerprint sketch with all closed curves or the fingerprint sketch with closed and non-closed curves; you can also obtain the full fingerprint first. It is a fingerprint sketch with a closed curve or a fingerprint sketch with a closed curve and a non-closed curve, and then obtains the complete fingerprint image or the incomplete fingerprint image; it can also obtain the fingerprint image and the fingerprint sketch at the same time. This application will not do it here. limited. In one of the embodiments, the fingerprint sample image includes a fingerprint image and a fingerprint sketch, and step S110 may include: obtaining the fingerprint image from the fingerprint logger or from the fingerprint library; obtaining the fingerprint image from the fingerprint logger or making it through drawing software Or automatically generate fingerprint sketches. In the process of collecting fingerprints, some fingerprints may be incomplete. At this time, it is necessary to simulate and restore a complete fingerprint image based on the incomplete and incomplete part of the fingerprint image. Therefore, inputting the incomplete fingerprint image as a sample into the fingerprint generator can prompt the fingerprint generator to learn the ability of fingerprint restoration. Step S120: Input the fingerprint sample image into a machine learning model, and the machine learning model outputs the generated fingerprint model.

2 is a detailed description of step S120 in the method for generating a fingerprint model based on a confrontation network according to the corresponding embodiment of FIG. 1. In the method for generating a fingerprint model based on a confrontation network, the machine learning model includes a first machine learning sub-model and The second machine learning sub-model, step S120 may include the following steps: step S121, the fingerprint sample image is input to the first machine learning sub-model, and the first machine learning sub-model outputs a fingerprint simulation image; wherein, the The fingerprint simulation image refers to a fingerprint image generated by simulation based on the fingerprint sample image and according to the feature distribution of the fingerprint sample image. Step S122: The fingerprint simulation image is input to the second machine learning sub-model, and the second machine learning sub-model outputs a judgment result of whether the fingerprint simulation image is output, so that the first machine learning sub-model can judge according to the judgment. As a result, the parameters are adjusted so that the simulated fingerprint image output by the first machine learning sub-model is more similar to the fingerprint sample image; step S123, the recognition rate of the second machine learning sub-model is calculated, wherein the recognition rate includes all The second machine learning sub-model outputs the judgment result of the fingerprint simulation image for the positive sample and the judgment result of the negative sample output is not the fingerprint simulation image to the proportion of all judgment results output by the second machine learning sub-model; step S124, If the recognition rate of the second machine learning sub-model reaches the predetermined recognition threshold, output the fingerprint simulation image as the generated fingerprint model. In one of the embodiments, the training method of the machine learning model may be: input the fingerprint sample image into the first machine learning sub-model, the first machine learning sub-model outputs a fingerprint simulation image, and then The fingerprint simulation image is input to the second machine learning sub-model, and the second machine learning sub-model outputs the judgment result of whether it is a fingerprint simulation image, and the second machine learning sub-model is trained based on this. When the second machine When the recognition rate of the learning sub-model for the fingerprint simulation image reaches the first preset threshold, the training of the second machine learning sub-model is stopped. In this process, the second machine learning sub-model's ability to judge fingerprint simulation images is improved. In this way, when the recognition rate of the simulated fingerprint image by the second machine learning sub-model reaches the first preset threshold, the judgment accuracy of the simulated fingerprint image by the second machine learning sub-model reaches a high degree . Then continue to input the fingerprint simulation image into the second machine learning sub-model, and the second machine learning sub-model will feed back the output judgment result to the first machine learning sub-model, and the first machine learning sub-model can be based on the feedback The information adjusts its own model parameters to improve the similarity between the output fingerprint simulation image and the fingerprint sample image. The second machine learning sub-model continuously feeds back identification information to the first machine learning sub-model, and the first machine learning sub-model continuously adjusts its own model parameters. At this time, the recognition rate of the second machine learning sub-model for the fingerprint simulation image is increased with The similarity between the simulated fingerprint image and the fingerprint sample image increases and decreases. When the recognition rate of the fingerprint simulation image output by the first machine learning sub-model is reduced to a second preset threshold by the second machine learning sub-model, the training of the first machine learning sub-model is stopped. At this time, the similarity between the simulated fingerprint image and the fingerprint sample image reaches a high level. At this time, continue to input the fingerprint simulation image into the second machine learning sub-model, and sequentially loop the training of the second machine learning sub-model of the fingerprint in question and the training of the first machine learning sub-model Until the second machine learning sub-model's recognition rate of the fingerprint simulation image output by the first machine learning sub-model is reduced to a predetermined recognition threshold and cannot be further improved, the entire training process ends and the fingerprint model is obtained. In this way, the first machine learning sub-model and the second machine learning sub-model form an adversarial network to learn from each other, so that the fingerprint simulation image output by the first machine learning sub-model is similar to the fingerprint sample image and the second machine learning The correct rate of the sub-model judgment is gradually improved due to the improvement of the opponent in the confrontation learning, until it reaches a higher level.

In one of the embodiments, as shown in FIG. 3, the machine learning model is trained as follows: Step S101, the fingerprint sample image is input to the first machine learning sub-model, and the first machine learning sub-model outputs Fingerprint simulation image; step S102, use the primary fingerprint simulation image as a positive sample, and use the fingerprint sample image as a negative sample to form a first fingerprint image sample set; step S103, combine each of the first fingerprint image samples A fingerprint image sample is input into the second machine learning sub-model one by one for learning. The second machine learning sub-model outputs the judgment result of fingerprint simulation image. If the output of the positive sample is not the judgment result of fingerprint simulation image, or for negative The sample output is the judgment result of the fingerprint simulation image. The first machine learning sub-model is adjusted so that the second machine learning sub-model outputs the opposite judgment result; step S104, whether the output of the second machine learning sub-model is the fingerprint simulation image The judgment result is input into the first machine learning sub-model, so that the first machine learning sub-model adjusts the first machine learning sub-model according to the judgment result of whether the output of the second machine learning sub-model is a fingerprint simulation image or not, so that the The similarity between the fingerprint simulation image output by the first machine learning sub-model and the fingerprint sample image is improved; step S105, the recognition rate of the second machine learning sub-model is calculated; step S106, if the second machine learning sub-model When the recognition rate reaches the predetermined recognition threshold, the fingerprint simulation image is output as the generated fingerprint model.

In one of the embodiments, the method for training the machine learning model is training based on maximizing the value of the loss function, where the loss function is:

L _GAN ＝E _x～Pdata(x) [log D(x)+E _z～Pz(z) [log(1-D(G(z)))]

Among them: LGAN is the loss function value, x is the true fingerprint image, D is the second machine learning sub-model, EX～Pdata(x) is the loss expectation for the second machine learning sub-model, and z is the first machine learning sub-model of the input fingerprint. The fingerprint sample image of the model, G(z) is the fingerprint simulation image output by the first machine learning sub-model, D(G(z))) is the processing function of inputting the fingerprint simulation image into the second machine learning sub-model, Ez～Pz(z) are the expectations for the loss of the first machine learning sub-model.

Among them, the loss function refers to a function that measures the degree of difference between the data obtained by the model and the actual data, that is, a function used to evaluate the quality of the model. Here it refers to the degree of difference between the fingerprint mimic generated by the fingerprint generator and the real fingerprint image. If the fingerprint generator is used to calculate the degree of difference, it is concluded that the smaller the degree of difference (loss function value) is, the higher the similarity between the fingerprint pseudo image generated by the fingerprint generator and the true fingerprint image is, so The fingerprint generator is trained based on the value of the minimized loss function.

In one of the embodiments, the specific steps of the first machine learning sub-model to output the fingerprint simulation image may include: obtaining pixel change values of adjacent areas in the fingerprint sample image, and judging whether the pixel change value is less than a preset change threshold If the pixel change value is less than the preset change threshold, adjust the pixels in the adjacent area of the fingerprint sample image. If the amount of pixel change in the adjacent area in the fingerprint sample image is large, it means that the pixels in the adjacent area in the fingerprint sample image are not smooth. By calculating the amount of change between pixels in adjacent areas in the fingerprint sample image, fine-tuning the pixels in this area, so that the image remains smooth. Specifically, the change value is compared with the preset change threshold. If the change value is greater than the preset change threshold, it means that the pixels in the area are discontinuous and therefore need to be adjusted.

In one of the embodiments, the change value of each pixel in the adjacent area in the fingerprint sample image can be obtained, and the sum of the change value of each pixel can be calculated to obtain the total change value of each pixel in the adjacent area. The total change value is fed back to the loss function, so that each pixel in the adjacent area is adjusted, and the fingerprint generator generates a smoother fingerprint image.

Wherein, when calculating the change value of each pixel and the sum of the change value of each pixel based on a one-dimensional angle, the sum of the change value of each pixel in the adjacent area is:

The adding the sum of the change values of the pixels in the adjacent area to the loss function is:

L _GAN-TV ＝E _x～Pdata(x) [log D(x)+E _z～Pz(z) [log(1-D(G(z)))]+λTV(G(z))

Among them, y _n+1 -y _n represents the change value of two adjacent pixels,

Represents the sum of the change values of all two adjacent pixels in the adjacent area, and λ is a constant term coefficient.

Wherein, when calculating the change value of each pixel and the sum of the change values of all pixels in the adjacent area based on a two-dimensional angle, the sum of the change values of each pixel in the adjacent area is:

L _GAN-TV ＝E _x～Pdata(x) [log D(x)+E _z～Pz(z) [log(1-D(G(z)))]+λTV(G(z))

Among them, y _i+1,j -y _i,j represents the change value of two adjacent pixels in the i direction in the two-dimensional plane, and y _i,j+1 -y _i,j represents the phase up in the j direction in the two-dimensional plane. The change value of two adjacent pixels, |y _i+1,j -y _i,j |+|y _i,j+1 -y _i,j | The sum of the absolute values of the pixel change values in each direction, ∑ _i,j |y _i+1,j -y _i,j |+|y _i,j+1 -y _i,j | The sum of the absolute values of the pixel change values in the i and j directions of two adjacent pixels, and λ is a constant term coefficient.

In this way, the change value between pixels is evaluated from the horizontal and vertical directions, and the optimized fingerprint simulation image is smoother in all angles.

In one of the embodiments, the formula when the machine learning model outputs the fingerprint model is:

G ^* = arg min _G max _D L _GAN

Where: where: G ^* represents the first machine learning sub-model, min _G represents the minimization of the first machine learning sub-model, max _D is the maximization of the second machine learning sub-model, and L _GAN represents the loss function.

As shown in FIG. 4, in one embodiment, a fingerprint model generation device based on a confrontation network is provided. The fingerprint model generation device based on a confrontation network can be integrated into the above-mentioned fingerprint generation device, and can specifically include sample image acquisition. The unit 110 and the machine learning output unit 120. The sample image obtaining unit 110 is used to obtain a fingerprint sample image; the machine learning output unit 120 is used to input the fingerprint sample image into a machine learning model, and the machine learning model outputs a fingerprint model; wherein, the machine learning model includes a first A machine learning sub-model and a second machine learning sub-model, the machine learning output unit 120 includes: a simulation image output unit 121, configured to input the fingerprint sample image into the first machine learning sub-model, the first The machine learning sub-model outputs a fingerprint simulation image; the judgment result output unit 122 is configured to input the fingerprint simulation image into the second machine learning sub-model, and the second machine learning sub-model outputs a judgment result of whether the fingerprint simulation image is The recognition rate calculation unit 123 is used to calculate the recognition rate of the second machine learning sub-model, where the recognition rate includes the judgment result of the second machine learning sub-model for the positive sample output is a fingerprint simulation image and the The negative sample output is not the ratio of the judgment result of the fingerprint simulation image to all the judgment results output by the second machine learning sub-model; the fingerprint model output unit 124 is used for if the recognition rate of the second machine learning sub-model reaches the When the recognition threshold is predetermined, the fingerprint simulation image is output as a fingerprint model. For the implementation process of the functions and roles of each module in the above device, please refer to the implementation process of the corresponding steps in the fingerprint model generation method based on the confrontation network for details, which will not be repeated here. It should be noted that although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above may be embodied in one module or unit. Conversely, the features and functions of a module or unit described above can be further divided into multiple modules or units to be embodied. In addition, although the various steps of the method in the present application are described in a specific order in the drawings, this does not require or imply that these steps must be performed in the specific order, or that all the steps shown must be performed to achieve the desired result. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step for execution, and/or one step may be decomposed into multiple steps for execution, etc. Through the description of the above embodiments, those skilled in the art can easily understand that the example embodiments described here can be implemented by software, or can be implemented by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, U disk, mobile hard disk, etc.) or on the network , Including several instructions to make a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) execute the method according to the embodiment of the present application.

In the exemplary embodiment of the present application, a computer device capable of implementing the above method is also provided. 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, each aspect of the present application can be specifically implemented in the following forms, namely: complete hardware implementation, complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which can be collectively referred to herein as "Circuit", "Module" or "System". Hereinafter, a computer device 500 according to this embodiment of the present application will be described with reference to FIG. 5. The computer device 500 shown in FIG. 5 is only an example, and should not bring any limitation to the function and use range of the embodiments of the present application. As shown in FIG. 5, the computer device 500 is represented in the form of a general-purpose computing device. The components of the computer device 500 may include, but are not limited to: the aforementioned at least one processing unit 510, the aforementioned at least one storage unit 520, and a bus 530 connecting different system components (including the storage unit 520 and the processing unit 510). Wherein, the storage unit stores program code, and the program code can be executed by the processing unit 510, so that the processing unit 510 executes the various exemplary methods described in the “Exemplary Method” section of this specification. Steps of implementation. For example, the processing unit 510 may perform step S110 as shown in FIG. 1 to obtain a fingerprint sample image. Step S120: Input the fingerprint sample image into a machine learning model, and the machine learning model outputs a fingerprint model. The storage unit 520 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 5201 and/or a cache storage unit 5202, and may further include a read-only storage unit (ROM) 5203. The storage unit 520 may also include a program/utility tool 5204 having a set (at least one) program module 5205. Such program module 5205 includes but is not limited to: an operating system, one or more application programs, other program modules, and program data, Each of these examples or some combination may include the implementation of a network environment. The bus 530 may represent one or more of several types of bus structures, including a storage unit bus or a storage unit controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local area using any bus structure among multiple bus structures. bus. The computer device 500 may also communicate with one or more external devices 700 (such as keyboards, pointing devices, Bluetooth devices, etc.), and may also communicate with one or more devices that enable a user to interact with the computer device 500, and/or communicate with Any device (such as a router, modem, etc.) that enables the computer device 500 to communicate with one or more other computing devices. Such communication can be performed through an input/output (I/O) interface 550. In addition, the computer device 500 may also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and/or a public network, such as the Internet) through the network adapter 560. As shown in the figure, the network adapter 560 communicates with other modules of the computer device 500 through the bus 530. It should be understood that although not shown in the figure, other hardware and/or software modules can be used in conjunction with the computer device 500, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives And data backup storage system, etc. Through the description of the above embodiments, those skilled in the art can easily understand that the example embodiments described here can be implemented by software, or can be implemented by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, U disk, mobile hard disk, etc.) or on the network , Including several instructions to make a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) execute the method according to the embodiment of the present application. In the exemplary embodiment of the present application, a computer-readable storage medium is also provided, on which is stored a program product capable of implementing the above-mentioned method in this specification. 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 code. When the program product runs on a terminal device, the program code is used to make the The terminal device executes the steps according to various exemplary embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.

Referring to FIG. 6, a program product 600 for implementing the above method according to an embodiment of the present application is described. It can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be installed in a terminal device, For example, running on a personal computer. However, the program product of this application is not limited to this. In this document, the readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, device, or device. The above-mentioned computer-readable storage medium may be a non-volatile readable storage medium, such as a non-volatile readable storage medium stored in a CD-ROM, U disk, or mobile hard disk device, and may include several instructions to make a computer A device (which may be a personal computer, a server, a terminal device, or a network device, etc.) executes the method according to the embodiment of the present application. The program product can use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or a combination of any of the above. More specific examples (non-exhaustive list) of readable storage media include: electrical connections with one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable Type programmable read only memory (EPROM or flash memory), optical fiber, portable compact disk read only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above. The computer-readable signal medium may include a data signal propagated in baseband or as a part of a carrier wave, and readable program code is carried therein. This propagated data signal can take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. The readable signal medium may also be any readable medium other than a readable storage medium, and the readable medium may send, propagate, or transmit a program for use by or in combination with the instruction execution system, apparatus, or device. The program code contained on the readable medium can be transmitted by any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above. The program code used to perform the operations of the present application can be written in any combination of one or more programming languages. The programming languages include object-oriented programming languages—such as Java, C++, etc., as well as conventional procedural programming languages. Programming language-such as "C" language or similar programming language. The program code can be executed entirely on the user's computing device, partly on the user's device, executed as an independent software package, partly on the user's computing device and partly executed on the remote computing device, or entirely on the remote computing device or server Executed on. In the case of a remote computing device, the remote computing device can be connected to a user computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (for example, using Internet service providers). Business to connect via the Internet). In addition, the above-mentioned drawings are merely schematic illustrations of the processing included in the method according to the exemplary embodiments of the present application, and are not intended for limitation. It is easy to understand that the processing shown in the above drawings does not indicate or limit the time sequence of these processings. In addition, it is easy to understand that these processes can be executed synchronously or asynchronously in multiple modules, for example.

The above content is only a preferred exemplary embodiment of the present application, and is not intended to limit the implementation of the present application. According to the main concept and spirit of the present application, those of ordinary skill in the art can easily make corresponding modifications or modifications. Therefore, the protection scope of this application shall be subject to the protection scope required by the claims.

Claims

A fingerprint model generation method based on a confrontation network, including:

Obtain a fingerprint sample image;

Inputting the fingerprint sample image into a machine learning model, and the machine learning model outputs the generated fingerprint model;

Wherein, the machine learning model includes a first machine learning sub-model and a second machine learning sub-model, the fingerprint sample image is input into the machine learning model, and the fingerprint model generated by the output of the machine learning model includes:

Inputting the fingerprint sample image into the first machine learning sub-model, and the first machine learning sub-model outputs a fingerprint simulation image;

The fingerprint simulation image is input to the second machine learning sub-model, and the second machine learning sub-model outputs a judgment result of whether the fingerprint simulation image is output, so that the first machine learning sub-model adjusts parameters according to the judgment result , Making the output fingerprint simulation image of the first machine learning sub-model more similar to the fingerprint sample image;

Calculate the recognition rate of the second machine learning sub-model, where the recognition rate includes the judgment result of the second machine learning sub-model that the output of the positive sample is a fingerprint simulation image and the judgment that the output of the negative sample is not a fingerprint simulation image The ratio of the result to all the judgment results output by the second machine learning sub-model;

If the recognition rate of the second machine learning sub-model reaches the predetermined recognition threshold, output the fingerprint simulation image as the generated fingerprint model.
The method of claim 1, wherein the machine learning model is trained as follows:

Inputting the fingerprint sample image into the first machine learning sub-model, and the first machine learning sub-model outputs a fingerprint simulation image;

Taking the primary fingerprint simulation image as a positive sample and taking the fingerprint sample image as a negative sample to form a first fingerprint image sample set;

Input each fingerprint image sample in the first fingerprint image sample set one by one into the second machine learning sub-model for learning, and whether the second machine learning sub-model output is the judgment result of the fingerprint simulation image, if the output is for the positive sample If it is not the judgment result of the fingerprint simulation image, or the judgment result of the fingerprint simulation image for the negative sample output, adjust the first machine learning sub-model so that the second machine learning sub-model outputs the opposite judgment result;

The judgment result of whether the output of the second machine learning sub-model is a fingerprint simulation image is input to the first machine learning sub-model, so that the first machine learning sub-model outputs a fingerprint simulation image according to the second machine learning sub-model. Adjusting the first machine learning sub-model to increase the similarity between the fingerprint simulation image output by the first machine learning sub-model and the fingerprint sample image;

Calculating the recognition rate of the second machine learning sub-model;

If the recognition rate of the second machine learning sub-model reaches the predetermined recognition threshold, output the fingerprint simulation image as the generated fingerprint model.
The method according to claim 1, wherein the fingerprint sample image includes a true fingerprint image and a fingerprint sketch, and the acquiring the fingerprint sample image includes:

Obtain the fingerprint image from the fingerprint logger or retrieve the fingerprint image from the fingerprint database;

Obtain from the fingerprint logger or make fingerprint sketches through drawing software or automatically generate fingerprints.
The method according to claim 1, wherein the fingerprint sample image includes a true fingerprint image and a fingerprint sketch, and the acquiring the fingerprint sample image includes:

Obtain the complete fingerprint image or the incomplete fingerprint image;

Obtain fingerprint sketches that are all closed curves or fingerprint sketches that contain closed and unclosed curves.
The method according to claim 1, wherein the method of training the machine learning model is training based on the value of a maximization loss function, wherein the maximization loss function is:

L GAN ＝E x～Pdata(x) [log D(x)+E z～Pz(z) [log(1-D(G(z)))]

Among them: L GAN is the maximum loss function value, x is the true fingerprint image, D is the second machine learning sub-model E x～Pdata(x) is the expected loss of the second machine learning sub-model, z is the first input fingerprint The fingerprint sample image of the machine learning sub-model, G(z) is the fingerprint simulation image output by the first machine learning sub-model, D(G(z)) is the process of inputting the fingerprint simulation image into the second machine learning sub-model The function, E z～Pz(z) is the expectation for the loss of the first machine learning sub-model.
The method of claim 1, wherein the step of outputting a fingerprint simulation image by the first machine learning sub-model comprises:

Acquiring pixel change values of adjacent areas in the fingerprint sample image, and judging whether the pixel change value is less than a preset change threshold;

If the pixel change value is less than the preset change threshold value, the pixels in the adjacent area in the fingerprint sample image are adjusted.
8. The method of claim 6, wherein adjusting pixels in adjacent regions in the fingerprint sample image comprises:

Obtain the change value of each pixel in the adjacent area in the fingerprint sample image;

Calculating the sum of the change values of the pixels in the adjacent area to obtain the total change value of the pixels in the adjacent area;

Feedback the total change value of each pixel in the adjacent area to a pixel loss function, and adjust each pixel in the adjacent area;

Wherein, the pixel loss function is:

L GAN-TV ＝E x～Pdata(x) [log D(x)+E z～Pz(z) [log(1-D(G(z)))]+λTV(G(z))

Among them, L GAN-TV is the loss function value, x is the true fingerprint image, D is the second machine learning sub-model, E x ~ Pdata(x) is the expectation of the loss of the second machine learning sub-model, and z is the first fingerprint of the input A fingerprint sample image of a machine learning sub-model, G(z) is the fingerprint simulation image output by the first machine learning sub-model, D(G(z)) is the fingerprint simulation image input into the second machine learning sub-model The processing function, Ez～Pz(z) is the expectation of the loss of the first machine learning sub-model, TV(G(z)) is the total change value of each pixel in the adjacent area, and λ is the constant term coefficient.
A fingerprint model generation device based on a confrontation network, including:

Sample image acquisition unit for acquiring fingerprint sample images;

A machine learning output unit, configured to input the fingerprint sample image into a machine learning model, and the machine learning model outputs the generated fingerprint model;

Wherein, the machine learning model includes a first machine learning sub-model and a second machine learning sub-model, and the machine learning output unit includes:

A simulation image output unit configured to input the fingerprint sample image into the first machine learning sub-model, and the first machine learning sub-model outputs a fingerprint simulation image;

A judgment result output unit, configured to input the fingerprint simulation image into the second machine learning sub-model, and the second machine learning sub-model outputs a judgment result of whether the fingerprint simulation image is a fingerprint simulation image;

The recognition rate calculation unit is configured to calculate the recognition rate of the second machine learning sub-model, where the recognition rate includes the judgment result of the second machine learning sub-model that the output of the positive sample is a fingerprint simulation image and the negative sample The ratio of the judgment result output not being a fingerprint simulation image to all the judgment results output by the second machine learning sub-model;

The fingerprint model output unit is configured to output the fingerprint simulation image as the generated fingerprint model if the recognition rate of the second machine learning sub-model reaches the predetermined recognition threshold.
8. The device of claim 8, wherein the machine learning model is trained as follows:

Inputting the fingerprint sample image into the first machine learning sub-model, and the first machine learning sub-model outputs a fingerprint simulation image;

Taking the primary fingerprint simulation image as a positive sample and taking the fingerprint sample image as a negative sample to form a first fingerprint image sample set;

Input each fingerprint image sample in the first fingerprint image sample set one by one into the second machine learning sub-model for learning, and whether the second machine learning sub-model output is the judgment result of the fingerprint simulation image, if the output is for the positive sample If it is not the judgment result of the fingerprint simulation image, or the judgment result of the fingerprint simulation image for the negative sample output, adjust the first machine learning sub-model so that the second machine learning sub-model outputs the opposite judgment result;

The judgment result of whether the output of the second machine learning sub-model is a fingerprint simulation image is input to the first machine learning sub-model, so that the first machine learning sub-model outputs a fingerprint simulation image according to the second machine learning sub-model. Adjusting the first machine learning sub-model to increase the similarity between the fingerprint simulation image output by the first machine learning sub-model and the fingerprint sample image;

Calculating the recognition rate of the second machine learning sub-model;

If the recognition rate of the second machine learning sub-model reaches the predetermined recognition threshold, output the fingerprint simulation image as the generated fingerprint model.
8. The device according to claim 8, wherein the fingerprint sample image includes a true fingerprint image and a fingerprint sketch, and the sample image acquisition unit is configured to:

Obtain the fingerprint image from the fingerprint logger or retrieve the fingerprint image from the fingerprint database;

Obtain from the fingerprint logger or make fingerprint sketches through drawing software or automatically generate fingerprints.
8. The device according to claim 8, wherein the fingerprint sample image includes a true fingerprint image and a fingerprint sketch, and the sample image acquisition unit is configured to:

Obtain the complete fingerprint image or the incomplete fingerprint image;

Obtain fingerprint sketches that are all closed curves or fingerprint sketches that contain closed and unclosed curves.
8. The apparatus of claim 8, wherein the machine learning output unit is configured to perform training based on the value of a maximum loss function, wherein the maximum loss function is:

L GAN ＝E x～Pdata(x) [log D(x)+E z～Pz(z) [log(1-D(G(z)))]

Among them: L GAN is the maximum loss function value, x is the true fingerprint image, D is the second machine learning sub-model E x～Pdata(x) is the expected loss of the second machine learning sub-model, z is the first input fingerprint The fingerprint sample image of the machine learning sub-model, G(z) is the fingerprint simulation image output by the first machine learning sub-model, D(G(z)) is the process of inputting the fingerprint simulation image into the second machine learning sub-model The function, E z～Pz(z) is the expectation for the loss of the first machine learning sub-model.
The device according to claim 8, wherein the analog image output unit is configured to:

Acquiring pixel change values of adjacent areas in the fingerprint sample image, and judging whether the pixel change value is less than a preset change threshold;

If the pixel change value is less than the preset change threshold value, the pixels in the adjacent area in the fingerprint sample image are adjusted.
The device of claim 13, wherein the adjustment of pixels in adjacent areas in the fingerprint sample image comprises:

Obtain the change value of each pixel in the adjacent area in the fingerprint sample image;

Calculating the sum of the change values of the pixels in the adjacent area to obtain the total change value of the pixels in the adjacent area;

Feedback the total change value of each pixel in the adjacent area to a pixel loss function, and adjust each pixel in the adjacent area;

Wherein, the pixel loss function is:

L GAN-TV ＝E x～Pdata(x) [log D(x)+E z～Pz(z) [log(1-D(G(z)))]+λTV(G(z))

Among them, L GAN-TV is the loss function value, x is the true fingerprint image, D is the second machine learning sub-model, E x ~ Pdata(x) is the expectation of the loss of the second machine learning sub-model, and z is the first fingerprint of the input A fingerprint sample image of a machine learning sub-model, G(z) is the fingerprint simulation image output by the first machine learning sub-model, D(G(z)) is the fingerprint simulation image input into the second machine learning sub-model The processing function, E z～Pz(z) is the expectation of the loss of the first machine learning sub-model, TV(G(z)) is the total change value of each pixel in the adjacent area, and λ is the constant term coefficient.
A computer device includes a processor and a memory, and computer-readable instructions are stored on the memory. When the computer-readable instructions are executed by the processor, the processor is configured to perform the following processing:

Obtain a fingerprint sample image;

Inputting the fingerprint sample image into a machine learning model, and the machine learning model outputs the generated fingerprint model;

Wherein, the machine learning model includes a first machine learning sub-model and a second machine learning sub-model, the fingerprint sample image is input into the machine learning model, and the fingerprint model generated by the output of the machine learning model includes:

Inputting the fingerprint sample image into the first machine learning sub-model, and the first machine learning sub-model outputs a fingerprint simulation image;

The fingerprint simulation image is input to the second machine learning sub-model, and the second machine learning sub-model outputs a judgment result of whether the fingerprint simulation image is output, so that the first machine learning sub-model adjusts parameters according to the judgment result , Making the output fingerprint simulation image of the first machine learning sub-model more similar to the fingerprint sample image;

Calculate the recognition rate of the second machine learning sub-model, where the recognition rate includes the judgment result of the second machine learning sub-model that the output of the positive sample is a fingerprint simulation image and the judgment that the output of the negative sample is not a fingerprint simulation image The ratio of the result to all the judgment results output by the second machine learning sub-model;

If the recognition rate of the second machine learning sub-model reaches the predetermined recognition threshold, output the fingerprint simulation image as the generated fingerprint model.
The computer device according to claim 15, wherein when the computer-readable instructions are executed by the processor, the processor is further configured to perform the following processing:

Inputting the fingerprint sample image into the first machine learning sub-model, and the first machine learning sub-model outputs a fingerprint simulation image;

Taking the primary fingerprint simulation image as a positive sample and taking the fingerprint sample image as a negative sample to form a first fingerprint image sample set;

Input each fingerprint image sample in the first fingerprint image sample set one by one into the second machine learning sub-model for learning, and whether the second machine learning sub-model output is the judgment result of the fingerprint simulation image, if the output is for the positive sample If it is not the judgment result of the fingerprint simulation image, or the judgment result of the fingerprint simulation image for the negative sample output, adjust the first machine learning submodel so that the second machine learning submodel outputs the opposite judgment result;

The judgment result of whether the output of the second machine learning sub-model is a fingerprint simulation image is input to the first machine learning sub-model, so that the first machine learning sub-model outputs a fingerprint simulation image according to the second machine learning sub-model. Adjusting the first machine learning sub-model to increase the similarity between the fingerprint simulation image output by the first machine learning sub-model and the fingerprint sample image;

Calculating the recognition rate of the second machine learning sub-model;

If the recognition rate of the second machine learning sub-model reaches the predetermined recognition threshold, output the fingerprint simulation image as the generated fingerprint model.
15. The computer device according to claim 15, wherein the fingerprint sample image comprises a true fingerprint image and a fingerprint sketch, and said acquiring the fingerprint sample image comprises:

Obtain the fingerprint image from the fingerprint logger or retrieve the fingerprint image from the fingerprint database;

Obtain from the fingerprint logger or make fingerprint sketches through drawing software or automatically generate fingerprints.
15. The computer device according to claim 15, wherein the fingerprint sample image comprises a true fingerprint image and a fingerprint sketch, and said acquiring the fingerprint sample image comprises:

Obtain the complete fingerprint image or the incomplete fingerprint image;

Obtain fingerprint sketches that are all closed curves or fingerprint sketches that contain closed and unclosed curves.
The computer device according to claim 15, wherein the method for training the machine learning model is training based on the value of a maximization loss function, wherein the maximization loss function is:

L GAN ＝E x～Pdata(x) [log D(x)+E z～Pz(z) [log(1-D(G(z)))]

Among them: L GAN is the maximum loss function value, x is the true fingerprint image, D is the second machine learning sub-model E x～Pdata(x) is the expected loss of the second machine learning sub-model, z is the first input fingerprint The fingerprint sample image of the machine learning sub-model, G(z) is the fingerprint simulation image output by the first machine learning sub-model, D(G(z)) is the process of inputting the fingerprint simulation image into the second machine learning sub-model The function, E z～Pz(z) is the expectation for the loss of the first machine learning sub-model.
15. The computer device of claim 15, wherein the step of outputting a fingerprint simulation image by the first machine learning sub-model comprises:

Acquiring pixel change values of adjacent areas in the fingerprint sample image, and judging whether the pixel change value is less than a preset change threshold;

If the pixel change value is less than the preset change threshold value, the pixels in the adjacent area in the fingerprint sample image are adjusted.
22. The computer device of claim 20, wherein adjusting pixels in adjacent regions in the fingerprint sample image comprises:

Obtain the change value of each pixel in the adjacent area in the fingerprint sample image;

Calculating the sum of the change values of the pixels in the adjacent area to obtain the total change value of the pixels in the adjacent area;

Feedback the total change value of each pixel in the adjacent area to a pixel loss function, and adjust each pixel in the adjacent area;

Wherein, the pixel loss function is:

L GAN-TV ＝E x～Pdata(x) [log D(x)+E z～Pz(z) [log(1-D(G(z)))]+λTV(G(z))

Among them, L GAN-TV is the loss function value, x is the true fingerprint image, D is the second machine learning sub-model, E x ~ Pdata(x) is the expectation of the loss of the second machine learning sub-model, and z is the first fingerprint of the input A fingerprint sample image of a machine learning sub-model, G(z) is the fingerprint simulation image output by the first machine learning sub-model, D(G(z)) is the fingerprint simulation image input into the second machine learning sub-model The processing function, E z～Pz(z) is the expectation of the loss of the first machine learning sub-model, TV(G(z)) is the total change value of each pixel in the adjacent area, and λ is the constant term coefficient.
A computer-readable storage medium having a computer program stored thereon, wherein, when the computer program is executed by a processor, the processor is used to execute the method according to claims 1 to 7.