Fingerprint image generation method and device

Abstract

The invention provides a fingerprint image generation method and device. The fingerprint image generation method comprises the following steps: acquiring M sample fingerprint images; on the basis of each sample fingerprint image in the M sample fingerprint images, N approximate images are generated according to N target similarities, and each target similarity is the similarity between the corresponding approximate image and the corresponding sample fingerprint image; analyzing the difference between each sample fingerprint image in the M sample fingerprint images and the corresponding approximate image in the N approximate images to obtain (M * N) difference parameters; respectively counting M difference parameters respectively associated with P similarities in the N target similarities to obtain difference statistical parameters respectively corresponding to the P similarities; acquiring a reference fingerprint image; and based on the reference fingerprint image and the difference statistical parameters corresponding to the P similarities, generating simulated fingerprint images corresponding to the P similarities.

Description

Technical Field

[0001] The present disclosure generally relates to fingerprint recognition, and more particularly, to a fingerprint image generation method and apparatus. Background Art

[0002] Everyone's fingerprint is unique. Fingerprint recognition technology uses this feature to achieve identity authentication by collecting, processing, analyzing and comparing fingerprint images. However, with the development of technology, malicious attackers have gradually emerged to create fake fingerprints through forgery, imitation or other means to deceive fingerprint recognition systems. For this reason, PAD (Presentation Attack Detection) algorithms have emerged in the field to identify fake fingerprints before formally verifying the fingerprint identity, and can only authenticate fingerprints that have passed the identification (for example, fingerprints that are determined to be real) to resist presentation attacks.

[0003] The PAD algorithm usually uses a Liveness Score to measure the possible activities (such as forgery, attack, etc.) in the fingerprint image. The Liveness Score is usually between 0 and 1, or between 0 and 100. The higher the value, the more likely the fingerprint image is to be authentic and valid. In order to test the PAD algorithm, it may be necessary or advantageous to use a large number of fingerprint images with different Liveness Scores. The traditional method is to construct PAI (Presentation Attack Instrument) materials based on real fingerprint samples collected from volunteers. This process has many steps, is time-consuming, and may need to comply with relevant laws and ethical standards. Therefore, it is costly and difficult to cover the entire LivenessScore value range.

[0004] Some methods for directly generating electronic fingerprint images have been proposed in the related art, such as generating adversarial patches by training an adversarial patch generator and inserting them into sample images to generate artificial adversarial images. This method can generate fingerprint images with realistic visual perception, but the results of PAD detection are not predictable, and fingerprint images with a specified Liveness Score cannot be generated. At the same time, this method requires a large number of samples to train the adversarial network, and still has the problem of high cost.

[0005] Therefore, how to quickly and inexpensively obtain fingerprint images with different Liveness Scores is a problem that urgently needs to be solved or is conducive to being solved in this field. Summary of the invention

[0006] Some exemplary embodiments of the present disclosure provide a fingerprint image generation method and device, which can obtain fingerprint images with different Liveness Scores quickly and at low cost.

[0007] According to some example embodiments, a fingerprint image generation method is provided, comprising: acquiring M sample fingerprint images; generating N approximate images based on each sample fingerprint image in the M sample fingerprint images according to N target similarities, wherein each target similarity in the N target similarities is a similarity value between “a corresponding approximate image in the N approximate images” and “a corresponding sample fingerprint image in the M sample fingerprint images, on which the corresponding approximate image in the N approximate images is based”, and M and N are integers greater than 1; analyzing the difference between each sample fingerprint image in the M sample fingerprint images and each approximate image in the N approximate images to obtain (M×N) difference parameters, wherein each difference parameter in the (M×N) difference parameters is The difference between each sample fingerprint image in the corresponding M sample fingerprint images and the corresponding N approximate images is quantified, and each difference parameter in the (M×N) difference parameters is associated with the similarity value between the corresponding sample fingerprint image in the M sample fingerprint images and the corresponding N approximate images, so that each target similarity in the N target similarities is associated with the M difference parameters; the M difference parameters associated with each of the P similarities in the N target similarities are respectively counted to obtain the difference statistical parameters corresponding to each of the P similarities, where P is an integer greater than or equal to 1; a reference fingerprint image is obtained; and based on the reference fingerprint image and the difference statistical parameters corresponding to each of the P similarities, a simulated fingerprint image corresponding to each of the P similarities is generated.

[0008] According to some example embodiments, the generating of N approximate images based on each sample fingerprint image in the M sample fingerprint images according to N target similarities includes: taking each sample fingerprint image in the M sample fingerprint images as a benchmark, performing iterative calculation using an evolutionary algorithm to obtain the N approximate images corresponding to the M sample fingerprint images, wherein each approximate image in the N approximate images corresponds to an iteration number; wherein each target similarity in the N target similarities is represented by the iteration number of the corresponding approximate image in the N approximate images corresponding to the associated difference parameter in the M difference parameters.

[0009] According to some example embodiments, in the evolutionary algorithm, the structural similarity between the multiple candidate images generated in each iteration and the benchmark is the fitness value of the corresponding candidate image among the multiple candidate images, and the corresponding approximate image among the N approximate images of the corresponding number of iterations is screened out from the multiple candidate images based on the fitness value.

[0010] According to some example embodiments, the evolutionary algorithm comprises a genetic algorithm.

[0011] According to some example embodiments, the analyzing the difference between each sample fingerprint image in the M sample fingerprint images and each approximate image in the N approximate images to obtain (M×N) difference parameters includes: taking each sample fingerprint image in the M sample fingerprint images and the corresponding approximate image in the N approximate images as an image pair to form (M×N) image pairs; calculating the difference in pixel values ​​of each image pair in the (M×N) image pairs at corresponding pixel points as the difference value; for each image pair in the (M×N) image pairs, analyzing the distribution law of the difference values ​​at each pixel point to obtain parameters describing the distribution law as the difference parameters of the corresponding image pair in the (M×N) image pairs to obtain the (M×N) difference parameters.

[0012] According to some example embodiments, for each of the (M×N) image pairs, analyzing the distribution law of the difference values ​​on each pixel point to obtain parameters describing the distribution law as difference parameters of the corresponding image pairs in the (M×N) image pairs, and obtaining the (M×N) difference parameters includes: for each of the (M×N) image pairs, performing frequency statistical analysis on the difference values ​​on each pixel point, fitting to obtain probability distribution parameters, and using them as difference parameters of the corresponding image pairs in the (M×N) image pairs to obtain the (M×N) difference parameters.

[0013] According to some example embodiments, the probability distribution parameters comprise normal distribution parameters.

[0014] According to some example embodiments, the fingerprint image generation method further includes: respectively counting the M difference parameters associated with each of the N target similarities to obtain difference statistical parameters of the N target similarities; acquiring multiple verification fingerprint images; for each verification fingerprint image in the multiple verification fingerprint images, generating a simulated fingerprint image corresponding to each of the N target similarities based on the difference statistical parameters corresponding to the verification fingerprint image and the N target similarities; determining the activity score of the simulated fingerprint image corresponding to each of the N target similarities; and filtering out the P similarities from the N target similarities based on the activity score and the filtering condition.

[0015] According to some example embodiments, P is greater than 1, and the P similarities are a continuous interval in the similarity interval constituted by the N target similarities; the screening condition includes: for at least part of the verification fingerprint image, the activity scores of the simulated fingerprint images corresponding to the P similarities are continuously distributed in the activity score interval.

[0016] According to some example embodiments, a fingerprint image generating device is provided, comprising: a processing circuit, the processing circuit being configured to: acquire M sample fingerprint images; based on each sample fingerprint image in the M sample fingerprint images, generate N approximate images according to N target similarities, each of the N target similarities being a similarity value between “a corresponding approximate image in the N approximate images” and “a corresponding sample fingerprint image in the M sample fingerprint images, on which the corresponding approximate image in the N approximate images is based”, and M and N are integers greater than 1; analyze the difference between each sample fingerprint image in the M sample fingerprint images and each approximate image in the N approximate images to obtain (M×N) difference parameters, each of the (M×N) difference parameters being a similarity value between “a corresponding approximate image in the N approximate images” and “a corresponding sample fingerprint image in the M sample fingerprint images, on which the corresponding approximate image in the N approximate images is based”. The method comprises the following steps: using a difference parameter to quantify the difference between each sample fingerprint image in the corresponding M sample fingerprint images and the corresponding N approximate images, and each difference parameter in the (M×N) difference parameters is associated with a similarity value between a corresponding sample fingerprint image in the M sample fingerprint images and a corresponding approximate image in the N approximate images, so that each target similarity in the N target similarities is associated with the M difference parameters; respectively counting the M difference parameters associated with each of P similarities in the N target similarities to obtain a difference statistical parameter corresponding to each of the P similarities, where P is an integer greater than or equal to 1; obtaining a reference fingerprint image; and generating a simulated fingerprint image corresponding to each of the P similarities based on the reference fingerprint image and the difference statistical parameters corresponding to each of the P similarities.

[0017] According to some example embodiments, the processing circuit is configured to: use each sample fingerprint image of the M sample fingerprint images as a reference, perform iterative calculations using an evolutionary algorithm, and obtain the N approximate images corresponding to the M sample fingerprint images, wherein each approximate image of the N approximate images corresponds to an iteration number; wherein each target similarity of the N target similarities is represented by the iteration number of the corresponding approximate image of the N approximate images corresponding to the associated difference parameter of the M difference parameters.

[0018] According to some example embodiments, in the evolutionary algorithm, the structural similarity between the multiple candidate images generated in each iteration and the benchmark is the fitness value of the corresponding candidate image among the multiple candidate images, and the corresponding approximate image among the N approximate images of the corresponding number of iterations is screened out from the multiple candidate images based on the fitness value.

[0019] According to some example embodiments, the evolutionary algorithm comprises a genetic algorithm.

[0020] According to some example embodiments, the processing circuit is configured to: respectively treat each sample fingerprint image in the M sample fingerprint images and the corresponding approximate image in the N approximate images as an image pair to form (M×N) image pairs; calculate the difference in pixel values ​​at corresponding pixel points of each image pair in the (M×N) image pairs as a difference value; for each image pair in the (M×N) image pairs, analyze the distribution law of the difference values ​​at each pixel point to obtain parameters describing the distribution law as difference parameters of the corresponding image pair in the (M×N) image pairs, and obtain the (M×N) difference parameters.

[0021] According to some example embodiments, the processing circuit is configured to: for each image pair in the (M×N) image pairs, perform frequency statistical analysis on the difference values ​​at each pixel point, and obtain probability distribution parameters by fitting, as the difference parameters of the corresponding image pair in the (M×N) image pairs, to obtain the (M×N) difference parameters.

[0022] According to some example embodiments, the probability distribution parameters comprise normal distribution parameters.

[0023] According to some example embodiments, the processing circuit is configured to: respectively count the M difference parameters associated with each of the N target similarities to obtain the difference statistical parameters of the N target similarities; obtain multiple verification fingerprint images; for each verification fingerprint image in the multiple verification fingerprint images, generate a simulated fingerprint image corresponding to each of the N target similarities based on the difference statistical parameters corresponding to the verification fingerprint image and the N target similarities; determine the activity score of the simulated fingerprint image corresponding to each of the N target similarities; and filter out the P similarities from the N target similarities based on the activity score and the filtering condition.

[0024] According to some example embodiments, P is greater than 1, and the P similarities are a continuous interval in the similarity interval constituted by the N target similarities; the screening condition includes: for at least part of the verification fingerprint image, the activity scores of the simulated fingerprint images corresponding to the P similarities are continuously distributed in the activity score interval.

[0025] According to some example embodiments, there is provided a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, is configured to implement a fingerprint image generating method according to some example embodiments.

[0026] According to some example embodiments, a computing device is provided, comprising: at least one processor; at least one memory configured to store a computer program, wherein when the computer program is executed by the at least one processor, the computing device is configured to: obtain M sample fingerprint images; based on each sample fingerprint image in the M sample fingerprint images, generate N approximate images according to N target similarities, wherein each target similarity in the N target similarities is a similarity value between “a corresponding approximate image in the N approximate images” and “a corresponding sample fingerprint image in the M sample fingerprint images, on which the corresponding approximate image in the N approximate images is based”, and M and N are integers greater than 1; analyze the difference between each sample fingerprint image in the M sample fingerprint images and each approximate image in the N approximate images, and obtain (M×N) difference images. Parameters, each of the (M×N) difference parameters quantifies the difference between each sample fingerprint image in the corresponding M sample fingerprint images and the corresponding N approximate images, and each of the (M×N) difference parameters is associated with a similarity value between a corresponding sample fingerprint image in the M sample fingerprint images and a corresponding approximate image in the N approximate images, so that each target similarity in the N target similarities is associated with the M difference parameters; respectively counting the M difference parameters associated with each of P similarities in the N target similarities to obtain a difference statistical parameter corresponding to each of the P similarities, where P is an integer greater than or equal to 1; obtaining a reference fingerprint image; and generating a simulated fingerprint image corresponding to each of the P similarities based on the reference fingerprint image and the difference statistical parameters corresponding to each of the P similarities.

[0027] According to some example embodiments, there is provided a computer program product comprising instructions, wherein the instructions, when executed by a processor of a computer device, are configured to implement a fingerprint image generation method according to some example embodiments.

[0028] According to the fingerprint image generation method and device of some example embodiments of the present disclosure, it is found through preliminary analysis that the similarity between the sample fingerprint image as a benchmark and its approximate image shows an obvious correlation with the activity score of the approximate image. Based on this, by respectively generating approximate images with the same target similarity for multiple sample fingerprint images, and then counting the differences between each approximate image and the corresponding sample fingerprint image, a difference statistical parameter is obtained, which can be used to represent the difference between any benchmark image and its approximate image under the target similarity. The target similarity and the activity score of the approximate image show an obvious correlation, which makes the difference statistical parameter represent the difference between any benchmark image and its approximate image under the relevant activity score. On this basis, according to some example embodiments, for the reference fingerprint image, combined with the difference statistical parameter corresponding to the specified similarity, a new fingerprint image is generated, which is recorded as a simulated fingerprint image, and the simulated fingerprint image can have an activity score related to the specified similarity. Since the method according to some example embodiments of the present disclosure can obtain difference statistical parameters corresponding to different target similarities (or different activity scores) with the help of a small number of sample fingerprint images, and can directly reuse these difference statistical parameters to quickly generate simulated fingerprint images with the required activity scores based on any reference fingerprint image, it is possible to quickly and inexpensively obtain simulated fingerprint images with different activity scores.

[0029] Additional aspects and / or advantages of the present disclosure will be set forth in part in the description which follows and in part will be apparent from the description, or may be learned through practice of the general inventive concept of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The above and other objects and features of the exemplary embodiments of the present disclosure will become more apparent through the following description in conjunction with the accompanying drawings which exemplarily illustrate the embodiments, in which:

[0031] Figure 1 A flowchart illustrating a fingerprint image generation method according to some example embodiments;

[0032] Figure 2 A graph showing the variation trend of similarity and activity score with iterations in the approximate image generation process according to some example embodiments;

[0033] Figure 3 A schematic diagram illustrating a principle of difference data according to some example embodiments;

[0034] Figure 4 A structural diagram showing a fingerprint generation method according to some example embodiments;

[0035] Figure 5 A schematic diagram showing statistical distribution of difference data according to some example embodiments;

[0036] Figure 6 A schematic diagram showing activity score distribution of a simulated fingerprint image generated by a Gauss sampler according to some example embodiments;

[0037] Figure 7 A schematic flow chart showing a fingerprint generation method according to some example embodiments;

[0038] Figure 8 A comparison diagram showing verification results at multiple iterations of a fingerprint generation method according to some example embodiments;

[0039] Fig. 9 A schematic diagram showing distribution of matching scores of a simulated fingerprint image and a reference fingerprint image according to some example embodiments;

[0040] Fig.10 A diagram showing verification results of a fingerprint generation method according to some example embodiments on an ALGINATE material data set;

[0041] Fig.11 A block diagram showing a fingerprint image generating apparatus according to some example embodiments;

[0042] Fig.12 A schematic diagram illustrating a computing device according to some example embodiments. DETAILED DESCRIPTION

[0043] Reference will now be made in detail to some example embodiments of the present disclosure, examples of which are shown in the accompanying drawings, wherein like reference numerals refer to like parts throughout. The example embodiments will be described below with reference to the accompanying drawings in order to explain the present disclosure.

[0044] Figure 1 A flowchart illustrating a fingerprint image generating method according to some example embodiments is shown.

[0045] The fingerprint image generation method comprises at least two stages. Figure 1 Operations S101 to S104 are the preliminary data preparation stage, which are used to prepare the difference statistical parameters corresponding to different specified similarities or optional expected similarities. Operations S105 and S106 are the formal fingerprint generation stage, which are used to generate a simulated fingerprint image based on the reference fingerprint image and the previously prepared difference statistical parameters.

[0046] The data preparation phase according to some example embodiments may include:

[0047] In operation S101, M sample fingerprint images are acquired.

[0048] The M sample fingerprint images obtained in operation S101 may be used as reference data for preliminary data preparation operations according to some example embodiments. M may be an integer greater than 1, that is, it is not appropriate to use only one sample fingerprint image, so as to improve the universality of the difference statistical parameters obtained subsequently according to some example embodiments of the present disclosure.

[0049] In operation S102, N approximate images are generated based on each of the M sample fingerprint images according to N target similarities.

[0050] According to some example embodiments, the target similarity may be the similarity between the approximate image and the sample fingerprint image on which it is based, for example, SSIM (Structure SIMilarity), PSNR (Peak Signal-to-Noise Ratio), and MSE (Mean Squared Error) may be used. For example, in some example embodiments, the target similarity value between the approximate image and the sample fingerprint image on which the approximate image is based may be defined using structural similarity (SSIM), peak signal-to-noise ratio (PSNR), and mean square error (MSE), but example embodiments are not limited thereto. It should be understood that the "target" here refers to the similarity value being the target value, and does not mean the similarity using the target form. Similarly, the specified similarity in the following text refers to the similarity value being the specified value or the optional expected value.

[0051] After preliminary analysis, it is found that the similarity and the activity score of the corresponding approximate image show an obvious correlation, so different or changing similarities can be used to correspond to activity scores of different sizes. According to some example embodiments, N can also be an integer greater than 1, so that multiple similarities can be obtained to correspond to multiple activity scores.

[0052] It should be noted that, after preliminary analysis, it was found that when the same similarity corresponds to an activity score, it does not correspond to an exact activity score value, but to an activity score range, that is, there is a certain error range, but the activity score range is sufficient to meet the needs of practical applications. Specifically, in practical applications, taking the case where the activity score range is [0,100] as an example, an activity score above 80 is considered or can be considered a true fingerprint.

[0053] In operation S103, the difference between each sample fingerprint image and each correspondingly generated approximate image is analyzed to obtain (M×N) difference parameters.

[0054] According to some example embodiments, each difference parameter is used or can be used to quantify the difference between a sample fingerprint image and an approximate image, so that a total of (M×N) difference parameters can be obtained (M is the number of sample fingerprint images, and N is the number of approximate images generated corresponding to each sample fingerprint image). The difference parameter is associated with the similarity between the represented sample fingerprint image and the approximate image, such that each of the N target similarities, in turn, is associated with M difference parameters (corresponding to M sample fingerprint images).

[0055] In operation S104, the M difference parameters respectively associated with P specified similarities among the N target similarities are statistically counted to obtain the difference statistical parameters corresponding to each of the P specified similarities.

[0056] According to some example embodiments, the specified similarity is a similarity determined from the N target similarities and is used to perform subsequent formal fingerprint generation. P is an integer greater than or equal to 1 and less than or equal to N, so that 1 to N different simulated fingerprint images can be generated based on the same reference fingerprint image.

[0057] It should be understood that this operation emphasizes the fingerprint generation requirements for the subsequent operation S106, and it is necessary to obtain or facilitate obtaining the difference statistical parameters corresponding to each of the P specified similarities, but it does not necessarily mean that only the difference statistical parameters of these P specified similarities can be obtained. For example, when the number P of specified similarities is less than the number N of target similarities (P < N), in this operation S104, the difference statistical parameters corresponding to other target similarities among the N target similarities except the P specified similarities required for use in operation S106 can also be obtained.

[0058] The fingerprint generation stage according to some example embodiments may include:

[0059] In operation S105, a reference fingerprint image is obtained.

[0060] According to some example embodiments, the reference fingerprint image serves as a benchmark for generating a new fingerprint image, and the number thereof may be at least one, but the example embodiments are not limited thereto. For example, the number of reference fingerprint images can be increased as needed or optionally as desired, without specific limitation.

[0061] In operation S106, according to the reference fingerprint image and the difference statistical parameters corresponding to each of the P specified similarities, P simulated fingerprint images corresponding to each of the P specified similarities are generated.

[0062] Referring back to operations S103 and S104, it can be seen that the difference statistical parameter is a statistic of the difference parameter and can be regarded as a representative of the difference parameter. In some example embodiments, according to the path of obtaining the difference parameter from the difference in operation S103, the difference corresponding to the difference statistical parameter can be obtained by reverse calculation, and then the difference can be superimposed on the reference fingerprint image to obtain the simulated fingerprint image.

[0063] According to the fingerprint image generation method of some example embodiments, it is possible to obtain difference statistical parameters corresponding to different target similarities (or different activity scores) with the help of a small number of sample fingerprint images in the early data preparation stage, and these difference statistical parameters can be directly reused in the formal fingerprint generation stage to quickly generate a simulated fingerprint image with a required activity score or optionally a desired activity score based on any reference fingerprint image. Therefore, the fingerprint image generation method can achieve rapid and low-cost acquisition of simulated fingerprint images with different activity scores.

[0064] It should be noted that, in some example embodiments, both the sample fingerprint image and the reference fingerprint image are fake fingerprint images, so that multiple fake fingerprints with different activity scores can be generated based on an existing fake fingerprint.

[0065] Next, the fingerprint image generation method according to some example embodiments of the present disclosure is further introduced.

[0066] In some example embodiments, regarding the generation of approximate images, in an exemplary embodiment of the present disclosure, operation S102 may include: using each sample fingerprint image in the M sample fingerprint images as a reference, performing iterative calculations using an evolutionary algorithm, and obtaining N approximate images corresponding to the sample fingerprint image, wherein each approximate image in the N approximate images corresponds to an iteration number; wherein the target similarity is represented by the iteration number of the approximate image corresponding to the associated difference parameter. In some example embodiments, by performing iterative calculations using an evolutionary algorithm, each iteration uses the iteration result as an approximate image, and the properties of the evolutionary algorithm itself can be used so that as the iteration proceeds, the approximate image gradually approaches the sample fingerprint image as a reference, for example, the similarity gradually increases, thereby achieving convenient and reliable generation of approximate images. At this time, each iterative calculation can be regarded as corresponding to a target similarity, so that the target similarity is represented by the iteration number of the approximate image, so that a simple and easy-to-implement unified standard is provided when approximate images are generated based on different sample fingerprint images.

[0067] According to some example embodiments, in the evolutionary algorithm, the structural similarity between the multiple candidate images generated in each iteration and the sample fingerprint image as a benchmark is used as the fitness of the corresponding candidate image, and the approximate image of the corresponding number of iterations is screened out from the multiple candidate images according to the fitness. In some example embodiments, the fitness may represent a fitness metric or a fitness value. According to some example embodiments, using structural similarity as the similarity has been proven to improve the quality of the simulated fingerprint image finally generated. According to some example embodiments, the evolutionary algorithm can be combined with the similarity, and the similarity is used as the screening criterion in the evolutionary algorithm, which fully ensures that the similarity gradually increases as the iteration proceeds.

[0068] As an example, selecting an approximate image of a corresponding number of iterations from multiple candidate images based on fitness includes: selecting a candidate image with the largest fitness from multiple candidate images as the approximate image of the corresponding number of iterations; or determining a target similarity corresponding to a current number of iterations based on a pre-configured target similarity corresponding to each number of iterations, and selecting a candidate image with a fitness closest to the target similarity from multiple candidate images as the approximate image of the corresponding number of iterations.

[0069] In some example embodiments of the present disclosure, the evolutionary algorithm includes a genetic algorithm (GA), which has been verified to be able to improve the quality of the simulated fingerprint image finally generated. In some example embodiments, other evolutionary algorithms may also be used, such as a particle swarm optimization algorithm (PSO), an ant colony optimization algorithm (ACO), a differential evolution algorithm (DE), etc., but the example embodiments of the present disclosure are not limited thereto.

[0070] It should be understood that, according to some example embodiments, in addition to using an evolutionary algorithm for iterative calculation, other methods may also be used to generate an approximate image, and the present disclosure is not limited thereto.

[0071] Figure 2 A graph showing the variation trend of similarity and activity scores with iterations in the approximate image generation process according to some example embodiments of the present disclosure is obtained by calculating and analyzing multiple sample fingerprint graphs. Figure 2, the horizontal axis represents the number of iterations (e.g., 1 to 5937), the vertical axis on the left represents the activity score (e.g., an activity score between 1 and 100), and the vertical axis on the right represents the structural similarity (as a similarity). According to some example embodiments, as the iterations proceed, the similarity gradually increases, and each iteration corresponds to a smaller activity score interval. In some example embodiments, the activity score as a whole first fluctuates greatly, then shows a trend of gradually increasing, and then decreases. For example, Figure 2 As shown, the changing trends of the activity scores and the similarities are not always consistent. Based on the changing trends of the activity scores and the similarities, a portion of the entire target similarities may be selected as a specified similarity or, optionally, a desired similarity.

[0072] Regarding the determination of the specified similarity, in some example embodiments, the fingerprint image generation method according to some example embodiments of the present disclosure may further include: respectively counting the M difference parameters associated with each of the N target similarities to obtain the difference statistical parameters of the N target similarities; obtaining multiple verification fingerprint images; for each verification fingerprint image in the multiple verification fingerprint images, generating a simulated fingerprint image corresponding to each of the N target similarities according to the difference statistical parameters corresponding to the verification fingerprint image and the N target similarities; determining the activity score of the simulated fingerprint image corresponding to each of the N target similarities; and filtering out P specified similarities from the N target similarities according to the activity score and preset filtering conditions. According to some example embodiments, by using multiple verification fingerprint images to perform trial generation of simulated fingerprint images, for example, based on the multiple verification fingerprint images and the difference statistical parameters prepared in addition, according to the fingerprint generation stage Figure 1 The operation S106 shown in the figure generates a corresponding plurality of simulated fingerprint images, and then analyzes the activity scores of the generated simulated fingerprint images to understand whether these activity scores meet the needs of actual applications, and then selects P specified similarities that meet the needs (for example, meet the preset filtering conditions) in combination with the preset filtering conditions, which helps to ensure that the simulated fingerprint images generated in the formal fingerprint generation stage meet the needs of actual applications.

[0073] It should be understood that in order to ensure that the needs of actual applications are met, a larger range of value intervals can be determined for the target similarity first. According to some example embodiments, if the simulated fingerprint image is generated directly based on all target similarities, it can theoretically meet the needs of actual applications, but there are disadvantages of large amount of calculation and many invalid calculations. By combining the preset screening conditions according to some example embodiments, P specified similarities are screened out from N target similarities, which can further reduce the amount of calculation and improve the efficiency of generating simulated fingerprint images. In some example embodiments, if a large number of simulated fingerprint images need to be generated, or if it is beneficial to generate a large number of simulated fingerprint images, difference statistical parameters corresponding to other target similarities other than the specified similarity can also be used, and the present disclosure does not limit this.

[0074] In some example embodiments of the present disclosure, P is greater than 1, and the P specified similarities are a continuous interval in the similarity interval composed of the N target similarities, so that a large number of simulated fingerprint images can be generated. According to some example embodiments, the preset screening conditions may include: for at least part of the verification fingerprint image, the activity scores of the simulated fingerprint images corresponding to the P specified similarities are continuously distributed and cover the specified activity score interval or the optionally expected activity score interval. In some example embodiments, for the same verification fingerprint image, by requiring the activity scores of the simulated fingerprint image to be continuously distributed and cover the specified activity score interval, it is helpful to ensure that the activity scores of the subsequently formally generated simulated fingerprint images continuously cover the specified activity score interval, thereby meeting the needs of practical applications. For example, in some example embodiments, for the same verification fingerprint image, when the activity scores of the simulated fingerprint image are continuously distributed and cover the specified activity score interval or the optionally expected activity score interval, it can be ensured that the activity scores of the subsequently formally generated simulated images continuously cover the specified activity score interval or the optionally expected activity score interval, so that in some example embodiments, the needs of practical applications can be met. For example, the specified activity score interval can be the entire value interval of the activity score to achieve full coverage. Depending on the value standard of the activity score, the specified activity score interval is, for example, [0, 1] or [0, 100]. Therefore, in some example embodiments, by requiring or screening at least part of the verification fingerprint images to meet the aforementioned requirements, rather than all verification fingerprint images meeting the aforementioned requirements, a certain or optionally expected error can be allowed to exist to improve the applicability of the present disclosure. For example, a preset proportion of verification fingerprint images that meet the aforementioned requirements can be configured. As long as the proportion of verification fingerprint images that meet the aforementioned requirements reaches the preset proportion, it can be considered that the preset screening condition is met.

[0075] In some example embodiments, regarding the analysis of difference parameters, in some example embodiments of the present disclosure, operation S103 includes: treating (e.g., receiving, acquiring, etc.) a sample fingerprint image and a correspondingly generated approximate image as an image pair to form (M×N) image pairs; each image has a large number of pixels, calculating the difference in pixel values ​​of each image pair at corresponding pixels as a difference value, and obtaining a set of difference values ​​at each pixel in each image pair; for each image pair, analyzing the distribution law of the difference values ​​at each pixel, obtaining parameters describing the distribution law as the difference parameters of the image pair, and obtaining (M×N) difference parameters. Therefore, in some example embodiments, by using the parameters describing the distribution law of the difference values ​​as the difference parameters, a large number of difference values ​​in each image pair can be concisely described.

[0076] In some example embodiments of the present disclosure, in operation S103, for each image pair, analyzing the distribution law of the difference values ​​at each pixel point, obtaining parameters describing the distribution law, and using them as the difference parameters of the image pair, the operation of obtaining (M×N) difference parameters may include: for each image pair, performing frequency statistical analysis on the difference values ​​at each pixel point, fitting to obtain probability distribution parameters, and using them as the difference parameters of the image pair, and obtaining (M×N) difference parameters. In some example embodiments, it has been verified that when performing frequency statistical analysis, a relatively stable distribution law can be obtained, and as the target similarity increases, the distribution of the difference values ​​presents a trend of gradually concentrating. Such a general law is more conducive to generalization to other images, so that the activity score of the simulated fingerprint image generated based on any reference fingerprint image is more likely to meet expectations, thereby improving the quality of the generated simulated fingerprint image.

[0077] In an example embodiment of the present disclosure, the probability distribution parameters may include normal distribution parameters, such as expected μ and standard deviation σ. By fitting the difference values ​​at each pixel point in the image pair with a normal distribution, a difference parameter with high representativeness can be obtained, which helps to improve the quality of the generated simulated fingerprint image.

[0078] Next, a fingerprint image generating method according to some example embodiments of the present disclosure is introduced through an example embodiment.

[0079] In this example embodiment, the fingerprint generation method can generate a corresponding simulated fingerprint image according to a specified activity score or optionally a desired activity score, which is used to solve the problem that the cost of collecting fingerprint images with PAI materials is high and fingerprint images cannot cover the entire activity score value range. Figure 3 As shown in Figure 2, this method is based on analyzing the Gap Data (difference data, reference) between the approximate image (generated by the traditional iterative method) and the sample fingerprint image as a benchmark. Figure 3), and model GapData to generate new simulated fingerprint images. This method requires less data in the data preparation stage, and the speed of generating simulated fingerprint images in the later stage is fast. The generated simulated fingerprint images can cover the entire domain of activity scores (for example, [0,1], [0,100]), and this method has universal applicability.

[0080] Figure 4 A structural diagram showing a fingerprint generation method according to some example embodiments.

[0081] Reference Figure 4 , the whole method can be composed of the following three (but not limited to three) main modules.

[0082] 1. Genetic Algorithm (GA) Iterator. The Genetic Algorithm (GA) Iterator can be used to construct a set of approximate images with SSIM (Structure SIMilarity) as the iteration index. For example, through iteration, an approximate image can be generated whose similarity with the sample fingerprint image used as a benchmark gradually increases under the SSIM index. The larger the SSIM, the higher the similarity between the generated approximate image and the sample fingerprint image used as a benchmark, and thus the more fingerprint detail information is retained. According to some example embodiments, the activity score of the generated approximate image will also change in a certain pattern with the change of the SSIM value, for example, Figure 2 shown.

[0083] 2. Gap Data (GD) Analyzer. The GD (Gap Data, for example, difference data) analyzer can be used to analyze the difference in pixel values ​​at corresponding pixel points between the approximate image generated by each iteration and the sample fingerprint image used as the benchmark. For example, by performing a frequency statistical analysis of pixel values, it can be confirmed whether these GDs approximately obey the Gaussian distribution, and further fit the corresponding Gaussian distribution parameters μ and σ (as difference parameters), such as Figure 5 As shown. Each approximate image and its corresponding sample fingerprint image can fit a Gaussian distribution, thereby obtaining a set of Gaussian distribution parameters. Each iteration can obtain multiple sets of Gaussian distribution parameters corresponding to multiple sample fingerprint images. In other words, in some example embodiments, for each iteration, the number of sets of Gaussian distribution parameters obtained is equal to the number of sample fingerprint images used. In some example embodiments, GD analysis can confirm the degree of influence of different GDs on the activity score of the approximate image.

[0084] 3. Gauss sampler. The Gauss sampler can be used with a specified number of iterations or optionally with a desired iteration interval. start ,Iteration end The mean of the Gaussian distribution parameters μ and σ with the specified number of iterations is used as the new Gaussian sampling parameter and Sampling can be performed based on a reference fingerprint image to generate a simulated fingerprint image.

[0085] According to some example embodiments, the GA iterator, the GD analyzer, and the Gauss sampler may include one or more instances of a processing circuit, may be included in one or more instances of a processing circuit, and / or may be implemented by one or more instances of a processing circuit, such as hardware including a logic circuit, a hardware / software combination such as a processor that executes software, or a combination thereof. For example, the processing circuit may more specifically include, but is not limited to, a central processing unit (CPU), an arithmetic logic unit (ALU), a graphics processing unit (GPU), an application processor (AP), a digital signal processor (DSP), a microcomputer, a field programmable gate array (FPGA) and a programmable logic unit, a microprocessor, an application specific integrated circuit (ASIC), a neural network processing unit (NPU), an electronic control unit (ECU), an image signal processor (ISP), etc. In some example embodiments, the processing circuit may include a non-transitory computer-readable storage device (e.g., a memory) storing an instruction program, such as a solid-state drive (SSD), and a processor (e.g., a CPU) configured to execute the instruction program to implement the functions and / or methods performed by the GA iterator, the GD analyzer, and the Gauss sampler.

[0086] According to some example embodiments, since GD approximately obeys Gaussian distribution, the simulated data GD of GD can be generated by Gauss sampler. i (the subscript i denotes the number of iterations), for example, the mean of the Gaussian distribution parameter of the specified number of iterations or, alternatively, the desired number of iterations as described above and According to the GD analysis results, different GDs superimposed on the fingerprint image as a reference will obtain new images with different activity scores. Therefore, in some example embodiments, the GD generated by the Gauss sampler simulation is superimposed on the reference fingerprint image as a reference. i A new simulated fingerprint image with a changed activity score can be generated. For example, based on the simulated data GD i A difference data sequence is generated, and then each difference data in the difference data sequence is randomly assigned to different pixel points in the reference fingerprint image, and the obtained new image is the simulated fingerprint image.

[0087] According to some example embodiments, a specified iteration number interval or an optionally desired iteration number interval may be determined as follows: the GD simulation data GD corresponding to a larger iteration number interval is iIt is applied to multiple verification fingerprint images to generate a set of simulated fingerprint images for each of these verification fingerprint images, and then the activity score variation pattern of each set of simulated fingerprint images is analyzed to find out the iteration number interval in which the activity score gradually increases and covers the entire activity score domain, as the specified iteration number interval or the optionally expected iteration number interval. In this way, in some example embodiments, the activity scores of the generated simulated fingerprint images are covered within the range of 0 to 1 or 0 to 100. When it comes to the formal fingerprint generation stage, the Gauss sampler can make the activity score distribution form of the simulated fingerprint image generated by the simulated sampling consistent with the morphological trend of the GA iteration within the specified iteration number interval or the optionally expected iteration number interval, based on the GD corresponding to the GA iteration data, such as Figure 6 It should be noted that, compared Figure 2 and Figure 6 It can be found that the number of iterations of the two is not exactly the same, because Figure 6 There is hysteresis, and different fingerprint images basically meet the requirements when simulating new images. Figure 6 Therefore, according to some example embodiments, the simulation data GD of GD i There are general rules that can be used to generate simulated fingerprint images based on other existing fingerprint images (eg, reference fingerprint images).

[0088] In some example embodiments, since a set of (μ, σ) parameters can represent a Gaussian distribution, a functional relationship of f(μ, σ) = Liveness Score can be further established to more accurately associate the difference parameter with the activity score, and then analyze the general regularity of the functional relationship on different images.

[0089] Figure 7 A schematic flow chart showing a fingerprint generation method according to some example embodiments.

[0090] Reference Figure 7 According to some example embodiments, the fingerprint generation method may include a data preparation stage on the left (e.g., data preparation operations S701 to S709) and a fingerprint generation stage on the right (e.g., fingerprint generation operations S710 to S714), and in some example embodiments, the step of determining a specified iteration number interval or optionally a desired iteration number interval may be omitted. According to some example embodiments, the method may directly generate a corresponding simulated fingerprint image for all iterations of iterative calculations performed in the data preparation stage using the corresponding Gaussian distribution parameter mean.

[0091] In the data conversion stage, in operation S701, the number m of sample fingerprint images that require or are conducive to GA iteration and the number k of GA iterations are first set.

[0092] Next, in operation S702, a sample fingerprint image Image to be iterated by GA is input. i (The superscript i represents the serial number of the sample fingerprint image). For this image, in operation S703, an approximate image is generated using SSIM as the GA iteration index. (The subscript iter indicates the number of iterations). In operation S704, the difference data is calculated. In operation S705, statistics For example, in operation S705, the method counts the frequency of each value in GD. In operation S706, fitting The corresponding Gaussian distribution parameters and Accordingly, the above iterative calculation to the operation S703 to S706 of fitting the Gaussian distribution parameters is repeated a total of k times to implement all iterative calculations for a sample fingerprint image. For example, in operation S707, the method determines whether operations S703 to S706 have been repeated k times to implement all iterative calculations for a sample fingerprint image. If the method determines in operation S707 that operations S703 to S706 have not been repeated k times ("No"), the method returns to operations S703 to S706. If the method determines in operation S707 that operations S703 to S706 have been repeated k times ("Yes"), the method continues to operation S708.

[0093] Thereafter, in operation S708, in order to obtain the mean value of the Gaussian distribution parameters corresponding to each iteration, the mean value of the Gaussian distribution parameters may be calculated once after calculating each sample fingerprint image (refer to Figure 7 ), or you can wait until all m sample fingerprint images are iteratively calculated and then calculate the mean of the Gaussian distribution parameters corresponding to each iteration, and the result is the same. The specific calculation formula is:

[0094]

[0095] Output Gaussian distribution parameter list

[0096] In operation S709, the method according to some example embodiments determines whether a new sample fingerprint image is input. For example, in operation S709, when no new sample fingerprint image is input, the method enters the fingerprint generation phase (e.g., fingerprint generation operations S710 to S714). In the fingerprint generation phase, in operation S710, first input a reference fingerprint image Image to be generated for the simulated fingerprint image. j Then, in operation S711, and Generate a sample for the Gaussian distribution parameters Where iter = (1, ..., k), in operation S712, a simulated fingerprint image is generated For example, in operation S711, the method performs Gaussian sampling by taking the average value of the Gaussian distribution parameter as a parameter and generates a GD. In operation S712, the method superimposes the generated GD on the reference fingerprint image to generate a simulated fingerprint image. Repeated generation and The operations (operations S711 and S712) are performed for a total of k times, and in operation S714, a list of simulated fingerprint images generated based on the reference fingerprint image is output. For example, in operation S713, the method determines whether operations S711 and S712 are repeated k times, if "no", operations S711 and S712 continue to be repeated (e.g., iterated), if "yes", in operation S714, the method outputs a list of generated simulated fingerprint images.

[0097] Reference Figure 8 , use the fingerprint algorithm applied on the bank card to verify the fingerprint image generation method of the above example embodiment. Figure 8 As shown in the figure, the experiment verifies the effect of activity score quantization and generalization on the data set of ALGINATE (sodium alginate) material. The image size of this data set is 114*114, and the PAD recognition model used in the experiment is a CNN (convolutional neural network) model, specifically the MobilNetV1 model. The number of sample fingerprint images of GA iterations is m=10, and the number of GA iterations is k=10000.

[0098] To evaluate the effectiveness of the proposed method, the generated simulated fingerprint images were visually inspected. Figure 8 The comparison of the results of the reference fingerprint image made of ALGINATE material and the generated simulated fingerprint image at different iterations (e.g., 1000, 3000, 5000, 6000) is shown. The generated simulated fingerprint image has a relatively constant ridge thickness and appropriate contrast while maintaining the shape of the original image fingerprint. The results show that the generated simulated fingerprint image is judged as a different activity score result by the PAD recognition model while maintaining its original shape. Compared with the original reference fingerprint image, the Match Score is greater than the similarity threshold of 170000.

[0099] It should be noted here that the conventional process of fingerprint recognition is to first determine whether the fingerprint is true or false, and then perform fingerprint matching after confirming that it is true to see whether it is your own fingerprint. Therefore, in conventional practice, the fingerprint that is forged is often your own fingerprint. If the matching score of the simulated fingerprint image generated by the method according to some example embodiments of the present disclosure is lower than the similarity threshold, it means that the simulated fingerprint image and the fingerprint of the person do not match, and there is no point in forging. The simulated fingerprint image generated in this way is also meaningless, so the matching score needs to be large enough.

[0100] In some example embodiments, in order to evaluate the actual effect of the method, a quantitative evaluation is also performed on the generated simulated fingerprint images. The activity scores of the simulated fingerprint images synthesized by GD generated by Gaussian sampling are distributed in the entire activity score value range of 0 to 100, such as Figure 6 shown.

[0101] According to some example embodiments, in the statistics of the matching score comparison results of the generated simulated fingerprint image and the reference fingerprint image as a benchmark, 87.86% (4393 / 5000) of the images are greater than the similarity threshold (170000), such as Fig. 9 shown.

[0102] The method according to some example embodiments randomly selected 1000 images from the ALGINATE material dataset, verifying the universal applicability of the fingerprint image generation method according to some example embodiments. Fig.10 The distribution of activity scores of simulated fingerprint images generated by this method based on multiple fake fingerprint images made of untrained materials is shown. It should be understood that there are many kinds of PAI materials. For materials used in GA iterations (e.g., trained materials), using fake fingerprints made of the materials as reference fingerprint images, the generated simulated fingerprint images often meet the requirements. Figure 6 The law of obtaining a batch of simulated fingerprint images that are continuously distributed and cover the entire activity score value domain is obtained. However, in some example embodiments, for untrained materials (such as the ALGINATE material here), this law may not be met, for example Fig.10 Based on this, we can count all the reference fingerprint images involved in fingerprint generation that meet Figure 6 The proportion of regular reference fingerprint images is used as the effectiveness of the method. After calculation, the effectiveness of the method in this verification is 95.2% (952 / 1000).

[0103] The above has been combined Figures 1 to 10 A fingerprint image generation method according to some example embodiments of the present disclosure is described. Fig.11 A fingerprint image generating apparatus and units (eg, components) thereof according to some example embodiments of the present disclosure are described.

[0104] Fig.11 A block diagram of a fingerprint image generating device according to some example embodiments of the present disclosure is shown.

[0105] Reference Fig.11 The fingerprint image generating device may include a sample acquiring unit 111 , an approximate generating unit 112 , a difference analyzing unit 113 , a difference statistics unit 114 , a reference acquiring unit 115 and a simulation generating unit 116 .

[0106] According to some example embodiments, the sample acquisition unit 111, the approximate generation unit 112, the difference analysis unit 113, the difference statistics unit 114, the reference acquisition unit 115, and the simulation generation unit 116 of the fingerprint generation device may include one or more instances of a processing circuit, may be included in one or more instances of a processing circuit, and / or may be implemented by one or more instances of a processing circuit, such as hardware including a logic circuit, a hardware / software combination such as a processor that executes software, or a combination thereof. In some example embodiments, the processing circuit may also include a non-transitory computer-readable storage device (e.g., a memory) storing an instruction program, such as a solid-state drive (SSD), and a processor (e.g., a CPU) configured to execute the instruction program to implement the functions and / or methods performed by some or all of any device, electronic device, module, unit, and / or part thereof according to any example embodiment.

[0107] The sample acquisition unit 111 is configured to acquire M sample fingerprint images.

[0108] The approximate generation unit 112 is configured to generate N approximate images based on each sample fingerprint image in the M sample fingerprint images according to N target similarities, wherein the target similarity is the similarity between the approximate image and the sample fingerprint image on which it is based.

[0109] In some example embodiments, the approximate generation unit 112 is configured to: use each sample fingerprint image in the M sample fingerprint images as a reference, perform iterative calculations using an evolutionary algorithm, and obtain N approximate images corresponding to the sample fingerprint image, wherein each approximate image in the N approximate images corresponds to an iteration number; wherein the target similarity is represented by the iteration number of the approximate image corresponding to the associated difference parameter.

[0110] In some example embodiments, in the evolutionary algorithm, the structural similarity between multiple candidate images generated in each iteration and the sample fingerprint image used as a benchmark is used as the fitness of the corresponding candidate image, and the approximate image of the corresponding iteration number is screened out from the multiple candidate images according to the fitness.

[0111] In some example embodiments, the evolutionary algorithm comprises a genetic algorithm.

[0112] The difference analysis unit 113 is configured to analyze the difference between each sample fingerprint image and each correspondingly generated approximate image to obtain (M×N) difference parameters, wherein each difference parameter is used to quantitatively represent the difference between a sample fingerprint image and an approximate image, and the difference parameter is associated with the similarity between the represented sample fingerprint image and the approximate image, so that each target similarity is associated with M difference parameters.

[0113] In some example embodiments, the difference analysis unit 113 is configured to: take (for example, acquire, receive, etc.) a sample fingerprint image and a correspondingly generated approximate image as an image pair, to form (M×N) image pairs; calculate the difference in pixel values ​​at corresponding pixel points of each image pair as a difference value; for each image pair, analyze the distribution law of the difference values ​​at each pixel point, obtain parameters describing the distribution law, as difference parameters of the image pair, to obtain (M×N) difference parameters.

[0114] In some example embodiments, the difference analysis unit 113 is configured to: for each image pair, perform frequency statistical analysis on the difference values ​​at each pixel point, obtain probability distribution parameters by fitting, and use them as difference parameters of the image pair, thereby obtaining (M×N) difference parameters.

[0115] In some example embodiments, the probability distribution parameters include normal distribution parameters.

[0116] The difference statistics unit 114 is configured to respectively count the M difference parameters associated with each of the P designated similarities among the N target similarities, and obtain the difference statistics parameters corresponding to each of the P designated similarities.

[0117] In some example embodiments, the difference statistics unit 114 is configured to: respectively count the M difference parameters associated with each of the N target similarities to obtain the difference statistical parameters of the N target similarities; the fingerprint image generation device also includes a verification acquisition unit, a verification generation unit, a score determination unit and a verification screening unit (all not shown in the figure), the verification acquisition unit is configured to obtain multiple verification fingerprint images; the verification generation unit is configured to generate, for each verification fingerprint image in the multiple verification fingerprint images, a simulated fingerprint image corresponding to each of the N target similarities according to the difference statistical parameters corresponding to the verification fingerprint image and the N target similarities; the score determination unit is configured to determine the activity score of the simulated fingerprint image corresponding to each of the N target similarities; the verification screening unit is configured to screen out P specified similarities from the N target similarities according to the activity score and preset screening conditions.

[0118] In some example embodiments, P is greater than 1, and the P specified similarities are a continuous interval in the similarity interval formed by the N target similarities; the preset screening conditions include: for at least part of the verified fingerprint image, the activity scores of the simulated fingerprint images corresponding to the P specified similarities are continuously distributed and cover the specified activity score interval.

[0119] The reference acquiring unit 115 is configured to acquire a reference fingerprint image.

[0120] The simulation generation unit 116 is configured to generate simulated fingerprint images corresponding to P designated similarities, respectively, based on the reference fingerprint image and the difference statistical parameters corresponding to the P designated similarities, respectively.

[0121] M and N are both integers greater than 1, and P is an integer greater than or equal to 1.

[0122] In addition, according to some example embodiments of the present disclosure, there is also provided a computer-readable storage medium (eg, a memory) on which a computer program is stored. When the computer program is executed, the fingerprint image generation method according to some example embodiments of the present disclosure is implemented.

[0123] In some example embodiments of the present disclosure, the computer-readable storage medium (e.g., memory) may carry (e.g., may be configured to store) one or more programs, and when the computer program is executed, the following steps may be implemented: obtaining M sample fingerprint images; based on each sample fingerprint image in the M sample fingerprint images, generating N approximate images according to N target similarities, wherein the target similarity is the similarity between the approximate image and the sample fingerprint image on which it is based; analyzing the difference between each sample fingerprint image and each correspondingly generated approximate image to obtain (M×N) difference parameters, wherein each difference parameter is used to quantify a The method comprises the following steps: calculating the difference between a sample fingerprint image and an approximate image, wherein the difference parameter is associated with the similarity between the represented sample fingerprint image and the approximate image, so that each target similarity is associated with M difference parameters; respectively counting the M difference parameters associated with each of the P specified similarities among the N target similarities, and obtaining the difference statistical parameters corresponding to each of the P specified similarities; obtaining a reference fingerprint image; and generating simulated fingerprint images corresponding to each of the P specified similarities according to the reference fingerprint image and the difference statistical parameters corresponding to each of the P specified similarities, wherein M and N are both integers greater than 1, and P is an integer greater than or equal to 1.

[0124] According to some example embodiments, the computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, 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 disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In some example embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a computer program that may be used by or in conjunction with an instruction execution system, device or device. The computer program contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination thereof. The computer-readable storage medium may be contained in any device; it may also exist alone without being assembled into the device.

[0125] In addition, according to some example embodiments of the present disclosure, a computer program product is also provided, and instructions in the computer program product can be executed by a processor of a computer device to complete the fingerprint image generation method according to some example embodiments of the present disclosure.

[0126] The above has been combined Fig.11 The fingerprint image generating apparatus according to some example embodiments of the present disclosure is described. Fig.12 A computing device according to some example embodiments of the present disclosure is described.

[0127] Fig.12 A schematic diagram illustrating a computing device according to some example embodiments.

[0128] Reference Fig.12 According to some example embodiments, the computing device 12 may include a memory 121 and a processor 122, wherein the memory 121 may store or may be configured to store a computer program, and when the computer program is executed by the processor 122, a fingerprint image generation method according to some example embodiments of the present disclosure is implemented.

[0129] In some example embodiments, the memory 121 may be a nonvolatile memory such as a flash memory, a phase-change random access memory (PRAM), a magnetoresistive RAM (MRAM), a resistive RAM (ReRAM), or a ferroelectric RAM (FRAM), or may be a volatile memory such as a static RAM (SRAM), a dynamic RAM (DRAM), or a synchronous DRAM (SDRAM).

[0130] In some example embodiments of the present disclosure, when the computer program is executed by the processor 122, the following operations may be implemented: obtaining M sample fingerprint images; based on each sample fingerprint image in the M sample fingerprint images, generating N approximate images according to N target similarities, wherein the target similarity is the similarity between the approximate image and the sample fingerprint image on which it is based; analyzing the difference between each sample fingerprint image and each correspondingly generated approximate image to obtain (M×N) difference parameters, wherein each difference parameter is used to quantitatively represent the difference between one sample fingerprint image and one approximate image. The difference between the two images is calculated, and the difference parameter is associated with the similarity between the represented sample fingerprint image and the approximate image, so that each of the target similarities is associated with M difference parameters; respectively counting the M difference parameters associated with each of the P specified similarities among the N target similarities to obtain the difference statistical parameters corresponding to each of the P specified similarities; obtaining a reference fingerprint image; and generating simulated fingerprint images corresponding to each of the P specified similarities according to the reference fingerprint image and the difference statistical parameters corresponding to each of the P specified similarities, wherein M and N are both integers greater than 1, and P is an integer greater than or equal to 1.

[0131] The computing device according to some example embodiments may include, but is not limited to, devices such as mobile phones, laptop computers, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, and the like. Fig.12 The computing device shown is merely an example and should not bring any limitation to the functionality and scope of use of the example embodiments of the present disclosure.

[0132] The above has been referred to Figures 1 to 12 The fingerprint image generation method and apparatus according to some example embodiments of the present disclosure are described. However, it should be understood that: Fig.11 The fingerprint image generating device and its units shown in the figure may be respectively configured as software, hardware, firmware or any combination of the above items to perform specific functions. Fig.12 The computing device shown in is not limited to including the components shown above, but some components may be added or deleted as needed, and the above components may also be combined.

[0133] According to the fingerprint image generation method and device of some example embodiments of the present disclosure, it is found through preliminary analysis that the similarity between the sample fingerprint image as a benchmark and its approximate image shows an obvious correlation with the activity score of the approximate image. Based on this, by respectively generating approximate images with the same target similarity for multiple sample fingerprint images, and then counting the differences between each approximate image and the corresponding sample fingerprint image, a difference statistical parameter is obtained, which can be used to represent the difference between any benchmark image and its approximate image under the target similarity. At the same time, in some example embodiments, the target similarity and the activity score of the approximate image show an obvious correlation, which makes the difference statistical parameter represent the difference between any benchmark image and its approximate image under the relevant activity score. On this basis, for the reference fingerprint image, combined with the difference statistical parameter corresponding to the specified similarity, a new fingerprint image is generated, which is recorded as a simulated fingerprint image, and the simulated fingerprint image can have an activity score related to the specified similarity. Since the method according to some example embodiments can obtain difference statistical parameters corresponding to different target similarities (or different activity scores) with the help of a small number of sample fingerprint images, and these difference statistical parameters can be directly reused to quickly generate simulated fingerprint images with the required activity scores based on any reference fingerprint image, it is possible to quickly and inexpensively obtain simulated fingerprint images with different activity scores.

[0134] While the present disclosure has been shown and described with reference to example embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present disclosure as defined by the following claims.

Claims

1. A fingerprint image generation method, It is characterized in that The fingerprint image generation method comprises: Obtain M sample fingerprint images; Based on each sample fingerprint image in the M sample fingerprint images, N approximate images are generated according to N target similarities, each of the N target similarities is the similarity between "a corresponding approximate image in the N approximate images" and "a corresponding sample fingerprint image in the M sample fingerprint images, on which the corresponding approximate image in the N approximate images is based", and M and N are integers greater than 1; Analyzing the difference between each sample fingerprint image in the M sample fingerprint images and each approximate image in the N approximate images to obtain (M×N) difference parameters, each difference parameter in the (M×N) difference parameters quantifies the difference between each sample fingerprint image in the corresponding M sample fingerprint images and the corresponding N approximate images, each difference parameter in the (M×N) difference parameters is associated with the similarity between the corresponding sample fingerprint image in the M sample fingerprint images and the corresponding approximate image in the N approximate images, so that each target similarity in the N target similarities is associated with the M difference parameters; Respectively counting M difference parameters associated with P similarities among the N target similarities to obtain difference statistical parameters corresponding to the P similarities, where P is an integer greater than or equal to 1; Obtaining a reference fingerprint image; Based on the reference fingerprint image and the difference statistical parameters corresponding to the P similarities, simulated fingerprint images corresponding to the P similarities are generated.

2. The fingerprint image generation method according to claim 1, It is characterized in that The step of generating N approximate images based on each sample fingerprint image in the M sample fingerprint images and according to N target similarities comprises: Taking each sample fingerprint image in the M sample fingerprint images as a reference, performing iterative calculation using an evolutionary algorithm to obtain the N approximate images corresponding to the M sample fingerprint images, wherein each approximate image in the N approximate images corresponds to one iteration number; Each of the N target similarities is represented by the number of iterations of a corresponding approximate image in the N approximate images corresponding to an associated difference parameter in the M difference parameters.

3. The fingerprint image generation method according to claim 2, It is characterized in that In the evolutionary algorithm, the structural similarity between the benchmarks of the multiple candidate images generated in each iteration is the fitness value of the corresponding candidate image among the multiple candidate images, and the corresponding approximate image among the N approximate images of the corresponding number of iterations is screened out from the multiple candidate images based on the fitness value.

4. The fingerprint image generation method according to claim 2, It is characterized in that The evolutionary algorithm includes a genetic algorithm.

5. The fingerprint image generation method according to claim 1, It is characterized in that The analyzing the difference between each sample fingerprint image in the M sample fingerprint images and each approximate image in the N approximate images to obtain (M×N) difference parameters includes: Respectively taking each sample fingerprint image in the M sample fingerprint images and a corresponding approximate image in the N approximate images as an image pair to form (M×N) image pairs; Calculate the difference in pixel values ​​at corresponding pixel points of each image pair in the (M×N) image pairs as a difference value; For each of the (M×N) image pairs, the distribution law of the difference values ​​on each pixel point is analyzed to obtain parameters describing the distribution law as the difference parameters of the corresponding image pair in the (M×N) image pairs, thereby obtaining the (M×N) difference parameters.

6. The fingerprint image generation method according to claim 5, It is characterized in that For each of the (M×N) image pairs, analyzing the distribution law of the difference values ​​on each pixel point, obtaining parameters describing the distribution law as difference parameters of the corresponding image pair in the (M×N) image pairs, and obtaining the (M×N) difference parameters, including: For each of the (M×N) image pairs, frequency statistical analysis is performed on the difference values ​​at each pixel point, and probability distribution parameters are obtained by fitting, which are used as difference parameters of the corresponding image pair in the (M×N) image pairs, thereby obtaining the (M×N) difference parameters.

7. The fingerprint image generation method according to claim 6, It is characterized in that The probability distribution parameters include normal distribution parameters.

8. The fingerprint image generation method according to claim 1, It is characterized in that The fingerprint image generation method further comprises: Respectively counting the M difference parameters associated with the N target similarities to obtain the difference statistical parameters of the N target similarities; Acquire multiple verification fingerprint images; For each verification fingerprint image among the multiple verification fingerprint images, generating a simulated fingerprint image corresponding to each of the N target similarities according to the difference statistical parameters corresponding to each of the verification fingerprint image and the N target similarities; Determine the activity score of the simulated fingerprint image corresponding to each of the N target similarities; The P similarities are screened out from the N target similarities according to the activity score and the screening condition.

9. The fingerprint image generation method according to claim 8, It is characterized in that P is greater than 1, and the P similarities are a continuous interval in the similarity interval formed by the N target similarities; The screening condition includes: for at least part of the verification fingerprint images, the activity scores of the simulated fingerprint images corresponding to the P similarities are continuously distributed in the activity score interval.

10. A fingerprint image generating device, It is characterized in that The fingerprint image generating device comprises: A processing circuit, the processing circuit being configured to: Obtain M sample fingerprint images; Based on each sample fingerprint image in the M sample fingerprint images, N approximate images are generated according to N target similarities, each of the N target similarities is a similarity value between "a corresponding approximate image in the N approximate images" and "a corresponding sample fingerprint image in the M sample fingerprint images, on which the corresponding approximate image in the N approximate images is based", and M and N are integers greater than 1; Analyze the difference between each sample fingerprint image in the M sample fingerprint images and each approximate image in the N approximate images to obtain (M×N) difference parameters, each difference parameter in the (M×N) difference parameters quantifies the difference between each sample fingerprint image in the corresponding M sample fingerprint images and the corresponding N approximate images, and each difference parameter in the (M×N) difference parameters is associated with a similarity value between a corresponding sample fingerprint image in the M sample fingerprint images and a corresponding approximate image in the N approximate images, so that each target similarity in the N target similarities is associated with the M difference parameters; Respectively counting M difference parameters associated with P similarities among the N target similarities to obtain difference statistical parameters corresponding to the P similarities, where P is an integer greater than or equal to 1; Obtaining a reference fingerprint image; Based on the reference fingerprint image and the difference statistical parameters corresponding to the P similarities, simulated fingerprint images corresponding to the P similarities are generated.

11. A non-transitory computer-readable storage medium storing a computer program, It is characterized in that When the computer program is executed by a processor, it is configured to implement the fingerprint image generation method according to any one of claims 1 to 9.

12. A computing device, It is characterized in that The computing device comprises: at least one processor; At least one memory is configured to store a computer program, wherein the computer program, when executed by the at least one processor, is configured to: Obtain M sample fingerprint images; Based on each sample fingerprint image in the M sample fingerprint images, N approximate images are generated according to N target similarities, each of the N target similarities is the similarity between "a corresponding approximate image in the N approximate images" and "a corresponding sample fingerprint image in the M sample fingerprint images, on which the corresponding approximate image in the N approximate images is based", and M and N are integers greater than 1; Analyzing the difference between each sample fingerprint image in the M sample fingerprint images and each approximate image in the N approximate images to obtain (M×N) difference parameters, each difference parameter in the (M×N) difference parameters quantifies the difference between each sample fingerprint image in the corresponding M sample fingerprint images and the corresponding N approximate images, each difference parameter in the (M×N) difference parameters is associated with the similarity between the corresponding sample fingerprint image in the M sample fingerprint images and the corresponding approximate image in the N approximate images, so that each target similarity in the N target similarities is associated with the M difference parameters; Respectively counting M difference parameters associated with P similarities among the N target similarities to obtain difference statistical parameters corresponding to the P similarities, where P is an integer greater than or equal to 1; Obtaining a reference fingerprint image; Based on the reference fingerprint image and the difference statistical parameters corresponding to the P similarities, simulated fingerprint images corresponding to the P similarities are generated.

13. A computer program product, It is characterized in that The instructions in the computer program product can be executed by a processor of a computer device to implement the fingerprint image generation method according to any one of claims 1 to 9.

Images