Training Method of Fingerprint Repair Model, Fingerprint Recognition Method, Medium and Device

Abstract

The present disclosure relates to a training method, a fingerprint recognition method, a medium and a device for a fingerprint restoration model, and relates to the technical field of machine learning. The method includes: obtaining a first standard fingerprint, and inputting the first standard fingerprint into a preset dry fingerprint generation model to obtain a first simulated dry fingerprint; using a first feature extraction model to extract first local fingerprint features of the first simulated dry fingerprint, and using the first local fingerprint features to perform fingerprint restoration on the first simulated dry fingerprint to obtain a first output result; constructing a first loss function according to the first output result and the first standard fingerprint, and using the first loss function to adjust parameters in the first feature extraction model to obtain a fingerprint restoration model. The present disclosure improves the accuracy of the fingerprint restoration model.

Description

Background Art

[0002] Fingerprint recognition technology is a relatively mature identity information recognition technology and is widely used in fields such as public security and personal information protection. In the field of fingerprint recognition, due to the partial loss of texture information in dry fingerprints and the inability to repair them with classical image processing methods, the recognition accuracy of dry fingerprints is relatively low.

[0003] The training method of the existing fingerprint repair model can be implemented through deep learning technology. Among them, deep learning technology is a popular technology field in the current field of image processing and can achieve effects that classical algorithms cannot reach in image restoration and reconstruction tasks. In the existing training method of the fingerprint repair model based on deep learning, a large number of dry fingerprints and their corresponding normal high-definition fingerprints need to be collected, and then a dataset is constructed based on them to train the fingerprint repair model.

[0004] However, it is very difficult to collect dry fingerprints and their corresponding normal high-definition fingerprints, and the data volume of the constructed dataset is small. During the training process of the fingerprint repair model, there is an overfitting problem, which further reduces the accuracy of the fingerprint repair model.

[0005] Therefore, there is an urgent need to provide a new training method for the fingerprint repair model to overcome the above problems.

[0006] It should be noted that the information disclosed in the above background art is only used to strengthen the understanding of the background of the present disclosure, and thus may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0007] The purpose of the present disclosure is to provide a training method for a fingerprint repair model, a fingerprint recognition method, a computer-readable storage medium, and an electronic device, so as to at least to some extent overcome the problem of relatively low accuracy of the fingerprint repair model caused by the limitations and defects of related technologies.

[0008] According to one aspect of the present disclosure, a training method for a fingerprint repair model is provided, including:

[0009] Obtain a first standard fingerprint and input the first standard fingerprint into a preset dry fingerprint generation model to obtain a first simulated dry fingerprint;

[0010] Use a first feature extraction model to extract the first local fingerprint feature of the first simulated dry fingerprint, and use the first local fingerprint feature to repair the first simulated dry fingerprint to obtain a first output result;

[0011] Construct a first loss function based on the first output result and the first standard fingerprint, and use the first loss function to adjust the parameters in the first feature extraction model to obtain a fingerprint repair model.

[0012] In an exemplary embodiment of the present disclosure, the first feature extraction model includes multiple layers of first convolutional layers, multiple layers of first downsampling layers, and multiple layers of first upsampling layers, and the number of layers of the multiple layers of first downsampling layers and the multiple layers of first upsampling layers is the same;

[0013] Among them, using the first feature extraction model to extract the first local fingerprint features of the first simulated dry fingerprint, and using the first local fingerprint features to repair the first simulated dry fingerprint to obtain a first output result, including:

[0014] Perform convolution processing on the first simulated dry fingerprint using the multiple layers of first convolutional layers included in the first feature extraction model to obtain multiple local features of different scales;

[0015] Perform downsampling on the multiple local features of different scales using the multiple layers of first downsampling layers to obtain regional features of multiple local regions;

[0016] Use the multiple layers of first upsampling layers to fill the regional features of the multiple local regions into the fingerprint missing regions in the first simulated dry fingerprint to obtain the first output result.

[0017] In an exemplary embodiment of the present disclosure, constructing a first loss function according to the first output result and the first standard fingerprint includes:

[0018]

[0019] where loss is the first loss function, I gt is the first standard fingerprint, is the first output result; is the absolute value between the first standard fingerprint and the first output result.

[0020] In an exemplary embodiment of the present disclosure, before inputting the first standard fingerprint into a preset dry fingerprint generation model to obtain a first simulated dry fingerprint, the training method of the fingerprint repair model further includes:

[0021] Train a second feature extraction model using a second standard fingerprint and a first actual dry fingerprint to obtain the preset dry fingerprint generation model.

[0022] In an exemplary embodiment of the present disclosure, training the second feature extraction model using the second standard fingerprint and the first actual dry fingerprint to obtain the preset dry fingerprint generation model includes:

[0023] Input the second standard fingerprint into the second feature extraction model to obtain a second output result, and configure a first label and a second label for the first actual dry fingerprint and the second output result respectively;

[0024] Input the first actual dry fingerprint and the second output result into the standard stacked classification network respectively to obtain a third output result corresponding to the first actual dry fingerprint and a fourth output result corresponding to the second output result;

[0025] Construct a second loss function corresponding to the standard stacked classification network according to the third output result and the fourth output result, and construct a third loss function corresponding to the second feature extraction model according to the second output result and the fourth output result;

[0026] Construct an adversarial generation training loss function according to the second loss function and the third loss function, and use the adversarial generation training loss function to train the second feature extraction model and the standard stacked classification network to obtain the preset dry fingerprint generation model and the fingerprint discrimination model.

[0027] In an exemplary embodiment of the present disclosure, constructing a second loss function corresponding to the standard stacked classification network according to the third output result and the fourth output result includes:

[0028]

[0029] Where V 1 (D, G) is the second loss function corresponding to the standard stacked classification network, is an arbitrary sampling of the dataset data composed of the first actual dry fingerprints, x is the first actual dry fingerprint, D(x) is the third output result, D(G(z)) is the fourth output result, G(z) is the second output result, and z is the second standard fingerprint.

[0030] In an exemplary embodiment of the present disclosure, constructing a third loss function corresponding to the second feature extraction model according to the second output result and the fourth output result includes:

[0031]

[0032] Where V 2 (D, G) is the third loss function corresponding to the second feature extraction model, For randomly sampling the data set composed of the second standard fingerprints, D(G(z)) is the fourth output result, G(z) is the second output result, z is the second standard fingerprint, and ||z - G(z)|| 1 is the absolute value between the second standard fingerprint and the second output result.

[0033] According to one aspect of the present disclosure, there is provided a fingerprint recognition method, including:

[0034] Obtaining a fingerprint to be recognized, and classifying the fingerprint to be recognized based on a preset fingerprint classification model to obtain a fingerprint classification result;

[0035] Determining whether the fingerprint to be recognized is a fingerprint to be repaired according to the fingerprint classification result;

[0036] When it is determined that the fingerprint to be recognized is a fingerprint to be repaired, performing fingerprint repair on the fingerprint to be recognized based on a preset fingerprint repair model to obtain a standard repaired fingerprint; wherein, the fingerprint repair model is obtained by training a first feature extraction model through the training method of the fingerprint repair model described in any one of the above;

[0037] Recognizing the standard repaired fingerprint based on a preset fingerprint recognition model to obtain a fingerprint recognition result.

[0038] In an exemplary embodiment of the present disclosure, before classifying the fingerprint to be recognized based on a preset fingerprint classification model to obtain a fingerprint classification result, the fingerprint recognition method further includes:

[0039] Obtaining a third standard fingerprint and a second actual dry fingerprint, and training a classification network model to be trained by using the third standard fingerprint and the second actual dry fingerprint to obtain the preset fingerprint classification model.

[0040] In an exemplary embodiment of the present disclosure, training a classification network model to be trained by using the third standard fingerprint and the second actual dry fingerprint to obtain the preset fingerprint classification model includes:

[0041] Inputting the third standard fingerprint and the second actual dry fingerprint into the classification network model to be trained to obtain a fifth output result;

[0042] Constructing a fourth loss function according to the fifth output result, the third label of the third standard fingerprint, and the fourth label of the second actual dry fingerprint;

[0043] Adjusting the parameters included in the classification network model to be trained based on the fourth loss function to obtain the preset fingerprint classification model.

[0044] In an exemplary embodiment of the present disclosure, the classification network model to be trained is a binary classification network model, and the classification network model to be trained includes any one of a ResNet model, a VGG model, and a DenseNet model.

[0045] According to one aspect of the present disclosure, there is provided a training apparatus for a fingerprint repair model, including:

[0046] A first dry fingerprint generation module, configured to obtain a first standard fingerprint and input the first standard fingerprint into a preset dry fingerprint generation model to obtain a first simulated dry fingerprint;

[0047] A first dry fingerprint repair module, configured to extract first local fingerprint features of the first simulated dry fingerprint by using a first feature extraction model, and perform fingerprint repair on the first simulated dry fingerprint by using the first local fingerprint features to obtain a first output result;

[0048] A first parameter adjustment module, configured to construct a first loss function according to the first output result and the first standard fingerprint, and adjust parameters in the first feature extraction model by using the first loss function to obtain a fingerprint repair model.

[0049] According to one aspect of the present disclosure, there is provided a fingerprint recognition apparatus, including:

[0050] A fingerprint classification module, configured to obtain a fingerprint to be recognized and classify the fingerprint to be recognized based on a preset fingerprint classification model to obtain a fingerprint classification result;

[0051] A fingerprint to be repaired determination module, configured to determine whether the fingerprint to be recognized is a fingerprint to be repaired according to the fingerprint classification result;

[0052] A second fingerprint repair module, configured to, when determining that the fingerprint to be recognized is a fingerprint to be repaired, perform fingerprint repair on the fingerprint to be recognized based on a preset fingerprint repair model to obtain a standard repaired fingerprint; wherein, the fingerprint repair model is obtained by training a first feature extraction model through the training method of the fingerprint repair model described in any one of the above;

[0053] A fingerprint recognition module, configured to recognize the standard repaired fingerprint based on a preset fingerprint recognition model to obtain a fingerprint recognition result.

[0054] According to one aspect of the present disclosure, there is provided a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the training method of the fingerprint repair model described in any one of the above, and the fingerprint recognition method described in any one of the above are implemented.

[0055] According to one aspect of the present disclosure, there is provided an electronic device, including:

[0056] a processor; and

[0057] a memory for storing executable instructions of the processor;

[0058] wherein the processor is configured to execute the training method of the fingerprint repair model described in any one of the above, and the fingerprint recognition method described in any one of the above by executing the executable instructions.

[0059] A training method of a fingerprint repair model provided by an embodiment of the present disclosure. On the one hand, by inputting a first standard fingerprint into a preset dry fingerprint generation model, a first simulated dry fingerprint is obtained; then, a first local fingerprint feature of the first simulated dry fingerprint is extracted by using a first feature extraction model, and the first simulated dry fingerprint is fingerprint-repaired by using the first local fingerprint feature to obtain a first output result; finally, a first loss function is constructed according to the first output result and the first standard fingerprint, and the parameters in the first feature extraction model are adjusted by using the first loss function to obtain a fingerprint repair model. Since the first simulated dry fingerprint can be directly generated by the preset dry fingerprint generation model, and then the first feature extraction model is trained by the first simulated dry fingerprint to obtain a fingerprint repair model, the problem in the prior art that it is very difficult to collect dry fingerprints and their corresponding normal high-definition fingerprints, the data volume of the constructed data set is small, and there is an overfitting problem in the training process of the fingerprint repair model, resulting in a low accuracy of the fingerprint repair model is solved, and the accuracy of the fingerprint repair model is improved; on the other hand, since it is not necessary to collect a large number of dry fingerprints and their corresponding normal high-definition fingerprints, the training difficulty of the fingerprint repair model is simplified, and the training speed of the fingerprint repair model is improved.

[0060] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The accompanying drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present disclosure, and are used together with the specification to explain the principles of the present disclosure. Obviously, the accompanying drawings in the following description are only some embodiments of the present disclosure, and those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

[0062] Figure 1 A flowchart schematically showing a training method of a fingerprint repair model according to an exemplary embodiment of the present disclosure.

[0063] Figure 2Schematic diagram showing an example of a first standard fingerprint according to an exemplary embodiment of the present disclosure.

[0064] Figure 3 Schematic diagram showing an example of a first simulated dry fingerprint according to an exemplary embodiment of the present disclosure.

[0065] Figure 4 Schematic diagram of a method flow for extracting first local fingerprint features of the first simulated dry fingerprint using a first feature extraction model and performing fingerprint repair on the first simulated dry fingerprint using the first local fingerprint features to obtain a first output result according to an exemplary embodiment of the present disclosure.

[0066] Figure 5 Schematic diagram showing a structural example of a first feature extraction model according to an exemplary embodiment of the present disclosure.

[0067] Figure 6 Schematic diagram showing an example of a fingerprint missing area of a first simulated dry fingerprint according to an exemplary embodiment of the present disclosure.

[0068] Figure 7 Schematic diagram showing an example of a first simulated dry fingerprint after repair according to an exemplary embodiment of the present disclosure.

[0069] Figure 8 Schematic diagram of a method flow for training a second feature extraction model using a second standard fingerprint and a first actual dry fingerprint to obtain the preset dry fingerprint generation model according to an exemplary embodiment of the present disclosure.

[0070] Figure 9 Schematic diagram of a flowchart of a fingerprint recognition method according to an exemplary embodiment of the present disclosure.

[0071] Figure 10 Schematic diagram of a block diagram of a training device for a fingerprint repair model according to an exemplary embodiment of the present disclosure.

[0072] Figure 11 Schematic diagram of a block diagram of a fingerprint recognition device according to an exemplary embodiment of the present disclosure.

[0073] Figure 12 Schematic diagram showing an electronic device for implementing the above-mentioned fingerprint repair model training method and fingerprint recognition method according to an exemplary embodiment of the present disclosure. Detailed implementation manners

[0074] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of the embodiments of the present disclosure. However, those skilled in the art will recognize that the technical solutions of the present disclosure may be practiced without one or more of the specific details, or may be implemented using other methods, components, devices, steps, etc. In other instances, well-known technical solutions are not shown or described in detail to avoid obscuring the various aspects of the present disclosure.

[0075] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0076] Deep learning technology has made rapid progress nowadays. Some large enterprises have invested huge capital and manpower in the research of deep learning technology, and continuously launched their unique products and technologies. Other enterprises such as IBM, Microsoft, Amazon, etc. are also constantly entering the field of deep learning and have achieved certain results. Deep learning technology has made breakthrough progress in the field of human data perception, such as describing image content, recognizing objects in complex environments in images, and performing speech recognition in noisy environments. At the same time, deep learning technology can also solve the problems of image generation and fusion.

[0077] Fingerprint recognition is a relatively mature technology in biometric pattern recognition in recent years. This technology requires the extraction of features of fingerprint images and applications such as fingerprint matching and identity recognition. Although fingerprint recognition algorithms are relatively mature in application, for relatively common abnormal fingerprints such as dry fingerprints, due to serious local fingerprint information loss, the existing fingerprint recognition algorithms have very low recognition rates for dry fingerprints; and for the texture repair of dry fingerprint images, it is beneficial to improve the recognition accuracy of fingerprint recognition algorithms for dry fingerprints.

[0078] In this exemplary embodiment, a training method for a fingerprint restoration model is first provided. This method can run on a server, a server cluster, a cloud server, etc. Of course, those skilled in the art can also run the method of the present disclosure on other platforms according to requirements, and no special limitation is made in this exemplary embodiment. Refer to Figure 1 As shown, the training method of the fingerprint restoration model may include the following steps:

[0079] Step S110. Obtain a first standard fingerprint and input the first standard fingerprint into a preset dry fingerprint generation model to obtain a first simulated dry fingerprint;

[0080] Step S120. Use a first feature extraction model to extract the first local fingerprint feature of the first simulated dry fingerprint, and use the first local fingerprint feature to perform fingerprint restoration on the first simulated dry fingerprint to obtain a first output result;

[0081] Step S130. Construct a first loss function according to the first output result and the first standard fingerprint, and use the first loss function to adjust the parameters in the first feature extraction model to obtain a fingerprint restoration model.

[0082] In the above training method of the fingerprint restoration model, on the one hand, by inputting the first standard fingerprint into the preset dry fingerprint generation model, a first simulated dry fingerprint is obtained; then, the first local fingerprint feature of the first simulated dry fingerprint is extracted by using the first feature extraction model, and the first simulated dry fingerprint is restored by using the first local fingerprint feature to obtain a first output result; finally, a first loss function is constructed according to the first output result and the first standard fingerprint, and the parameters in the first feature extraction model are adjusted by using the first loss function to obtain a fingerprint restoration model. Since the first simulated dry fingerprint can be directly generated by the preset dry fingerprint generation model, and then the first feature extraction model is trained by the first simulated dry fingerprint to obtain a fingerprint restoration model, the problem in the prior art that it is very difficult to collect dry fingerprints and their corresponding normal high-definition fingerprints, the data volume of the constructed data set is small, and there is an overfitting problem in the training process of the fingerprint restoration model, resulting in a low accuracy of the fingerprint restoration model is solved, and the accuracy of the fingerprint restoration model is improved; on the other hand, since it is not necessary to collect a large number of dry fingerprints and their corresponding normal high-definition fingerprints, the training difficulty of the fingerprint restoration model is simplified, and the training speed of the fingerprint restoration model is improved.

[0083] Hereinafter, the training method and fingerprint recognition method of the fingerprint restoration model in the exemplary embodiment of the present disclosure will be explained and described in detail with reference to the accompanying drawings.

[0084] First, the invention objectives of the exemplary embodiments of the present disclosure are explained and described. Specifically, the exemplary embodiments of the present disclosure use deep learning methods to restore the missing details of dry fingerprints; at the same time, to solve the problem of difficult acquisition of training data, the exemplary embodiments of the present disclosure adopt an unsupervised data generation network to generate dry fingerprints using normal high-definition fingerprints, and then use the generated dry fingerprints and the data pairs of high-definition fingerprints to train the fingerprint restoration network so that it can repair the missing information of dry fingerprints.

[0085] Secondly, in a method for training a fingerprint restoration model provided by the exemplary embodiments of the present disclosure:

[0086] In step S110, a first standard fingerprint is obtained, and the first standard fingerprint is input into a preset dry fingerprint generation model to obtain a first simulated dry fingerprint.

[0087] In the exemplary embodiments of the present disclosure, first, the first standard fingerprint can be obtained from a fingerprint data acquisition library. The first standard fingerprint can be a fingerprint with complete high-definition fingerprint features, and the first standard fingerprint can specifically refer to Figure 2 as shown; secondly, the first standard fingerprint can be input into a preset dry fingerprint generation model to obtain a first simulated dry fingerprint. Among them, the specific processing process of the first simulated dry fingerprint can include: first, performing convolutional processing on the first standard fingerprint using the dry fingerprint generation model to obtain multiple local features of different scales; then performing downsampling on the multiple local features of different scales to obtain regional features of multiple local regions; finally, deleting the regional features of multiple local regions in the second standard fingerprint to obtain the first simulated dry fingerprint; among them, a specific example diagram of the first simulated dry fingerprint can refer to Figure 3 as shown.

[0088] In step S120, a first local fingerprint feature of the first simulated dry fingerprint is extracted using a first feature extraction model, and the first simulated dry fingerprint is fingerprint-restored using the first local fingerprint feature to obtain a first output result; wherein, the first feature extraction model includes multiple first convolutional layers, multiple first downsampling layers, and multiple first upsampling layers, and the number of layers of the multiple first downsampling layers and the multiple first upsampling layers is the same.

[0089] Specifically, referring to Figure 4 as shown, extracting the first local fingerprint feature of the first simulated dry fingerprint using the first feature extraction model and fingerprint-restoring the first simulated dry fingerprint using the first local fingerprint feature to obtain a first output result may include the following steps:

[0090] Step S410, perform convolution processing on the first simulated dry fingerprint using multiple first convolutional layers included in the first feature extraction model to obtain local features of multiple different scales;

[0091] Step S440, perform downsampling on the local features of multiple different scales using the multiple first downsampling layers to obtain regional features of multiple local regions;

[0092] Step S430, use the multiple first upsampling layers to fill the regional features of the multiple local regions into the fingerprint missing regions in the first simulated dry fingerprint to obtain the first output result.

[0093] Next, steps S410 - S430 will be explained and described. First, the first feature extraction model will be explained and described. Specifically, the first feature extraction module can be a Unet model, and of course it can also be a SIFT (Scale-invariant feature transform) feature extraction model. This example does not make special restrictions on this. Specifically, taking the first feature extraction model as a Unet model as an example, the first feature extraction model will be explained and described.

[0094] Reference Figure 5 As shown, the first feature extraction model may include a first input layer 501, multiple first convolutional layers 502, multiple first downsampling layers 503, multiple first upsampling layers 504, and a first output layer 505; among them, the first input layer is connected to the input of the first convolutional layer among the multiple first convolutional layers, the output of the last convolutional layer among the multiple first convolutional layers is connected to the input of the first downsampling layer among the multiple first downsampling layers, the output of the last first downsampling layer among the multiple first downsampling layers is connected to the input of the first upsampling layer among the multiple first upsampling layers on the right, and the output of the last first upsampling layer among the multiple first upsampling layers is connected to the output layer; and there are, the number of layers of the multiple upsampling layers and the multiple downsampling layers is the same.

[0095] Among them, in the process of using the first feature extraction model to process the first simulated dry fingerprint to obtain the first output result: First, perform convolution processing on the first simulated dry fingerprint using multiple first convolutional layers to obtain local features of multiple different scales; then perform downsampling on the local features of multiple different scales using multiple first downsampling layers to obtain regional features of multiple local regions; finally, use multiple first upsampling layers to fill the regional features of multiple local regions into the fingerprint missing regions in the first simulated dry fingerprint (this fingerprint missing region can be referred to, for example, Figure 6As shown, the first output result can be obtained, which is also the repaired fingerprint after being repaired by the first feature extraction model. This repaired fingerprint can be referred to, for example, as Figure 7 as shown.

[0096] In step S130, a first loss function is constructed based on the first output result and the first standard fingerprint, and the parameters in the first feature extraction model are adjusted using the first loss function to obtain a fingerprint repair model.

[0097] In the present exemplary embodiment, after obtaining the first output result, a first loss function can be constructed based on the first output result and the first standard fingerprint, and then the first feature extraction model can be trained using this first loss function to obtain a fingerprint repair model. Among them, the first loss function can be specifically shown as the following formula (1):

[0098]

[0099] where loss is the first loss function, I gt is the first standard fingerprint, is the first output result; is the absolute value between the first standard fingerprint and the first output result.

[0100] Hereinafter, the dry fingerprint generation model involved in the exemplary embodiment of the present disclosure will be explained and described. Among them, the dry fingerprint generation model can be trained in the following manner: The second feature extraction model is trained using the second standard fingerprint and the first actual dry fingerprint to obtain the preset dry fingerprint generation model. Specifically, referring to Figure 8 as shown, training the second feature extraction model using the second standard fingerprint and the first actual dry fingerprint to obtain the preset dry fingerprint generation model may include the following steps:

[0101] Step S810: Input the second standard fingerprint into the second feature extraction model to obtain a second output result, and configure a first label and a second label for the first actual dry fingerprint and the second output result respectively;

[0102] Step S820: Input the first actual dry fingerprint and the second output result into the standard stacked classification network respectively to obtain a third output result corresponding to the first actual dry fingerprint and a fourth output result corresponding to the second output result;

[0103] Step S830: Construct a second loss function corresponding to the standard stacked classification network based on the third output result and the fourth output result, and construct a third loss function corresponding to the second feature extraction model based on the second output result and the fourth output result;

[0104] Step S840: Construct an adversarial generation training loss function according to the second loss function and the third loss function, and use the adversarial generation training loss function to train the second feature extraction model and the standard stacked classification network to obtain the preset dry fingerprint generation model and the fingerprint discrimination model.

[0105] Next, steps S810 - S840 will be explained and described. First, the second feature extraction model will be explained and described. Specifically, the second feature extraction model can be a Unet model, or it can also be a SIFT (Scale-invariant feature transform) feature extraction model. This example does not make special restrictions on this. Specifically, taking the second feature extraction model as a Unet model as an example, the second feature extraction model will be explained and described. Among them, the second feature extraction model can include a second input layer, multiple second convolutional layers, multiple second downsampling layers, multiple second upsampling layers, and a second output layer; among them, the second input layer is connected to the input of the first second convolutional layer among the multiple second convolutional layers, the output of the last convolutional layer among the multiple second convolutional layers is connected to the input of the first second downsampling layer among the multiple second downsampling layers, the output of the last second downsampling layer among the multiple second downsampling layers is connected to the input of the first second upsampling layer among the multiple second upsampling layers on the right, and the output of the last second upsampling layer among the multiple second upsampling layers is connected to the output layer; and there is, the number of layers of the multiple second upsampling layers and the multiple second downsampling layers is the same. It should be supplemented and explained here that the structure of the second feature extraction model is generally similar to the structure of the first feature extraction model. The number of layers of the second upsampling layer and the first upsampling layer can be the same or different. This example does not make special limitations on this.

[0106] Among them, in the process of using the second feature extraction model to process the second standard fingerprint to obtain the second output result, multiple second convolutional layers can be first used to perform convolutional processing on the second standard fingerprint to obtain local features of multiple different scales; then multiple second downsampling layers are used to downsample the local features of multiple different scales to obtain regional features of multiple local regions; then multiple second upsampling layers are used to delete the regional features of multiple local regions in the second standard fingerprint to obtain the second output result; then, a first label 1 and a second label 0 are respectively configured for the first actual dry fingerprint and the second output result; further, the first actual dry fingerprint and the second output result are respectively input into a standard stacked classification network (such as a VGG classification network) to obtain a third output result corresponding to the first actual dry fingerprint and a fourth output result corresponding to the second output result; furthermore, a second loss function corresponding to the standard stacked classification network is constructed according to the third output result and the fourth output result, and a third loss function corresponding to the second feature extraction model is constructed according to the second output result and the fourth output result. Among them, the second loss function and the third loss function can be respectively shown in the following formulas (2) and (3):

[0107]

[0108] Among them, V 1 (D, G) is the second loss function corresponding to the standard stacked classification network, is an arbitrary sampling of the dataset data composed of the first actual dry fingerprints, x is the first actual dry fingerprint, D(x) is the third output result, D(G(z)) is the fourth output result, G(z) is the second output result, and z is the second standard fingerprint;

[0109]

[0110] Among them, V 2 (D, G) is the third loss function corresponding to the second feature extraction model, is an arbitrary sampling of the dataset composed of the second standard fingerprints, D(G(z)) is the fourth output result, G(z) is the second output result, z is the second standard fingerprint, and ||z - G(z)|| 1 is the absolute value between the second standard fingerprint and the second output result.

[0111] Finally, an adversarial generation training loss function is constructed according to the second loss function and the third loss function, and the second feature extraction model and the standard stacked classification network are trained using the adversarial generation training loss function to obtain a preset dry fingerprint generation model and a fingerprint discrimination model. Among them, the adversarial generation training loss function L can be specifically shown in the following formula (4):

[0112] L=maxV 1 (D,G)+minV 2 (D,G); Formula (4)

[0113] At this point, the preset dry fingerprint generation model and fingerprint discrimination model can be obtained. This method avoids the problem that the normal state image and the dry state image of the same fingerprint need to be pressed twice on the collection device; and during the two pressings, it is easy to have problems such as fingerprint pressing position offset, inconsistent pressing force, and inconsistent pressing angle, resulting in the inability of the collected normal fingerprint and dry fingerprint data pairs to maintain consistent spatial information. Therefore, the data collected in this way cannot be used for the training of the dry fingerprint repair model. In addition, by using unpaired normal fingerprints and dry fingerprints to train the generation network, the network has the ability to learn dry fingerprint features, and then converts the normal fingerprint image into a dry fingerprint image to form a usable data pair, which can improve the generation efficiency of the data pair and thus improve the training efficiency of the model.

[0114] The exemplary embodiment of the present disclosure also provides a fingerprint recognition method, which can be run on a server, a server cluster or a cloud server, etc. Of course, those skilled in the art can also run the method of the present disclosure on other platforms as required, and this exemplary embodiment does not specifically limit this. Figure 9 As shown, the fingerprint recognition method may include the following steps:

[0115] Step S910, obtaining a fingerprint to be identified, and classifying the fingerprint to be identified based on a preset fingerprint classification model to obtain a fingerprint classification result.

[0116] In this example embodiment, first, the fingerprint to be identified collected by the fingerprint collection device can be obtained; secondly, the fingerprint to be identified can be classified based on the fingerprint classification model to obtain a fingerprint classification result; wherein the fingerprint classification result can be the probability of it being a dry fingerprint, or the probability of it being a normal fingerprint.

[0117] It should be noted here that in order to classify the fingerprint to be recognized, it is first necessary to train the classification network model to be trained, and then obtain the fingerprint classification model. Among them, the specific training process may include: obtaining the third standard fingerprint and the second actual dry fingerprint, and using the third standard fingerprint and the second actual dry fingerprint to train the classification network model to be trained to obtain the preset fingerprint classification model. Among them, using the third standard fingerprint and the second actual dry fingerprint to train the classification network model to be trained to obtain the preset fingerprint classification model may include: first, inputting the third standard fingerprint and the second actual dry fingerprint into the classification network model to be trained to obtain a fifth output result; secondly, constructing a fourth loss function according to the fifth output result, the third label of the third standard fingerprint, and the fourth label of the second actual dry fingerprint; finally, adjusting the parameters included in the classification network model to be trained based on the fourth loss function to obtain the preset fingerprint classification model. And, the classification network model to be trained is a binary classification network model, and the classification network model to be trained includes a ResNet model, a VGG model, and a DenseNet model. Of course, it may also include other binary classification models, and this example does not make special restrictions on this.

[0118] It should be further noted here that since the above fingerprint classification model may be a binary classification model, the fourth loss function used in the training process may be a cross-entropy loss function; by setting this loss function, the training speed of the classification network model to be trained can be improved, and at the same time, the burden on the system can be reduced; of course, other loss functions can also be selected according to actual needs, and this example does not make special restrictions on this.

[0119] Step S920, determine whether the fingerprint to be recognized is a fingerprint to be repaired according to the fingerprint classification result.

[0120] Specifically, if the probability of belonging to a dry fingerprint is greater than a preset threshold (such as 0.6 or 0.7, etc.), it can be considered that it belongs to a fingerprint to be repaired; otherwise, it is considered that the fingerprint to be recognized is a normal fingerprint; when the fingerprint to be recognized is a normal fingerprint, it can be directly recognized by the fingerprint recognition model.

[0121] Step S930, when it is determined that the fingerprint to be recognized is a fingerprint to be repaired, perform fingerprint repair on the fingerprint to be recognized based on a preset fingerprint repair model to obtain a standard repaired fingerprint; wherein, the fingerprint repair model is obtained by training the first feature extraction model through the training method of the aforementioned fingerprint repair model.

[0122] Step S940, recognize the standard repaired fingerprint based on a preset fingerprint recognition model to obtain a fingerprint recognition result.

[0123] Specifically, the preset fingerprint recognition model can be, for example, a convolutional neural network model (Convolutional Neural Networks, CNN), or a recurrent neural network model (Recurrent Neural Network, RNN). This example does not impose special restrictions; in the specific recognition process, the fingerprint recognition model can be used to extract the features of the fingerprint to be recognized (standard repaired fingerprint), and then match the extracted fingerprint features in the template library to obtain the fingerprint recognition result (success or failure).

[0124] In Figure 9 In the fingerprint recognition method shown, when it is determined that the fingerprint to be recognized is a fingerprint to be repaired, the fingerprint to be recognized can be repaired and then fingerprint recognition can be performed, thereby improving the accuracy of the fingerprint recognition result and the probability of successful fingerprint recognition; for example, applying this method to scenarios such as fingerprint attendance or fingerprint passwords can avoid the problem of low user experience caused by failed fingerprint recognition. At the same time, the dry fingerprint repair network (fingerprint repair model) can be embedded in the fingerprint recognition system at the terminal to repair the dry fingerprint and then send it to the fingerprint recognition system, thereby improving the accuracy of fingerprint recognition.

[0125] The exemplary embodiment of the present disclosure also provides a training device for a fingerprint repair model. Referring to Figure 10 shown, the training device for the fingerprint repair model may include a first dry fingerprint generation module 1010, a first dry fingerprint repair module 1020, and a first parameter adjustment module 1030. Among them:

[0126] The first dry fingerprint generation module 1010 can be used to obtain a first standard fingerprint and input the first standard fingerprint into a preset dry fingerprint generation model to obtain a first simulated dry fingerprint;

[0127] The first dry fingerprint repair module 1020 can be used to extract the first local fingerprint features of the first simulated dry fingerprint by using a first feature extraction model, and use the first local fingerprint features to repair the first simulated dry fingerprint to obtain a first output result;

[0128] The first parameter adjustment module 1030 can be used to construct a first loss function according to the first output result and the first standard fingerprint, and use the first loss function to adjust the parameters in the first feature extraction model to obtain a fingerprint repair model.

[0129] In an exemplary embodiment of the present disclosure, the first feature extraction model includes multiple layers of first convolutional layers, multiple layers of first downsampling layers, and multiple layers of first upsampling layers, and the number of layers of the multiple layers of first downsampling layers and the multiple layers of first upsampling layers is the same;

[0130] Among them, using the first feature extraction model to extract the first local fingerprint features of the first simulated dry fingerprint, and using the first local fingerprint features to repair the first simulated dry fingerprint to obtain a first output result, including:

[0131] Performing convolution processing on the first simulated dry fingerprint using the multiple layers of first convolutional layers included in the first feature extraction model to obtain local features of multiple different scales;

[0132] Performing downsampling on the local features of multiple different scales using the multiple layers of first downsampling layers to obtain regional features of multiple local regions;

[0133] Using the multiple layers of first upsampling layers to fill the regional features of the multiple local regions into the fingerprint missing regions in the first simulated dry fingerprint to obtain the first output result.

[0134] In an exemplary embodiment of the present disclosure, constructing a first loss function according to the first output result and the first standard fingerprint, including:

[0135]

[0136] Among them, loss is the first loss function, I gt is the first standard fingerprint, is the first output result; is the absolute value between the first standard fingerprint and the first output result.

[0137] In an exemplary embodiment of the present disclosure, the training device of the fingerprint repair model further includes:

[0138] A dry fingerprint generation model training module, which can be used to train the second feature extraction model using the second standard fingerprint and the first actual dry fingerprint to obtain the preset dry fingerprint generation model.

[0139] In an exemplary embodiment of the present disclosure, training the second feature extraction model using the second standard fingerprint and the first actual dry fingerprint to obtain the preset dry fingerprint generation model, including:

[0140] Inputting the second standard fingerprint into the second feature extraction model to obtain a second output result, and respectively configuring a first label and a second label for the first actual dry fingerprint and the second output result;

[0141] Input the first actual dry fingerprint and the second output result into the standard stacked classification network respectively, to obtain a third output result corresponding to the first actual dry fingerprint and a fourth output result corresponding to the second output result;

[0142] Construct a second loss function corresponding to the standard stacked classification network according to the third output result and the fourth output result, and construct a third loss function corresponding to the second feature extraction model according to the second output result and the fourth output result;

[0143] Construct an adversarial generation training loss function according to the second loss function and the third loss function, and use the adversarial generation training loss function to train the second feature extraction model and the standard stacked classification network, to obtain the preset dry fingerprint generation model and fingerprint discrimination model.

[0144] In an exemplary embodiment of the present disclosure, constructing a second loss function corresponding to the standard stacked classification network according to the third output result and the fourth output result includes:

[0145]

[0146] where V(D,G) is the second loss function corresponding to the standard stacked classification network, is an arbitrary sampling of the data set data composed of the first actual dry fingerprints, x is the first actual dry fingerprint, D(x) is the third output result, D(G(z)) is the fourth output result, G(z) is the second output result, and z is the second standard fingerprint.

[0147] In an exemplary embodiment of the present disclosure, constructing a third loss function corresponding to the second feature extraction model according to the second output result and the fourth output result includes:

[0148]

[0149] where V(D,G) is the third loss function corresponding to the second feature extraction model, is an arbitrary sampling of the data set composed of the second standard fingerprints, D(G(z)) is the fourth output result, G(z) is the second output result, z is the second standard fingerprint, ||z - G(z)|| 1 is the absolute value between the second standard fingerprint and the second output result.

[0150] The present disclosure also provides a fingerprint recognition device. Refer to Figure 11As shown, the fingerprint recognition device may include a fingerprint classification module 1110, a fingerprint to be repaired determination module 1120, a second fingerprint repair module 1130, and a fingerprint recognition module 1140. Among them:

[0151] The fingerprint classification module 1110 can be used to obtain a fingerprint to be recognized and classify the fingerprint to be recognized based on a preset fingerprint classification model to obtain a fingerprint classification result;

[0152] The fingerprint to be repaired determination module 1120 can be used to determine whether the fingerprint to be recognized is a fingerprint to be repaired according to the fingerprint classification result;

[0153] The second fingerprint repair module 1130 can be used to, when determining that the fingerprint to be recognized is a fingerprint to be repaired, perform fingerprint repair on the fingerprint to be recognized based on a preset fingerprint repair model to obtain a standard repaired fingerprint; wherein, the fingerprint repair model is obtained by training a first feature extraction model through the training method of any one of the above fingerprint repair models;

[0154] The fingerprint recognition module 1140 can be used to recognize the standard repaired fingerprint based on a preset fingerprint recognition model to obtain a fingerprint recognition result.

[0155] In an exemplary embodiment of the present disclosure, the fingerprint recognition device further includes:

[0156] The fingerprint classification model training module can be used to obtain a third standard fingerprint and a second actual dry fingerprint, and use the third standard fingerprint and the second actual dry fingerprint to train a classification network model to be trained to obtain the preset fingerprint classification model.

[0157] In an exemplary embodiment of the present disclosure, training the classification network model to be trained with the third standard fingerprint and the second actual dry fingerprint to obtain the preset fingerprint classification model includes:

[0158] Inputting the third standard fingerprint and the second actual dry fingerprint into the classification network model to be trained to obtain a fifth output result;

[0159] Constructing a fourth loss function according to the fifth output result, the third label of the third standard fingerprint, and the fourth label of the second actual dry fingerprint;

[0160] Adjusting the parameters included in the classification network model to be trained based on the fourth loss function to obtain the preset fingerprint classification model.

[0161] In an exemplary embodiment of the present disclosure, the classification network model to be trained is a binary classification network model, and the classification network model to be trained includes any one of a ResNet model, a VGG model, and a DenseNet model.

[0162] The specific details of each module in the above fingerprint repair model training device and fingerprint recognition device have been described in detail in the corresponding fingerprint recognition model training method and fingerprint recognition method, and thus will not be elaborated herein.

[0163] It should be noted that although several modules or units of a 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 disclosure, the features and functions of the two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0164] In addition, although the steps of the methods in the present disclosure are described in a specific order in the drawings, this does not require or imply that these steps must be executed in that specific order, or that all the steps shown must be executed 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.

[0165] In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.

[0166] Those skilled in the art to which the present disclosure pertains can understand that various aspects of the present disclosure can be implemented as a system, a method, or a program product. Therefore, various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation manner, a complete software implementation manner (including firmware, microcode, etc.), or an implementation manner combining hardware and software aspects, which can be collectively referred to herein as "circuit", "module", or "system".

[0167] The following refers to Figure 12 to describe the electronic device 1200 according to this embodiment of the present disclosure. Figure 12 The shown electronic device 1200 is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present disclosure.

[0168] As Figure 12As shown, the electronic device 1200 is presented in the form of a general-purpose computing device. The components of the electronic device 1200 may include, but are not limited to: at least one of the above-mentioned processing units 1210, at least one of the above-mentioned storage units 1220, a bus 1230 connecting different system components (including the storage unit 1220 and the processing unit 1210), and a display unit 1240.

[0169] Among them, the storage unit stores program code, and the program code can be executed by the processing unit 1210, so that the processing unit 1210 executes the steps according to various exemplary embodiments of the present disclosure described in the "Exemplary Method" section above of this specification. For example, the processing unit 1210 can execute steps such as Figure 1 shown in: Step S110: Obtain a first standard fingerprint, and input the first standard fingerprint into a preset dry fingerprint generation model to obtain a first simulated dry fingerprint; Step S120: Use a first feature extraction model to extract the first local fingerprint feature of the first simulated dry fingerprint, and use the first local fingerprint feature to repair the first simulated dry fingerprint to obtain a first output result; Step S130: Construct a first loss function according to the first output result and the first standard fingerprint, and use the first loss function to adjust the parameters in the first feature extraction model to obtain a fingerprint repair model.

[0170] Again, for example, the processing unit 1210 can execute steps such as Figure 9 shown in: Step S910: Obtain a fingerprint to be recognized, and classify the fingerprint to be recognized based on a preset fingerprint classification model to obtain a fingerprint classification result; Step S920: Determine whether the fingerprint to be recognized is a fingerprint to be repaired according to the fingerprint classification result; Step S930: When it is determined that the fingerprint to be recognized is a fingerprint to be repaired, repair the fingerprint to be recognized based on a preset fingerprint repair model to obtain a standard repaired fingerprint; wherein, the fingerprint repair model is obtained by training the first feature extraction model through the training method of the fingerprint repair model described above; Step S940: Recognize the standard repaired fingerprint based on a preset fingerprint recognition model to obtain a fingerprint recognition result.

[0171] The storage unit 1220 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 12201 and / or a cache storage unit 12202, and may further include a read-only storage unit (ROM) 12203.

[0172] The storage unit 1220 may also include a program / utilities 12204 having a set (at least one) of program modules 12205. Such program modules 12205 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment.

[0173] The bus 1230 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus structures.

[0174] The electronic device 1200 may also communicate with one or more external devices 1300 (such as a keyboard, a pointing device, a Bluetooth device, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 1200, and / or may communicate with any device that enables the electronic device 1200 to communicate with one or more other computing devices (such as a router, a modem, etc.). Such communication may be carried out through the input / output (I / O) interface 1250. Moreover, the electronic device 1200 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 1260. As shown in the figure, the network adapter 1260 communicates with other modules of the electronic device 1200 through the bus 1230. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 1200, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0175] Through the description of the above embodiments, those skilled in the art can easily understand that the example embodiments described herein can be implemented by software, or can be implemented by a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure 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, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0176] In an exemplary embodiment of the present disclosure, there is also provided a computer-readable storage medium, on which a program product capable of implementing the above methods in this specification is stored. In some possible implementation manners, various aspects of the present disclosure 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 cause the terminal device to execute the steps according to various exemplary embodiments of the present disclosure described in the above "Exemplary Method" section of this specification.

[0177] The program product for implementing the above method according to an embodiment of the present disclosure may be a portable compact disc read-only memory (CD-ROM) and includes program code, and can run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto. In this document, the readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0178] The program product may adopt any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable 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 disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0179] The computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries the readable program code. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable signal medium may also be any readable medium other than the readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0180] The program code contained on the readable medium can be transmitted by any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination of the above.

[0181] Program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and also including conventional procedural programming languages such as the "C" language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user's computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0182] In addition, the above-mentioned drawings are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, rather than for limiting purposes. It is easy to understand that the processes shown in the above-mentioned drawings do not indicate or limit the chronological order of these processes. Additionally, it is also easy to understand that these processes may be executed, for example, synchronously or asynchronously in multiple modules.

[0183] Other embodiments of the present disclosure will be readily envisioned by those of ordinary skill in the art upon consideration of the specification and practice of the invention herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include known common knowledge or conventional technical means in the art that are not invented by the present disclosure. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of the present disclosure are pointed out by the claims.

Claims

1. A training method for a fingerprint restoration model, characterized in that, it includes: Obtain a first standard fingerprint, and input the first standard fingerprint into a preset dry fingerprint generation model to obtain a first simulated dry fingerprint; Use a first feature extraction model to extract the first local fingerprint features of the first simulated dry fingerprint, and use the first local fingerprint features to perform fingerprint restoration on the first simulated dry fingerprint to obtain a first output result; Construct a first loss function according to the first output result and the first standard fingerprint, and use the first loss function to adjust the parameters in the first feature extraction model to obtain a fingerprint restoration model; Among them, the preset dry fingerprint generation model is obtained by the following method: input a second standard fingerprint into a second feature extraction model to obtain a second output result, and configure a first label and a second label for the first actual dry fingerprint and the second output result respectively; input the first actual dry fingerprint and the second output result into a standard stacked classification network respectively to obtain a third output result corresponding to the first actual dry fingerprint and a fourth output result corresponding to the second output result; construct a second loss function corresponding to the standard stacked classification network according to the third output result and the fourth output result, and construct a third loss function corresponding to the second feature extraction model according to the second output result and the fourth output result; construct an adversarial generation training loss function according to the second loss function and the third loss function, and use the adversarial generation training loss function to train the second feature extraction model and the standard stacked classification network to obtain the preset dry fingerprint generation model and a fingerprint discrimination model.

2. The training method for a fingerprint restoration model according to claim 1, characterized in that, The first feature extraction model includes multiple layers of first convolutional layers, multiple layers of first downsampling layers, and multiple layers of first upsampling layers, and the number of layers of the multiple layers of first downsampling layers and the multiple layers of first upsampling layers is the same; Among them, using a first feature extraction model to extract the first local fingerprint features of the first simulated dry fingerprint, and using the first local fingerprint features to perform fingerprint restoration on the first simulated dry fingerprint to obtain a first output result includes: Perform convolutional processing on the first simulated dry fingerprint using the multiple layers of first convolutional layers included in the first feature extraction model to obtain multiple local features of different scales; Use the multiple layers of first downsampling layers to perform downsampling on the multiple local features of different scales to obtain regional features of multiple local regions; Use the multiple layers of first upsampling layers to fill the regional features of the multiple local regions into the fingerprint missing regions in the first simulated dry fingerprint to obtain the first output result.

3. The training method for a fingerprint restoration model according to claim 1, characterized in that, Constructing a first loss function according to the first output result and the first standard fingerprint includes: where loss is the first loss function, and I gt is the first standard fingerprint, is the first output result; is the absolute value between the first standard fingerprint and the first output result.

4. The training method for a fingerprint restoration model according to claim 1, characterized in that, Construct a second loss function corresponding to the standard stacked classification network according to the third output result and the fourth output result, including: Among them, V 1 (D, G) is the second loss function corresponding to the standard stacked classification network, is to randomly sample the dataset data composed of the first actual dry fingerprints, x is the first actual dry fingerprint, D(x) is the third output result, D(G(z)) is the fourth output result, G(z) is the second output result, and z is the second standard fingerprint; is to randomly sample the dataset composed of the second standard fingerprints.

5. The training method of the fingerprint repair model according to claim 1, wherein, Construct a third loss function corresponding to the second feature extraction model according to the second output result and the fourth output result, including: Among them, V 2 (D, G) is the third loss function corresponding to the second feature extraction model, is any sampling of the dataset composed of the second standard fingerprints, D(G(z)) is the fourth output result, G(z) is the second output result, z is the second standard fingerprint, ||z - G(z)|| 1 is the absolute value between the second standard fingerprint and the second output result.

6. A fingerprint recognition method, wherein, including: Obtain a fingerprint to be recognized, and classify the fingerprint to be recognized based on a preset fingerprint classification model to obtain a fingerprint classification result; Determine whether the fingerprint to be recognized is a fingerprint to be repaired according to the fingerprint classification result; When it is determined that the fingerprint to be recognized is a fingerprint to be repaired, perform fingerprint repair on the fingerprint to be recognized based on a preset fingerprint repair model to obtain a standard repaired fingerprint; wherein, the fingerprint repair model is obtained by training a first feature extraction model through the training method of the fingerprint repair model according to any one of claims 1-5; Recognize the standard repaired fingerprint based on a preset fingerprint recognition model to obtain a fingerprint recognition result.

7. The fingerprint recognition method according to claim 6, wherein, Before classifying the fingerprint to be recognized based on a preset fingerprint classification model to obtain a fingerprint classification result, the fingerprint recognition method further includes: Obtain a third standard fingerprint and a second actual dry fingerprint, and use the third standard fingerprint and the second actual dry fingerprint to train a classification network model to be trained to obtain the preset fingerprint classification model.

8. The fingerprint recognition method according to claim 7, wherein, Using the third standard fingerprint and the second actual dry fingerprint to train a classification network model to be trained to obtain the preset fingerprint classification model, including: Input the third standard fingerprint and the second actual dry fingerprint into the classification network model to be trained to obtain a fifth output result; Construct a fourth loss function according to the fifth output result, the third label of the third standard fingerprint, and the fourth label of the second actual dry fingerprint; Adjust the parameters included in the classification network model to be trained based on the fourth loss function to obtain the preset fingerprint classification model.

9. The fingerprint recognition method according to claim 8, wherein, The classification network model to be trained is a binary classification network model, and the classification network model to be trained includes any one of a ResNet model, a VGG model, and a DenseNet model.

10. A computer-readable storage medium, on which a computer program is stored, wherein, When the computer program is executed by a processor, it implements the training method of the fingerprint repair model according to any one of claims 1-5, and the fingerprint recognition method according to any one of claims 6-9.

11. An electronic device, wherein, including: a processor; and a memory for storing executable instructions of the processor; Wherein, the processor is configured to execute the training method of the fingerprint repair model according to any one of claims 1-5 and the fingerprint recognition method according to any one of claims 6-9 by executing the executable instructions.

Images