A finger vein recognition method and device, electronic equipment and storage medium

Abstract

The application provides a finger vein recognition method and device, electronic equipment and storage medium. The method comprises the following steps: obtaining a to-be-recognized image; performing feature extraction on the to-be-recognized image through a preset finger vein feature extraction model to obtain a finger vein feature; wherein the finger vein feature extraction model comprises a nonlinear feature extraction network; the nonlinear feature extraction network is obtained based on a nonlinear convolution kernel; performing finger vein recognition according to the finger vein feature to obtain a recognition result. The nonlinear feature network with less to-be-optimized parameters is used to replace the conventional linear feature network, thereby reducing the calculation cost and training time and making the model easier to train. Moreover, the nonlinear convolution network is applied to the finger vein recognition, the nonlinear parameter dependent convolution kernel provides additional nonlinear features for the finger vein recognition, the feature extraction capability of the network is stronger, and the recognition accuracy is higher.

Description

Technical Field

[0001] This application relates to the field of computer vision, and more specifically, to a finger vein recognition method, device, electronic device, and storage medium. Background Technology

[0002] With the development of deep learning technology, deep learning has been widely applied in fields such as computer vision, language processing, and image semantic analysis. Finger vein recognition extracts features through deep convolutional networks, but as the number of layers and parameters in convolutional neural networks continues to increase, the computational cost also increases, resulting in low inference efficiency for convolutional neural networks. Summary of the Invention

[0003] The purpose of this invention is to provide a server routing method, apparatus, electronic device, and storage medium that effectively reduces the number of parameters and improves efficiency by replacing conventional convolution kernels with convolution kernels of nonlinear functions.

[0004] In a first aspect, embodiments of this application provide a finger vein recognition method, comprising: obtaining an image to be recognized; extracting features from the image to be recognized using a preset finger vein feature extraction model to obtain finger vein features; wherein the finger vein feature extraction model includes a nonlinear feature extraction network; the nonlinear feature extraction network is obtained based on a nonlinear convolution kernel; and performing finger vein recognition based on the finger vein features to obtain a recognition result.

[0005] In the above implementation process, a pre-defined finger vein feature extraction model is used to extract features from the image to be recognized. This model includes a nonlinear feature extraction network. By replacing the conventional linear feature network with a nonlinear feature network that has fewer parameters to be optimized, computational costs and training time are reduced, making the model easier to train. Furthermore, the application of a nonlinear convolutional network to finger vein recognition, with its nonlinear parameters dependent on the convolutional kernel, provides additional nonlinear features for finger vein recognition, enhancing the network's feature extraction capabilities and improving recognition accuracy.

[0006] Optionally, in this embodiment of the application, before extracting features from the image to be identified using a preset finger vein feature extraction model to obtain finger vein features, the method further includes: obtaining a preset nonlinear function; the nonlinear function includes a first independent variable and a second independent variable; generating a nonlinear convolution kernel based on the nonlinear function, a first value of the first independent variable, and a second value of the second independent variable; and generating a finger vein feature extraction model based on the nonlinear convolution kernel.

[0007] In the above implementation process, a nonlinear convolution kernel is generated by a preset nonlinear function, which in turn constitutes a nonlinear feature extraction network. The nonlinear feature extraction network with fewer parameters to be optimized is trained, which reduces the training time of the model and improves efficiency. At the same time, it provides additional nonlinear features for finger vein recognition, making the network's feature extraction ability stronger and the recognition accuracy higher.

[0008] Optionally, in this embodiment, generating a finger vein feature extraction model based on a nonlinear convolution kernel includes: obtaining a preset feature extraction network; determining a convolutional layer to be replaced from the feature extraction network, the convolutional layer to be replaced including the original convolutional kernel; replacing the original convolutional kernel in the convolutional layer to be replaced with a nonlinear convolutional kernel; replacing the original convolutional kernel in the convolutional layer to be replaced with a nonlinear convolutional kernel to obtain a nonlinear feature extraction network; and training the nonlinear feature extraction network with pre-collected training data to generate a finger vein feature extraction model.

[0009] In the above implementation process, a nonlinear feature extraction network is obtained based on the feature extraction network and the nonlinear convolution kernel. The nonlinear feature extraction network with fewer parameters to be optimized is trained to reduce the model training time and improve the model generation efficiency.

[0010] Optionally, in this embodiment of the application, feature extraction of the image to be identified is performed using a preset finger vein feature extraction model to obtain finger vein features, including: calculating the linear convolution kernel corresponding to the nonlinear convolution kernel in the finger vein feature extraction model according to a nonlinear function; obtaining a linear finger vein feature extraction model based on the linear convolution kernel; and extracting features of the image to be identified using the linear finger vein feature extraction model to obtain finger vein features.

[0011] In the above implementation process, a linear finger vein feature extraction model is obtained by calculating the regular convolution kernel corresponding to each nonlinear convolution kernel and assigning the parameters of all layers to a preset feature extraction network. This reduces training time when the original model is used to process the image to be recognized.

[0012] Optionally, in this embodiment, the preset nonlinear function includes:

[0013]

[0014]

[0015] Among them, F oval (x,y,θ,λ,ψ,σ,γ) is a nonlinear function, where x is the first independent variable, y is the second independent variable, θ is the tilt angle, λ is the wavelength, ψ is the phase offset, σ is the standard deviation of the function, and γ is the aspect ratio. This is coordinate data.

[0016] In the above implementation process, the number of parameters that need to be trained in the model is reduced by using nonlinear functions, and the model is easier to train by using functions with simple structures, thereby improving the efficiency of model training. In addition, parameter-dependent convolutional kernels provide additional nonlinear features, making the network's feature extraction ability stronger and the recognition accuracy higher.

[0017] Optionally, in this embodiment, generating a nonlinear convolution kernel based on a nonlinear function, a first value of a first independent variable, and a second value of a second independent variable includes: determining pixel coordinate data based on the first value of the first independent variable and the second value of the second independent variable; and inputting the pixel coordinate data into a nonlinear function to generate a nonlinear convolution kernel.

[0018] In the above implementation process, the parameters to be optimized in the conventional convolution kernel are represented by a nonlinear function with fewer parameters, resulting in a nonlinear convolution kernel. This provides additional nonlinear features for finger vein recognition, improving the model's recognition ability and accuracy.

[0019] Optionally, in this embodiment of the application, finger vein recognition is performed based on finger vein features to obtain recognition results, including: obtaining corresponding registration features based on finger vein features; obtaining distance data between finger vein features and registration features through a metric algorithm; and obtaining recognition results based on the distance data and a preset threshold.

[0020] In the above implementation process, conventional convolutional networks are relatively large with a large number of parameters. When applied to finger vein recognition tasks, this parameter redundancy makes the model difficult to train and computationally expensive. Nonlinear feature networks reduce computational costs and training time, making the model easier to train. Furthermore, applying nonlinear convolutional networks to finger vein recognition, with nonlinear parameters dependent on the convolutional kernel, provides additional nonlinear features for finger vein recognition, enhancing the network's feature extraction capabilities and improving recognition accuracy.

[0021] Secondly, embodiments of this application also provide a finger vein recognition device, comprising: an image acquisition module for acquiring an image to be recognized; a feature extraction module for extracting features from the image to be recognized using a preset finger vein feature extraction model to obtain finger vein features; wherein the finger vein feature extraction model includes a nonlinear feature extraction network; the nonlinear feature extraction network is obtained based on a nonlinear convolution kernel; and a recognition module for recognizing finger veins based on the finger vein features to obtain a recognition result.

[0022] Thirdly, embodiments of this application also provide an electronic device, including: a processor and a memory, the memory storing machine-readable instructions executable by the processor, which, when executed by the processor, perform the method described above.

[0023] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the methods described above.

[0024] The finger vein recognition method, apparatus, electronic device, and storage medium provided in this application extract features from the image to be recognized using a pre-defined finger vein feature extraction model, which includes a nonlinear feature extraction network. By replacing the conventional linear feature network with a nonlinear feature network that has fewer parameters to be optimized, computational costs and training time are reduced, making the model easier to train. Furthermore, the application of a nonlinear convolutional network to finger vein recognition, with its nonlinear parameters dependent on the convolutional kernel, provides additional nonlinear features for finger vein recognition, enhancing the network's feature extraction capabilities and improving recognition accuracy. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A flowchart illustrating a finger vein recognition method provided in an embodiment of this application;

[0027] Figure 2 This is a schematic diagram of a first alternative convolutional layer provided in an embodiment of this application;

[0028] Figure 3 This is a schematic diagram of a second alternative convolutional layer provided in an embodiment of this application;

[0029] Figure 4 This is a partial schematic diagram of a curved surface provided in an embodiment of this application;

[0030] Figure 5 This is a schematic diagram of the structure of the finger vein recognition device provided in the embodiments of this application;

[0031] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0032] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0033] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this application.

[0034] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0035] Please see Figure 1 The illustration shows a flowchart of a finger vein recognition method provided in an embodiment of this application. The finger vein recognition method provided in this application can be applied to electronic devices, which may include a terminal and a server; the terminal may specifically be a smartphone, tablet computer, computer, personal digital assistant (PDA), etc.; the server may specifically be an application server or a web server. The finger vein recognition method may include the following steps:

[0036] Step S110: Obtain the image to be recognized.

[0037] The image to be identified is an image that includes finger vein information. Specifically, it can be acquired through a finger vein acquisition device or a fingerprint acquisition device, or it can be retrieved from a pre-stored image to be identified from a storage device.

[0038] Step S120: Extract features from the image to be identified using a preset finger vein feature extraction model to obtain finger vein features; wherein, the finger vein feature extraction model includes a nonlinear feature extraction network; the nonlinear feature extraction network is obtained based on a nonlinear convolution kernel.

[0039] The image to be identified is input into a pre-defined finger vein feature extraction model. The model extracts features from the image to obtain the corresponding finger vein features. The finger vein feature extraction model includes a non-linear feature extraction network, which is trained using pre-defined sample set data and test set data to obtain the finger vein feature extraction model.

[0040] The nonlinear feature extraction network is generated based on a nonlinear convolutional kernel. This kernel is determined by a pre-defined nonlinear function, and its region size can be determined by the value of the independent variable in the nonlinear function. Specifically, for example, the convolution of a predetermined region in a specified layer of a pre-defined convolutional network is replaced with the aforementioned nonlinear convolutional kernel to obtain the nonlinear feature extraction network. The pre-defined convolutional network can be any convolutional neural network, such as SqueezeNet, ShuffleNet, or MobileNet.

[0041] Step S130: Perform finger vein recognition based on finger vein characteristics and obtain the recognition result.

[0042] The obtained finger vein features are used for finger vein recognition. For example, the finger vein features are retrieved in a preset database using distance metrics, or the extracted finger vein features are matched with pre-stored finger vein features to obtain recognition results. The recognition results can be the degree of consistency of the matching, or they can directly represent whether the matching results are consistent.

[0043] In the above implementation process, a pre-defined finger vein feature extraction model is used to extract features from the image to be recognized. This model includes a nonlinear feature extraction network. Conventional convolutional networks are relatively large with many parameters, leading to parameter redundancy in applications. Nonlinear feature networks have fewer parameters to be optimized, thus reducing computational cost and training time, making the model easier to train. Furthermore, applying the nonlinear convolutional network to finger vein recognition, with its nonlinear parameters dependent on the convolutional kernel, provides additional nonlinear features for finger vein recognition, enhancing the network's feature extraction capabilities and improving recognition accuracy.

[0044] Optionally, in this embodiment of the application, before extracting features from the image to be identified using a preset finger vein feature extraction model to obtain finger vein features, the method further includes: obtaining a preset nonlinear function; the nonlinear function includes a first independent variable and a second independent variable; generating a nonlinear convolution kernel based on the nonlinear function, a first value of the first independent variable, and a second value of the second independent variable; and generating a finger vein feature extraction model based on the nonlinear convolution kernel.

[0045] In the specific implementation process: the nonlinear function can be a predefined two-dimensional nonlinear function, such as the Schmid filter function or the product of a trigonometric function and a Gaussian function. A nonlinear function with better recognition performance can be selected based on a test method to generate the nonlinear convolution kernel. The nonlinear function can include two independent variables, a first independent variable and a second independent variable, which together determine the coordinates of a pixel in the image.

[0046] Obtain the first value of the first independent variable and the second value of the second independent variable. Substitute these values ​​into the nonlinear function to generate a nonlinear convolution kernel. Specifically, for example, determine multiple pixel coordinates based on the value ranges of the first and second independent variables. The value ranges of the first and second independent variables correspond to the local surface of the nonlinear function, and these multiple pixel coordinates constitute the nonlinear convolution kernel.

[0047] Based on the nonlinear convolutional kernels generated above, a nonlinear feature extraction network is generated. The nonlinear feature extraction network is then trained to generate a finger vein feature extraction model.

[0048] In the above implementation process, a nonlinear convolution kernel is generated by a preset nonlinear function, which in turn constitutes a nonlinear feature extraction network. Only the nonlinear feature extraction network with fewer parameters to be optimized needs to be trained, which reduces the training time of the model and improves efficiency. At the same time, it provides additional nonlinear features for finger vein recognition, improves the feature extraction capability of the network, and improves the recognition accuracy of the model.

[0049] Optionally, in this embodiment, generating a finger vein feature extraction model based on a nonlinear convolution kernel includes: obtaining a preset feature extraction network; determining a convolutional layer to be replaced from the feature extraction network, the convolutional layer to be replaced including the original convolutional kernel; replacing the original convolutional kernel in the convolutional layer to be replaced with a nonlinear convolutional kernel; replacing the original convolutional kernel in the convolutional layer to be replaced with a nonlinear convolutional kernel to obtain a nonlinear feature extraction network; and training the nonlinear feature extraction network with pre-collected training data to generate a finger vein feature extraction model.

[0050] The method for generating a finger vein feature extraction model based on a nonlinear convolution kernel includes determining the convolutional layer to be replaced from a preset feature extraction network. The convolutional layer to be replaced can be all layers of the feature extraction network or only some layers of the feature extraction network.

[0051] Please see Figure 2 The diagram shown is a schematic of the first alternative convolutional layer provided in the embodiments of this application;

[0052] Please see Figure 3 The diagram shown illustrates a second alternative convolutional layer provided in an embodiment of this application.

[0053] like Figure 2 As shown, if the preset feature extraction network is MobileNet, then the replacement convolutional layer can be the first 16 layers of MobileNet; for example... Figure 3As shown, if the preset feature extraction network is SqueezeNet, then the replacement convolutional layer can be any layer in SqueezeNet. The convolutional layer to be replaced includes the original convolutional kernel, which is the convolution in the layer to be replaced that needs to be replaced. The size of the original convolutional kernel is the same as the size of the non-linear convolutional kernel.

[0054] The original convolutional kernels in the convolutional layer to be replaced are replaced with non-linear convolutional kernels. Taking a non-linear convolutional kernel size of 3*3 and a feature extraction network of SqueezeNet as an example, this replacement step can be to replace the 3*3 convolutional kernels of all layers of SqueezeNet with 3*3 non-linear convolutional kernels, so as to obtain a non-linear feature extraction network based on the feature extraction network and the non-linear convolutional kernel.

[0055] A finger vein feature extraction model is generated by training a nonlinear feature extraction network using pre-collected training data. The training data can include sample data and test data. The nonlinear function is continuously differentiable, and the training process can utilize open-source frameworks to train the nonlinear feature extraction network.

[0056] In the above implementation process, a nonlinear feature extraction network is obtained based on the feature extraction network and the nonlinear convolution kernel. The nonlinear feature extraction network is then trained to reduce the model training time and improve the model generation efficiency.

[0057] Optionally, in this embodiment of the application, feature extraction of the image to be identified is performed using a preset finger vein feature extraction model to obtain finger vein features, including: calculating the linear convolution kernel corresponding to the nonlinear convolution kernel in the finger vein feature extraction model according to a nonlinear function; obtaining a linear finger vein feature extraction model based on the linear convolution kernel; and extracting features of the image to be identified using the linear finger vein feature extraction model to obtain finger vein features.

[0058] In the specific implementation process: After training the nonlinear feature extraction network, a finger vein feature extraction model is obtained, which includes a nonlinear convolution kernel. Based on the nonlinear function, the corresponding linear convolution kernel in the finger vein feature extraction model can be calculated. The linear convolution kernel can be a regular convolution kernel, which includes nine parameters to be optimized.

[0059] Based on the linear convolution kernel, a linear finger vein feature extraction model is obtained. Taking SqueezeNet as an example, the parameters of all layers of SqueezeNet are calculated based on the obtained linear convolution kernel, and the parameters of all layers are assigned to a new regular SqueezeNet to obtain the linear finger vein feature extraction model.

[0060] The linear finger vein feature extraction model can be deployed, transmitted, and inferred normally. It extracts finger vein features from the image to be recognized.

[0061] In one optional embodiment, the model can be applied in the following ways: First, for general-purpose AI (artificial intelligence) computing hardware, if the hardware has sufficient storage space, the finger vein feature extraction model or a linear finger vein feature extraction model can be deployed directly on the hardware. The model can then be used in the conventional manner for transmission, loading, and inference. Second, if the hardware storage space is limited, the finger vein feature extraction model is deployed on the hardware side. Each time the device starts, it is initialized to generate a linear finger vein feature extraction model. This linear model is then used to process the image to be recognized, and no further initialization is required during subsequent model use. The finger vein feature extraction model saves more space than the linear model; the additional initialization time is minimal and has no significant impact on usage; and transmission efficiency is higher when the model is remotely updated. Third, for directly programmable hardware, a flexible combination of non-linear and linear convolutional kernels can be selected to optimize cost, inference efficiency, and battery life.

[0062] In the above implementation process, a linear finger vein feature extraction model is obtained by calculating the regular convolution kernel corresponding to each nonlinear convolution kernel and assigning the parameters of all layers to a preset feature extraction network. This reduces training time when the original model is used to process the image to be recognized.

[0063] In one optional implementation, the preset nonlinear function includes:

[0064]

[0065]

[0066] Among them, F oval (x,y,θ,λ,ψ,σ,γ) is a nonlinear function, where x is the first independent variable, y is the second independent variable, θ is the tilt angle, λ is the wavelength, ψ is the phase offset, σ is the standard deviation of the function, and γ is the aspect ratio. This is coordinate data.

[0067] A conventional convolution kernel includes 9 parameters to be optimized, while a nonlinear function includes 5 parameters to be optimized, effectively reducing the number of parameters.

[0068] In the above implementation process, the number of parameters that need to be trained in the model is reduced by using nonlinear functions, and the model is easier to train by using functions with simple structures, thereby improving the efficiency of model training. In addition, parameter-dependent convolutional kernels provide additional nonlinear features, making the network's feature extraction ability stronger and the recognition accuracy higher.

[0069] Optionally, in this embodiment, generating a nonlinear convolution kernel based on a nonlinear function, a first value of a first independent variable, and a second value of a second independent variable includes: determining pixel coordinate data based on the first value of the first independent variable and the second value of the second independent variable; and inputting the pixel coordinate data into a nonlinear function to generate a nonlinear convolution kernel.

[0070] Please see Figure 4 The diagram shows a partial schematic of the curved surface provided in an embodiment of this application.

[0071] In the specific implementation process: First, obtain the first value of the first independent variable and the second value of the second independent variable. Then, determine the pixel coordinate data based on the first value of the first independent variable and the second value of the second independent variable.

[0072] The following is a specific example. The first independent variable x takes the value [-1, 0, 1]; the second independent variable y takes the value [-1, 0, 1]. We take nine coordinate points corresponding to the first independent variable x = [-1, 0, 1] and the second independent variable y = [-1, 0, 1], which are [-1, -1], ...

[0073] [-1, 0], [-1, 1], [0, -1], [0, 0], [0, 1], [1, -1], [1, 0], and [1, 1]. For example... Figure 2 As shown, the nine points in the figure constitute the nonlinear convolution kernel.

[0074] In the above implementation process, the parameters to be optimized in the conventional convolution kernel are represented by a nonlinear function with fewer parameters, resulting in a nonlinear convolution kernel. This provides additional nonlinear features for finger vein recognition, improving the model's recognition ability and accuracy.

[0075] Optionally, in this embodiment, finger vein identification is performed based on finger vein characteristics.

[0076] Obtaining the recognition result includes: obtaining the corresponding registration feature based on the finger vein feature; obtaining the distance data between the finger vein feature and the registration feature through a metric algorithm; and obtaining the recognition result based on the distance data and a preset threshold.

[0077] In the specific implementation process: Based on the finger vein feature, the corresponding registration feature is obtained. For example, based on the finger vein feature, a search is performed in a pre-set registration feature database to obtain the registration feature that is closest to the finger vein feature. The closest registration feature is the registration feature in the registration feature database that is most similar to the finger vein feature.

[0078] The distance data between finger vein features and registered features is obtained through a metric algorithm. Based on the distance data and a preset threshold, the recognition result is obtained. It is understandable that the process of comparing finger vein features with registered features in this step can also be implemented using a recognition model or other methods.

[0079] The recognition results can characterize whether the finger vein feature is consistent with the registered feature, and can also indicate the degree of consistency between the finger vein feature and the registered feature. For example, if the distance data is less than a preset threshold, it is considered that the finger corresponding to the finger vein feature is the same finger as the finger corresponding to the found registered feature; if the distance data is greater than the preset threshold, it is considered that the finger corresponding to the finger vein feature is not the same finger as the finger corresponding to the found registered feature, so as to complete the finger vein recognition.

[0080] In an optional embodiment, a finger vein device identifies the user's finger veins to determine if they match pre-stored finger vein features, thereby performing identity verification. The finger vein device may include a light source, an image acquisition device, a feature extraction network, and a calculation unit for a feature distance metric. The light source and image acquisition device are used to acquire images containing finger vein information; the feature extraction network is used to extract finger vein features; and the calculation unit is used to compare the extracted features with registered features using a feature distance metric to obtain the recognition result.

[0081] Before performing finger vein recognition, a feature extraction network and a feature distance metric computation unit need to be deployed on the hardware. Then, the registration mode is activated. The registration module is used to pre-store finger vein features. For example, when a user places their finger in a designated position, the image acquisition device is triggered. After acquiring an image to be registered, including finger vein features, the feature extraction network is run to extract features from the image to be registered, obtain registration features, and save the registration features to the data storage area and automatically number them.

[0082] After registering the finger vein features, an operating mode can be implemented. For example, when the user places their finger in a designated position, the image acquisition device is triggered, acquires the image to be identified, and extracts features through the network to obtain the finger vein features.

[0083] The feature distance metric calculation unit calculates the nearest registered feature in the data storage area to the finger vein feature based on the finger vein feature. A metric algorithm then obtains the distance data between the finger vein feature and the registered feature. Based on the distance data and a preset threshold, a recognition result is obtained. Specifically, if the distance data is less than the preset threshold, the finger corresponding to the finger vein feature is considered to be the same finger as the finger corresponding to the found registered feature; if the distance data is greater than the preset threshold, the finger corresponding to the finger vein feature is considered not to be the same finger as the finger corresponding to the found registered feature, thus achieving identity verification.

[0084] In the above implementation process, conventional convolutional networks are relatively large with a large number of parameters. When applied to finger vein recognition tasks, this parameter redundancy leads to difficulties in model training and excessively high computational costs. Nonlinear feature networks reduce computational costs and training time, making the model easier to train. Furthermore, applying nonlinear convolutional networks to finger vein recognition, with nonlinear parameters dependent on the convolutional kernel, provides additional nonlinear features for finger vein recognition, enhancing the network's feature extraction capabilities and improving recognition accuracy.

[0085] Please see Figure 5 The diagram shown is a structural schematic of a finger vein recognition device provided in an embodiment of this application; this application provides a finger vein recognition device 200, including:

[0086] Image acquisition module 210 is used to acquire the image to be recognized;

[0087] The feature extraction module 220 is used to extract features from the image to be recognized using a preset finger vein feature extraction model to obtain finger vein features; wherein, the finger vein feature extraction model includes a nonlinear feature extraction network; the nonlinear feature extraction network is obtained based on a nonlinear convolution kernel;

[0088] The recognition module 230 is used to perform finger vein recognition based on finger vein characteristics and obtain recognition results.

[0089] Optionally, in this embodiment of the application, the finger vein recognition device further includes: a model generation module, used to obtain a preset nonlinear function; the nonlinear function includes a first independent variable and a second independent variable; a nonlinear convolution kernel is generated based on the nonlinear function, a first value of the first independent variable and a second value of the second independent variable; and a finger vein feature extraction model is generated based on the nonlinear convolution kernel.

[0090] Optionally, in this embodiment of the application, the finger vein recognition device and the model generation module are further configured to obtain a preset feature extraction network; determine a convolutional layer to be replaced from the feature extraction network, the convolutional layer to be replaced including the original convolutional kernel; replace the original convolutional kernel in the convolutional layer to be replaced with a nonlinear convolutional kernel; replace the original convolutional kernel in the convolutional layer to be replaced with a nonlinear convolutional kernel to obtain a nonlinear feature extraction network; and train the nonlinear feature extraction network with pre-collected training data to generate a finger vein feature extraction model.

[0091] Optionally, in this embodiment of the application, the finger vein recognition device, feature extraction module 220, is specifically used to calculate the linear convolution kernel corresponding to the nonlinear convolution kernel in the finger vein feature extraction model according to the nonlinear function; obtain the linear finger vein feature extraction model according to the linear convolution kernel; and extract features from the image to be recognized through the linear finger vein feature extraction model to obtain finger vein features.

[0092] Optionally, in this embodiment of the application, the finger vein recognition device includes a preset nonlinear function comprising:

[0093]

[0094] Among them, F oval (x,y,θ,λ,ψ,σ,γ) is a nonlinear function, where x is the first independent variable, y is the second independent variable, θ is the tilt angle, λ is the wavelength, ψ is the phase offset, σ is the standard deviation of the function, and γ is the aspect ratio. This is coordinate data.

[0095] Optionally, in this embodiment of the application, the finger vein recognition device and the model generation module are further configured to determine pixel coordinate data based on the first value of the first independent variable and the second value of the second independent variable; and input the pixel coordinate data into a nonlinear function to generate a nonlinear convolution kernel.

[0096] Optionally, in this embodiment of the application, the finger vein recognition device, the recognition module 230, is specifically used to obtain the corresponding registration feature based on the finger vein feature; obtain the distance data between the finger vein feature and the registration feature through a measurement algorithm; and obtain the recognition result based on the distance data and a preset threshold.

[0097] It should be understood that this device corresponds to the above-described finger vein recognition method embodiment and is capable of performing the various steps involved in the above method embodiment. The specific functions of this device can be found in the description above, and detailed descriptions are omitted here to avoid repetition. The device includes at least one software functional module that can be stored in memory or embedded in the device's operating system (OS) in the form of software or firmware.

[0098] Please see Figure 6 The diagram shows a structural schematic of an electronic device provided in an embodiment of this application. An electronic device 300 provided in this application includes a processor 310 and a memory 320. The memory 320 stores machine-readable instructions executable by the processor 310. When the machine-readable instructions are executed by the processor 310, the method described above is performed.

[0099] This application also provides a storage medium storing a computer program, which is executed by a processor to perform the above-described method.

[0100] The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0101] It should be understood that the disclosed apparatus and methods can also be implemented in other ways, given the several embodiments provided in this application. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0102] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0103] The above description is only an optional implementation of the embodiments of this application, but the protection scope of the embodiments of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the embodiments of this application should be covered within the protection scope of the embodiments of this application.

Claims

1. A method for recognizing finger veins, characterized in that, include: Obtain the image to be recognized; The image to be identified is subjected to feature extraction using a preset finger vein feature extraction model to obtain finger vein features; wherein, the finger vein feature extraction model includes a nonlinear feature extraction network; the nonlinear feature extraction network is obtained based on a nonlinear convolution kernel; Finger vein identification is performed based on the described finger vein characteristics to obtain identification results; Before extracting features from the image to be identified using a preset finger vein feature extraction model to obtain finger vein features, the method further includes: Obtain a preset nonlinear function; the nonlinear function includes a first independent variable and a second independent variable; The nonlinear convolution kernel is generated based on the nonlinear function, the first value of the first independent variable, and the second value of the second independent variable. The finger vein feature extraction model is generated based on the nonlinear convolution kernel. The nonlinear convolution kernel is generated based on the nonlinear function, the first value of the first independent variable, and the second value of the second independent variable, including: Pixel coordinate data are determined based on the first value of the first independent variable and the second value of the second independent variable; The pixel coordinate data is input into the nonlinear function to generate the nonlinear convolution kernel.

2. The method according to claim 1, characterized in that, The finger vein feature extraction model is generated based on the nonlinear convolution kernel, including: Obtain the preset feature extraction network; The convolutional layer to be replaced is determined from the feature extraction network, the convolutional layer to be replaced including the original convolutional kernel; The original convolutional kernel in the convolutional layer to be replaced is replaced by the nonlinear convolutional kernel; The original convolutional kernel in the convolutional layer to be replaced is replaced with the nonlinear convolutional kernel to obtain a nonlinear feature extraction network; The nonlinear feature extraction network is trained using pre-collected training data to generate the finger vein feature extraction model.

3. The method according to claim 1, characterized in that, The step of extracting features from the image to be identified using a preset finger vein feature extraction model to obtain finger vein features includes: Calculate the linear convolution kernel corresponding to the nonlinear convolution kernel in the finger vein feature extraction model based on the nonlinear function; Based on the linear convolution kernel, a linear finger vein feature extraction model is obtained; The linear finger vein feature extraction model is used to extract features from the image to be identified, thereby obtaining the finger vein features.

4. The method according to claim 1, characterized in that, in, The preset nonlinear functions include: in, It is a nonlinear function. As the first independent variable, It is the second independent variable; The tilt angle, For wavelength, This is the phase offset. Let the standard deviation of the function be . Aspect ratio, This is coordinate data.

5. The method according to any one of claims 1-4, characterized in that, Finger vein recognition is performed based on the described finger vein characteristics to obtain recognition results, including: Based on the described finger vein features, the corresponding registration features are obtained; The distance data between the finger vein feature and the registered feature is obtained through a metric algorithm; The recognition result is obtained based on the distance data and the preset threshold.

6. A finger vein recognition device, characterized in that, include: The image acquisition module is used to acquire the image to be recognized; The feature extraction module is used to extract features from the image to be identified using a preset finger vein feature extraction model to obtain finger vein features; wherein, the finger vein feature extraction model includes a nonlinear feature extraction network; the nonlinear feature extraction network is obtained based on a nonlinear convolution kernel; The recognition module is used to recognize finger veins based on the finger vein characteristics and obtain recognition results; It also includes a model generation module for obtaining a preset nonlinear function; the nonlinear function includes a first independent variable and a second independent variable; the nonlinear convolution kernel is generated based on the nonlinear function, a first value of the first independent variable and a second value of the second independent variable; and the finger vein feature extraction model is generated based on the nonlinear convolution kernel. It is also used to determine pixel coordinate data based on the first value of the first independent variable and the second value of the second independent variable; and to input the pixel coordinate data into the nonlinear function to generate the nonlinear convolution kernel.

7. An electronic device, characterized in that, include: A processor and a memory, the memory storing machine-readable instructions executable by the processor, which, when executed by the processor, perform the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the method as described in any one of claims 1 to 5.

Images