An anti-counterfeiting identification system, method and related device supporting multiple image types

Abstract

The application discloses an anti-counterfeiting identification system and method supporting multiple image types and related devices. The system includes an image type discrimination module for discriminating the image type of an input image, which includes near-infrared images, visible light images, and binocular paired images; a network branch selection module for selecting the corresponding network branch in the anti-counterfeiting model according to the image type of the input image to perform anti-counterfeiting identification on the input image and obtain the live body category probability of the input image. The anti-counterfeiting identification system in the application can discriminate the image type of the input image, and then select the corresponding network branch according to the image type to perform anti-counterfeiting identification on the input image, supporting anti-counterfeiting identification of multiple image types, and realizing three input modes of customer self-selection of monocular near-infrared image input, monocular visible light image input, and binocular image paired input, thereby meeting the customer's multiple image input type requirements.

Description

Technical Field

[0001] The present application relates to the field of anti-counterfeiting identification technology, and in particular to an anti-counterfeiting identification system, method and related devices that support multiple image types. Background Art

[0002] The existing biometric anti-counterfeiting model structure only allows a single type of input image, such as only allowing near-infrared images or only allowing visible light images. However, the anti-counterfeiting model with a single type of input image is prone to the following deficiencies when processing downstream tasks: (1) A single type of image cannot provide comprehensive information about the target object and has the problem of limited feature expression ability; (2) A single type of image may be more sensitive to specific types of disturbances, has poor anti-interference ability, and is limited in application scenarios; (3) Since different machines are equipped with different cameras, if the model design only allows a single type of image input, it cannot simultaneously meet the customer's multiple customization needs. Summary of the Invention

[0003] The present application provides an anti-counterfeiting identification system, method and related devices that support multiple image types, which are used to solve the technical problem that the existing anti-counterfeiting model can only perform anti-counterfeiting identification on a single type of input image and cannot simultaneously meet the customer's multiple customized needs.

[0004] In view of this, the first aspect of the present application provides an anti-counterfeiting recognition system that supports multiple image types, including:

[0005] An image type discrimination module is used to discriminate the image type of the input image, including near-infrared image, visible light image and binocular paired image;

[0006] The network branch selection module is used to select a corresponding network branch in the anti-counterfeiting model according to the image type of the input image to perform anti-counterfeiting identification on the input image to obtain the living body category probability of the input image.

[0007] Optionally, the image type discrimination module includes:

[0008] a channel feature extraction submodule, configured to extract channel features of an input image and determine a first type of the input image according to the channel features of the input image;

[0009] a spectrum feature extraction submodule, configured to perform Fourier transform on an input image, calculate a spectrum amplitude to obtain spectrum features of the input image, and determine a second type of the input image based on the spectrum features of the input image;

[0010] The first discrimination subunit is configured to determine a final image type of the input image according to the first type and the second type of the input image.

[0011] Optionally, the image type discrimination module includes:

[0012] The second discrimination submodule is used to discriminate the image type of the input image through an image type discrimination model.

[0013] Optionally, the channel feature extraction submodule is specifically used to:

[0014] Extracting the number of channels of the input image to obtain channel features of the input image;

[0015] If the value of the channel feature of the input image is 1, it is determined that the input image is a near-infrared image;

[0016] If the value of the channel feature of the input image is 3, it is determined that the input image is a visible light image;

[0017] If the value of the channel feature of the input image is 4, it is determined that the input image is a binocular paired image.

[0018] Optionally, the spectrum feature extraction submodule is specifically used to:

[0019] Performing Fourier transform on the input image, setting the zero-frequency feature as the center of the spectrum, and calculating the spectrum amplitude to obtain the spectrum feature of the input image;

[0020] If the proportion of the spectral features of the input image in the low-frequency region exceeds a first threshold, determining that the input image is a near-infrared image;

[0021] If the proportion of the spectral features of the input image in the high-frequency region exceeds a second threshold, determining that the input image is a visible light image;

[0022] If the proportion of the spectral features of the input image in the low-frequency region exceeds a third threshold, and the proportion of the spectral features of the input image in the high-frequency region exceeds a fourth threshold, the input image is determined to be a binocular paired image.

[0023] Optionally, the training process of the image type discrimination model includes:

[0024] Obtain various types of face images;

[0025] Extracting channel features and spectral features of the face image;

[0026] The channel features and spectral features of the face image are used as input, and the image type of the face image is used as output to train the target network to obtain an image type discrimination model.

[0027] Optionally, the anti-counterfeiting model includes a first network branch, a second network branch and a third network branch, the first network branch includes a near-infrared feature extraction module and a fully connected layer, the second network branch includes a visible light feature extraction module and a fully connected layer; the third network branch includes a near-infrared feature extraction module, a visible light feature extraction module, a feature fusion module and a fully connected layer.

[0028] The second aspect of the present application provides an anti-counterfeiting identification method supporting multiple image types, including:

[0029] Determine the image type of the input image, which includes near-infrared image, visible light image and binocular paired image;

[0030] According to the image type of the input image, a corresponding network branch in the anti-counterfeiting model is selected to perform anti-counterfeiting identification on the input image to obtain a living body category probability of the input image.

[0031] A third aspect of the present application provides an electronic device, the device comprising a processor and a memory;

[0032] The memory is used to store program code and transmit the program code to the processor;

[0033] The processor is configured to execute the anti-counterfeiting identification method supporting multiple image types described in the second aspect according to instructions in the program code.

[0034] In a fourth aspect, the present application provides a computer-readable storage medium for storing program code. When the program code is executed by a processor, the anti-counterfeiting identification method supporting multiple image types described in the second aspect is implemented.

[0035] It can be seen from the above technical solutions that this application has the following advantages:

[0036] The anti-counterfeiting recognition system that supports multiple image types provided by this application can determine the image type of the input image, and then select the corresponding network branch to perform anti-counterfeiting recognition on the input image according to the image type. It supports anti-counterfeiting recognition of multiple image types and allows customers to independently choose from three input modes: monocular near-infrared image input, monocular visible light image input, and binocular image paired input, thereby meeting customers' needs for multiple image input types. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0038] Figure 1 A schematic diagram of the structure of an anti-counterfeiting recognition system supporting multiple image types provided in an embodiment of the present application;

[0039] Figure 2 A flowchart of an anti-counterfeiting identification method supporting multiple image types provided in an embodiment of the present application;

[0040] Figure 3 Another flowchart of an anti-counterfeiting identification method supporting multiple image types provided in an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to help those skilled in the art better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0042] For easier understanding, please refer to Figure 1 , the embodiment of the present application provides an anti-counterfeiting recognition system that supports multiple image types, including:

[0043] An image type discrimination module is used to discriminate the image type of the input image, including near-infrared image, visible light image and binocular paired image;

[0044] The network branch selection module is used to select the corresponding network branch in the anti-counterfeiting model according to the image type of the input image to perform anti-counterfeiting identification on the input image and obtain the living class probability of the input image.

[0045] In one embodiment, the image type discrimination module includes:

[0046] a channel feature extraction submodule, configured to extract channel features of an input image and determine a first type of the input image according to the channel features of the input image;

[0047] a spectrum feature extraction submodule, configured to perform Fourier transform on the input image, calculate the spectrum amplitude to obtain the spectrum feature of the input image, and determine the second type of the input image based on the spectrum feature of the input image;

[0048] The first discrimination subunit is configured to determine a final image type of the input image according to the first type and the second type of the input image.

[0049] Furthermore, the channel feature extraction submodule is specifically used to:

[0050] extracting a channel number of the input image to obtain a channel feature of the input image; if a value of the channel feature of the input image is 1, determining that the input image is a near-infrared image; if the value of the channel feature of the input image is 3, determining that the input image is a visible light image; if the value of the channel feature of the input image is 4, determining that the input image is a binocular paired image.

[0051] Further, the spectrum feature extraction submodule is specifically configured to:

[0052] performing Fourier transform on the input image, setting a zero-frequency feature as a spectrum center, and calculating a spectrum amplitude to obtain a spectrum feature of the input image;

[0053] if a proportion of the spectrum feature of the input image in a low-frequency region exceeds a first threshold value, determining that the input image is a near-infrared image;

[0054] if a proportion of the spectrum feature of the input image in a high-frequency region exceeds a second threshold value, determining that the input image is a visible light image;

[0055] if the proportion of the spectrum feature of the input image in the low-frequency region exceeds a third threshold value and the proportion of the spectrum feature of the input image in the high-frequency region exceeds a fourth threshold value, determining that the input image is a binocular paired image.

[0056] The spectrum feature extraction submodule performs Fourier transform on the input image, defines a zero-frequency feature as a spectrum center, and then calculates a spectrum amplitude to obtain a spectrum feature of the input image. If the spectrum feature of the input image is mainly concentrated in a low-frequency region, it is determined that the input image is a near-infrared image; if the spectrum feature of the input image is mainly concentrated in a high-frequency region, it is determined that the input image is a visible light image; if the spectrum feature of the input image shows a combination mode of high-frequency and low-frequency features, it is determined that the input image is a binocular paired image.

[0057] In another embodiment, the image type discrimination module comprises:

[0058] The second discrimination submodule is configured to discriminate the image type of the input image through the image type discrimination model.

[0059] The image type discrimination model is trained by taking the channel feature and the spectrum feature of the face image as input and taking the image type of the face image as output.

[0060] a plurality of types of face images are obtained;

[0061] the channel feature and the spectrum feature of the face image are extracted;

[0062] the image type discrimination model is obtained by training a target network by taking the channel feature and the spectrum feature of the face image as input and taking the image type of the face image as output.

[0063] Acquire multiple facial images of three image types: near-infrared, visible light, and binocular pairs. Extract the channel and spectral features of each type of facial image. Train a target network using these features as input and the image type of the facial image as output. Calculate a loss value based on the predicted image category output by the target network and the true image category. Update the target network parameters using this loss value until the target network converges. This trained target network is then used as the image type discrimination model. The target network can employ an existing convolutional neural network.

[0064] The image type discrimination model is used to discriminate the image type of the input image and output the predicted probabilities of each image type. The sum of the predicted probabilities of the three image types is 1. The image type with the highest predicted probability is selected as the category of the current input image. This output category determines the network branch path selected in the next stage of the network branch selection module.

[0065] After the image type discrimination module determines the image type of the input image, it sends the image type of the input image to the network branch selection module. The network branch selection module selects the corresponding network branch in the anti-counterfeiting model based on the image type of the input image to perform anti-counterfeiting identification on the input image and obtain the probability of the input image being a living object. The anti-counterfeiting model includes a first network branch, a second network branch, and a third network branch. The first network branch includes a near-infrared feature extraction module and a fully connected layer, the second network branch includes a visible light feature extraction module and a fully connected layer, and the third network branch includes a near-infrared feature extraction module, a visible light feature extraction module, a feature fusion module, and a fully connected layer. When the image type of the input image is a near-infrared image, the network branch selection module selects the first network branch to perform anti-counterfeiting identification on the input image; when the image type of the input image is a visible light image, the network branch selection module selects the second network branch to perform anti-counterfeiting identification on the input image; when the image type of the input image is a binocular paired image, the network branch selection module selects the third network branch to perform anti-counterfeiting identification on the input image.

[0066] For details, please refer to Figure 2 If the image type of the current input image is a near-infrared image, the network branch selection module closes all paths ① in the entire network path, and paths ② and ③ are disconnected. The near-infrared image is input to the feature extraction module to extract near-infrared image features. After deep learning feature training operations and passing through the fully connected layer, the final image is output with the probability of living body category.

[0067] If the current input image type is a visible light image, the network branch selection module closes all paths ② in the entire network path, leaving paths ① and ③ disconnected. The visible light image is input to the feature extraction module to extract visible light image features. After deep learning feature training, it passes through the fully connected layer and outputs the final image liveness category probability.

[0068] If the image type of the current input image is a binocular paired image, the network branch selection module closes all paths ③ in the entire network path, and paths ① and ② are disconnected. The binocular paired image is input to the feature extraction module to extract near-infrared and visible light binocular paired image features respectively. The binocular features are then concat-fused through the feature fusion module. After deep learning feature training operations, the fully connected layer is passed to output the final image live category probability.

[0069] The anti-counterfeiting recognition system that supports multiple image types provided by the present application can determine the image type of the input image, and then select the corresponding network branch to perform anti-counterfeiting recognition on the input image according to the image type. It supports anti-counterfeiting recognition of multiple image types and allows customers to independently choose between three input modes: monocular near-infrared image input, monocular visible light image input, and binocular image paired input, thereby meeting customers' needs for multiple image input types; and, in the binocular paired image input mode, binocular feature fusion can enhance feature expression capabilities, which helps to improve anti-counterfeiting recognition accuracy.

[0070] The above is an embodiment of an anti-counterfeiting identification system supporting multiple image types provided by the present application. The following is an embodiment of an anti-counterfeiting identification method supporting multiple image types provided by the present application.

[0071] Please refer to Figure 3 , an embodiment of the present application provides an anti-counterfeiting recognition method supporting multiple image types, including:

[0072] Step 301: Determine the image type of the input image.

[0073] In one embodiment, a channel feature of the input image is extracted, and a first type of the input image is determined according to the channel feature of the input image;

[0074] Performing Fourier transform on the input image, calculating the spectrum amplitude to obtain spectrum characteristics of the input image, and determining the second type of the input image based on the spectrum characteristics of the input image;

[0075] A final image type of the input image is determined according to the first type and the second type of the input image.

[0076] Specifically, the number of channels of the input image can be extracted as the channel feature of the input image. If the number of channels of the input image is 1, the input image is judged to be a near-infrared image with a single-channel input; if the number of channels of the input image is 3, the input image is judged to be a visible light image with a three-channel input; if the number of channels of the input image is 4, the input image is judged to be a binocular paired image with a mixed-channel input.

[0077] Perform a Fourier transform on the input image, define the zero-frequency feature as the center of the spectrum, and then calculate the spectrum amplitude to obtain the input image's spectral characteristics. If the input image's spectral characteristics are primarily concentrated in the low-frequency region, the input image is identified as a near-infrared image; if the input image's spectral characteristics are primarily concentrated in the high-frequency region, the input image is identified as a visible light image; if the input image's spectral characteristics exhibit a combination of high- and low-frequency features, the input image is identified as a binocular paired image. Finally, the input image's final image type is determined by combining the first and second types of the input image.

[0078] In another embodiment, the image type of the input image can be determined by an image type discrimination model. The training process of the image type discrimination model includes:

[0079] Obtain various types of face images;

[0080] Extract channel features and spectral features of face images;

[0081] The channel features and spectral features of the face image are used as input, and the image type of the face image is used as output to train the target network to obtain an image type discrimination model.

[0082] Acquire multiple facial images of three image types: near-infrared, visible light, and binocular pairs. Extract the channel and spectral features of each type of facial image. Train a target network using these features as input and the image type of the facial image as output. Calculate a loss value based on the predicted image category output by the target network and the true image category. Update the target network parameters using this loss value until the target network converges. This trained target network is then used as the image type discrimination model. The target network can employ an existing convolutional neural network.

[0083] The image type discrimination model is used to discriminate the image type of the input image and output the predicted probabilities of each image type. The sum of the predicted probabilities of the three image types is 1. The image type with the highest predicted probability is selected as the category of the current input image. This output category determines the network branch path selected in the next stage of the network branch selection module.

[0084] Step 302: Select a corresponding network branch in the anti-counterfeiting model according to the image type of the input image to perform anti-counterfeiting recognition on the input image to obtain a liveness category probability of the input image.

[0085] After determining the image type of the input image, the corresponding network branch in the anti-counterfeiting model is selected according to the image type of the input image to perform anti-counterfeiting identification on the input image, thereby obtaining the probability of the input image being a living object. The anti-counterfeiting model includes a first network branch, a second network branch, and a third network branch. The first network branch includes a near-infrared feature extraction module and a fully connected layer, the second network branch includes a visible light feature extraction module and a fully connected layer, and the third network branch includes a near-infrared feature extraction module, a visible light feature extraction module, a feature fusion module, and a fully connected layer. When the image type of the input image is a near-infrared image, the first network branch is selected to perform anti-counterfeiting identification on the input image; when the image type of the input image is a visible light image, the second network branch is selected to perform anti-counterfeiting identification on the input image; and when the image type of the input image is a binocular paired image, the third network branch is selected to perform anti-counterfeiting identification on the input image.

[0086] For details, please refer to Figure 2 If the image type of the current input image is a near-infrared image, all paths ① in the entire network path are closed, and paths ② and ③ are disconnected. The near-infrared image is input to the feature extraction module to extract near-infrared image features. After deep learning feature training operations and passing through the fully connected layer, the final image is output with a probability of living body category.

[0087] If the image type of the current input image is a visible light image, all paths ② in the entire network path are closed, and paths ① and ③ are disconnected. The visible light image is input to the feature extraction module to extract visible light image features. After deep learning feature training, it passes through the fully connected layer and outputs the final image liveness category probability;

[0088] If the image type of the current input image is a binocular paired image, all paths ③ in the entire network path are closed, and paths ① and ② are disconnected. The binocular paired image is input into the feature extraction module to extract near-infrared and visible light binocular paired image features respectively. The binocular features are then concat-fused through the feature fusion module. After deep learning feature training, the fully connected layer is passed to output the final image live category probability.

[0089] The anti-counterfeiting recognition system that supports multiple image types provided by the present application can determine the image type of the input image, and then select the corresponding network branch to perform anti-counterfeiting recognition on the input image according to the image type. It supports anti-counterfeiting recognition of multiple image types and allows customers to independently choose between three input modes: monocular near-infrared image input, monocular visible light image input, and binocular image paired input, thereby meeting customers' needs for multiple image input types; and, in the binocular paired image input mode, binocular feature fusion can enhance feature expression capabilities, which helps to improve anti-counterfeiting recognition accuracy.

[0090] An embodiment of the present application further provides an electronic device, the device including a processor and a memory;

[0091] The memory is used to store program codes and transmit the program codes to the processor;

[0092] The processor is configured to execute the anti-counterfeiting identification method supporting multiple image types in the aforementioned method embodiment according to instructions in the program code.

[0093] An embodiment of the present application further provides a computer-readable storage medium, characterized in that the computer-readable storage medium is used to store program code, and when the program code is executed by a processor, the anti-counterfeiting identification method supporting multiple image types in the aforementioned method embodiment is implemented.

[0094] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific steps of the embodiment of the method described above can refer to the corresponding process in the aforementioned device embodiment and will not be repeated here.

[0095] In the specification of this application and the above-mentioned drawings, the terms "first," "second," "third," "fourth," etc. (if any) are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. In addition, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or elements is not necessarily limited to those steps or elements explicitly listed, but may include other steps or elements not explicitly listed or inherent to such process, method, product, or apparatus.

[0096] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or plural.

[0097] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0098] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0099] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0100] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the method described in each embodiment of the present application through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (full name: Read-Only Memory, English abbreviation: ROM), random access memory (full name: Random Access Memory, English abbreviation: RAM), disk or optical disk, and other media that can store program code.

[0101] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An anti-counterfeiting identification system that supports multiple image types, characterized in that: include: An image type discrimination module is used to discriminate the image type of the input image, including near-infrared image, visible light image and binocular paired image; The image type discrimination module includes: a channel feature extraction submodule, configured to extract channel features of an input image and determine a first type of the input image according to the channel features of the input image; a spectrum feature extraction submodule, configured to perform Fourier transform on an input image, calculate a spectrum amplitude to obtain spectrum features of the input image, and determine a second type of the input image based on the spectrum features of the input image; a first discriminating subunit, configured to determine a final image type of the input image according to the first type and the second type of the input image; Or, a second discrimination submodule, configured to discriminate the image type of the input image by using an image type discrimination model; A network branch selection module is used to select a corresponding network branch in the anti-counterfeiting model according to the image type of the input image to perform anti-counterfeiting identification on the input image to obtain the living category probability of the input image; the anti-counterfeiting model includes a first network branch, a second network branch and a third network branch, the first network branch includes a near-infrared feature extraction module and a fully connected layer, the second network branch includes a visible light feature extraction module and a fully connected layer; the third network branch includes a near-infrared feature extraction module, a visible light feature extraction module, a feature fusion module and a fully connected layer.

2. The anti-counterfeiting identification system supporting multiple image types according to claim 1, characterized in that: The channel feature extraction submodule is specifically used to: Extracting the number of channels of the input image to obtain channel features of the input image; If the value of the channel feature of the input image is 1, it is determined that the input image is a near-infrared image; If the value of the channel feature of the input image is 3, it is determined that the input image is a visible light image; If the value of the channel feature of the input image is 4, it is determined that the input image is a binocular paired image.

3. The anti-counterfeiting identification system supporting multiple image types according to claim 1, characterized in that: The spectrum feature extraction submodule is specifically used to: Performing Fourier transform on the input image, setting the zero-frequency feature as the center of the spectrum, and calculating the spectrum amplitude to obtain the spectrum feature of the input image; If the proportion of the spectral features of the input image in the low-frequency region exceeds a first threshold, determining that the input image is a near-infrared image; If the proportion of the spectral features of the input image in the high-frequency region exceeds a second threshold, determining that the input image is a visible light image; If the proportion of the spectral features of the input image in the low-frequency region exceeds a third threshold, and the proportion of the spectral features of the input image in the high-frequency region exceeds a fourth threshold, the input image is determined to be a binocular paired image.

4. The anti-counterfeiting identification system supporting multiple image types according to claim 1, characterized in that: The training process of the image type discrimination model includes: Obtain various types of face images; Extracting channel features and spectral features of the face image; The channel features and spectral features of the face image are used as input, and the image type of the face image is used as output to train the target network to obtain an image type discrimination model.

5. An anti-counterfeiting identification method supporting multiple image types, characterized in that: include: Determine the image type of the input image, which includes near-infrared image, visible light image and binocular paired image; The step of determining the image type of the input image includes: Extracting channel features of an input image, and determining a first type of the input image according to the channel features of the input image; Performing a Fourier transform on the input image, calculating a spectrum amplitude to obtain a spectrum feature of the input image, and determining a second type of the input image based on the spectrum feature of the input image; determining a final image type of the input image according to the first type and the second type of the input image; Alternatively, the image type of the input image is discriminated by an image type discrimination model; According to the image type of the input image, the corresponding network branch in the anti-counterfeiting model is selected to perform anti-counterfeiting identification on the input image to obtain the living category probability of the input image; the anti-counterfeiting model includes a first network branch, a second network branch and a third network branch, the first network branch includes a near-infrared feature extraction module and a fully connected layer, the second network branch includes a visible light feature extraction module and a fully connected layer; the third network branch includes a near-infrared feature extraction module, a visible light feature extraction module, a feature fusion module and a fully connected layer.

6. An electronic device, characterized in that: The device includes a processor and a memory; The memory is used to store program code and transmit the program code to the processor; The processor is configured to execute the anti-counterfeiting identification method supporting multiple image types according to claim 5 according to instructions in the program code.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium is used to store program codes, and when the program codes are executed by a processor, the anti-counterfeiting identification method supporting multiple image types as claimed in claim 5 is implemented.

Images