Identity information identification method and device, storage medium and electronic equipment

Abstract

The application discloses an identity information recognition method and device, a storage medium and an electronic device. The method comprises the following steps: obtaining a target image of identity information to be recognized, and a target part of a target object for identity recognition is present in the target image; removing a reference frequency band in the target image to obtain a key frequency band image, the reference frequency band being determined based on a frequency band preprocessing network, the frequency band preprocessing network being used to determine the difference between a sample image and a de-frequency sample image, the de-frequency sample image being an image obtained by filtering the sample image in the frequency band; and recognizing the identity information of the target object based on the part features of the target part extracted from the key frequency band image. The application solves the technical problem of inaccurate identity information recognition in the identity information recognition method provided by the prior art.

Description

Technical Field

[0001] The present application relates to the field of computers, and more specifically, to a method and device for identifying identity information, a storage medium, and an electronic device. Background Art

[0002] At present, in order to identify the personal identity information of different users, it is usually necessary to obtain the user's authorization and then use a feature extraction model to extract the biometric image features of the target object presented in the image to be identified. These biometric image features are then used to identify and determine the user identity information of the target object.

[0003] However, when using the identity information recognition methods provided by the above-mentioned prior art, the image features extracted by the feature extraction model include not only the aforementioned biometric image features but also often include interference features unrelated to the user's identity information, such as muscle features and skin features. These interference features are not uniquely related to the user's identity and therefore cannot represent the user's true identity, resulting in inaccurate user identity information.

[0004] In other words, the identification method of identity information provided in the prior art still has the technical problem of inaccurate identity information identification.

[0005] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0006] The embodiments of the present application provide a method and device for identifying identity information, a storage medium, and an electronic device to at least solve the technical problem of inaccurate identity information identification in the identity information identification methods provided in the prior art.

[0007] According to one aspect of an embodiment of the present application, a method for identifying identity information is provided, including: obtaining a target image of identity information to be identified, wherein the target image presents a target part of a target object used for identity identification; removing a reference frequency band in the target image to obtain a key frequency band image, wherein the reference frequency band is determined based on a frequency band preprocessing network, the frequency band preprocessing network is used to determine the difference between a sample image and a de-frequencyed sample image, and the de-frequencyed sample image is an image obtained after frequency band filtering of the sample image; and identifying the identity information of the target object based on part features of the target part extracted from the key frequency band image.

[0008] According to another aspect of an embodiment of the present application, an identity information recognition device is also provided, including: an acquisition unit, used to acquire a target image of identity information to be recognized, wherein the target image presents a target part of a target object for identity recognition; a removal unit, used to remove a reference frequency band in the target image to obtain a key frequency band image, wherein the reference frequency band is determined based on a frequency band preprocessing network, the frequency band preprocessing network is used to determine the difference between a sample image and a de-frequencyed sample image, and the de-frequencyed sample image is an image obtained after frequency band filtering of the sample image; an identification unit, used to identify the identity information of the target object based on the part features of the target part extracted from the key frequency band image.

[0009] Optionally, in this embodiment, the above-mentioned removal unit includes: a transformation processing module, which is used to transform the target image to obtain the frequency components of the target image in each frequency band; a determination module, which is used to determine the reference frequency components that match the reference frequency band from the frequency components in each frequency band; a removal module, which is used to remove the reference frequency components to obtain the key frequency band; and an inverse transformation processing module, which is used to perform inverse transformation processing on the frequency components in the key frequency band to obtain the key frequency band image.

[0010] Optionally, in this embodiment, the above-mentioned device also includes: a first acquisition unit, used to acquire multiple groups of sample image pairs, wherein each group of sample image pairs includes a sample image and a de-frequencyed sample image; an extraction unit, used to extract sample image features corresponding to the sample image, and de-frequencyed sample image features corresponding to the de-frequencyed sample image in the frequency band preprocessing network; a first determination unit, used to determine the de-frequencying confidence between the sample image and the de-frequencyed sample image based on the feature similarity between the sample image features and the de-frequencyed sample image features; a second determination unit, used to determine the reference frequency band according to the sorting result of the de-frequencying confidence when the de-frequencying confidence corresponding to each of the multiple groups of sample image pairs is acquired.

[0011] Optionally, in this embodiment, the above-mentioned first acquisition unit includes: a first transformation processing module, used to transform the current sample image to obtain sample frequency components of the current sample image in M ​​frequency bands; a filtering processing module, used to filter the sample frequency components of the current sample image through the i-th frequency domain filter used to filter the i-th frequency band to obtain the i-th filtered sample frequency component, where i is a natural number greater than or equal to 1 and less than or equal to M; a first inverse transformation processing module, used to inverse transform the i-th filtered sample frequency component to obtain the i-th de-frequencyed reference sample image; a combination module, used to combine the current sample image with the M de-frequencyed reference sample images corresponding to each of the M frequency bands, respectively, to obtain M groups of sample image pairs corresponding to the current sample image; and an acquisition module, used to obtain sample image pairs corresponding to each of the N sample images, to obtain multiple groups of sample image pairs.

[0012] Optionally, in this embodiment, the above-mentioned first determination unit includes: a first acquisition module, used to obtain the cosine distance between the sample image feature and the de-frequencyed sample image feature as the feature similarity; a first determination module, used to determine the difference between the target threshold and the feature similarity as the de-frequencying confidence.

[0013] Optionally, in this embodiment, the second determination unit includes: a sorting module, configured to sort the determined de-frequencying confidences in descending order to obtain a sorted sequence; and a second determination module, configured to determine, as a reference frequency band, a frequency band corresponding to the de-frequencying sample image corresponding to each of the first k de-frequencying confidences in the sorted sequence, where k is a natural number greater than or equal to 1 and less than or equal to P, and P is the number of the determined de-frequencying confidences.

[0014] Optionally, in this embodiment, the above-mentioned recognition unit includes: a feature extraction module, which is used to use a feature extraction network to extract features from the key frequency band image to obtain the part features of the target part, wherein the feature extraction network is trained using the key frequency band located outside the reference frequency band in the sample image, and the frequency domain features of the key frequency band carry the identity information of the object; a search module, which is used to search for the object identity information that matches the part features in the feature database as the identity information of the identified target object.

[0015] Optionally, in this embodiment, the above-mentioned device also includes: a training unit, which is used to determine each sample image as the current sample image in turn, and perform the following steps: removing the reference frequency band in the current sample image to obtain the current training sample image; using the initialized feature extraction network to extract features of the current training sample image to obtain the current training sample features; obtaining the training loss value between the current training sample features and the label sample features, wherein the label sample features are the part features of the target part extracted in advance from the current sample image; when the training loss value reaches the convergence condition, the currently trained feature extraction network is determined as the feature extraction network.

[0016] Optionally, in this embodiment, the above-mentioned search module is also used to determine the matching similarities between Q identity features stored in the feature database and the part features, where Q is a natural number greater than or equal to 1; sort the matching similarities to obtain a sorting result; and determine the object identity information indicated by the identity feature corresponding to the maximum matching similarity in the sorting result as the identity information of the target object.

[0017] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, in which a computer program is stored. The computer program is configured to execute the above-mentioned identity information identification method when running.

[0018] According to another aspect of an embodiment of the present application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the above-described method for identifying identity information.

[0019] According to another aspect of the embodiments of the present application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above-mentioned identity information identification method through the computer program.

[0020] In an embodiment of the present application, a target image for identifying identity information is obtained, wherein the target image contains a target part of a target object for identification. A reference frequency band is then removed from the target image to obtain a key frequency band image. The reference frequency band is determined based on a frequency band preprocessing network, which determines the difference between a sample image and a de-spectrized sample image. The de-spectrized sample image is obtained by frequency band filtering the sample image. Next, the identity information of the target object is identified based on the part features of the target part extracted from the key frequency band image. In other words, in this embodiment of the present application, the target image for identifying identity information is processed by removing frequency bands that carry little to no identity information to obtain a key frequency band image. The part features of the target part extracted from the key frequency band image are then used to identify the identity information of the target object. Interference features in the target image that are unrelated to the user's identity information are removed, thereby making the determined identity information of the target object more accurate. This solves the technical problem of inaccurate identity information identification in existing identity information identification methods, achieving the technical effect of improving the accuracy of identity information identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0022] Figure 1 is a schematic diagram of an application environment of an optional identity information recognition method according to an embodiment of the present application;

[0023] Figure 2 is a flowchart of an optional method for identifying identity information according to an embodiment of the present application;

[0024] Figure 3 is a schematic diagram of an optional method for identifying identity information according to an embodiment of the present application;

[0025] Figure 4 is a schematic diagram of an optional method for identifying identity information according to an embodiment of the present application;

[0026] Figure 5 is a schematic diagram of an optional method for identifying identity information according to an embodiment of the present application;

[0027] Figure 6 is a flowchart of an optional method for identifying identity information according to an embodiment of the present application;

[0028] Figure 7is a flowchart of an optional method for identifying identity information according to an embodiment of the present application;

[0029] Figure 8 is a schematic diagram of an optional method for identifying identity information according to an embodiment of the present application;

[0030] Figure 9 This is a schematic structural diagram of an optional identity information recognition device according to an embodiment of the present application;

[0031] Figure 10 It is a schematic structural diagram of an optional electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to 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 drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] According to one aspect of the embodiment of the present application, a method for identifying identity information is provided. Optionally, as an optional implementation, the identification of the identity information can be applied to, but is not limited to, Figure 1 In the environment shown. Figure 1As shown, terminal device 102 includes a memory 104 for storing various data generated during the operation of terminal device 102, a processor 106 for processing and computing the aforementioned data, and a display 108 for displaying a target image. Terminal device 102 can exchange data with server 112 via network 110. Server 112 is connected to database 114, which is used to store various data. Terminal device 102 can run an application for performing identity information recognition.

[0035] Furthermore, the above method Figure 1 The specific application process in the environment shown is as follows:

[0036] Steps S102 to S104 are executed, and the terminal device 102 obtains a target image of identity information to be recognized, wherein the target image presents a target part of the target object for identity recognition, and sends the target image to the server 112 via the network 110 .

[0037] Then, steps S106-S108 are executed. Server 112 removes the reference frequency band from the target image to obtain a key frequency band image. The reference frequency band is determined based on a frequency band preprocessing network, which is used to determine the difference between the sample image and the de-frequencyed sample image. The de-frequencyed sample image is an image obtained by frequency band filtering the sample image. Server 112 identifies the identity information of the target object based on the part features of the target part extracted from the key frequency band image.

[0038] Then, step S110 is executed, and the server 112 sends the identity information of the target object to the terminal device 102 via the network 110 .

[0039] In an embodiment of the present application, a target image for identifying identity information is obtained, wherein the target image contains a target part of a target object for identification. A reference frequency band is then removed from the target image to obtain a key frequency band image. The reference frequency band is determined based on a frequency band preprocessing network, which determines the difference between a sample image and a de-spectrized sample image. The de-spectrized sample image is obtained by frequency band filtering the sample image. Next, the identity information of the target object is identified based on the part features of the target part extracted from the key frequency band image. In other words, in this embodiment of the present application, the target image for identifying identity information is processed by removing frequency bands that carry little to no identity information to obtain a key frequency band image. The part features of the target part extracted from the key frequency band image are then used to identify the identity information of the target object. Interference features in the target image that are unrelated to the user's identity information are removed, thereby making the determined identity information of the target object more accurate. This solves the technical problem of inaccurate identity information identification in existing identity information identification methods, achieving the technical effect of improving the accuracy of identity information identification.

[0040] Optionally, in this embodiment, the above-mentioned terminal device can be a terminal device configured with a target client, which can include but is not limited to at least one of the following: a mobile phone (such as an Android phone, an iOS phone, etc.), a laptop computer, a tablet computer, a PDA, an MID (Mobile Internet Devices), a PAD, a desktop computer, a smart TV, etc. The target client can be a video client, an instant messaging client, a browser client, an education client, etc. The above-mentioned network can include but is not limited to: a wired network, a wireless network, wherein the wired network includes: a local area network, a metropolitan area network and a wide area network, and the wireless network includes: Bluetooth, WIFI and other networks that realize wireless communication. The above-mentioned server can be a single server, or it can be a server cluster composed of multiple servers, or a cloud server. The above is only an example, and this embodiment does not impose any limitation on this.

[0041] Alternatively, as an alternative, Figure 2 As shown, the above-mentioned identification method of identity information includes:

[0042] S202 : Acquire a target image of identity information to be identified, wherein the target image presents a target part of a target object for identity identification.

[0043] Optionally, the above-mentioned identity information recognition method can be applied, but is not limited to, to scenarios where the identity of an object is identified using target part features (i.e., biometric features) of the object, including palm print recognition, fingerprint recognition, facial recognition, eye print recognition, etc., which are not limited in this embodiment. For example: unlocking an electronic mobile device, where a smart mobile device uses the user's palm print, fingerprint, facial, eye print, pupil, etc. to identify the user's identity to unlock the function; access control systems, where some companies and institutions use palm prints, fingerprints, facial, eye print, pupil, etc. to identify employees; banks and financial institutions, where some banks and financial institutions require palm prints, fingerprints, facial, eye print, pupil, etc. to confirm the identity of customers when conducting transactions or handling business; payment systems, where payment systems use palm prints, fingerprints, facial, eye prints, pupil, etc. to confirm the identity of users for payment; and vehicle startup, where vehicles use palm prints, fingerprints, facial, eye prints, pupil, etc. to confirm the owner's identity to start the vehicle. It should be noted that the several scenarios illustrated above are optional scenarios provided for the convenience of explaining this embodiment. The above-mentioned identity information recognition method can also be applied to other forms of scenarios in which the identity of an object is identified by utilizing the target part features of the object (i.e., biometric features). This is not limited in this embodiment.

[0044] Furthermore, the above-mentioned target image can be, but is not limited to, an image used to indicate biometric information provided by the target object that can identify the identity of the target object. Specifically, the type of the above-mentioned target image is related to the application scenario of the above-mentioned identity information recognition method. For example, assuming that the identity information recognition method is applied to a palmprint recognition scenario, then the above-mentioned target image can be the palm image of the target object; assuming that the identity information recognition method is applied to a fingerprint recognition scenario, then the above-mentioned target image can be the finger image of the target object; assuming that the identity information recognition method is applied to a facial recognition scenario, then the above-mentioned target image can be the facial image of the target object; assuming that the identity information recognition method is applied to an eyeprint recognition scenario, then the above-mentioned target image can be the eye image of the target object.

[0045] Accordingly, the above-mentioned target part can be but is not limited to the part used to indicate the biometric information included in the above-mentioned target image that can identify the identity of the target object, such as the palm print, fingerprint, facial features, eye pattern, etc. of the target object. In this embodiment, the object is not limited in any way.

[0046] S204: Remove the reference frequency band in the target image to obtain a key frequency band image, wherein the reference frequency band is determined based on a frequency band preprocessing network, which is used to determine the difference between the sample image and the de-frequencyed sample image, and the de-frequencyed sample image is an image obtained by frequency band filtering the sample image.

[0047] It should be noted that in this embodiment, the reference frequency band can be, but is not limited to, used to indicate frequency bands corresponding to a target image containing a large amount of interfering image content. For example, assuming the identity information recognition method is applied to a palm print recognition scenario, and the target image is the palm print image of the target subject, the interfering image content can be, but is not limited to, used to indicate other image content in the target image besides the target subject's palm print, such as the target subject's muscle image, a background image captured within the target image, and so on.

[0048] Furthermore, the above-mentioned frequency band preprocessing network may include, but is not limited to: a feature recognition network and a loss function. Among them, the above-mentioned feature recognition network may be, but is not limited to, used to obtain the image features corresponding to the sample image and the de-frequency sample image, and the above-mentioned loss function may be, but is not limited to, used to determine the difference between the sample image and the de-frequency sample image. The above-mentioned feature recognition network may be, but is not limited to, a lightweight convolutional neural network (MobileFaceNet), or other feature recognition networks with similar functions, and this is not limited in this embodiment. The above-mentioned loss function may be, but is not limited to, a loss function based on angle cosine (Arcface), or other loss functions with similar functions, and this is not limited in this embodiment.

[0049] Optionally, in this embodiment, the frequency domain characteristics of the above-mentioned reference frequency band can be obtained based on but not limited to the following steps: obtaining multiple groups of sample image pairs, wherein each group of sample image pairs includes a sample image and a de-frequencyed sample image; performing the following operations on each group of sample image pairs in sequence: extracting sample image features corresponding to the sample image and de-frequencyed sample image features corresponding to the de-frequencyed sample image in the frequency band preprocessing network; determining the de-frequencying confidence between the sample image and the de-frequencyed sample image based on the feature similarity between the sample image features and the de-frequencyed sample image features; and determining the frequency domain characteristics of the reference frequency band according to the sorting results of the de-frequencying confidences when the de-frequencying confidences corresponding to the multiple groups of sample image pairs are obtained.

[0050] S206 , identifying the identity information of the target object based on the part features of the target part extracted from the key frequency band image.

[0051] Optionally, in this embodiment, the above-mentioned identification of the identity information of the target object based on the part features of the target part extracted from the key frequency band image may include, but is not limited to: using a feature extraction network to extract features from the key frequency band image to obtain the part features of the target part; then, searching the feature database for object identity information that matches the part features as the identity information of the identified target object.

[0052] It should be noted that in this embodiment, the key frequency band can be, but is not limited to, used to indicate a valid frequency band. Specifically, the valid frequency band is a frequency band in an image that is sufficient to identify the identity of an object. The reference frequency band is a frequency band in an image that contains a large amount of image content that is irrelevant to the identity of the object.

[0053] As an optional implementation, assuming that the above-mentioned identification method of identity information is applied to the palmprint recognition scenario, it can be based on but not limited to the following: Figure 3 The following steps are shown to illustrate the above method:

[0054] Get Figure 3 The target image of the identity information to be identified is shown in (a), wherein the target image presents the palm of the target object for identity identification. Then, the frequency band that matches the reference frequency band in the target image is removed to obtain the following: Figure 3 (b) shows a critical frequency band image containing only the target object's fingerprint. The reference frequency band is determined based on a frequency band preprocessing network, which is used to determine the difference between the sample image and the de-frequencyed sample image. The de-frequencyed sample image is obtained by frequency band filtering the sample image. Furthermore, the target object's identity information is identified based on the target part features extracted from the critical frequency band image by a feature extraction network. The feature extraction network is trained using the critical frequency bands in the sample image that are outside the reference frequency band.

[0055] Optionally, as another embodiment, assuming that the above-mentioned identity information recognition method is applied to a facial recognition scenario for unlocking a mobile phone, the above-mentioned method can be explained by way of example based on, but not limited to, the following steps:

[0056] Upon receiving a request from the user to unlock the phone, the phone captures an image of the user's face, which displays the user's facial features, muscles, and skin. Next, any image content that could interfere with identifying the user (such as muscles and skin texture) is removed, retaining only the facial features. The phone then uses the image features extracted from the facial features to determine whether the user currently unlocking the phone is the phone's owner. If the user currently unlocking the phone is the owner, the phone is unlocked for the user's use.

[0057] Optionally, as another embodiment, assuming that the above-mentioned identity information recognition method is applied to a fingerprint recognition scenario of an access control system, the above-mentioned method can be explained by way of example based on, but not limited to, the following steps:

[0058] Upon receiving a user request to open the access control device, the access control device captures an image of the user's finger, which displays the user's fingerprint, muscles, skin, and other features. Next, the device removes any image content that could interfere with identifying the user (e.g., muscles, skin texture, etc.), retaining only the user's fingerprint. The device then uses the image features extracted from the fingerprint image to determine whether the user has permission to enter the area protected by the access control device. If the user has permission, the access control device is opened.

[0059] Optionally, as another embodiment, assuming that the above-mentioned identity information recognition method is applied to a pupil recognition scenario in business processing, the above-mentioned method can be explained by example based on, but not limited to, the following steps:

[0060] When a user requests a service, the service processing system captures an eye image showing the user's pupil, eyeball, vitreous body, and surrounding eye muscles. Next, the system removes any image content that could interfere with identifying the user (e.g., the eyeball, vitreous body, and surrounding eye muscles), retaining only the pupil image. The system then uses the image features extracted from the pupil image to determine whether the user has permission to process the service. If the user has permission, the service is processed.

[0061] Optionally, as another embodiment, assuming that the above-mentioned identity information recognition method is applied to a palm print recognition scenario for vehicle startup, the above-mentioned method can be explained by way of example based on, but not limited to, the following steps:

[0062] Upon receiving a user request to start the vehicle, the vehicle system captures an image of the user's palm, which displays the user's palm prints, muscles, skin, and other features. Next, the system removes any image content that could interfere with user identification (e.g., muscles and skin texture), retaining only the palm print. The system then uses the image features extracted from the palm print image to determine whether the user has permission to start the vehicle. If the user has permission, the system initiates the vehicle start.

[0063] In an embodiment of the present application, a target image for identifying identity information is obtained, wherein the target image contains a target part of a target object for identification. A reference frequency band is then removed from the target image to obtain a key frequency band image. The reference frequency band is determined based on a frequency band preprocessing network, which determines the difference between a sample image and a de-spectrized sample image. The de-spectrized sample image is obtained by frequency band filtering the sample image. Next, the identity information of the target object is identified based on the part features of the target part extracted from the key frequency band image. In other words, in this embodiment of the present application, the target image for identifying identity information is processed by removing frequency bands that carry little to no identity information to obtain a key frequency band image. The part features of the target part extracted from the key frequency band image are then used to identify the identity information of the target object. Interference features in the target image that are unrelated to the user's identity information are removed, thereby making the determined identity information of the target object more accurate. This solves the technical problem of inaccurate identity information identification in existing identity information identification methods, achieving the technical effect of improving the accuracy of identity information identification.

[0064] Optionally, as an optional solution, removing the frequency bands in the target image that match the frequency domain characteristics of the reference frequency band to obtain the key frequency band image includes:

[0065] S1, transform the target image to obtain the frequency components of the target image in each frequency band;

[0066] S2, determining a reference frequency component that matches the reference frequency band from the frequency components in each frequency band;

[0067] S3, remove the reference frequency component to obtain the key frequency band;

[0068] S4, performing inverse transformation processing on the frequency components in the key frequency band to obtain a key frequency band image.

[0069] Optionally, in this embodiment, the above-mentioned change processing of the target image can be but is not limited to indicating the change processing from time domain representation to frequency domain representation. Specifically, the target image is converted from pixels in the original time domain dimension to frequency components in the frequency domain dimension.

[0070] It should be noted that, in this embodiment, the above-mentioned removal of the reference frequency component to obtain the key frequency band may include, but is not limited to: determining the frequency component in the above-mentioned each frequency band that overlaps with the reference frequency band as the above-mentioned reference frequency component; or, determining the frequency component in the above-mentioned each frequency band whose similarity between the frequency domain characteristics and the frequency domain characteristics of the reference frequency band is greater than a predetermined threshold as the above-mentioned reference frequency component.

[0071] Furthermore, filters can be used, but are not limited to, to remove reference frequency components from the frequency components of the target image in each frequency band. It should be noted that a filter can be a function that processes the input signal to filter out unwanted frequency components or enhance desired frequency components. These functions can be various mathematical functions or algorithms used to filter the signal. Specifically, low-pass, high-pass, band-pass, or band-stop filters can be used, but are not limited to, to filter out reference frequency components from the frequency components of the target image in each frequency band. In addition, in addition to using filters, the above-mentioned reference frequency components can also be removed by other methods such as digital signal processing (i.e., certain frequency components in the frequency band can be filtered through digital signal processing algorithms), modulation and demodulation (i.e., certain frequency components in the frequency band can be transferred to other frequency bands through modulation and demodulation technology, thereby achieving removal), phase modulation (i.e., the removal of certain frequency components in the frequency band can be achieved by changing the phase of the signal), frequency mixing (i.e., certain frequency components in the frequency band can be transferred to other frequency bands through frequency mixing technology, thereby achieving removal), and adaptive filtering (i.e., dynamic removal of certain frequency components in the frequency band can be achieved through adaptive filters). No limitation is imposed on this in this embodiment.

[0072] In an embodiment of the present application, a target image is transformed to obtain the frequency components of the target image in each frequency band. Then, from the frequency components in each frequency band, a reference frequency component that matches the reference frequency band is determined. Next, the reference frequency component is removed to obtain a key frequency band. Furthermore, the frequency components in the key frequency band are inversely transformed to obtain a key frequency band image. In other words, using an embodiment of the present application, a key frequency band image is obtained by removing the frequency domain features that do not carry the identity information of the object from the target image for which identity information is to be identified, and then the location features of the target location extracted from the key frequency band image are used to identify the identity information of the target object. Interference features in the target image that are not related to the user's identity information are removed, thereby making the determined identity information of the target object more accurate. This solves the technical problem of inaccurate identity information identification in the identity information identification method provided in the prior art, and achieves the technical effect of improving the accuracy of identity information identification.

[0073] Optionally, as an optional solution, before obtaining the target image of the identity information to be identified, the method further includes:

[0074] Acquire multiple groups of sample image pairs, wherein each group of sample image pairs includes a sample image and a sample image;

[0075] Optionally, in this embodiment, the de-frequencyed sample images included in each set of sample image pairs may be, but are not limited to, images obtained by filtering the sample images in each set of sample image pairs according to the sample frequency band. The sample frequency band may be, but is not limited to, randomly generated frequency bands.

[0076] Furthermore, the above-mentioned acquisition of multiple groups of sample images may include, but is not limited to, the following steps: taking the N acquired sample images as the current sample image in sequence, and performing the following operations: transforming the current sample image to obtain the sample frequency components of the current sample image in M ​​frequency bands; repeating the following operations until the M frequency bands corresponding to the current sample image are traversed: filtering the sample frequency components of the current sample image using the i-th frequency domain filter for filtering the i-th frequency band to obtain the i-th filtered sample frequency component, where i is a natural number greater than or equal to 1 and less than or equal to M; inversely transforming the i-th filtered sample frequency component to obtain the i-th de-frequencyed reference sample image; combining the current sample image with the M de-frequencyed reference sample images corresponding to each of the M frequency bands to obtain M groups of sample image pairs corresponding to the current sample image; obtaining the sample image pairs corresponding to each of the N sample images to obtain multiple groups of sample image pairs.

[0077] For each set of sample image pairs, perform the following operations in sequence:

[0078] Extracting sample image features corresponding to the sample image and de-frequency sample image features corresponding to the de-frequency sample image in the frequency band preprocessing network;

[0079] Optionally, extracting sample image features corresponding to the sample image and de-frequencyed sample image features corresponding to the de-frequencyed sample image in the frequency band preprocessing network may include, but is not limited to, extracting sample image features corresponding to the sample image and de-frequencyed sample image features corresponding to the de-frequencyed sample image using a feature recognition network included in the frequency band preprocessing network. The feature recognition network may be, but is not limited to, a lightweight convolutional neural network (MobileFaceNet) or other feature recognition networks with similar functionality, and this is not limited in this embodiment.

[0080] For example, assuming that x represents the sample image and x_removeb represents the de-frequencyed sample image, the sample image features corresponding to the sample image and the de-frequencyed sample image features corresponding to the de-frequencyed sample image can be extracted in the following manner, but is not limited to:

[0081] embeddings_remove=F_pretrain(x_remove)(1)

[0082] embeddings = F_pretrain(x)(2)

[0083] Among them, the above embeddings_remove is used to represent the de-frequency sample image features, the above embeddings is used to represent the sample image features, and the above F_pretrain is used to represent the feature recognition network.

[0084] Determining the de-frequencying confidence between the sample image and the de-frequencying sample image based on the feature similarity between the sample image feature and the de-frequencying sample image feature;

[0085] Furthermore, in this embodiment, the feature similarity between the above-mentioned sample image features and the de-frequencyed sample image features can be obtained by, but is not limited to, image feature similarity comparison, such as using Euclidean distance, cosine similarity, correlation coefficient, feature matching, etc., which is not limited in this embodiment.

[0086] Furthermore, the above-mentioned determination of the de-frequencying confidence between the sample image and the de-frequencying sample image based on the feature similarity between the sample image features and the de-frequencying sample image features may include, but is not limited to: determining the difference between the target threshold and the feature similarity as the de-frequencying confidence.

[0087] When the frequency removal confidences corresponding to the plurality of sample image pairs are obtained, the reference frequency band is determined according to the sorting results of the frequency removal confidences.

[0088] Optionally, in this embodiment, determining the reference frequency band based on the sorting results of the frequency removal confidences may include, but is not limited to, sorting the determined frequency removal confidences in descending order to obtain a sorted sequence. Then, determining the frequency band corresponding to the frequency removal sample image corresponding to the first k frequency removal confidences in the sorted sequence as the reference frequency band, where k is a natural number greater than or equal to 1 and less than or equal to P, and P is the number of determined frequency removal confidences.

[0089] As an optional implementation, it can be based on but not limited to Figure 4 The following steps are used to illustrate the identification method of the above identity information:

[0090] Execute step S402 to input the sample images and de-frequencyed sample images in each sample image pair in the multiple groups of sample image pairs into the feature recognition network respectively. Then execute step S404 to use the feature recognition network to perform feature extraction on the sample images and de-frequencyed sample images in each sample image pair respectively to obtain the sample image features of the sample images in each sample image pair and the de-frequencyed sample image features of the de-frequencyed sample images. Next, execute step S406 to perform a similarity comparison on the sample image features and the de-frequencyed sample image features to obtain the feature similarity corresponding to each sample image pair. Next, execute step S408 to obtain the de-frequencyed confidence corresponding to each sample image pair based on the feature similarity corresponding to each sample image pair. Then execute step S410 to determine the reference frequency band based on the de-frequencyed confidence corresponding to each sample image pair.

[0091] In an embodiment of the present application, multiple groups of sample image pairs are obtained, wherein each group of sample image pairs includes a sample image and a de-frequencyed sample image. Then, the following operations are performed on each group of sample image pairs in sequence: extracting sample image features corresponding to the sample image and de-frequencyed sample image features corresponding to the de-frequencyed sample image in the frequency band preprocessing network; determining the de-frequencying confidence between the sample image and the de-frequencyed sample image based on the feature similarity between the sample image features and the de-frequencyed sample image features; and determining the reference frequency band according to the sorting results of the de-frequencying confidence when the de-frequencying confidence corresponding to each of the multiple groups of sample image pairs is obtained. In other words, by adopting an embodiment of the present application, by pre-determining the content that will interfere with identity recognition (i.e., the reference frequency band) before identifying the identity information of the target image, when identifying the identity information of the target image, the image content that may interfere with identity recognition can be directly removed according to the above-mentioned reference frequency band. This solves the technical problem of inaccurate identity information recognition in the identity information recognition method provided in the prior art, and achieves the technical effect of improving the accuracy of identity information recognition.

[0092] Optionally, as an optional solution, obtaining multiple groups of sample image pairs includes:

[0093] The obtained N sample images are used as the current sample images in sequence, and the following operations are performed:

[0094] Perform transformation processing on the current sample image to obtain the sample frequency components of the current sample image in M ​​frequency bands;

[0095] It should be noted that the above-mentioned transformation processing performed on the current sample image can be but is not limited to indicating the change from time domain representation to frequency domain representation. Specifically, the current sample image is converted from pixels in the original time domain dimension to frequency components in the frequency domain dimension.

[0096] Optionally, in this embodiment, the above-mentioned transformation processing of the current sample image to obtain the sample frequency components of the current sample image in M ​​frequency bands may include, but is not limited to: transforming the current sample image to obtain the frequency components corresponding to the current sample image, and then dividing the frequency components corresponding to the current sample image into sample frequency components in M ​​frequency bands.

[0097] Optionally, the above transformation processing is performed on the current sample image to obtain the frequency components corresponding to the current sample image, which may be implemented, but is not limited to, in the following manner:

[0098] F(u, v) = FFT(I(x, y)) (3)

[0099] Among them, the above F(u, v) is used to represent the frequency component corresponding to the current sample image, the above I(x, y) is used to represent the current sample image, and the above FFT is used to represent the Fourier transform formula.

[0100] Repeat the following steps until the M frequency bands corresponding to the current sample image are traversed:

[0101] Performing filtering processing on the sample frequency component of the current sample image by using the i-th frequency domain filter for filtering the i-th frequency band to obtain the i-th filtered sample frequency component, where i is a natural number greater than or equal to 1 and less than or equal to M;

[0102] Optionally, in this embodiment, the above-mentioned i-th frequency band can be, but is not limited to, used to indicate the sample frequency band mentioned above, that is, a randomly generated frequency band. It should be noted that, in this embodiment, it is possible but not limited to generate multiple sample frequency bands, so as to remove the above-mentioned multiple sample frequency bands in each sample image respectively, and obtain multiple de-frequencyed sample images corresponding to the sample image. If the frequency domain characteristics of the reference frequency band cannot be obtained according to the sorting result of the de-frequencying confidence, multiple sample frequency bands can be randomly generated again, and then the above-mentioned multiple sample frequency bands can be removed in each sample image respectively, and obtain multiple de-frequencyed sample images corresponding to the sample image. And so on, until the frequency domain characteristics of the reference frequency band can be obtained according to the sorting result of the de-frequencying confidence.

[0103] Performing inverse transformation on the frequency component of the i-th filtered sample to obtain the i-th de-frequency reference sample image;

[0104] It should be noted that, in this embodiment, the inverse transform process described above may be used, but is not limited to, to indicate a time domain transform process for the i-th filtered sample frequency component. That is, the i-th filtered sample frequency component is converted into an image represented by pixels (i.e., the i-th de-frequency reference sample image).

[0105] Combine the current sample image with the M frequency-demultiplexed reference sample images corresponding to each of the M frequency bands to obtain M groups of sample image pairs corresponding to the current sample image;

[0106] Obtain sample image pairs corresponding to each of the N sample images to obtain multiple groups of sample image pairs.

[0107] Optionally, as an optional embodiment, taking the above M as 3 as an example, F(u1, v1), F(u2, v2), and F(u3, v3) are used to represent the above M frequency bands. Figure 5 The following steps are shown to illustrate the above method:

[0108] The obtained N sample images are used as the current sample images in sequence, and the following operations are performed:

[0109] like Figure 5 As shown, step S502 is executed to perform frequency domain transformation processing on the current sample image to obtain sample frequency components of the current sample image in M ​​frequency bands, including: F(u1, v1), F(u2, v2), and F(u3, v3). Then, step S504 is executed to filter F(u1, v1), F(u2, v2), and F(u3, v3) using the i-th frequency domain filter used to filter the i-th frequency band to obtain G(u1, v1), G(u2, v2), and G(u3, v3). Next, step S506 is executed to perform time domain transformation processing on G(u1, v1), G(u2, v2), and G(u3, v3) to obtain de-frequency reference sample image 1, de-frequency reference sample image 2, and de-frequency reference sample image 3 corresponding to the current sample image. Then, sample image pairs corresponding to each of the N sample images are obtained to obtain multiple groups of sample image pairs.

[0110] Optionally, in the above example, taking F(u1, v1) as an example, the above, through the i-th frequency domain filter for filtering the i-th frequency band, is to filter F(u1, v1) to obtain G(u1, v1), which can be but not limited to the following implementation:

[0111] G(u1,v1)=H(u i , v i )*F(u1,v1)(4)

[0112] Among them, (u i , v i ) is used to represent the above-mentioned i-th frequency band, and the above-mentioned H is the i-th frequency domain filter. It should be noted that in this embodiment, other methods can also be used to filter F(u1, v1) using the i-th frequency domain filter for filtering the i-th frequency band to obtain G(u1, v1), and this is not limited in this embodiment.

[0113] Optionally, in the above example, taking G(u1, v1) as an example, performing time domain transformation processing on G(u1, v1) to obtain the frequency-removed reference sample image 1 corresponding to the current sample image may be, but is not limited to, implemented in the following manner:

[0114] x_remove 1=IFFT(G(u1,v1))(5)

[0115] Here, x_remove 1 is used to represent the above-mentioned frequency-removed reference sample image 1, and the above-mentioned IFFT is used to represent the inverse Fourier transform formula.

[0116] In an embodiment of the present application, the N acquired sample images are sequentially used as current sample images, and the following operations are performed: the current sample image is transformed to obtain sample frequency components of the current sample image in M ​​frequency bands; the following operations are repeated until the M frequency bands corresponding to the current sample image are traversed: the sample frequency components of the current sample image are filtered using the i-th frequency domain filter for filtering the i-th frequency band to obtain the i-th filtered sample frequency component, where i is a natural number greater than or equal to 1 and less than or equal to M; the i-th filtered sample frequency component is inversely transformed to obtain the i-th de-frequencyed reference sample image; the current sample image is respectively combined with the M de-frequencyed reference sample images corresponding to each of the M frequency bands to obtain M groups of sample image pairs corresponding to the current sample image; and the sample image pairs corresponding to each of the N sample images are obtained to obtain multiple groups of sample image pairs. In other words, using the embodiments of the present application, by filtering the sample image, a de-frequencyed sample image corresponding to the sample image is obtained, thereby forming multiple groups of sample image pairs using the sample image and the de-frequencyed sample image corresponding to the sample image. Furthermore, using the multiple groups of sample image pairs, the content that may interfere with identity recognition (i.e., the frequency domain characteristics of the reference frequency band) is pre-determined. This allows, when identifying the identity information of the target image, to directly remove image content that may interfere with identity recognition based on the frequency domain characteristics of the reference frequency band. This solves the technical problem of inaccurate identity information recognition in the identity information recognition methods provided in the prior art, achieving the technical effect of improving the accuracy of identity information recognition.

[0117] Alternatively, as an optional solution, in this embodiment, not only can the de-frequency reference sample images be obtained by filtering the M frequency bands of the sample image, but the de-frequency reference sample images can also be obtained by filtering directly based on the frequency components corresponding to the sample image without performing frequency band processing on the sample image. Specifically, in the case of obtaining the de-frequency reference sample images by filtering based on the frequency components corresponding to the sample image, the above-mentioned acquisition of multiple groups of sample images includes:

[0118] The obtained N sample images are used as the current sample images in sequence, and the following operations are performed:

[0119] Performing transformation processing on the current sample image to obtain the sample frequency component of the current sample image;

[0120] Performing filtering processing on the sample frequency component of the current sample image by using the i-th frequency domain filter for filtering the i-th frequency band to obtain the i-th filtered sample frequency component, where i is a positive integer;

[0121] Performing inverse transformation on the frequency component of the i-th filtered sample to obtain the i-th de-frequency reference sample image;

[0122] Combine the current sample image with the i-th de-frequencyed reference sample image to obtain i groups of sample image pairs corresponding to the current sample image;

[0123] Obtain sample image pairs corresponding to each of the N sample images to obtain multiple groups of sample image pairs.

[0124] In an embodiment of the present application, the N acquired sample images are sequentially used as current sample images, and the following operations are performed: the current sample image is transformed to obtain the sample frequency component of the current sample image; the sample frequency component of the current sample image is filtered using the i-th frequency domain filter for filtering the i-th frequency band to obtain the i-th filtered sample frequency component, where i is a positive integer; the i-th filtered sample frequency component is inversely transformed to obtain the i-th de-frequencyed reference sample image; the current sample image is combined with the i-th de-frequencyed reference sample image to obtain i groups of sample image pairs corresponding to the current sample image; and the sample image pairs corresponding to each of the N sample images are obtained to obtain multiple groups of sample image pairs. In other words, using the embodiments of the present application, by filtering the sample image, a de-frequencyed sample image corresponding to the sample image is obtained, thereby forming multiple groups of sample image pairs using the sample image and the de-frequencyed sample image corresponding to the sample image. Furthermore, using the multiple groups of sample image pairs, the content that may interfere with identity recognition (i.e., the frequency domain characteristics of the reference frequency band) is pre-determined. This allows, when identifying the identity information of the target image, to directly remove image content that may interfere with identity recognition based on the frequency domain characteristics of the reference frequency band. This solves the technical problem of inaccurate identity recognition in the identity information recognition methods provided in the prior art, achieving the technical effect of improving the accuracy of identity recognition.

[0125] Optionally, as an optional solution, determining the de-frequencying confidence between the sample image and the de-frequencying sample image based on the feature similarity between the sample image feature and the de-frequencying sample image feature includes:

[0126] The cosine distance between the sample image feature and the de-frequency sample image feature is obtained as the feature similarity.

[0127] The difference between the target threshold and the feature similarity is determined as the de-frequency confidence.

[0128] For example, the cosine distance between the sample image feature and the de-frequency sample image feature can be obtained as the feature similarity by, but not limited to, the following method:

[0129] Similarity=Cosine(embeddings_remove,embeddings)(6)

[0130] The Similarity parameter is used to represent feature similarity, the embeddings_remove parameter is used to represent de-frequency sample image features, the embeddings parameter is used to represent sample image features, and the Cosine parameter is used to represent the cosine similarity function.

[0131] For further example, taking the target threshold as 1, the difference between the target threshold and the feature similarity can be determined as the de-frequency confidence by, but not limited to, the following method:

[0132] Loss_remove = 1 - cosine(embeddings_remove, embeddings) (7)

[0133] The above-mentioned Loss_remove is used to represent the frequency removal confidence.

[0134] In an embodiment of the present application, the cosine distance between the sample image feature and the de-frequencyed sample image feature is obtained as the feature similarity. Then, the difference between the target threshold and the feature similarity is determined as the de-frequencying confidence. In other words, by adopting the embodiment of the present application, by adopting the method of determining the difference between the target threshold and the feature similarity as the de-frequencying confidence, the de-frequencying confidence can more fully reflect the difference between the sample image feature and the de-frequencyed sample image feature, thereby making the frequency domain characteristics of the reference frequency band determined based on the de-frequencying confidence more accurate.

[0135] Optionally, as an optional solution, determining the frequency domain features of the reference frequency band according to the sorting results of the frequency removal confidences includes:

[0136] Sort the determined frequency removal confidences in descending order to obtain a sorted sequence;

[0137] The frequency domain features corresponding to the de-frequencying sample images corresponding to the first k de-frequencying confidence levels in the sorting sequence are determined as the frequency domain features of the reference frequency band, where k is a natural number greater than or equal to 1 and less than or equal to P, and P is the number of determined de-frequencying confidence levels.

[0138] Optionally, in this embodiment, the frequency domain features corresponding to the de-frequency sample images corresponding to the de-frequency confidences at the first k positions in the sorting sequence are determined as the frequency domain features of the reference frequency band, which may include but are not limited to:

[0139] The de-frequency sample images corresponding to the first k de-frequency confidence levels are determined as the first de-frequency sample images. Next, the de-frequency sample image with the highest de-frequency confidence level in the first de-frequency sample image is determined as the second de-frequency sample image. Furthermore, the sample frequency band used to generate the second de-frequency sample image is determined as the second frequency component, and the sample frequency band used to generate the third de-frequency sample image is determined as the third frequency component, wherein the third de-frequency sample image is the de-frequency sample image in the first de-frequency sample image excluding the second de-frequency sample image. Furthermore, if the second frequency component and the third frequency component are the same frequency component, the second de-frequency sample image is determined as the target de-frequency sample image. Thus, the frequency band corresponding to the target de-frequency sample image is determined as the reference frequency band, and the frequency domain features corresponding to the target de-frequency sample image are determined as the frequency domain features of the reference frequency band.

[0140] As an optional implementation, assuming that the de-frequency sample images corresponding to the first k de-frequency confidence levels are de-frequency sample image 1, de-frequency sample image 2, de-frequency sample image 3, and de-frequency sample image 4, the following steps are used to illustrate the above-mentioned identity information recognition method:

[0141] S1. Determine the de-frequency confidence levels corresponding to de-frequency sample image 1, de-frequency sample image 2, de-frequency sample image 3, and de-frequency sample image 4, respectively, as de-frequency confidence level 1, de-frequency confidence level 2, de-frequency confidence level 3, and de-frequency confidence level 4, where de-frequency confidence level 1 is greater than de-frequency confidence level 2, de-frequency confidence level 2 is greater than de-frequency confidence level 3, and de-frequency confidence level 3 is greater than de-frequency confidence level 4.

[0142] S2, determining a sample frequency band for generating the de-frequencyed sample image 1, and determining the sample frequency band for generating the de-frequencyed sample image 1, the sample frequency band for generating the de-frequencyed sample image 2, and the sample frequency band for generating the de-frequencyed sample image 3 respectively.

[0143] S3. When the sample frequency band used to generate the de-frequencyed sample image 1, the sample frequency band used to generate the de-frequencyed sample image 1, the sample frequency band used to generate the de-frequencyed sample image 2, and the sample frequency band used to generate the de-frequencyed sample image 3 are all the same, the frequency band corresponding to the de-frequencyed sample image 1 is determined as the reference frequency band, and the frequency domain feature corresponding to the de-frequencyed sample image 1 is determined as the reference frequency band.

[0144] In an embodiment of the present application, the determined frequency removal confidences are sorted in descending order to obtain a sorted sequence. Then, the frequency bands corresponding to the frequency removal sample images corresponding to the first k frequency removal confidences in the sorted sequence are determined as reference frequency bands. This achieves the purpose of determining the frequency band corresponding to the frequency removal sample image with the largest difference between the frequency removal sample image and the original image as the frequency domain feature of the reference frequency band that may interfere with identity recognition. Specifically, the greater the difference between the frequency removal sample image and the original image, the less image content that can identify the user's identity is included in the frequency removal sample image, and accordingly, the more interference content. In other words, determining the frequency band corresponding to the frequency removal sample image with the largest difference between the frequency removal sample image and the original image as the reference frequency band that may interfere with identity recognition is to determine the corresponding frequency band in the frequency removal sample image that includes interference content as the reference frequency band. This thereby improves the accuracy of the frequency domain feature of the reference frequency band.

[0145] Optionally, as an optional solution, identifying the identity information of the target object based on the part features of the target part extracted from the key frequency band image includes:

[0146] S1. Use a feature extraction network to extract features from the key frequency band image to obtain the location features of the target location. The feature extraction network is trained using the key frequency bands in the sample image that are outside the reference frequency band. The frequency domain features of the key frequency bands carry the identity information of the object.

[0147] It should be noted that in this embodiment, the feature extraction network can be, but is not limited to, a lightweight convolutional neural network (MobileFaceNet) or other feature recognition networks with similar functions, and this is not limited in this embodiment. Among them, MobileFaceNet is a lightweight convolutional neural network specifically for face recognition tasks. It adopts the structure of a deep convolutional neural network, but designs some lightweight modules and techniques to reduce the computational complexity and memory usage of the model while maintaining high accuracy. MobileFaceNet uses lightweight convolution operations such as depthwise separable convolution to reduce the number of model parameters and computational complexity. In addition, it also uses lightweight residual connections and channel attention mechanisms to further improve the accuracy and efficiency of the model. Due to its small model size and low computational complexity, MobileFaceNet is very suitable for face recognition applications on mobile devices and embedded systems. The model has achieved good performance in face recognition tasks and can quickly and accurately perform face recognition in resource-constrained environments.

[0148] S2, searching the feature database for object identity information that matches the part feature as the identity information of the identified target object.

[0149] Optionally. In this embodiment, the feature database may include, but is not limited to, real part features of multiple objects, including the target object. Furthermore, the above-mentioned search for object identity information that matches the part features in the feature database as the identity information of the identified target object may include, but is not limited to: determining the matching similarity between Q identity features stored in the feature database and the part features, respectively, where Q is a natural number greater than or equal to 1. Then, the matching similarities are sorted to obtain a sorting result. Next, the object identity information indicated by the identity feature corresponding to the maximum matching similarity in the sorting result is determined as the identity information of the target object.

[0150] In an embodiment of the present application, a feature extraction network is used to extract features from a key frequency band image to obtain the location features of the target location, wherein the feature extraction network is trained using a key frequency band outside the reference frequency band in the sample image, and the frequency domain features of the key frequency band carry the identity information of the object. Then, the object identity information that matches the location features is searched in the feature database as the identity information of the identified target object. In other words, by using an embodiment of the present application, the efficiency of identifying identity information is improved by using a feature extraction network to determine the identity information of the target object.

[0151] Optionally, as an optional solution, before obtaining the target image of the identity information to be identified, the method further includes:

[0152] Each sample image is determined as the current sample image in turn, and the following steps are performed:

[0153] Remove the reference frequency band in the current sample image to obtain the current training sample image;

[0154] Use the initialized feature extraction network to extract features from the current training sample image to obtain the current training sample features;

[0155] Obtaining a training loss value between the current training sample feature and the label sample feature, wherein the label sample feature is a part feature of the target part pre-extracted from the current sample image;

[0156] When the training loss value reaches the convergence condition, the currently trained feature extraction network is determined as the feature extraction network.

[0157] Optionally, in this embodiment, removing the reference frequency band from the current sample image to obtain the current training sample image may include, but is not limited to: performing a transformation on the current sample image to obtain frequency components of the current sample image in various frequency bands. Then, determining, from the frequency components in various frequency bands, a reference frequency component that matches the reference frequency band. Next, removing the reference frequency component to obtain a key frequency band. Furthermore, performing an inverse transformation on the frequency components in the key frequency band to obtain a key frequency band image.

[0158] It should be noted that in this embodiment, a loss function can be used but is not limited to determine the training loss value between the current training sample features and the actual sample features, wherein the above-mentioned loss function can be but is not limited to using a loss function based on the angle cosine (Arcface), or other loss functions with similar functions. This is not limited to any limitation in this embodiment.

[0159] Alternatively, as an optional implementation, Figure 6 The following steps are used to illustrate the identification of the above identity information:

[0160] Execute steps S602-S604 to determine the current sample image. Then remove the reference frequency band in the current sample image to obtain the current training sample image.

[0161] Then, steps S606-S608 are executed to extract features from the current training sample image using the initialized feature extraction network to obtain training sample features. The training loss value between the current training sample features and the actual sample features is obtained, where the actual sample features are the target part features pre-extracted from the current sample image.

[0162] Then, step S610 is executed to determine whether the training loss value reaches a predetermined threshold. If the training loss value reaches the predetermined threshold, step S612 is executed to determine the currently trained feature extraction network as the feature extraction network and stop training. If the training loss value does not reach the predetermined threshold, step S602 is executed again.

[0163] In an embodiment of the present application, a sample image is obtained. Then, the reference frequency band in the sample image is removed to obtain a training sample image. Next, the initialized feature extraction network is used to extract features from the training sample image to obtain training sample features. Furthermore, a training loss value is obtained between the training sample features and the actual sample features, wherein the actual sample features are the part features of the target part extracted in advance from the sample image. Thus, when the training loss value reaches a predetermined threshold, the currently trained feature extraction network is determined as the feature extraction network. In other words, using an embodiment of the present application, the feature extraction network is trained by using an image corresponding to a key frequency band outside the reference frequency band in the sample image (i.e., a training sample image), so that the feature extraction network obtained after training can focus on extracting image features corresponding to the key frequency band. Thereby, the purpose of obtaining an accurate feature extraction network is achieved even when there is less training sample data.

[0164] Optionally, as an optional solution, the object identity information matching the part feature is searched in the feature database, and the identity information of the identified target object includes:

[0165] The matching similarities between Q identity features stored in the feature database and the part features are determined, where Q is a natural number greater than or equal to 1.

[0166] Sort the matching similarities to obtain the sorting results;

[0167] The object identity information indicated by the identity feature corresponding to the maximum matching similarity in the sorting results is determined as the identity information of the target object.

[0168] Optionally, in this embodiment, the Q identity features stored in the feature database may be, but are not limited to, pre-stored identity features for identifying the real identities of Q objects, wherein the Q objects include the target object.

[0169] Furthermore, in this embodiment, the method used to determine the matching similarity between the Q identity features stored in the feature database and the part features may include, but is not limited to, Euclidean distance, cosine similarity, correlation coefficient, feature matching, etc., which is not limited in this embodiment. Optionally, the object identity information may be, but is not limited to, the name of the target object.

[0170] In an embodiment of the present application, the matching similarities between Q identity features stored in a feature database and the part features are determined, where Q is a natural number greater than or equal to 1. The matching similarities are sorted to obtain a sorting result. The object identity information indicated by the identity feature corresponding to the maximum matching similarity in the sorting result is determined as the identity information of the target object. In other words, in this embodiment, by comparing the part features with Q identity features stored in the feature database that can represent the true identity of the object, thereby determining the identity information of the target object, the accuracy of identity information recognition is improved.

[0171] Alternatively, as an optional implementation, it can be based on but not limited to the following Figure 7 The following steps are used to illustrate the overall pre-processing process of the above-mentioned identity information recognition method:

[0172] Step S702: Perform Fourier transform processing on the sample image to obtain frequency components corresponding to the sample image.

[0173] Step S704 : performing sample frequency band removal processing on the frequency components of the sample image to obtain a frequency-removed sample image, wherein the sample frequency band is a randomly determined frequency band.

[0174] Step S706 : Input the sample image and the de-frequencyed sample image into a feature recognition network to obtain sample image features of the sample image and de-frequencyed sample image features of the de-frequencyed sample image.

[0175] Step S708, obtaining a loss value between the sample image feature of the sample image and the de-frequency sample image feature of the de-frequency sample image;

[0176] Step S710, determining a reference frequency band based on a loss value between a sample image feature of the sample image and a de-frequencyed sample image feature of the de-frequencyed sample image;

[0177] Step S712, removing the reference frequency band in the sample image to obtain a training sample image;

[0178] Step S714: train the initialized feature extraction network using the training sample images until the feature extraction network reaches convergence.

[0179] By adopting the embodiment of the present application, by pre-determining the content that will interfere with identity recognition (i.e., the reference frequency band) before identifying the identity information of the target image, when identifying the identity information of the target image, the image content that may interfere with identity recognition can be directly removed according to the above-mentioned reference frequency band. This solves the technical problem of inaccurate identity information recognition in the identity information recognition method provided in the prior art, and achieves the technical effect of improving the accuracy of identity information recognition. Furthermore, by using the image corresponding to the key frequency band outside the reference frequency band in the sample image (i.e., the training sample image) to train the feature extraction network, the feature extraction network obtained after training can focus on extracting image features corresponding to the key frequency band. This achieves the purpose of obtaining an accurate feature extraction network even when there is less training sample data.

[0180] Optionally, as an optional implementation, assuming that the above-mentioned identification method of identity information is applied to the palm print recognition scenario of the access control system, it can be based on but not limited to the following: Figure 8 The following steps are used to illustrate the overall application process of the above-mentioned identity information recognition method:

[0181] Step S802: upon receiving a signal from a user to enter the area A protected by the access control device 802, the access control device 802 obtains a palm print image of the user, wherein the palm print image contains the user's palm print for identity recognition.

[0182] Step S804, the access control device 802 sends the user's palm print image to the server 804;

[0183] Step S806: the server 804 performs frequency domain transformation on the palmprint image to obtain the frequency components of the palmprint image in various frequency bands.

[0184] Step S808: The server 804 removes frequency bands (e.g., muscle frequency bands, skin frequency bands, etc.) that may interfere with identity recognition from the frequency components of the palmprint image in each frequency band, thereby obtaining palmprint frequency bands in the palmprint image.

[0185] Step S810: the server 804 performs time domain transformation processing on the frequency components in the palmprint frequency band to obtain a palmprint image;

[0186] Step S812: the server 804 uses a feature extraction network to extract features from the palmprint image to obtain palmprint features of the user's palmprint;

[0187] In step S814, the server 804 searches the feature database for a user identifier that matches the palm print feature of the user, and then determines whether the user has the authority to enter area A based on the user identifier.

[0188] Step S816: If the server 804 determines that the user has the right to enter the area A, it sends a prompt message for unlocking the access control device to the access control device 802;

[0189] Step S818: When the access control device 802 receives the prompt information, it opens and unlocks the access control device so that the user can enter area A.

[0190] The present invention uses an embodiment of the present invention to remove frequency domain features that do not carry the subject's identity information from a palm image to obtain a palm print image. The palm print features extracted from the palm print image are then used to identify the user's identity information, making the determined user identity information more accurate. This also improves the reliability of access control equipment.

[0191] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0192] According to another aspect of the embodiment of the present application, there is also provided an identity information identification device for implementing the above-mentioned identity information identification method. Figure 9 As shown, the device includes:

[0193] An acquisition unit 902 is configured to acquire a target image of identity information to be recognized, wherein the target image presents a target portion of a target object for identity recognition;

[0194] a removal unit 904 configured to remove a reference frequency band from the target image to obtain a key frequency band image, wherein the reference frequency band is determined based on a frequency band preprocessing network configured to determine a difference between a sample image and a de-frequencyed sample image, wherein the de-frequencyed sample image is an image obtained by frequency band filtering the sample image;

[0195] The identification unit 906 is configured to identify the identity information of the target object based on the part features of the target part extracted from the key frequency band image.

[0196] Optionally, in this embodiment, the above-mentioned removal unit includes: a transformation processing module, which is used to transform the target image to obtain the frequency components of the target image in each frequency band; a determination module, which is used to determine the reference frequency components that match the reference frequency band from the frequency components in each frequency band; a removal module, which is used to remove the reference frequency components to obtain the key frequency band; and an inverse transformation processing module, which is used to perform inverse transformation processing on the frequency components in the key frequency band to obtain the key frequency band image.

[0197] Optionally, in this embodiment, the above-mentioned device also includes: a first acquisition unit, used to acquire multiple groups of sample image pairs, wherein each group of sample image pairs includes a sample image and a de-frequencyed sample image; an extraction unit, used to extract sample image features corresponding to the sample image, and de-frequencyed sample image features corresponding to the de-frequencyed sample image in the frequency band preprocessing network; a first determination unit, used to determine the de-frequencying confidence between the sample image and the de-frequencyed sample image based on the feature similarity between the sample image features and the de-frequencyed sample image features; a second determination unit, used to determine the reference frequency band according to the sorting result of the de-frequencying confidence when the de-frequencying confidence corresponding to each of the multiple groups of sample image pairs is acquired.

[0198] Optionally, in this embodiment, the above-mentioned first acquisition unit includes: a first transformation processing module, used to transform the current sample image to obtain sample frequency components of the current sample image in M ​​frequency bands; a filtering processing module, used to filter the sample frequency components of the current sample image through the i-th frequency domain filter used to filter the i-th frequency band to obtain the i-th filtered sample frequency component, where i is a natural number greater than or equal to 1 and less than or equal to M; a first inverse transformation processing module, used to inverse transform the i-th filtered sample frequency component to obtain the i-th de-frequencyed reference sample image; a combination module, used to combine the current sample image with the M de-frequencyed reference sample images corresponding to each of the M frequency bands, respectively, to obtain M groups of sample image pairs corresponding to the current sample image; and an acquisition module, used to obtain sample image pairs corresponding to each of the N sample images, to obtain multiple groups of sample image pairs.

[0199] Optionally, in this embodiment, the above-mentioned first determination unit includes: a first acquisition module, used to obtain the cosine distance between the sample image feature and the de-frequencyed sample image feature as the feature similarity; a first determination module, used to determine the difference between the target threshold and the feature similarity as the de-frequencying confidence.

[0200] Optionally, in this embodiment, the second determination unit includes: a sorting module, configured to sort the determined de-frequencying confidences in descending order to obtain a sorted sequence; and a second determination module, configured to determine, as a reference frequency band, a frequency band corresponding to the de-frequencying sample image corresponding to each of the first k de-frequencying confidences in the sorted sequence, where k is a natural number greater than or equal to 1 and less than or equal to P, and P is the number of the determined de-frequencying confidences.

[0201] Optionally, in this embodiment, the above-mentioned recognition unit includes: a feature extraction module, which is used to use a feature extraction network to extract features from the key frequency band image to obtain the part features of the target part, wherein the feature extraction network is trained using the key frequency band located outside the reference frequency band in the sample image, and the frequency domain features of the key frequency band carry the identity information of the object; a search module, which is used to search for the object identity information that matches the part features in the feature database as the identity information of the identified target object.

[0202] Optionally, in this embodiment, the above-mentioned device also includes: a training unit, which is used to determine each sample image as the current sample image in turn, and perform the following steps: removing the reference frequency band in the current sample image to obtain the current training sample image; using the initialized feature extraction network to extract features of the current training sample image to obtain the current training sample features; obtaining the training loss value between the current training sample features and the label sample features, wherein the label sample features are the part features of the target part extracted in advance from the current sample image; when the training loss value reaches the convergence condition, the currently trained feature extraction network is determined as the feature extraction network.

[0203] Optionally, in this embodiment, the above-mentioned search module is also used to determine the matching similarities between Q identity features stored in the feature database and the part features, where Q is a natural number greater than or equal to 1; sort the matching similarities to obtain a sorting result; and determine the object identity information indicated by the identity feature corresponding to the maximum matching similarity in the sorting result as the identity information of the target object.

[0204] For specific embodiments, reference may be made to the examples shown in the above-mentioned method for identifying identity information, and this embodiment will not be described in detail here.

[0205] According to another aspect of the embodiment of the present application, an electronic device for implementing the above-mentioned identification method of identity information is also provided. The electronic device may be Figure 1 The terminal device or server shown in FIG. This embodiment is described by taking the electronic device as a terminal as an example. Figure 10As shown, the electronic device includes a memory 1002 and a processor 1004. The memory 1002 stores a computer program, and the processor 1004 is configured to execute the steps in any of the above method embodiments through the computer program.

[0206] Optionally, in this embodiment, the electronic device may be located in at least one network device among a plurality of network devices of a computer network.

[0207] Optionally, in this embodiment, the processor may be configured to execute the following steps through a computer program:

[0208] S1, obtaining a target image of identity information to be identified, wherein the target image presents a target part of a target object for identity identification;

[0209] S2, removing a reference frequency band from the target image to obtain a key frequency band image, wherein the reference frequency band is determined based on a frequency band preprocessing network, which is used to determine the difference between the sample image and the de-frequencyed sample image, and the de-frequencyed sample image is an image obtained by frequency band filtering the sample image;

[0210] S3, identifying the identity information of the target object based on the part features of the target part extracted from the key frequency band image.

[0211] Alternatively, those skilled in the art will appreciate that Figure 10 The structure shown is for illustration only, and the electronic device may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (MID), a PAD, or other terminal devices. Figure 10 It does not limit the structure of the above electronic device. For example, the electronic device may also include Figure 10 More or fewer components (such as network interfaces, etc.) as shown in, or with Figure 10 Different configurations shown.

[0212] Among them, the memory 1002 can be used to store software programs and modules, such as the program instructions / modules corresponding to the identity information identification method and device in the embodiment of the present application. The processor 1004 executes various functional applications and data processing by running the software programs and modules stored in the memory 1002, that is, realizes the above-mentioned identity information identification method. The memory 1002 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 1002 may further include a memory remotely located relative to the processor 1004, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned networks include but are not limited to the Internet, corporate intranets, local area networks, mobile communication networks and combinations thereof. Among them, the memory 1002 can be used for storing, but is not limited to, target images of objects, etc. As an example, Figure 10 As shown, the above-mentioned memory 1002 may include but is not limited to the acquisition unit 902, removal unit 904, and identification unit 906 in the above-mentioned identity information identification device. In addition, it may also include but is not limited to other module units in the above-mentioned identity information identification device, which will not be repeated in this example.

[0213] Optionally, the transmission device 1006 is configured to receive or send data via a network. Specific examples of the network may include a wired network and a wireless network. In one embodiment, the transmission device 1006 includes a network interface controller (NIC), which can be connected to other network devices and a router via a network cable to communicate with the Internet or a local area network. In one embodiment, the transmission device 1006 is a radio frequency (RF) module, which is configured to communicate with the Internet wirelessly.

[0214] In addition, the electronic device further includes: a connection bus 1008 for connecting various module components in the electronic device.

[0215] In other embodiments, the terminal device or server may be a node in a distributed system, wherein the distributed system may be a blockchain system, and the blockchain system may be a distributed system formed by connecting multiple nodes via network communication. The nodes may form a point-to-point network, and any computing device, such as a server, terminal, or other electronic device, may become a node in the blockchain system by joining the point-to-point network.

[0216] According to one aspect of the present application, a computer program product is provided, comprising a computer program / instructions containing program code for executing the above-described method. In such an embodiment, the computer program can be downloaded and installed from a network via a communication component and / or installed from a removable medium. When the computer program is executed by a central processing unit, the various functions provided in the embodiments of the present application are performed.

[0217] According to one aspect of the present application, a computer-readable storage medium is provided. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the above method.

[0218] Optionally, in this embodiment, the computer-readable storage medium may be configured to store a computer program for performing the following steps:

[0219] S1, obtaining a target image of identity information to be identified, wherein the target image presents a target part of a target object for identity identification;

[0220] S2, removing a reference frequency band from the target image to obtain a key frequency band image, wherein the reference frequency band is determined based on a frequency band preprocessing network, which is used to determine the difference between the sample image and the de-frequencyed sample image, and the de-frequencyed sample image is an image obtained by frequency band filtering the sample image;

[0221] S3, identifying the identity information of the target object based on the part features of the target part extracted from the key frequency band image.

[0222] It should be noted that in this application, when the facial (or other biometric) recognition technology involved is applied to specific products or technologies in the above embodiments of this application, the relevant data collection, use and processing processes should comply with national laws and regulations. Before collecting facial information, the information processing rules should be informed and the target object's separate consent should be obtained. Facial information should be processed strictly in accordance with the requirements of laws and regulations and personal information processing rules, and technical measures should be taken to ensure the security of relevant data.

[0223] Optionally, in the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program that has a predetermined function and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0224] Optionally, in this embodiment, a person of ordinary skill in the art may understand that all or part of the steps in the various methods of the above embodiments may be completed by instructing the hardware related to the terminal device through a program, and the program may be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0225] If the integrated units in the above embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in the above-mentioned 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, which is stored in a storage medium and includes several instructions for enabling one or more computer devices (which can be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application.

[0226] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0227] In the several embodiments provided in this application, it should be understood that the disclosed client can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. 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. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, and can be electrical or other forms.

[0228] 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.

[0229] 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.

[0230] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for identifying identity information, characterized in that: include: Acquire a target image of identity information to be identified, wherein the target image presents a target part of a target object for identity identification; removing a reference frequency band from the target image to obtain a key frequency band image, wherein the reference frequency band is determined based on a frequency band preprocessing network, the reference frequency band includes a frequency band corresponding to an interfering image content, the frequency band preprocessing network is used to determine a difference between a sample image and a de-frequencyed sample image, the de-frequencyed sample image is an image obtained by removing a plurality of sample frequency bands from the sample image, the plurality of sample frequency bands being frequency bands generated based on sample frequency components, the sample frequency components being frequency components obtained by transforming the sample image; The identity information of the target object is identified based on the part features of the target part extracted from the key frequency band image.

2. The method according to claim 1, characterized in that The removing of the reference frequency band in the target image to obtain the key frequency band image includes: Performing transformation processing on the target image to obtain frequency components of the target image in various frequency bands; Determining, from the frequency components in each frequency band, a reference frequency component that matches the reference frequency band; Removing the reference frequency component to obtain a key frequency band; An inverse transformation process is performed on the frequency components in the key frequency band to obtain the key frequency band image.

3. The method according to claim 1, characterized in that Before obtaining the target image of the identity information to be identified, the method further includes: Acquire multiple groups of sample image pairs, wherein each group of the sample image pairs includes the sample image and the de-frequencyed sample image; The following operations are performed on each set of sample image pairs in sequence: Extracting, in the frequency band preprocessing network, sample image features corresponding to the sample image and de-frequency sample image features corresponding to the de-frequency sample image; Determining a de-frequencying confidence between the sample image and the de-frequencying sample image based on a feature similarity between the sample image feature and the de-frequencying sample image feature; When the frequency removal confidences corresponding to the plurality of groups of sample image pairs are obtained, the reference frequency band is determined according to a ranking result of the frequency removal confidences.

4. The method according to claim 3, characterized in that The acquiring of multiple groups of sample image pairs comprises: The obtained N sample images are sequentially used as current sample images, and the following operations are performed: Performing transformation processing on the current sample image to obtain sample frequency components of the current sample image in M ​​frequency bands; Repeat the following steps until the M frequency bands corresponding to the current sample image are traversed: performing filtering processing on the sample frequency component of the current sample image using an i-th frequency domain filter for filtering the i-th frequency band to obtain an i-th filtered sample frequency component, where i is a natural number greater than or equal to 1 and less than or equal to M; Performing inverse transformation on the i-th filtered sample frequency component to obtain an i-th de-frequency reference sample image; Combining the current sample image with the M frequency-demultiplexed reference sample images corresponding to the M frequency bands, respectively, to obtain M groups of sample image pairs corresponding to the current sample image; Obtain sample image pairs corresponding to each of the N sample images to obtain the multiple groups of sample image pairs.

5. The method according to claim 3, characterized in that The determining of the de-frequencying confidence between the sample image and the de-frequencying sample image based on the feature similarity between the sample image feature and the de-frequencying sample image feature includes: Obtaining a cosine distance between the sample image feature and the de-frequency sample image feature as the feature similarity; The difference between the target threshold and the feature similarity is determined as the frequency removal confidence.

6. The method according to claim 5, characterized in that Determining the frequency domain features of the reference frequency band according to the sorting results of the frequency removal confidences includes: Sorting the determined frequency removal confidences in descending order to obtain a sorted sequence; The frequency band corresponding to the de-frequencying sample image corresponding to each of the first k de-frequencying confidences in the sorting sequence is determined as the reference frequency band, where k is a natural number greater than or equal to 1 and less than or equal to P, and P is the number of the determined de-frequencying confidences.

7. The method according to claim 1, characterized in that The identifying of the identity information of the target object based on the part features of the target part extracted from the key frequency band image includes: Extracting features from the key frequency band image using a feature extraction network to obtain part features of the target part, wherein the feature extraction network is trained using key frequency bands in the sample image that are outside the reference frequency band, and the frequency domain features of the key frequency bands carry identity information of the object; The object identity information that matches the part feature is searched in the feature database as the identity information of the identified target object.

8. The method according to claim 7, characterized in that Before obtaining the target image of the identity information to be identified, the method further includes: Each sample image is determined as the current sample image in turn, and the following steps are performed: removing the reference frequency band in the current sample image to obtain a current training sample image; Performing feature extraction on the current training sample image using the initialized feature extraction network to obtain current training sample features; Obtaining a training loss value between the current training sample feature and the label sample feature, wherein the label sample feature is a part feature of the target part pre-extracted from the current sample image; When the training loss value reaches a convergence condition, the feature extraction network currently being trained is determined as the feature extraction network.

9. The method according to claim 7, characterized in that The object identity information matching the part feature is found in the feature database as the identified target object identity information, including: Determining the matching similarity between Q identity features stored in the feature database and the part features, where Q is a natural number greater than or equal to 1; Sorting the matching similarities to obtain a sorting result; The object identity information indicated by the identity feature corresponding to the maximum matching similarity in the sorting result is determined as the identity information of the target object.

10. An identity information recognition device, characterized in that: include: an acquiring unit, configured to acquire a target image of identity information to be identified, wherein the target image presents a target portion of a target object for identity identification; a removal unit configured to remove a reference frequency band from the target image to obtain a key frequency band image, wherein the reference frequency band is determined based on a frequency band preprocessing network, the reference frequency band includes a frequency band corresponding to an interference image content, the frequency band preprocessing network is configured to determine a difference between a sample image and a de-frequencyed sample image, the de-frequencyed sample image is an image obtained by removing a plurality of sample frequency bands from the sample image, the plurality of sample frequency bands being frequency bands generated based on sample frequency components, the sample frequency components being frequency components obtained by transforming the sample image; The identification unit is configured to identify the identity information of the target object based on the part features of the target part extracted from the key frequency band image.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program is executed by a processor to perform the method according to any one of claims 1 to 9.

12. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

13. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 9 through the computer program.

Images