IDENTIFICATION METHOD, APPARATUS, COMPUTER DEVICE AND COMPUTER PROGRAM

Abstract

The identity identification method includes a step (202) of acquiring a biological part image obtained for a target part of a user to be identified; a step (204) of identifying a feature morphology type of the target part and extracting morphological features of each pixel from the biological part image in accordance with an image feature morphology that matches the feature morphology type, where the image feature morphology is a morphology formed by combining the distribution positions of each of the feature extraction coverage pixels, and the feature extraction coverage pixels are pixels in the biological part image that is the target at the time of each feature extraction; a step (206) of obtaining biological part features of the user to be identified based on the morphological features of each pixel; and a step (208) of matching the biological part features with registered part features of a registered user to obtain a feature matching result, and specifying an identity identification result of the user to be identified in accordance with the feature matching result.

Description

[Technical Field]

[0001] [CROSS-REFERENCE TO RELATED APPLICATIONS] This application claims the benefit of priority to Chinese Patent Application No. 2023111498217, entitled "IDENTIFICATION METHOD, APPARATUS, COMPUTER DEVICE AND STORAGE MEDIUM," filed with the State Intellectual Property Office of the People's Republic of China on September 7, 2023, the entire contents of which are incorporated herein by reference.

[0002] [Technical field] The present application relates to the field of computer technology, and in particular to an identification method, apparatus, computer device, storage medium and computer program product. [Background technology]

[0003] With the development of computer technology, increasingly mature identity authentication technology has been widely used in various fields, such as business collaboration, payment services, social media, security systems, etc. Among these, the use of human inherent biometric features, such as hand shape, fingerprints, face shape, retina, ear pinna, and other localized parts for identity authentication, has become the development direction of identity authentication technology.

[0004] Currently, in identity authentication technologies based on biometric features of local body parts, when a user authenticates themselves based on biometric features such as hand shape, face, fingerprints, or palm prints, they typically collect images of the local body parts and extract biometric features from the collected images to authenticate the user. However, there has traditionally been a problem in that the accuracy of extracting biometric parts is low, which also reduces the accuracy of identity authentication. Summary of the Invention [Means for solving the problem]

[0005] According to various embodiments of the present application, an identity identification method, apparatus, computer device, computer readable storage medium and computer program product are provided.

[0006] According to one aspect of the present application, there is provided an identity identification method performed by a computing device, the method comprising: acquiring a body part image obtained for a target part of a user to be identified; identifying a feature morphology type of the target region, and extracting a morphology feature for each pixel from the biological region image according to at least one image feature morphology that matches the feature morphology type, wherein the image feature morphology is a morphology formed by combining the distribution positions of each feature extraction coverage pixel, and the feature extraction coverage pixel is a pixel in the biological region image that is a target for each feature extraction; obtaining a body part feature of the user to be identified based on the morphological feature of each pixel; The method includes a step of matching the biological part features with the registered part features to obtain a feature matching result, and specifying the identity identification result of the user to be identified according to the feature matching result, wherein the registered part features are biological part features obtained by registering the registered user's identity on a biological part image corresponding to a target part of the registered user.

[0007] According to another aspect of the present application, there is further provided an identification device adapted for use in a computer device, the device comprising: a body part image acquisition module configured to perform a step of acquiring a body part image obtained for a target body part of a user to be identified; a morphological feature extraction module configured to perform the steps of identifying a feature morphology type of the target region and extracting a morphological feature for each pixel from the image of the biological region according to at least one image feature morphology that matches the feature morphology type, wherein the image feature morphology is a morphology formed by combining the distribution positions of each of the feature extraction coverage pixels, and the feature extraction coverage pixels are pixels in the image of the biological region that are the target for each feature extraction; a body part feature acquisition module configured to perform a step of acquiring a body part feature of a user to be identified based on the morphological features of each pixel; and a feature matching module configured to execute a step of matching the biometric feature with the registered biometric feature to obtain a feature matching result, and specifying the identity identification result of the user to be identified according to the feature matching result, wherein the registered biometric feature is a biometric feature obtained by performing personal registration on a biometric feature image corresponding to a target body part of the registered user.

[0008] According to another aspect of the present application, there is further provided a computing device including a memory having computer-readable instructions stored thereon, and a processor, the computer-readable instructions, when executed, causing the processor to perform the steps of the above-described method for identity identification.

[0009] According to another aspect of the present application, there is further provided a computer-readable storage medium having stored thereon computer-readable instructions that, when executed, cause a processor to perform the steps of the above-described method for identity identification.

[0010] According to another aspect of the present application, there is further provided a computer program product including computer readable instructions that, when executed, cause a processor to perform the steps of the above-described identification method.

[0011] The details of one or more embodiments of the application are set forth in the drawings and description below. Other features and advantages of the application will become apparent from the description, drawings, and claims. [Brief explanation of the drawings]

[0012] In order to more clearly describe the embodiments of the present application or the technical methods of the prior art, the following briefly introduces the drawings used in the description of the embodiments or the prior art. It is obvious that the drawings described below are only the embodiments of the present application, and those skilled in the art can obtain other drawings from these disclosed drawings without exerting any labor equivalent to an inventive step.

[0013] [Figure 1] FIG. 1 is a diagram illustrating an application environment of an identity identification method according to an embodiment. [Figure 2] 1 is a flowchart of an identity identification method in one embodiment. [Figure 3] FIG. 1 is a schematic diagram illustrating the application of palm print authentication in one embodiment. [Figure 4] FIG. 2 is a schematic diagram of a circular feature extractor in one embodiment; [Figure 5] FIG. 2 is a schematic diagram of a linear feature extraction unit according to an embodiment. [Figure 6] FIG. 1 is a schematic diagram of a circular convolution kernel in one embodiment. [Figure 7] FIG. 1 is a schematic diagram of a linear convolution kernel in one embodiment. [Figure 8] 10 is a flowchart of an identity identification method according to another embodiment. [Figure 9] 1 is a flowchart illustrating region of interest identification in one embodiment. [Figure 10] 1 is a flowchart of a palm print authentication method according to an embodiment. [Figure 11] FIG. 1 is a schematic diagram illustrating region of interest identification in one embodiment. [Figure 12] FIG. 1 is a schematic diagram of a 9×9 linear convolution kernel in one embodiment. [Figure 13] FIG. 1 is a schematic diagram of a 16×16 linear convolution kernel in one embodiment. [Figure 14] 1 is a schematic diagram illustrating the configuration of an identification device according to an embodiment. [Figure 15] FIG. 2 is a diagram illustrating the internal structure of a computing device in one embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0014] Hereinafter, the technical methods in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. It is obvious that the embodiments described in this specification are only a part of the embodiments of the present application, not all of them. Based on the embodiments of the present application, all other embodiments that a person skilled in the art can obtain without exerting any labor equivalent to an inventive step shall fall within the scope of protection of the present application.

[0015] The identity identification method according to an embodiment of the present application can be applied to the application environment shown in FIG. 1 . In the application environment, a terminal 102 is communicatively connected to a server 104 via a network. A data storage system can store data to be processed by the server 104. The data storage system can be configured separately, integrated into the server 104, or located on a cloud or other server. The terminal 102 can collect images of a target part of a user to be identified and obtain a biological part image of the target part. Specifically, the terminal 102 can collect biological part images of the target part of the user to be identified in response to an identity authentication trigger event. For example, the terminal 102 can collect biological part images, such as images of the palm or fingers of the user to be identified. The terminal 102 transmits the collected biological part images to the server 104 and uses the server 104 to extract morphological features of each pixel from the biological part image according to at least one image feature type matching the target part feature type type. The image feature type is formed by combining the distribution positions of the target pixels within each pixel during feature extraction. The server 104 obtains the biological part features of the user to be identified based on the morphological features of each pixel, compares the biological part features with the registered part features of the registered users, and determines the identity identification result of the user to be identified based on the feature comparison result. By using the server 104 to return the identity identification result of the user to be identified to the terminal 102, the terminal 102 can perform processing based on the identity identification result, and can, for example, lift access restrictions on the user to be identified.

[0016] Alternatively, by using the server 104 to directly return the feature matching result to the terminal 102, the terminal 102 can determine the identity identification result of the target user according to the feature matching result returned from the server 104, thereby realizing the identity identification of the target user. In another selectable application example, the identity identification process may be performed solely by the terminal 102. That is, the terminal 102 extracts morphological features of each pixel from the body part image according to at least one image feature form matching the feature morphology type of the target part, obtains body part features of the user to be identified based on the morphological features of each pixel, matches the body part features with the registered body part features of the registered users, and determines the identity identification result of the user to be identified according to the feature matching result.

[0017] The terminal 102 may include, but is not limited to, various personal computers, laptops, smartphones, tablets, Internet of Things devices, portable wearable devices, etc. The Internet of Things devices may include smart speakers, smart TVs, smart air conditioners, smart car devices, etc. The portable wearable devices may include smart watches, smart bracelets, head-mounted devices, etc. To collect biometric features of the target area, the terminal 102 is equipped with a sensor device for collecting area images of the user's target area. The server 104 may be implemented as an independent server, a server cluster consisting of multiple servers, or based on a cloud server.

[0018] In one embodiment, an identity identification method is provided, as shown in Fig. 2. The method is executed by a computer device, and specifically, may be executed by a computer device such as a terminal or a server alone, or may be executed by a terminal and a server in cooperation with each other. In one embodiment of the present application, the method will be described by taking as an example that the method is applied to the server in Fig. 1. The method includes the following steps 202 to 208.

[0019] Step 202: A body part image obtained for a target part of the user to be identified is acquired.

[0020] Among these, identity verification is an authentication process that verifies whether a user's actual identity matches their claimed identity. With the development of identity verification technology, identity verification methods based on biometric characteristics have become widely used. Identity verification processes are applied to various scenarios and are triggered by identity verification trigger events. An identity verification trigger event refers to an event that triggers identity verification, including, but not limited to, an action or instruction that triggers the identity verification process. For example, in an access control system scenario, an identity verification event may be triggered when a user needs to pass through an automatic door with access control functionality, thereby authenticating the user. Another example is when a user pays at a payment terminal. As shown in Figure 3, the target body part may be a palm. The terminal may collect a palm image of the user's hand to obtain a biometric body part image. If the biometric body part image is specifically a palm image, identity verification processes can be performed based on the palm image. Furthermore, identity verification can also be applied to addiction prevention system scenarios. For example, in an online game addiction prevention system, if it is necessary to limit the amount of time a minor spends playing games, an identity authentication event for authenticating the game user may be triggered when the addiction prevention system is activated, such as when the game user's cumulative game play time reaches a preset time threshold. Identity authentication can identify whether the game user is an adult or the owner of a game account, thereby making it possible to limit the amount of time the minor spends playing games.

[0021] The identification process is realized based on collected biometric features. Biometric features are features of a user's measurable body parts, such as hand shape, fingerprint, face shape, iris, retina, palm, and various other types of biometric features. When performing the identification process using the biometric features of a user's measurable body parts, it is necessary to collect biometric data about the user's body parts, extract biometric features from the collected biometric data, and identify the user based on the extracted biometric features. For example, when performing identification based on fingerprint authentication, it is necessary to collect fingerprint data about the user's fingers and identify the user based on the collected fingerprint data, such as fingerprint images. For example, when performing identification based on the palm, it is necessary to collect palm data from the user's palm and identify the user based on the collected palm data.

[0022] A user to be identified refers to a user who needs to be identified, such as a user who triggers an identity authentication event. For example, a user can enter the data collection area of ​​an access control system by passing through an automatic door with access control functionality. When the access control system detects the user's presence in the data collection area, identity authentication is required, triggering identity authentication. The access control system then collects biometric data of the user to be identified in the data collection area (various biometric data, such as the face data, finger data, and palm data of the user to be identified). The target body part is the part of the human body corresponding to the biometric data to be collected. The target body part relates to the biometric data or biometric features related to identity authentication. For example, if identity authentication is based on face, the corresponding target body part is the facial part of the user to be identified who needs identity authentication, the collected biometric data is face data (specifically, a face image), and the biometric features required for identity identification are face features. For example, if personal authentication is based on the palm of the hand, the corresponding target part is the palm of the user to be identified, the collected biometric data is palm data (specifically, a palm image), and the biometric features required for identity identification are palm features. A biometric part image is an image collected of the target part, and includes the biometric features of the corresponding target part. For example, if the target part is a face, the biometric part image may be a face image. If the target part is a finger, the biometric part image may be a finger image. If the target part is a palm, the biometric part image may be a palm image.

[0023] Specifically, the server can acquire body part images. The body part images are collected for a target part of the user to be identified. For example, a camera can be used to collect body part images for the target part of the user to be identified. Body part images with different biometric features can be collected for different target parts. Specifically, the terminal can collect images for the target part of the user to be identified and transmit the collected body part images to the server.

[0024] Step 204: Identify a feature morphology type of the target part, and extract morphology features for each pixel from the biological part image according to at least one image feature morphology that matches the feature morphology type, where the image feature morphology is a morphology formed by combining the distribution positions of the feature extraction coverage pixels, and the feature extraction coverage pixels are pixels in the biological part image that are targeted for each feature extraction.

[0025] Among these, the feature shape type refers to the corresponding expression shape of the biometric feature of the target part. Different target parts have different expression shapes of biometric feature information. For example, in the case of a finger, the biometric feature information is mainly characterized by the fingerprint of the finger, so the feature shape type may be composed of the skin ridges that appear on the pads of the fingertips, specifically including shape types such as whorl (hole type), arch (arch type), and loop (loop type). In addition, in the case of a palm, the biometric feature information is characterized by the palm print or palm vein shape, so the corresponding feature shape type may be the palm print or palm vein shape, specifically a line shape type.

[0026] The image feature morphology refers to the feature morphology used when performing feature extraction on the target region's biological part image. The image feature morphology is compatible with the target region's feature morphology type, specifically, the expression morphology corresponding to the target region's biological feature, making it suitable for extracting the target region's biological feature. Different types of image feature morphology correspond to different feature extraction methods, so feature extraction can be performed using a feature extraction method compatible with the target region's feature morphology type, improving the feature extraction effect. Optionally, the adaptive correspondence between the image feature morphology and the feature morphology type can be established based on the results of multiple tests. In a specific embodiment, multiple candidate image feature morphologies can be identified in advance, and feature extraction can be performed individually for each candidate image feature morphology for each biological part's feature morphology type. Based on the feature extraction results corresponding to each candidate image feature morphology, the server can identify an image feature morphology suitable for the feature morphology type from each candidate image feature morphology, thereby establishing an adaptive correspondence between the image feature morphology and the feature morphology type.

[0027] An image feature morphology is a morphology formed by combining the respective distribution positions of target pixels each time feature extraction is performed. Specifically, each time feature extraction is performed on a pixel in a biological part image, the morphology formed by combining the respective distribution positions of the target pixels, i.e., feature extraction coverage pixels, in the biological part image matches the image feature morphology. The image feature morphology can be realized by a feature extraction means. For example, if the target part is a finger, the feature morphology type may include a spiral morphology, and the image feature morphology may be a spiral morphology. Specifically, feature extraction is performed by a spiral feature extraction means, i.e., the morphology formed by combining the respective distribution positions of feature extraction coverage pixels in the biological part image each time feature extraction is performed is a spiral morphology. In a specific application example, the morphology formed by combining the distribution positions of target pixels during each feature extraction is identified, and identification can be performed based on the formed morphology. For example, by comparing the formed morphology with each candidate image feature morphology, it can be determined whether the formed morphology matches the image feature morphology, and therefore it can be determined whether the morphology formed by combining the distribution positions of target pixels during each feature extraction is an image feature morphology that matches the feature morphology type. Since the image feature form can include at least one, feature extraction can be performed for one or more image feature forms on the biological part image, and morphological features of each pixel in different image feature forms can be obtained.

[0028] In a specific application example, the image feature shape is realized by a corresponding feature extraction means. That is, feature extraction is performed by a feature extraction means whose shape matches the image feature shape, thereby obtaining the respective morphological features of each pixel. As shown in FIG. 4, if the image feature shape is circular, feature extraction can be performed on the biological part image by circular feature extraction means 1 and feature extraction means 2. Feature extraction means 1 and feature extraction means 2 may have different sizes. That is, when feature extraction is performed on the biological part image, feature extraction can be performed only on pixels within the circular area covered by the circular feature extraction means. As shown in FIG. 5, if the image feature shape is linear, feature extraction can be performed on the biological part image by linear feature extraction means. For example, feature extraction can be performed on the biological part image by horizontal feature extraction means A. By rotating horizontal feature extraction means A, feature extraction means B, feature extraction means C, and feature extraction means D can be obtained in different directions depending on the rotation angle, thereby forming a multi-directional feature extraction means for feature extraction.

[0029] Morphological features are the results of feature extraction performed on a biological part image according to its image feature form. Morphological features correspond to pixels, meaning that a morphological feature can be extracted individually for each pixel in the biological part image. For pixels in the biological part image, corresponding morphological features can be extracted for each pixel with each image feature form. For example, if feature extraction is performed on a biological part image with three types of image feature forms, morphological features corresponding to each of the three image feature forms can be extracted for each pixel, meaning that three types of morphological features can be extracted for each pixel.

[0030] Specifically, the server identifies at least one image feature morphology for the body part image and performs feature extraction on the body part image based on the at least one image feature morphology to obtain morphological features for each pixel in the body part image. When feature extraction is performed on the body part image according to the image feature morphology, the corresponding image feature morphology can be formed by combining the distribution positions of the pixels targeted during each feature extraction, i.e., the feature extraction coverage pixels. Here, the at least one image feature morphology matches a target part feature morphology type, and the server can identify the target part feature morphology type and at least one image feature morphology that matches it according to the feature morphology type. In a specific embodiment, after the server identifies the target part feature morphology type, if there are multiple feature morphology types, the server can narrow down at least one feature morphology type from the multiple feature morphology types. For example, it can narrow down at least one feature morphology type that appears most frequently and identify at least one corresponding image feature morphology. The server can also directly identify each corresponding image feature morphology according to multiple feature morphology types.

[0031] In one exemplary application, the image feature morphology can be realized by a feature extraction means. The feature extraction means is specifically realized by a convolution kernel. That is, feature extraction is performed on the body part image using at least one convolution kernel that matches the feature morphology type of the target part, thereby obtaining each morphology feature of each pixel in the body part image. For example, if the target part is a palm, the feature morphology type may be a line morphology, so at least one linear convolution kernel can be specified. Each linear convolution kernel can have a different direction, so feature extraction can be performed on the palm image based on at least one linear convolution kernel, thereby effectively extracting palm print line features in the palm image.

[0032] As shown in Figure 6, in a 16x16 convolution kernel, by setting the convolution kernel units shaded with diagonal lines to enabled (e.g., setting the convolution weight to 1) and the convolution kernel units not shaded with diagonal lines to disabled (e.g., setting the convolution weight to 0), a circular convolution kernel can be formed, thereby realizing feature extraction according to the circular image feature morphology. As shown in Figure 7, in a 16x16 convolution kernel, by setting the convolution kernel units shaded with diagonal lines to enabled and the convolution kernel units not shaded with diagonal lines to disabled, a horizontal linear convolution kernel can be formed. By performing feature extraction on the biological part image using this horizontal linear convolution kernel, horizontal linear morphological features can be obtained.

[0033] Step 206: Based on the morphological features of each pixel, the body part features of the user to be identified are obtained.

[0034] The biometric feature is used to characterize the biometric features of the target body part of the user to be identified. Different target body parts of different users may correspond to different biometric feature, so that the user's identity can be identified based on the biometric feature.

[0035] Specifically, for each morphological feature of each pixel in the body part image, the server may identify the body part feature of the user to be identified based on the morphological feature of each pixel. For example, if the amount of the morphological feature of each pixel is 1, the server may directly connect the morphological features of each pixel to obtain the morphological feature of the body part image, and obtain the body part feature of the user to be identified based on the morphological feature of the body part image. For example, the server may directly use the morphological feature of the body part image as the body part feature of the user to be identified, and may further perform feature extraction on the morphological feature of the body part image. Specifically, the morphological feature of the body part image may be further extracted using a pre-trained feature extraction model to obtain the body part feature of the user to be identified. Furthermore, if the amount of the morphological feature of each pixel is 2 or more, this indicates that there are at least two image feature forms. For each pixel, the server may fuse multiple morphological features of each pixel, further connect the fused morphological features to obtain the morphological feature of the body part image, and obtain the body part feature of the user to be identified based on the morphological feature of the body part image.

[0036] Step 208: Match the biometric features with the registered biometric features to obtain a feature matching result, and identify the identity of the user to be identified for each feature matching result. Here, the registered biometric features are biometric features obtained by registering the biometric images corresponding to the target body parts of the registered user.

[0037] Among these, the registered body part features are those obtained by registering the registered user's body part image corresponding to the target body part. The registered body part features of the registered user can be used as reference features for identity identification. The registered body part features of the user to be identified are matched with the registered body part features. Based on the feature matching results, it is possible to determine the identity identification result of the user to be identified, such as determining whether the user to be identified is a registered user. If the user is a registered user, it is possible to further determine the specific user identifier of the user to be identified.

[0038] Specifically, the server can perform identity identification based on the biometric features of the user to be identified. Specifically, the server can acquire registered biometric features obtained by the registered user through prior personal registration. The server compares the biometric features with each registered biometric feature and obtains each feature comparison result. The server determines the identity identification result of the user to be identified based on the feature comparison result. Specifically, the server can determine whether the user to be identified is a registered user based on each feature comparison result. If the user to be identified is a registered user, the server can further determine the user identification information of the user to be identified. In a specific application example, the user can register in advance using a biometric image corresponding to the target body part. Specifically, the server extracts biometric features based on the biometric image corresponding to the target body part, and associates the extracted biometric features with a user identifier to realize user identity registration. Furthermore, the server can acquire registered biometric features based on the registered user's biometric features during identity registration, and perform identity identification using the registered biometric features of each registered user as reference features.

[0039] In one specific application example, as shown in FIG. 8, first, body part images of a target part of a user to be identified may be collected, and a server may acquire the collected body part images. The server may identify a feature morphology type corresponding to the target part and identify at least one image feature morphology matching the feature morphology type. The server may perform feature extraction on the body part image according to the at least one image feature morphology, thereby obtaining morphology features for each pixel in the body part image. The server may obtain body part features of the user to be identified based on the morphology features for each pixel. The server may obtain feature matching results by matching the body part features of the user to be identified with registered body part features, and may determine the identity of the user to be identified based on the feature matching results.

[0040] In the above-described identity identification method, a biological part image of a target part of a user to be identified is acquired, and morphological features of each pixel are extracted from the biological part image according to at least one image feature form that matches the feature form type of the target part. An image feature form is formed by combining the distribution positions of each pixel targeted for feature extraction within each pixel. The biological part features of the user to be identified are obtained based on the morphological features of each pixel. The biological part features are matched with registered biological part features of registered users, and the identity identification result of the user to be identified is determined based on the feature match result. During the identity identification process, by extracting the morphological features of each pixel from at least one image feature form that matches the feature form type of the target part, the accurate expression of the biological part features of the target part is improved, and the biological part features of the target part can be obtained with high accuracy, thereby improving the accuracy of personal authentication based on the biological part features of the target part.

[0041] In one exemplary embodiment, when the target area is a palm and the image feature form is a linear form, the step of extracting a respective morphological feature of each pixel from the image of the biological part according to at least one image feature form that matches the feature form type includes the step of extracting a respective morphological feature of each pixel from the image of the biological part according to the linear form in at least one direction.

[0042] In this method, the target part is a palm, and the biometric part image corresponding to the target part is a palm image. The palm feature morphology type includes palm print or palm vein feature morphology, specifically, a linear morphology. A palm print refers to the geometric information of the palm from the fingertips to the wrist, including various palm print features used for identity identification, such as main lines, wrinkles, fine lines, ridges, and intersections. A palm print feature refers to features reflected in palm print information that can be extracted from a palm image by capturing the palm. Different users typically have different palm print features. That is, different users' palms have different geometric features, and identity identification processing for different users can be realized based on the palm print features. Palm vein information refers to palm vein information, which is used to reflect the vein line information of a human palm and has the ability to identify biometrics. A palm vein image may be captured using an infrared camera. A palm vein feature refers to the vein features of the palm obtained by palm vein analysis. Generally, different users correspond to different palm vein features, that is, the palms of different users have different vein features, so that the identity identification process of different users can also be realized based on the palm vein features.

[0043] The biometric feature information of a palm print or palm vein is usually reflected by palm print lines or vein lines. If the palm feature type is a linear feature, the image feature matching the feature type is also a linear feature. Linear features can be extracted in various directions, so they can be classified into various types. For example, linear features include horizontal, vertical, and various directions at certain angles to the horizontal. Linear features in different directions can be used as different types of linear features, so feature extraction can be performed for each linear feature in different directions.

[0044] For example, in the case of a palm image, the server can identify at least one directional linear feature that matches the palm feature type. The server can perform feature extraction on the body part image according to the identified at least one directional linear feature to obtain the morphological features of each pixel in the body part image. For example, the server can identify a linear feature extraction means corresponding to the linear feature in each direction and perform feature extraction on the body part image using the linear feature extraction means to obtain the morphological features of each pixel in the body part image. As shown in FIG. 5, the server can specify feature extraction on the body part image using one of feature extraction means A, feature extraction means B, feature extraction means C, and feature extraction means D.

[0045] In this embodiment, the palm feature shape type for the palm biometric image is a linear shape. The server extracts features from the biometric image according to at least one linear shape in one direction, thereby obtaining the shape features of each pixel in the biometric image. This allows for feature extraction from the palm image according to at least one linear shape in one direction. This improves the accuracy of palm feature representation and allows palm feature acquisition with high accuracy, thereby improving the accuracy of palm feature-based identity authentication.

[0046] In one exemplary embodiment, the step of extracting respective morphological features of each pixel from the image of the biological part according to linear forms in at least one direction includes the steps of extracting directional morphological features from the image of the biological part according to linear forms in at least two directions, each corresponding to each pixel and the linear forms in at least two directions; obtaining, for each pixel, a directional fusion feature of the target pixel by fusing the target pixel with the directional morphological features corresponding to the linear forms in at least two directions; and obtaining respective morphological features of each pixel according to the directional fusion feature of each pixel.

[0047] Among these, linear features in different directions correspond to different types of linear features, and each can correspond to a different feature extraction method. When feature extraction is performed on a biological part image for at least two image feature features, directional morphological features corresponding to at least two image feature features can be obtained for each pixel. For example, when features are extracted from a biological part image according to horizontal and vertical linear features, a directional fusion feature corresponding to the horizontal linear feature and a directional fusion feature corresponding to the vertical linear feature can be obtained for each pixel in the biological part image. The directional fusion feature is a feature obtained by fusing directional morphological features corresponding to linear features in each direction of the same pixel. Specifically, fusion can be performed using a weighted fusion method. Based on the directional fusion feature, a morphological feature of the corresponding pixel can be obtained.

[0048] Specifically, the server identifies linear features in at least two directions for the palm image, and performs feature extraction on the body part image for each of the linear features in the at least two directions, thereby obtaining directional morphological features corresponding to each pixel in the body part image. Each pixel includes directional morphological features corresponding to each of the linear features in the at least two directions. For example, if there are five linear features, after performing feature extraction on the body part image for each of the five linear features, for each pixel in the body part image, such as pixel A, pixel A includes directional morphological features corresponding to the linear features in each direction, i.e., pixel A includes five directional morphological features. In a specific embodiment, the server identifies linear feature extraction means corresponding to each linear feature in each direction, and performs feature extraction on the body part image using the linear feature extraction means to obtain morphological features for each pixel in the body part image. As shown in FIG. 5, the server can specify at least two of feature extraction means A, feature extraction means B, feature extraction means C, and feature extraction means D to perform feature extraction on the body part image.

[0049] For each pixel from which directional morphological features are extracted from the biological part image, the server fuses the directional morphological features of the target pixel to obtain a directional fusion feature for the pixel, traverses each pixel, and obtains a respective directional fusion feature for each pixel. In a specific application example, for each pixel, the server performs average fusion or weighted fusion on the directional morphological features of the target pixel to obtain the directional fusion feature for the pixel. The server can obtain a respective morphological feature for each pixel based on the respective directional fusion feature for each pixel. Specifically, the server can directly use the respective directional fusion feature for each pixel as the respective morphological feature. The server can also perform further feature mapping processing on the respective directional fusion feature for each pixel to obtain a respective morphological feature for each pixel.

[0050] In this embodiment, for a palm biometric image, the palm feature morphology type is a linear morphology. The server performs feature extraction on the biometric image for at least two linear morphologies in order to obtain directional morphological features for each pixel in the biometric image. The server then fuses the directional morphological features for each pixel to obtain a directional fusion feature, and obtains morphological features based on the directional fusion features for each pixel. This allows feature extraction on the palm image for each linear morphology in multiple directions and fusing the extraction results corresponding to the linear morphologies in each direction, thereby obtaining highly accurate palm features and improving the accuracy of palm feature-based identity authentication.

[0051] In one exemplary embodiment, the step of extracting each morphological feature of each pixel from the image of the biological part according to at least one image feature form that matches the feature form type includes the steps of identifying at least one image feature form that matches the feature form type at at least one image feature scale, and extracting each morphological feature of each pixel from the image of the biological part according to the at least one image feature form at the at least one image feature scale.

[0052] Among them, the image feature scale characterizes the image pixel range covered during each feature extraction. Different image feature scales can cover different pixel ranges. The larger the image feature scale value, the more pixels are targeted each time feature extraction is performed, i.e., the more feature extraction coverage pixels are targeted during each feature extraction. The image feature scale value can be flexibly set according to actual needs, and multiple image feature scale values ​​can also be set. The image feature scale reflects the number of feature extraction coverage pixels targeted each time feature extraction is performed on a biological part image, while the image feature morphology reflects the distribution location of the feature extraction coverage pixels targeted each time feature extraction is performed on a biological part image. By combining the image feature scale with the image feature morphology, effective morphological features of biological part images can be fully extracted.

[0053] Specifically, the server identifies at least one image feature scale, and at each image feature scale, identifies at least one corresponding image feature form. The image feature form is compatible with the feature form type of the target region. For example, the image feature scales include three scales: 4x4, 8x8, and 16x16, and the image feature form may include linear forms in ten directions. The server performs image extraction on the image of the biological part according to the at least one image feature form under the identified at least one image feature scale, thereby obtaining a morphological feature for each pixel in the image of the biological part. In a specific embodiment, if there is one morphological feature corresponding to each pixel from which a morphological feature is extracted from the image of the biological part, feature extraction can be performed for each image feature form under one image feature scale to obtain a morphological feature corresponding to the image feature scale and image feature form for each pixel, and the server can identify the morphological feature as the morphological feature corresponding to the pixel. On the other hand, if there is more than one morphological feature corresponding to each pixel, for example, if there is more than one type of at least one of the image feature scale or image feature shape, multiple morphological features can be obtained for each pixel, and the server can combine the multiple morphological features to obtain the morphological feature corresponding to the pixel.

[0054] In a specific application example, the server identifies a feature extraction means corresponding to at least one image feature form at each image feature scale, and performs feature extraction on the body part image using the feature extraction means to obtain the respective morphological features of each pixel in the body part image. As shown in Figure 4, circular feature extraction means 1 and feature extraction means 2 correspond to different image feature scales. The server can identify at least one of feature extraction means 1 and feature extraction means 2 and perform feature extraction on the body part image.

[0055] In this embodiment, the server performs feature extraction on the biological part image according to at least one image feature morphology at at least one image feature scale, and when obtaining each morphological feature of each pixel in the biological part image, feature extraction can be performed according to the scale actually required. This improves the accuracy of expressing the target part features in combination with the feature scale, and the biological part features of the target part can be obtained with high accuracy, thereby improving the accuracy of personal authentication based on the biological part features of the target part.

[0056] In one exemplary embodiment, the step of extracting respective morphological features of each pixel from the biological part image according to at least one image feature form at at least one image feature scale includes the steps of: extracting scale morphological features corresponding to each pixel and the at least two image feature scales from the biological part image according to at least one image feature form at at least two image feature scales; for each pixel, obtaining a scale fusion feature of the target pixel by fusing the target pixel with the scale morphological features corresponding to the at least two image feature scales; and obtaining a respective morphological feature of each pixel according to the respective scale fusion feature of each pixel.

[0057] Among these, there are at least two image feature scales, and each image feature scale corresponds to at least one image feature form, enabling multi-scale feature extraction. For example, if there are M types of image feature scales and N types of image feature forms, M×N types of features can be extracted. In other words, M*N morphological features are obtained for each pixel for which feature extraction is performed in a biological part image. A scale morphological feature is a morphological feature extracted for each image feature form at one image feature scale, and each scale morphological feature corresponds to one image feature form at one image feature scale. For each pixel for which feature extraction is performed in a biological part image, the number of scale morphological features is the product of the number of types of image feature scales and the number of types of image feature forms. For each pixel for which feature extraction is performed in a biological part image, the scale fusion feature is a feature obtained by fusing multiple scale morphological features corresponding to that pixel, so the morphological feature of that pixel can be obtained based on the scale fusion feature.

[0058] Specifically, the server identifies at least two image feature scales, each of which can include at least one image feature form. The server performs feature extraction on the biological part image for each of the identified at least one image feature form at each image feature scale, thereby obtaining a scale morphology feature for each pixel. Here, each scale morphology feature corresponds to one image feature form at one image feature scale. For example, for image feature scale 1, image feature scale 2, and image feature scale 3, corresponding scale morphology features can be extracted for pixel A in the biological part image at each image feature scale. Specifically, the scale morphology features can include scale morphology feature a1 extracted at image feature scale 1, scale morphology feature a2 extracted at image feature scale 2, and scale morphology feature a3 extracted at image feature scale 3. Since feature extraction of at least one image feature form, such as image feature form 1, image feature form 2, and image feature form 3, can be performed at each image feature scale, the scale morphology features corresponding to the three image feature forms can be separately extracted for pixel A in the biological part image at each image feature scale. For example, scale morphological feature a1 extracted at image feature scale 1 may include scale morphological feature a11, scale morphological feature a12, and scale morphological feature a13, where scale morphological feature a11 is extracted for image feature form 1, scale morphological feature a12 is extracted for image feature form 2, and scale morphological feature a13 is extracted for image feature form 3. Since there are multiple image feature scales, each pixel may have multiple scale morphological features. The server can perform fusion on the scale morphological features of the target pixel, such as performing average fusion for each pixel, to obtain the scale fusion feature of the target pixel. After traversing each pixel and obtaining each scale fusion feature of each pixel, the server can obtain each morphological feature according to each scale fusion feature of each pixel.For example, the server may directly use the respective scale fusion features of each pixel as the respective morphological features, or may perform a further feature mapping process on the respective scale fusion features of each pixel to obtain the respective morphological features of each pixel.

[0059] In this embodiment, the server performs feature extraction on the biological part image according to at least one image feature morphology at at least two image feature scales, obtains a scale morphology feature for each pixel in the biological part image, fuses the scale morphology features for each pixel to obtain a scale fusion feature, and obtains each morphology feature according to the scale fusion feature for each pixel. This enables feature extraction at multiple scales on the biological part image and fusing the feature extraction results at multiple scales to obtain the biological part feature of the target part with high accuracy, thereby improving the accuracy of personal authentication based on the biological part feature of the target part.

[0060] In one exemplary embodiment, extracting each morphological feature of each pixel from the image of the biological part according to at least one image feature matching the feature type includes extracting each morphological feature of each pixel from the image of the biological part using a convolutional network in a pre-trained feature extraction model, wherein the convolutional network is used to perform image extraction on the image of the biological part for at least one image feature matching the feature type of the target part.

[0061] The feature extraction model may be pre-trained based on sample data, or may be trained based on various neural network algorithms. Examples of neural network algorithms include, but are not limited to, convolutional neural networks (CNNs), recurrent neural networks (RNNs), transformer algorithms, multilayer perceptrons (MLPs), and residual networks (ResNets). The feature extraction model may include a convolutional network used to perform feature extraction according to at least one image feature type matching the morphological type of the target region. That is, morphological feature extraction of the input image is achieved through a convolutional network in the feature extraction model to perform feature extraction on the input image for at least one image feature type.

[0062] For example, the server may obtain a pre-trained feature extraction model and use a convolutional network to extract morphological features for each pixel from the image of the biological part, using the convolutional network to perform feature extraction according to at least one image feature from the pre-trained feature extraction model that matches the morphological type of the target part. In a specific application example, each image feature corresponds to one convolutional kernel, and the feature extraction process for the image feature can be achieved using the corresponding convolutional kernel, i.e., the feature extraction of the image of the biological part can be achieved using at least one convolutional kernel. The convolutional network may include at least one convolutional kernel, and by inputting the image of the biological part into the convolutional network, the feature extraction process for the corresponding image feature can be performed via the at least one convolutional kernel, thereby obtaining morphological features for each pixel.

[0063] Furthermore, the step of obtaining biological part features of the user to be identified based on the respective morphological features of each pixel includes a step of extracting biological part features of the user to be identified for each morphological feature of each pixel using a part feature extraction network in the feature extraction model.

[0064] The feature extraction model also includes a subnetwork for the local feature extraction network, which performs further feature extraction on each morphological feature of each pixel to obtain the biological local features of the user to be identified. The local feature extraction network can be constructed by selecting an artificial neural algorithm based on actual needs and training it based on sample data. For example, the local feature extraction network can be constructed based on at least one of the following algorithms: CNN algorithm, RNN algorithm, Transformer algorithm, MLP algorithm, or ResNet algorithm.

[0065] Illustratively, the server can input the morphological features of each pixel to a part feature extraction network in the feature extraction model, which can then perform further feature extraction on the morphological features of each pixel to obtain the biological part features of the user to be identified. For example, the part feature extraction network connects the morphological features of each input pixel, specifically, connects the morphological features of each pixel according to the distribution position of each pixel in the biological part image to obtain the morphological features of the biological part image, and performs feature extraction on the morphological features of the biological part image to obtain the biological part features of the user to be identified.

[0066] In this embodiment, a convolutional network in a pre-trained feature extraction model extracts morphological features for each pixel from a biological part image according to at least one image feature morphology matching the feature morphology type of the target part. A biological part feature extraction network in the feature extraction model extracts biological part features of the user to be identified for each morphological feature of each pixel. This enables efficient and accurate feature extraction processing based on an artificial neural network model, contributing to improved processing efficiency and accuracy of personal authentication.

[0067] In one exemplary embodiment, the feature extraction model is obtained by a model training step, which includes the steps of: acquiring a plurality of biological part image samples; extracting respective sample morphological features of each sample pixel from the biological part image samples using a convolutional network in the feature extraction model to be trained; extracting biological part sample features for each sample morphological feature of each sample pixel using a part feature extraction network in the feature extraction model to be trained; identifying a training loss based on the biological part sample features and the sample morphological features; and updating the convolutional network and the part feature extraction network in the feature extraction model to be trained according to the training loss, and then continuing training until training is completed, thereby obtaining a trained feature extraction model.

[0068] Among them, the body part image samples are sample data for training the feature extraction model. Since the body part image samples are accompanied by identification tags, the feature extraction performance of the feature extraction model can be determined based on the identification tags. The sample morphological features are extracted based on the body part image samples by a convolutional network in the feature extraction model to be trained. The convolutional network can perform feature extraction on the input body part image samples according to at least one image feature morphology matching the target body part feature morphology type. The body part sample features are biometric features extracted by the body part feature extraction network in the feature extraction model to be trained. A training loss is obtained based on the body part sample features and the sample morphological features. The feature extraction performance of the feature extraction model can be evaluated based on the training loss, so that the feature extraction model can be updated and the feature extraction performance of the feature extraction model can be improved. Among them, the body part sample features can reflect the feature extraction performance of the body part feature extraction network, while the sample morphological features can reflect the feature extraction performance of the convolutional network.

[0069] For example, when training a feature extraction model, the server can obtain multiple body part image samples, which can be collected from a user's target body part. The server uses a convolutional network in the feature extraction model to perform feature extraction on the body part image samples, extracting sample morphological features for each sample pixel from the body part image samples. The server then uses a regional feature extraction network in the feature extraction model to perform further feature extraction on the sample morphological features for each sample pixel to obtain body part sample features. The server determines a training loss according to the sample morphological features and the body part image samples, and updates model parameters of the feature extraction model to be trained based on the training loss. Specifically, the server updates network parameters, such as weighting parameters, of the convolutional network and regional feature extraction network in the feature extraction model to be trained, and continues training using the updated feature extraction model until training is complete, thereby obtaining a trained feature extraction model. For example, when the training reaches a preset number of training times, when the feature extraction model satisfies the convergence condition, or when the feature extraction performance of the feature extraction model satisfies the performance requirements, the training termination condition is deemed to be satisfied and the training is terminated, and a trained feature extraction model can be obtained from the feature extraction model at the end of training.

[0070] In this embodiment, the server trains a feature extraction model including a convolutional network and a part feature extraction network based on biological part image samples, and identifies a training loss according to the sample morphological features that reflect the feature extraction ability of the convolutional network and the biological part sample features that reflect the feature extraction ability of the part feature extraction network.By updating the feature extraction model based on the training loss, the feature extraction performance of both the convolutional network and the part feature extraction network can be ensured, thereby improving the feature extraction performance of the feature extraction model and contributing to improved accuracy of personal authentication.

[0071] In one exemplary embodiment, the step of identifying a training loss based on the biological part sample features and the sample morphological features includes the steps of: obtaining a biological part feature extraction loss based on the biological part sample features; identifying negative sample pairs including biological part image samples with different identification tags; obtaining a sample pair loss based on the sample morphological features of each of the biological part image samples in the negative sample pairs; and obtaining a training loss according to the biological part feature extraction loss and the sample pair loss.

[0072] Among them, the body part feature extraction loss is obtained according to the body part sample features and is used to reflect the feature extraction performance of the body part feature extraction network. Specifically, it can be calculated using various loss algorithms such as the arcface (additive angular margin loss) algorithm, the mean squared error (MSE) algorithm, and the cross entropy algorithm. The negative sample pairs contain body part image samples with different identification tags. That is, the body part image samples in the negative sample pairs correspond to different users. The sample pair loss is obtained based on the sample morphology features of each body part image sample in the negative sample pairs. Specifically, it can be calculated using various loss algorithms such as the L1 loss (absolute value loss) and the L2 loss (mean squared error loss). The training loss is obtained based on the body part feature extraction loss and the sample pair loss, specifically by combining the body part feature extraction loss and the sample pair loss.

[0073] Specifically, the server can obtain a region feature extraction loss based on the body region sample features, for example, calculated using the arcface algorithm based on the body region sample features. The server identifies negative sample pairs including body region image samples with different identification tags and identifies sample morphology features of each of the body region image samples in the negative sample pairs. The server obtains a sample pair loss based on the sample morphology features of each of the body region image samples in the negative sample pairs, for example, calculated using the L1 loss algorithm according to the sample morphology features of each of the body region image samples in the negative sample pairs. The server can obtain a training loss based on the region feature extraction loss and the sample pair loss. Specifically, the server can obtain the training loss from the sum of the region feature extraction loss and the sample pair loss, and can update the model parameters of the feature extraction model based on the training loss.

[0074] In this embodiment, the server obtains a part feature extraction loss reflecting the feature extraction ability of the part feature extraction network according to the sample morphological features, obtains a sample pair loss reflecting the feature extraction ability of the convolutional network according to the sample morphological features of each of the biological part image samples in the negative sample pair, obtains a training loss based on the part feature extraction loss and the sample pair loss, and updates the model based on the training loss, thereby ensuring the feature extraction performance of both the convolutional network and the part feature extraction network, thereby improving the feature extraction performance of the feature extraction model and contributing to improved accuracy of identity authentication.

[0075] In one exemplary embodiment, the step of obtaining the biological part features of the user to be identified based on the respective morphological features of each pixel includes a step of connecting the respective morphological features of each pixel according to the respective distribution positions of each pixel in the biological part image to obtain the morphological features of the biological part image, and a step of performing feature extraction on the morphological features of the biological part image to obtain the biological part features of the user to be identified.

[0076] In this context, the distribution position refers to the spatial position of a pixel in the body part image. For example, the server identifies the distribution position of each pixel in the body part image and connects the morphological features of each pixel for each distribution position to obtain the morphological features of the body part image. The server performs further feature extraction based on the morphological features of the body part image to obtain the body part features of the user to be identified. In a specific application example, the server performs feature extraction on the morphological features of the body part image using a part feature extraction network in a pre-trained feature extraction model. For example, the morphological features of the body part image can be input into the part feature extraction network of the feature extraction model to obtain the body part features of the user to be identified.

[0077] In this embodiment, the server extracts features from the morphological features of the biological part image obtained by connecting the morphological features of each pixel, and obtains the biological part features of the user to be identified. This makes it possible to integrate the morphological features of each pixel, thereby improving the expressiveness of the biological part features and thereby improving the accuracy of personal authentication.

[0078] In one exemplary embodiment, the step of extracting each morphological feature of each pixel from the image of the biological part according to at least one image feature conforming to the feature type includes the steps of identifying a region of interest from the image of the biological part, identifying at least one image feature conforming to the feature type, and extracting each morphological feature of each pixel within the region of interest according to the at least one image feature.

[0079] The region of interest is an image region identified from the image of the body part from which feature extraction of the body part is required. For example, the server can identify the region of interest from the image of the body part by performing body part recognition on the image of the body part and identifying an area including the target part as the region of interest. The server can identify at least one image feature morphology that matches the feature morphology type of the target part, specifically, identify the feature morphology type of the target part and, based on the feature morphology type, identify at least one image feature morphology that matches the feature morphology. The server can perform feature extraction on the region of interest according to the identified at least one image feature morphology, thereby obtaining morphological features for each pixel in the region of interest.

[0080] In this embodiment, the server identifies an area of ​​interest from a biological part image, extracts features from the area of ​​interest, and performs personal authentication. This reduces the amount of data processing required during feature extraction, thereby improving the processing efficiency of personal authentication.

[0081] In one exemplary embodiment, the target site is the palm of the hand. As shown in Figure 9, the process of identifying the region of interest, ie, identifying the region of interest from the image of the body site, includes steps 902-906.

[0082] Step 902: Detect interdigital space feature points between different fingers on the palm from the body part image.

[0083] In this system, the target body part is a palm, the biometric image is a palm image, and the biometric features to be extracted are palm features, specifically, palm print features or palm vein features. The interfinger space feature points may be feature points for distinguishing each finger, specifically, the connection points between fingers in the palm print. Specifically, the server can identify palm feature points in the biometric image and identify each interfinger space feature point between fingers on the palm. For example, the server can specifically identify the connection points between adjacent fingers on the thumb, index finger, middle finger, ring finger, and little finger on the palm, and obtain each interfinger space feature point.

[0084] Step 904: Identify the focus of interest and area range parameters from the body part image based on the feature point positions of each inter-finger space feature point and the feature point distances between each inter-finger space feature point.

[0085] Among these, the feature point position refers to the spatial position of the interfinger space feature point in the body part image. The feature point distance refers to the distance between each feature point position of each interfinger space feature point. The focus of interest is a feature point of the region of interest to be identified, and specifically may be a vertex, a circle center, or a center point of the region of interest to be identified. The region range parameter is used to describe the range of the region of interest to be identified, and specifically may include parameters such as side length and radius. Specifically, the server can identify the feature point position of each interfinger space feature point in the body part image and determine the feature point distance between each interfinger space feature point based on each feature point position. The server can identify the focus of interest and the region range parameter in the body part image based on the feature point position and the feature point distance. For example, the circle center and radius can be identified based on the feature point position and the feature point distance, and the circular region of interest identified based on the circle center and radius can cover each interfinger space feature point and the palm area.

[0086] Step 906: Identify a region of interest in the image of the body part according to the focus of interest and region extent parameters.

[0087] Specifically, the server identifies a region of interest in the image of the biological part based on the focus of interest and the region range parameter. For example, if the focus of interest is a center point and the region range parameter is a side length, a polygonal region of interest can be constructed with the focus of interest as the center and the region range parameter as the geometric side length. For example, if the focus of interest is a center point and the region range parameter is a radius, a circular region of interest can be constructed with the focus of interest as a circle point and the region range parameter as the radius.

[0088] In this embodiment, the server detects each interfinger space feature point between the fingers of the palm in the biometric part image, identifies the focus of interest and area range parameters based on the feature point position of each interfinger space feature point and the feature point distance between each interfinger space feature point, and identifies the area of ​​interest in the biometric part image according to the focus of interest and area range parameters, thereby enabling the area of ​​interest to cover the palm area and reducing the amount of data processing during feature extraction while ensuring the accuracy of biometric feature extraction, thereby improving the processing efficiency of personal authentication.

[0089] In one exemplary embodiment, the step of matching the biological part features with the registered part features to obtain a feature matching result and determining an identity identification result of the user to be identified according to the feature matching result includes the steps of acquiring the registered part features of each registered user, separately determining feature similarities between the biological part features and each registered part feature, and determining an identity identification result of the user to be identified based on each feature similarity.

[0090] The registered body part features are obtained by enrolling the registered user in a body part image corresponding to the target body part of the registered user. The feature similarity can be calculated based on the body part features and the registered body part features, and specifically, can include at least one of various formats such as cosine similarity, Eucledian distance, Manhattan distance, and Minkowski distance.

[0091] Specifically, the server acquires pre-stored registered body part features of each registered user, calculates the similarity between the biometric body part features and each registered body part feature, and determines the feature similarity between the biometric body part features and each registered body part feature. The server obtains an identity identification result for the user to be identified based on each feature similarity. For example, the server can identify the registered user corresponding to the most similar registered body part feature as the identity identification result for the user to be identified. A similarity threshold may also be set. If the similarity exceeds the similarity threshold, the server determines that the features match, and identifies the registered user corresponding to the most similar registered body part feature as the identity identification result for the user to be identified. On the other hand, if the similarity is below the similarity threshold, the server determines that the features do not match, i.e., that the user to be identified is not a pre-registered registered user.

[0092] In this embodiment, the server determines the identity identification result of the user to be identified based on the feature similarity between the biometric body part features of the user to be identified and the registered body part features of each registered user, thereby enabling effective identity identification processing using the biometric body part features, thereby ensuring the accuracy of personal authentication.

[0093] This application further provides an application scenario for applying the above-mentioned identification method. Specifically, in this application scenario, the identification method is applied as follows:

[0094] Palm print recognition technology is a new generation of biometric authentication technology following facial recognition technology, and has been applied in areas such as mobile payment and identity authentication. Compared with facial recognition technology, palm print recognition technology identifies various user identities based on images of the palm print area. Palm prints are easily concealed, which is more advantageous for privacy protection and is not affected by wearing a mask, makeup, or sunglasses.

[0095] Conventional palmprint authentication technologies are generally divided into three types. The first is geometric feature-based palmprint authentication, which primarily identifies geometric shapes such as the spacing between fingers, the width of fingers, and the length of the palm. Specifically, geometric feature-based palmprint authentication involves image preprocessing, which involves processing the input palmprint image, such as noise reduction, enhancement, and binarization. Geometric feature extraction is then performed, which involves calculating the geometric features of the palm, such as the width, the spacing between fingers, and the length of the palm. Pattern matching is then performed, which involves using the extracted geometric features to match patterns with templates in a database to identify the subject's identity. Geometric feature-based palmprint authentication methods are more robust to low-resolution images and environmental interference, but their identification accuracy is limited.

[0096] The second is palmprint recognition based on dermatoglyphic features. In dermatoglyphic features, techniques such as directional fields and Gabor filters focus on the skin dermatoglyphic features found on the palm surface. When performing palmprint recognition based on dermatoglyphic features, image preprocessing is performed, specifically, processing such as noise removal, enhancement, and filtering is performed on the input palmprint image. Then, dermatoglyphic feature extraction is performed, specifically, using signal processing methods such as directional fields and Gabor filters to extract dermatoglyphic features from the preprocessed image. Then, feature matching is performed, specifically, by matching the input image with dermatoglyphic features in a database to perform palmprint recognition. Because palmprint recognition methods based on dermatoglyphic features use signal processing and machine learning techniques to extract and match palmprint features, these methods have good robustness and recognition accuracy in complex scenarios.

[0097] The third method is deep learning-based palmprint recognition, which uses modern deep learning techniques such as convolutional neural networks (CNNs) to perform end-to-end feature learning and recognition on palm images. Deep learning-based palmprint recognition involves image preprocessing, such as noise removal, enhancement, and normalization, on the input palmprint image. A deep learning model is trained, specifically a deep convolutional neural network is used to train labeled palmprint images and learn hierarchical palmprint features. Palmprint recognition involves preprocessing the input image and then inputting it into the pre-trained model for palmprint recognition. Deep learning-based palmprint recognition methods can automatically learn hierarchical feature representations of palmprint images, improving the accuracy and robustness of palmprint recognition.

[0098] However, in mobile payment scenarios, the user base is extremely large, and there may be a large number of highly similar samples. Traditional methods based on geometric features, print features, and deep learning all extract palm features using a square convolution kernel, but the palm contains many linear features, so the feature extraction effect is poor.

[0099] Because palm feature information is mainly concentrated in palm print lines, extracting palm print line features is a very important part of distinguishing between different palms. Regarding the problems of conventional methods, this embodiment proposes a palm print feature extraction method based on linear convolution, which can fully learn the linear features of palm print lines to distinguish between different palm print lines. Regarding the format of palm print line features, this embodiment proposes a palm print feature extraction method based on linear convolution kernels, which effectively improves the model's extraction effect for linear features by designing a linear convolution feature extractor, effectively improving the model's recognition ability for different palms, and thereby improving the accuracy of palm print recognition. This embodiment also proposes a palmprint authentication method based on linear convolution, including step 1: detecting keypoint locations in the interdigital spaces of the index finger, middle finger, and ring finger using a detection model; step 2: extracting a region of interest on the palm for each keypoint location in the interdigital spaces; step 3: extracting palmprint features from the region of interest using a multi-scale linear convolution kernel; step 4: fusing the linearly extracted features at different scales; and step 5: constraining the features using a pairwise loss function. This embodiment introduces multi-scale linear convolution to more effectively extract linear features from palmprints, and adds pixel-level orientation map contrast to better distinguish highly similar samples, thereby improving the accuracy of palmprint authentication.

[0100] Palmprint authentication technology is expected to be widely applied in business scenarios such as mobile payments and identity verification. In this embodiment, a palmprint feature extraction method based on linear convolution is provided, which can identify a user by matching features for each user's palm image. As shown in FIG. 10 , the palmprint authentication process for payment services includes step 1001 of collecting images using a terminal payment device; step 1002 of collecting a user's hand image; step 1003 of detecting interdigital space keypoints, specifically, detecting three interdigital space keypoints on the user's hand using a detection model; step 1004 of extracting palm regions of interest, specifically, extracting palm regions of interest based on the hand image and keypoint positions; and step 1005 of extracting multi-scale linear features, specifically, extracting palm regions of interest using a multi-scale linear model. The method includes step 1005 of extracting region features; step 1006 of fusing features, specifically fusing multi-scale linear features; step 1007 of extracting features by a recognition model, specifically inferring about multi-scale information by the recognition model; step 1009 of calculating similarity with base library features, specifically calculating cosine similarity between the stitched features and the base library features; and step 1009 of identifying a classification result based on the similarity, specifically identifying the identity information corresponding to the base library image with the highest similarity as the target user.

[0101] The palm region is detected from the collected images. Specifically, inter-finger points are identified using target detection technology, and palm region images are extracted from the images. Specifically, to extract the palm region of interest, as shown in Figure 11, the inter-finger space keypoint locations are first identified. Specifically, three inter-finger space keypoints—index finger A, middle finger B, and ring finger C—are detected using an inter-finger space point target detector based on YOLOv2 (You Only Look Once, a target detection algorithm). Next, a local coordinate system is established. Specifically, keypoints A and C form the x-axis of the local coordinate system, and the third point B forms the y-axis perpendicular to the x-axis. The center point D of the palm print is located at a distance A-C from the coordinate origin along the negative y-axis, with the distance D-E equal to 6 / 5 of the distance A-C. Finally, the ROI region is extracted. Specifically, the distance from point A to point C is multiplied by 3 / 2, which is set as half the side length d of the ROI region (Region of Interest), with point D as the center and a side length of 2d. The ROI region is extracted as an input for the recognition model. As shown in Figure 11, a square region with a side length of 3AC and centered at point D is identified as the ROI region.

[0102] During model training, each local region of interest extracted from each sample data set is resized (scaled) to 224x224. Furthermore, depending on the distribution of palm print lines, linear convolution kernels with different scales and orientations can be configured to extract palm print line features at different scales and orientations. As shown in Figure 12, for a 9x9 linear convolution, the center pixel of the convolution kernel is set as the origin, and linear convolution weights are constructed along directions a, 30°, c, 60°, d, 120°, e, and f, which are oriented at 150° relative to the horizontal. The width of the linear convolution weights is 1, and the remaining weights are 0. In other words, the weight of the shaded convolution units is 1, and the convolution result is valid, while the weight of the unshaded convolution units is 0, and the convolution result is also 0. As shown in Figure 13, for a 16x16 linear convolution, the kernel center pixel is set as the origin, and the linear convolution weights are constructed along the directions a, b, c, d, e, f, and f, which form 0 degrees, 30 degrees, 60 degrees, 90 degrees, 120 degrees, and 150 degrees, respectively, relative to the horizontal direction. The width of the linear convolution weights is 4, and the remaining weights are 0.

[0103] Furthermore, feature extraction is performed on the region of interest using linear convolution kernels at different scales and directions. The result of feature extraction is shown in the following equation.

[0104]

number

[0105] Furthermore, for each pixel, the results of the 12 linear convolutions are averaged to obtain the linear convolution fusion result of the pixel, specifically shown in the following equation:

[0106]

number

[0107] Furthermore, we connect the linear convolution results of all pixels at corresponding pixel positions to obtain an orientation map for the entire image, "orientation_map," which is then used as the input for the subsequent feature extraction network, which can be an Inception resnet50, and the output is a 512-dimensional feature, "feature_id."

[0108] Furthermore, arcface is used to calculate the loss function between palm image features with different IDs, and the difference in the pairwise pixel orientation maps of negative samples is added to this. Specifically, two palm images with different IDs are used as negative samples, and the L1 loss function between their respective orientation maps is calculated. The sum of the L1 losses of all negative samples is calculated as the pairwise loss, as shown in the following equation:

[0109]

number

[0110] The final loss function loss_final is calculated by adding the arcface loss and the pairwise loss as shown in the following equation:

[0111] [Number 4] loss_final=arcface_loss+pairwise_loss

[0112] After calculating the final loss function loss_final, we perform gradient backpropagation and continue training until training is complete, resulting in a trained feature extraction network.

[0113] When performing identity verification, an image of the user's hand may be collected through the camera of the payment terminal device, key points in the three interdigital spaces of the user's hand may be detected using a detection model, a region of interest on the palm may be extracted according to the hand image and the key point locations, multi-scale linear features may be extracted using multi-scale linear convolution, the multi-scale linear features may be averaged and fused, an encoding vector of the palmprint feature may be extracted using a feature extraction network, and the cosine similarity between the encoding vector of the palmprint feature and the base library feature may be calculated. The cosine similarity may be calculated using the following formula:

[0114]

number

[0115] In a specific application example, the effectiveness of the high-similarity palm print fine-grained recognition algorithm is verified. Table 1 shows the verification results for the twin dataset using the linear convolution method of this embodiment.

[0116] [Table 1]

[0117] To verify the effectiveness of linear convolution feature extraction, high-resolution palmprint images and blurred palmprint images of 40 groups of twins were measured as highly similar palmprint images. The measurement results are shown in Table 1. When left and right hands of twins from the same group were used as sample pairs, a total of 3,600 sample pairs were obtained. Of the high-resolution twin images, the conventional Arcface generated errors in the sample classification results for 37 groups, while the method of this embodiment generated no errors for any samples. On the other hand, of the blurred twin images, Arcface generated errors in the sample classification results for 46 groups, while the method of this embodiment generated no errors for any samples, demonstrating the high recognition accuracy of this method for highly similar samples.

[0118] Here, although the steps in the flowcharts according to the above embodiments are displayed in a sequential order as indicated by the arrows, it should be understood that these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps may be executed in other orders. Furthermore, at least some of the steps in the flowcharts according to the above embodiments may include multiple steps or multiple phases. These steps or phases do not necessarily have to be executed at the same time, but may be executed at different times. Furthermore, the execution order of these steps or phases is not necessarily consecutive, and they may be executed in order or alternately with other steps or at least some steps or phases.

[0119] Based on the same inventive concept, an embodiment of the present application further provides an identity identification device for realizing the above-mentioned identity identification method. The solution provided by the device is similar to the realization described in the above-mentioned method, so specific limitations on one or more embodiments of the identity identification device described below can be referred to the limitations on the above-mentioned identity identification method, and will not be further described here.

[0120] In one embodiment, an identification device 1400 is provided for application to a computing device, as shown in Figure 14. Such device 1400 includes: a body part image acquisition module 1402 configured to perform a step of acquiring a body part image obtained for a target part of a user to be identified; a morphological feature extraction module 1404 configured to perform the steps of identifying a feature morphology type of the target part and extracting a morphological feature for each pixel from the image of the biological part according to at least one image feature morphology that matches the feature morphology type, where the image feature morphology is a morphology formed by combining the distribution positions of each of the feature extraction coverage pixels, and the feature extraction coverage pixels are pixels in the image of the biological part that are the target for each feature extraction; a body part feature acquisition module 1406 configured to perform a step of acquiring a body part feature of the user to be identified based on the morphological feature of each pixel; and a feature matching module 1408 configured to execute a step of matching the biometric feature with the registered biometric feature to obtain a feature matching result, and specifying an identity identification result of the user to be identified according to the feature matching result, wherein the registered biometric feature is a biometric feature obtained by performing personal registration on a biometric feature image corresponding to a target body part of the registered user.

[0121] In one embodiment, when the target area is a palm and the image feature form is a linear form, the morphological feature extraction module 1404 is further configured to perform a step of extracting respective morphological features of each pixel from the biological area image according to the linear form in at least one direction.

[0122] In one embodiment, the morphological feature extraction module 1404 is further configured to perform the steps of extracting directional morphological features from the biological part image according to the linear features in at least two directions, each corresponding to a pixel and the at least two linear features; obtaining, for each pixel, a directional fusion feature of the target pixel by fusing the target pixel with the directional morphological features corresponding to the linear features in at least two directions; and obtaining each morphological feature of each pixel according to each directional fusion feature of each pixel.

[0123] In one embodiment, the morphological feature extraction module 1404 is further configured to perform the steps of identifying at least one image feature morphology that matches the feature morphology type at at least one image feature scale, and extracting respective morphological features for each pixel from the biological part image according to the at least one image feature morphology at the at least one image feature scale.

[0124] In one embodiment, the morphological feature extraction module 1404 is further configured to perform the steps of: extracting scale morphological features corresponding to each pixel and at least two image feature scales from the biological part image according to at least one image feature morphology at at least two image feature scales; obtaining, for each pixel, a scale fusion feature of the target pixel by fusing the target pixel with the scale morphological features corresponding to the at least two image feature scales; and obtaining a respective morphological feature of each pixel according to the respective scale fusion feature of each pixel.

[0125] In one embodiment, the morphological feature extraction module 1404 is further configured to perform the step of extracting a respective morphological feature for each pixel from the body part image using a convolutional network in the pre-trained feature extraction model, where the convolutional network is used to perform feature extraction on the body part image for at least one image feature matching the feature type. The body part feature acquisition module 1406 is further configured to perform the step of extracting a body part feature of the user to be identified for each morphological feature for each pixel using a body part feature extraction network in the feature extraction model.

[0126] In one embodiment, the model training module is configured to perform the steps of: acquiring a plurality of body part image samples; extracting respective sample morphological features for each sample pixel from the body part image samples using a convolutional network in the feature extraction model to be trained; extracting body part sample features for each sample morphological feature for each sample pixel using a region feature extraction network in the feature extraction model to be trained; determining a training loss based on the body part sample features and the sample morphological features; and updating the convolutional network and the region feature extraction network in the feature extraction model to be trained according to the training loss, and then continuing training until training is completed, thereby obtaining a trained feature extraction model.

[0127] In one embodiment, the model training module is further configured to perform the steps of obtaining a region feature extraction loss based on the body region sample features; identifying negative sample pairs including body region image samples with different identification tags; obtaining a sample pair loss based on each sample morphological feature of the body region image samples in the negative sample pairs; and obtaining a training loss according to the region feature extraction loss and the sample pair loss.

[0128] In one embodiment, the body part feature acquisition module 1406 is further configured to perform the steps of: connecting the morphological features of each pixel according to the respective distribution positions of each pixel in the body part image to obtain the morphological features of the body part image; and performing feature extraction on the morphological features of the body part image to obtain the body part features of the user to be identified.

[0129] In one embodiment, the morphological feature extraction module 1404 is further configured to perform the steps of identifying a region of interest from the image of the biological part, identifying at least one image feature that matches the feature type, and extracting a respective morphological feature for each pixel in the region of interest according to the at least one image feature.

[0130] In one embodiment, when the target part is a palm, the morphological feature extraction module 1404 is further configured to perform the steps of: detecting inter-finger space feature points between different fingers on the palm from the image of the biological part; determining a focus of interest and a region range parameter from the image of the biological part based on the feature point positions of the inter-finger space feature points and the feature point distances between the inter-finger space feature points; and determining a region of interest in the image of the biological part according to the focus of interest and the region range parameter.

[0131] In one embodiment, the feature matching module 1408 is further configured to perform the steps of obtaining respective enrollment body part features of each enrolled user, separately determining feature similarities between the biometric body part features and each enrollment body part feature, and determining an identity identification result of the user to be identified based on each feature similarity.

[0132] Each module in the above-mentioned identification device can be realized in whole or in part by software, hardware, or a combination thereof. Each module can be integrated into a processor in a computer device in the form of hardware, or can be independent from the processor, or can be stored in a memory in a computer device in the form of software so that it can be invoked by the processor to perform the operations corresponding to each module.

[0133] In one embodiment, a computer device is provided. The computer device may be a server or a terminal, and its internal structure is shown in FIG. 15. The computer device includes a processor, a memory, an input / output (I / O) interface, and a communication interface. The processor, memory, and I / O interface are connected via a system bus, and the communication interface is connected to the system bus via the I / O interface. The processor of the computer device is used to provide calculation and control functions. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer-readable instructions, and a database. The internal memory provides an operating environment for the operating system and the computer-readable instructions in the non-volatile storage medium. The database of the computer device is used to store data related to identity identification. The I / O interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer-readable instructions are executed by the processor, an identity identification method is realized.

[0134] Those skilled in the art will appreciate that the structure shown in Figure 15 is merely a block diagram of a partial configuration related to the embodiments of the present application, and does not limit the computer device to which the embodiments of the present application can be applied. A specific computer device may include more or fewer components than those shown, may combine some components, or may have components arranged differently.

[0135] In one embodiment, there is further provided a computing device including a memory having computer-readable instructions stored thereon, and a processor, the computer-readable instructions, when executed, causing the processor to perform the steps in each of the method embodiments described above.

[0136] In one embodiment, there is further provided a computer-readable storage medium having stored thereon computer-readable instructions that, when executed, cause a processor to perform the steps in each of the method embodiments described above.

[0137] In one embodiment, there is further provided a computer program product including computer readable instructions that, when executed, cause a processor to perform the steps in each of the method embodiments described above.

[0138] Furthermore, all user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) related to this application are information and data authorized by the user or fully authorized by all parties, and the collection, use and handling of related data must comply with relevant regulations.

[0139] Those skilled in the art will understand that all or part of the processes in the above method embodiments can be implemented by instructing associated hardware through computer-readable instructions. The computer-readable instructions may be stored in a non-volatile computer-readable storage medium. The execution of the computer-readable instructions may implement the processes in the above method embodiments. References to memory, databases, or other media in the embodiments of this application may include at least one of non-volatile memory and volatile memory. Non-volatile memory includes read-only memory (ROM), magnetic tape, floppy disks, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory includes random access memory (RAM) or external cache memory, etc. By way of example, the RAM may be in various forms, such as, but not limited to, static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the embodiments of the present application may include at least one of a relational database and a non-relational database. Non-relational databases include, but are not limited to, distributed databases based on blockchain.The processor in each embodiment of the present application may be, but is not limited to, a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, etc.

[0140] The technical features in the above embodiments can be combined in any way. For convenience of explanation, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered within the scope of this specification.

[0141] The above embodiments merely represent some implementation forms of the present application, and although the description is relatively specific and detailed, it should not be construed as limiting the scope of the present patent. Those skilled in the art can make some modifications and improvements without departing from the concept of the present invention, all of which shall fall within the scope of protection of the present application. Therefore, the scope of protection of the present application should be governed by the appended claims.

Claims

1. 1. A method of identity identification performed by a computing device, comprising: acquiring a body part image obtained for a target part of a user to be identified; a step of identifying a feature morphology type of the target part, and extracting a morphology feature of each pixel from the biological part image according to at least one image feature morphology that matches the feature morphology type, the image feature morphology being a morphology formed by combining distribution positions of feature extraction coverage pixels, and the feature extraction coverage pixels being pixels in the biological part image that are targeted at the time of feature extraction; obtaining a body part feature of the user to be identified based on the morphological feature of each pixel; a step of matching the biometric feature with a registered biometric feature to obtain a feature matching result, and specifying an identity identification result of the user to be identified according to the feature matching result, wherein the registered biometric feature is a biometric feature obtained by performing personal registration on a biometric feature image corresponding to a target body part of the registered user.

2. When the target portion is a palm and the image feature form is a linear form, extracting a morphological feature for each pixel from the image of the body part according to at least one image feature conforming to the feature type, The method of claim 1 , further comprising the step of extracting a morphological feature of each pixel from the image of the biological part according to a linear shape in at least one direction.

3. The step of extracting a morphological feature of each pixel from the image of the biological part according to a linear shape in at least one direction includes: extracting directional morphology features from the image of the biological part according to linear morphologies in at least two directions, the directional morphology features corresponding to each pixel and the linear morphologies in the at least two directions; For each of the pixels, fusing the pixel of interest with directional feature features corresponding to the at least two directional linear features, respectively, to obtain a directional fusion feature of the pixel of interest; The method of claim 1 or 2, further comprising: obtaining a morphological feature of each of the pixels according to the directional fusion feature of each of the pixels.

4. extracting a morphological feature for each pixel from the image of the body part according to at least one image feature conforming to the feature type, identifying at least one image feature that matches the feature type at at least one image feature scale; and extracting a respective morphological feature of each pixel from the image of the biological part according to the at least one image feature morphology at the at least one image feature scale.

5. extracting a respective morphological feature for each pixel from the image of the body part according to the at least one image feature morphology at the at least one image feature scale, extracting scale morphology features corresponding to each pixel and the at least two image feature scales from the image of the biological part according to the at least one image feature morphology at at least two image feature scales; For each of the pixels, fusing the pixel of interest with scale morphological features corresponding to the at least two image feature scales, respectively, to obtain a scale fusion feature of the pixel of interest; The method of any one of claims 1 to 4, further comprising the step of: obtaining a morphological feature of each of said pixels according to a scale fusion feature of each of said pixels.

6. extracting a morphological feature for each pixel from the image of the body part according to at least one image feature conforming to the feature type, extracting morphological features for each pixel from the image of the biological part using a convolutional network in a pre-trained feature extraction model; the convolutional network is used to perform feature extraction on the image of the body part according to at least one image feature that matches the feature type; The step of obtaining a biological part feature of the user to be identified based on the morphological feature of each of the pixels includes: The identity identification method according to any one of claims 1 to 5, further comprising a step of extracting biological part features of the user to be identified for each morphological feature of each pixel using a part feature extraction network in the feature extraction model.

7. The feature extraction model is obtained by a model training step, and the model training step includes: acquiring a plurality of sample images of a body part; extracting sample morphological features for each sample pixel from the sample images of the body part using a convolutional network in a feature extraction model to be trained; extracting a biological region sample feature for each sample morphological feature of each sample pixel by a region feature extraction network in the feature extraction model to be trained; determining a training loss based on the body part sample features and the sample morphological features; and updating the convolutional network and the local feature extraction network in the feature extraction model to be trained according to the training loss, and then continuing training until the training is completed, thereby obtaining the trained feature extraction model.

8. The step of identifying a training loss based on the body part sample features and the sample morphological features includes: obtaining a region feature extraction loss based on the body region sample features; identifying negative sample pairs including body part image samples with different identification tags; obtaining a sample pair loss based on the sample morphology features of each of the body part image samples in the negative sample pair; The method of any one of claims 1 to 7, further comprising: obtaining a training loss according to the region feature extraction loss and the sample pair loss.

9. The step of obtaining a biological part feature of the user to be identified based on the morphological feature of each of the pixels includes: acquiring a morphological feature of the image of the body part by connecting the morphological features of the pixels according to a distribution position of each of the pixels in the image of the body part; The identity identification method according to any one of claims 1 to 8, further comprising: extracting features from the morphological features of the body part image to obtain body part features of the user to be identified.

10. extracting a morphological feature for each pixel from the image of the body part according to at least one image feature conforming to the feature type, identifying a region of interest from the body part image; identifying at least one image feature that matches the feature type; The method of any one of claims 1 to 9, further comprising the step of: extracting a morphological feature of each pixel in said region of interest according to said at least one image feature morphology.

11. When the target part is a palm, the step of identifying a region of interest from the body part image includes: detecting interdigital space feature points between different fingers on the palm from the body part image; determining a focus of interest and an area range parameter from the image of the body part based on feature point positions of the inter-finger space feature points and feature point distances between the inter-finger space feature points; The method of any one of claims 1 to 10, further comprising the step of: identifying a region of interest in the image of the body part according to the focus of interest and the region range parameters.

12. the step of matching the biometric feature with a registered feature to obtain a feature matching result, and specifying an identity identification result of the user to be identified according to the feature matching result, acquiring registered body part features of each registered user; separately determining a feature similarity between the body part feature and each of the enrolled body part features; The identity identification method according to any one of claims 1 to 11, further comprising: a step of determining an identity identification result for the user to be identified based on each of the feature similarities.

13. An identification device applied to a computing device, comprising: a body part image acquisition module configured to perform a step of acquiring a body part image obtained for a target body part of a user to be identified; a morphological feature extraction module configured to execute a step of identifying a feature morphology type of the target part and extracting a morphological feature for each pixel from the image of the biological part according to at least one image feature morphology that matches the feature morphology type, the image feature morphology being a morphology formed by combining distribution positions of feature extraction coverage pixels, and the feature extraction coverage pixels being pixels in the image of the biological part that are targeted at the time of each feature extraction; a body part feature acquisition module configured to perform a step of acquiring a body part feature of the user to be identified based on the morphological features of each of the pixels; an identification device comprising: a feature matching module configured to execute a step of matching the biometric feature with a registered biometric feature to obtain a feature matching result, and specifying an identity identification result of the user to be identified according to the feature matching result, wherein the registered biometric feature is a biometric feature obtained by performing personal registration on a biometric feature image corresponding to a target body part of the registered user.

14. A computing device comprising a memory having computer readable instructions stored thereon, and a processor, the computer readable instructions, when executed, causing the processor to perform the steps of the method of any one of claims 1 to 12.

15. A computer-readable storage medium having stored thereon computer-readable instructions which, when executed, cause the processor to perform the steps of the method for identity identification according to any one of claims 1 to 12.

16. A computer program comprising computer readable instructions which, when executed, cause a processor to perform the steps of the method for identity identification according to any one of claims 1 to 12.

Images