Identity recognition method and apparatus, and computer device and storage medium

Abstract

An identity recognition method, comprising: acquiring a biological part image obtained for a target part of a user to be subjected to recognition (202); determining a feature morphology type of the target part, and according to an image feature morphology matching the feature morphology type, extracting from the biological part image respective morphological features of pixels, wherein the image feature morphology is a morphology formed after respective distribution positions of feature extraction coverage pixels are combined, and the feature extraction coverage pixels are pixels in the biological part image during each instance of feature extraction (204); on the basis of the respective morphological features of the pixels, obtaining a biological part feature of the user to be subjected to recognition (206); and performing feature matching on the biological part feature and registered part features of registered users, so as to obtain a feature matching result, and on the basis of the feature matching result, determining an identity recognition result for the user to be subjected to recognition (208).

Description

Identity identification method, device, computer equipment and storage medium

[0001] Related applications

[0002] This application claims priority to Chinese patent application number 2023111498217, filed on September 7, 2023, entitled “Identification Method, Apparatus, Computer Equipment and Storage Medium,” the entire text of which is hereby incorporated by reference. Technical Field

[0003] The present application relates to the field of computer technology, and in particular to an identity recognition method, apparatus, computer equipment, storage medium, and computer program product. Background Art

[0004] With the development of computer technology, increasingly mature identity recognition technologies have been widely used in various fields such as business cooperation, consumer payment, social media, and security. Among them, the use of inherent human biometrics, such as hand shape, fingerprints, face shape, retina, ear pinna, and other local biometric features for identity recognition has become a development trend in identity recognition technology.

[0005] Currently, in the technology of identity recognition based on biometric features of local parts, when users use biometric features such as hand shape, face, fingerprint, palm print, etc. for identity recognition, images of local parts are generally collected, and biometric features are extracted from the collected part images for identity recognition. However, the accuracy of extracted biometric features is currently low, resulting in low accuracy of identity recognition.

[0006] Summary of the Invention

[0007] According to various embodiments provided in this application, an identity recognition method, apparatus, computer device, computer-readable storage medium, and computer program product are provided.

[0008] In one aspect, the present application provides an identity recognition method, which is executed by a computer device and includes:

[0009] Acquire a biological part image obtained for a target part of the user to be identified;

[0010] Determining a characteristic pattern type corresponding to the target site, and extracting a pattern feature of each pixel from the biological site image according to at least one image feature pattern that matches the characteristic pattern type, wherein the image feature pattern is the pattern obtained by combining the distribution positions of pixels covered by feature extraction, and the pixels covered by feature extraction are the pixels in the biological site image targeted during each feature extraction;

[0011] Obtaining biological part features of the user to be identified based on the morphological features of each pixel; and

[0012] The biological part features are matched with the registered part features to obtain a feature matching result, and the identity recognition result for the user to be identified is determined based on the feature matching result, wherein the registered part features are the biological part features obtained by identity registration of the biological part image corresponding to the target part of the registered user.

[0013] On the other hand, the present application also provides an identity recognition device, applied to a computer device, comprising:

[0014] A body part image acquisition module, used to acquire a biological body part image obtained for a target body part of a user to be identified;

[0015] a morphological feature extraction module for determining a characteristic morphological type of the target part and extracting morphological features of each pixel from the biological part image according to at least one image characteristic pattern compatible with the characteristic morphological type, wherein the image characteristic pattern is the pattern obtained by combining the distribution positions of the pixels covered by the feature extraction, and the pixels covered by the feature extraction are the pixels in the biological part image targeted during each feature extraction;

[0016] A biological part feature acquisition module, configured to acquire the biological part feature of the user to be identified based on the morphological features of each pixel; and

[0017] The feature matching module is used to match the biological part features with the registered part features to obtain feature matching results, and determine the identity recognition results for the user to be identified based on the feature matching results. Among them, the registered part features are the biological part features obtained by identity registration of the biological part image corresponding to the target part of the registered user.

[0018] On the other hand, the present application also provides a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the above identity recognition method when executing the computer-readable instructions.

[0019] On the other hand, the present application also provides a computer-readable storage medium having computer-readable instructions stored thereon, which implement the steps of the above identity recognition method when executed by a processor.

[0020] On the other hand, the present application also provides a computer program product, comprising computer-readable instructions, which implement the steps of the above identity recognition method when executed by a processor.

[0021] The details of one or more embodiments of the present application are set forth in the accompanying drawings and the description below. Other features and advantages of the present application will become apparent from the description, drawings, and claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the conventional technology, the following briefly introduces the drawings required for use in the embodiments or the conventional technology descriptions. Obviously, the drawings described below are merely embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the disclosed drawings without any creative work.

[0023] FIG1 is a diagram illustrating an application environment of an identity recognition method according to an embodiment;

[0024] FIG2 is a schematic diagram of a flow chart of an identity recognition method in one embodiment;

[0025] FIG3 is a schematic diagram of an application of palmprint recognition in one embodiment;

[0026] FIG4 is a schematic diagram of a ring feature extraction unit in one embodiment;

[0027] FIG5 is a schematic diagram of a line feature extraction unit in one embodiment;

[0028] FIG6 is a schematic diagram of a ring-shaped convolution kernel in one embodiment;

[0029] FIG7 is a schematic diagram of a linear convolution kernel in one embodiment;

[0030] FIG8 is a schematic flow chart of an identity recognition method in yet another embodiment;

[0031] FIG9 is a schematic diagram of a process for determining a region of interest in one embodiment;

[0032] FIG10 is a schematic flow chart of a palmprint recognition method according to an embodiment;

[0033] FIG11 is a schematic diagram of determining a region of interest in one embodiment;

[0034] FIG12 is a schematic diagram of a 9*9 linear convolution kernel in one embodiment;

[0035] FIG13 is a schematic diagram of a 16*16 linear convolution kernel in one embodiment;

[0036] FIG14 is a block diagram of the structure of an identity recognition device in one embodiment;

[0037] FIG15 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0039] The identity recognition method provided in the embodiment of the present application can be applied in the application environment shown in Figure 1. Therein, the terminal 102 communicates with the server 104 through the network. The data storage system can store data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can capture images of the target part of the user to be identified and obtain a biological part image of the target part. Specifically, the terminal 102 can respond to the identity recognition trigger event and capture a biological part image of the target part of the user to be identified. For example, the terminal 102 can capture a palm part image, a finger part image, and other biological part images of the user to be identified. Terminal 102 transmits the captured biological part image to server 104, which extracts the morphological features of each pixel from the biological part image according to at least one image feature pattern that matches the characteristic pattern type of the target part. The distribution positions of the pixels targeted during each feature extraction process are combined to form the image feature pattern. Server 104 derives the biological part features of the user to be identified based on the morphological features of each pixel, performs feature matching on the biological part features with the registered part features of the registered users, and determines the identity recognition result for the user to be identified based on the feature matching result. Server 104 may return the identity recognition result for the user to be identified to terminal 102, so that terminal 102 can process the result based on the identity recognition result, such as releasing access control for the user to be identified.

[0040] In addition, the server 104 may directly return the feature matching results to the terminal 102, so that the terminal 102 can determine the identity recognition result for the target user based on the feature matching results returned by the server 104, thereby achieving the identity recognition of the target user. In other optional applications, the identity recognition process can also be implemented solely by the terminal 102. That is, the terminal 102 extracts the morphological features of each pixel from the biological part image according to at least one image feature pattern that is compatible with the characteristic morphological type of the target part. The terminal 102 obtains the biological part features of the user to be identified based on the morphological features of each pixel, performs feature matching on the biological part features with the registered part features of the registered user, and determines the identity recognition result for the user to be identified based on the feature matching results.

[0041] Terminal 102 may include, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices may include smart speakers, smart TVs, smart air conditioners, and smart car-mounted devices. Portable wearable devices may include smart watches, smart bracelets, and head-mounted devices. Terminal 102 may be equipped with a sensor device that captures images of target areas of the user to enable biometric collection of targeted areas. Server 104 may be implemented as a standalone server or a server cluster consisting of multiple servers, or it may be implemented as a cloud server.

[0042] In one embodiment, as shown in FIG2 , an identity recognition method is provided. The method is executed by a computer device. Specifically, the method can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to the server in FIG1 as an example for description. The method includes the following steps 202 to 208. Among them:

[0043] Step 202: Acquire a biological part image of a target part of the user to be identified.

[0044] Identity verification is the process of verifying whether a user's true identity matches their claimed identity. With the development of identity verification technology, biometric-based identity verification methods have gained widespread application. Identity verification processing can be applied in various scenarios and can be triggered based on identity verification trigger events. Identity verification trigger events refer to events that trigger identity verification and may include, but are not limited to, operations or instructions that trigger identity verification. For example, in an access control system, when a user needs to pass through a gate, an identity verification event can be triggered to perform identity verification processing on the user. Another example is when a user makes a payment at a payment terminal. As shown in Figure 3, the target part can be the palm. The terminal can capture an image of the palm of the user to be identified, obtaining a biometric part image, specifically a palm image, based on which identity verification processing can be performed. In addition, identity recognition can also be applied to anti-addiction system scenarios. For example, in an online game anti-addiction system, if it is necessary to limit the online gaming time of minors, then when the anti-addiction system is triggered, such as when the cumulative online gaming time of a game user reaches a preset time threshold, an identity recognition event can be triggered to identify the game user. Through identity recognition, it can be determined whether the game user is an adult, or whether he or she is the game account owner, thereby limiting the online gaming time of minors.

[0045] Identity recognition is performed based on collected biometrics, which are features of measurable body parts of the user, such as hand shape, fingerprints, facial shape, iris, retina, palm, and other types of biometric features. When performing identity recognition based on biometrics of measurable body parts of the user, it is necessary to collect biometric data for the user's body parts and extract biometric features from the collected biometric data, thereby performing identity recognition on the user based on the extracted biometric features. For example, if identity recognition is based on fingerprint recognition, it is necessary to collect fingerprint data from the user's thumb and perform identity recognition on the user based on the collected fingerprint data, such as a fingerprint image. For another example, if identity recognition is based on the palm of the hand, it is necessary to collect palm data from the user's palm and perform identity recognition on the user based on the collected palm data.

[0046] A user to be identified refers to a user who requires identity recognition, such as the user who triggers an identity recognition event. For example, when a user passes through an access control system, they may enter the access control system's data collection area. If the access control system detects the presence of a user in the data collection area, indicating that identity recognition is required, identity recognition is triggered. The access control system then collects biometric data of the user to be identified within the data collection area, such as facial data, finger data, or palm data, among other biometric data. A target part is the part of the human body for which biometric data is to be collected. The target part is related to the biometric data or biometric features involved in identity recognition. For example, if identity recognition is based on the face, the corresponding target part is the facial part of the user to be identified, the collected biometric data is facial data, specifically a facial image, and the biometric features for identity recognition are facial features. For another example, if identity recognition 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 for identity recognition are palm features. A biological part image is an image collected for a target part, and the biological part image carries the biological features of the corresponding target part. For example, when the target part is a face, the biological part image can be a face image; when the target part is a finger, the biological part image can be a finger image; when the target part is a palm, the biological part image can be a palm image.

[0047] Specifically, the server can obtain a biological part image. The biological part image is captured for a target part of the user to be identified. For example, the biological part image can be captured for the target part of the user to be identified using a camera. Different target parts can be captured to obtain biological part images with different biometric characteristics. The biological part image can be captured by the terminal for the target part of the user to be identified and sent to the server.

[0048] Step 204 determines the characteristic morphological type of the target part and extracts the morphological features of each pixel from the biological part image according to at least one image characteristic pattern that matches the characteristic morphological type. The image characteristic pattern is the pattern obtained by combining the distribution positions of the pixels covered by the feature extraction, and the pixels covered by the feature extraction are the pixels in the biological part image targeted during each feature extraction.

[0049] The characteristic pattern type refers to the corresponding manifestation of the biometric characteristics carried by the target part. For different target parts, the manifestation of biometric information varies. For example, for fingers, biometric information is mainly reflected through the lines on the fingers. The characteristic pattern type can be the pattern type formed by the uneven skin on the fingertips. Specifically, it can include patterns such as whorl, arch, and loop. For another example, for the palm, biometric information can be reflected through the palm lines or palm vein lines. The corresponding characteristic pattern type can be the palm line or palm vein pattern type, specifically a linear pattern type.

[0050] An image feature pattern refers to a feature pattern used when extracting features from an image of a biological part of a target part. The image feature pattern is compatible with the feature pattern type of the target part, and specifically can correspond to the corresponding manifestation of the biological features carried by the target part, thereby being suitable for extracting the biological features carried by the target part. Different types of image feature patterns can correspond to different feature extraction methods, so that feature extraction can be performed using a feature extraction method that is compatible with the feature pattern type of the target part, thereby improving the feature extraction effect. Optionally, the adaptive correspondence between the image feature pattern and the feature pattern type can be configured based on multiple test results. In specific implementation, multiple candidate image feature patterns can be pre-determined. For each feature pattern type of a biological part, feature extraction can be performed using various candidate image feature patterns. Based on the feature extraction results corresponding to each candidate image feature pattern, the server can determine an image feature pattern suitable for the feature pattern type from the various candidate image feature patterns, thereby constructing an adaptive correspondence between the image feature pattern and the feature pattern type.

[0051] The image feature pattern is the pattern formed by the combination of the distribution positions of the pixels targeted during each feature extraction. Specifically, when feature extraction is performed on pixels in a biological part image, the pattern formed by combining the distribution positions of the targeted pixels (i.e., the pixels covered by the feature extraction) in the biological part image conforms to the image feature pattern. The image feature pattern can be implemented based on a feature extraction unit. For example, when the target part is a finger, the feature pattern type can include a spiral pattern, and the image feature pattern can also be a spiral pattern. Specifically, feature extraction can be performed using a spiral feature extraction unit, that is, when feature extraction is performed, the pattern formed by the combination of the distribution positions of the pixels covered by the feature extraction in the biological part image is a spiral pattern. In specific applications, the pattern formed by the combination of the distribution positions of the pixels targeted during each feature extraction can be determined, and a judgment can be made based on the formed pattern. For example, the formed pattern can be matched with each candidate image feature pattern to determine whether the formed pattern matches the image feature pattern, thereby determining whether the pattern formed by the combination of the distribution positions of the pixels targeted during each feature extraction is an image feature pattern that is compatible with the feature pattern type. The image feature patterns may include at least one, so that feature extraction can be performed on the biological part image according to one or more image feature patterns to obtain the respective pattern features of each pixel under different image feature patterns.

[0052] In specific applications, image feature patterns can be implemented using corresponding feature extraction units. Specifically, feature extraction is performed using feature extraction units whose shapes match the image feature patterns, yielding the morphological features of each pixel. As shown in Figure 4 , if the image feature pattern is annular, feature extraction can be performed on the biological part image using annular feature extraction units 1 and 2. Feature extraction units 1 and 2 can have different sizes, meaning that when extracting features from the biological part image, only pixels within the annular region covered by the annular feature extraction unit are extracted. As shown in Figure 5 , if the image feature pattern is linear, feature extraction can be performed on the biological part image using linear feature extraction units. For example, horizontal feature extraction unit A can be used to extract features from the biological part image. By rotating horizontal feature extraction unit A, feature extraction units B, C, and D in different directions can be obtained depending on the rotation angle, thereby forming feature extraction units with multiple directions for feature extraction.

[0053] Morphological features are the feature extraction results obtained by extracting features from biological part images according to image feature patterns. Morphological features correspond to pixels, meaning that a morphological feature can be extracted for each pixel in the biological part image. For each pixel in the biological part image, a corresponding morphological feature can be extracted for each image feature pattern. For example, if feature extraction is performed on a biological part image using three image feature patterns, each pixel can extract a morphological feature corresponding to each of the three image feature patterns, resulting in three morphological features for each pixel.

[0054] Specifically, for a biological part image, the server can determine at least one image feature pattern and perform feature extraction on the biological part image based on the at least one image feature pattern to obtain the morphological features of each pixel in the biological part image. When performing feature extraction on the biological part image according to the image feature pattern, the pixels targeted during each feature extraction, i.e., the distribution positions of the pixels covered by the feature extraction, can be combined to form a corresponding image feature pattern. If the at least one image feature pattern matches the feature pattern type of the target part, the server can determine the feature pattern type of the target part and, based on the feature pattern type, determine at least one image feature pattern that matches it. In a specific implementation, after the server determines the feature pattern type of the target part, if there are multiple feature pattern types, the server can filter at least one feature pattern type from the multiple feature pattern types, such as filtering out at least one feature pattern type with the highest frequency, and determine at least one image feature pattern that matches it. Alternatively, the server can directly determine the image feature pattern that matches each of the multiple feature pattern types.

[0055] In one exemplary application, image feature patterns can be implemented based on a feature extraction unit, which can specifically be implemented based on a convolution kernel. Specifically, the feature extraction unit uses at least one convolution kernel adapted to the target part's feature pattern type to perform feature extraction on the biological part image, thereby obtaining the morphological features of each pixel in the biological part image. For example, if the target part is a palm, the feature pattern type can be linear. In this case, at least one linear convolution kernel can be determined, and each linear convolution kernel can have a different orientation. Thus, feature extraction can be performed on the palm image based on the linear convolution kernel in at least one orientation, effectively extracting palm line features from the palm image.

[0056] As shown in Figure 6, in a 16*16 convolution kernel, by setting the convolution kernel units filled with diagonal lines to valid, such as the convolution weight can be set to 1, and setting the convolution kernel units not filled with diagonal lines to invalid, such as the convolution weight can be set to 0, a ring-shaped convolution kernel can be formed, thereby enabling feature extraction based on the ring-shaped image feature pattern. As shown in Figure 7, in a 16*16 convolution kernel, by setting the convolution kernel units filled with diagonal lines to valid and the convolution kernel units not filled with diagonal lines to invalid, a horizontal linear convolution kernel can be formed. Based on this horizontal linear convolution kernel, feature extraction for biological part images can obtain horizontal linear morphological features.

[0057] Step 206 : Obtain the biological part feature of the user to be identified based on the morphological features of each pixel.

[0058] Among them, the biological part features are used to characterize the biological features carried by the target part of the user to be identified. The target parts of different users may correspond to different biological part features, so that identity recognition processing can be performed on the user based on the biological part features.

[0059] Specifically, the server can determine the biometric features of the user to be identified based on the morphological features of each pixel in the biometric image. For example, if each pixel has only one morphological feature, the server can directly concatenate the morphological features of each pixel to obtain the morphological features of the biometric image, and then obtain the biometric features of the user to be identified based on the morphological features of the biometric image. For example, the server can directly use the morphological features of the biometric image as the biometric features of the user to be identified, or further extract features from the morphological features of the biometric image. Specifically, the server can use a pre-trained feature extraction model to further extract features from the morphological features of the biometric image to obtain the biometric features of the user to be identified. Furthermore, if each pixel has at least two morphological features, indicating that the image has at least two feature types, the server can fuse the multiple morphological features of each pixel and then concatenate the fused morphological features to obtain the morphological features of the biometric image, and then obtain the biometric features of the user to be identified based on the morphological features of the biometric image.

[0060] In step 208, the biological part features are matched with the registered part features to obtain a feature matching result, and the identity recognition result of the user to be identified is determined based on the feature matching result, wherein the registered part features are the biological part features obtained by performing identity registration on the biological part image corresponding to the target part of the registered user.

[0061] Among them, the registered part feature is the biological part feature obtained by identity registration of the biological part image corresponding to the target part of the registered user. The registered part feature of the registered user can be used as the benchmark feature for identity recognition. The biological part feature of the user to be identified is matched with the registered part feature. The identity recognition result of the user to be identified can be determined based on the feature matching result, such as determining whether the user to be identified is a registered user. If it is a registered user, the specific user ID of the user to be identified can be further determined.

[0062] Specifically, the server can perform identity recognition based on the biological part features of the user to be identified. Specifically, the server can obtain the registered part features obtained by the registered user through identity registration in advance. The server performs feature matching on the biological part features with each registered part feature to obtain each feature matching result. The server determines the identity recognition result of the user to be identified based on each feature matching result. Specifically, it can determine whether the user to be identified is a registered user based on each feature matching result. When the user to be identified is a registered user, it can further determine the user identity information of the user to be identified. In specific applications, the user can use the biological part image corresponding to the target part to register the identity in advance. Specifically, the biological part features can be extracted based on the biological part image corresponding to the target part, and the extracted biological part features can be bound to the user identification to realize the user's identity registration. The registered part features are obtained based on the biological part features of the registered user when performing identity registration, so that the registered part features of each registered user are used as the reference features for identity recognition.

[0063] In a specific application, as shown in FIG8 , a biological part image of a target part of a user to be identified can be first captured, and the server can obtain the captured biological part image. The server can determine the characteristic pattern type corresponding to the target part and determine at least one image feature pattern compatible with the characteristic pattern type. The server can perform feature extraction on the biological part image according to the at least one image feature pattern to obtain the morphological features of each pixel in the biological part image. The server can then obtain the biological part features of the user to be identified based on the morphological features of each pixel. The server can perform feature matching on the biological part features of the user to be identified with the registered part features to obtain a feature matching result, and determine the identity recognition result for the user to be identified based on the feature matching result.

[0064] In the above-mentioned identity recognition method, a biological part image is obtained for the target part of the user to be identified. A morphological feature of each pixel is extracted from the biological part image according to at least one image feature pattern that matches the characteristic pattern type of the target part. The distribution positions of the pixels targeted during each feature extraction are combined to form an image feature pattern. Based on the morphological features of each pixel, the biological part features of the user to be identified are obtained. The biological part features are then feature-matched with the registered part features of the registered user, and an identity recognition result for the user to be identified is determined based on the feature matching results. During the identity recognition process, the morphological features of each pixel are extracted according to at least one image feature pattern that matches the characteristic pattern type of the target part. This enhances the targeted expression of the biological part features carried by the target part, enables accurate acquisition of the biological part features carried by the target part, and thus improves the accuracy of identity recognition based on the biological part features of the target part.

[0065] In an exemplary embodiment, the target part is a palm, and the image feature pattern is a linear pattern; according to at least one image feature pattern that is compatible with the feature pattern type, the morphological features of each pixel are extracted from the biological part image, including: extracting the morphological features of each pixel from the biological part image according to the linear pattern in at least one direction.

[0066] The target part is the palm, and the biological part image corresponding to the target part is a palm image. The palm feature type includes palm print or palm vein feature patterns, specifically linear patterns. Palm print refers to the texture information of the palm from the fingertips to the wrist, including various palm print features such as main lines, wrinkles, fine lines, ridge endings, and bifurcation points that can be used for identity recognition. Palm print features refer to the characteristics reflected by the palm texture information and can be extracted from the palm image by capturing the palm. Different users generally have different palm print features, that is, different users have different palm texture features. Based on palm print features, identity recognition can be performed for different users. Palm vein refers to the palm vein information, which reflects the venous lines in the human palm and has liveness recognition capabilities. Palm vein images can be captured by infrared cameras. Palm vein features are the venous features of the palm obtained through palm vein analysis. Different users generally have different palm vein features, that is, different users have different palm vein features. Based on palm vein features, identity recognition can also be performed for different users.

[0067] Biometric information of palm prints or palm veins is generally represented by palm lines or vein lines. If the palm's characteristic pattern type is a linear pattern, then the image characteristic pattern that matches this characteristic pattern type is also a linear pattern. Linear pattern feature extraction can have different directions, thereby classifying different types of line patterns. For example, line patterns can include horizontal, vertical, and various directions that form a certain angle with the horizontal direction. Line patterns in different directions can be considered different types of line patterns, allowing feature extraction to be performed based on the line patterns in different directions.

[0068] For example, for a palm image, the server can determine a line pattern in at least one direction, where the line pattern is compatible with a characteristic palm pattern type. The server can then perform feature extraction on the biological part image based on the determined line pattern in at least one direction, obtaining the morphological features of each pixel in the biological part image. For example, the server can determine a line feature extraction unit corresponding to each line pattern in each direction, and perform feature extraction on the biological part image based on the line feature extraction unit to obtain the morphological features of each pixel in the biological part image. As shown in Figure 5, the server can determine to use one of feature extraction unit A, feature extraction unit B, feature extraction unit C, and feature extraction unit D to perform feature extraction on the biological part image.

[0069] In this embodiment, for the biological part image of the palm, the characteristic pattern type of the palm is a linear pattern. The server extracts features of the biological part image according to the linear pattern in at least one direction to obtain the pattern features of each pixel in the biological part image. Thus, the palm image can be extracted according to the linear pattern in at least one direction, which enhances the targeted expression of palm features and can accurately obtain palm features, thereby improving the accuracy of identity recognition based on palm features.

[0070] In an exemplary embodiment, a morphological feature of each pixel of a biological part image is extracted according to a linear pattern in at least one direction, including: extracting directional morphological features corresponding to each pixel and the linear patterns in at least two directions according to the linear patterns in at least two directions; for each pixel, fusing the directional morphological features corresponding to the pixel and the linear patterns in at least two directions to obtain a directional fusion feature of the pixel; and obtaining a morphological feature of each pixel based on the directional fusion feature of each pixel.

[0071] Among them, line patterns in different directions correspond to different types of line patterns, which can correspond to different feature extraction methods. When feature extraction is performed on a biological part image using at least two image feature patterns, each pixel will obtain directional pattern features corresponding to at least two image feature patterns. For example, when feature extraction is performed on a biological part image according to line patterns in the horizontal and vertical directions, each pixel in the biological part image can obtain a directional fusion feature corresponding to the horizontal line pattern, and a directional fusion feature corresponding to the vertical line pattern. The directional fusion feature is a feature obtained by fusing the directional pattern features corresponding to each line pattern of the same pixel. Specifically, the fusion can be performed using a weighted fusion method, and the pattern feature of the corresponding pixel can be obtained based on the directional fusion feature.

[0072] Specifically, for a palm image, the server can determine line patterns in at least two directions and perform feature extraction on the biological part image based on the line patterns in the at least two directions, obtaining directional pattern features corresponding to each pixel in the biological part image. Each pixel includes a directional pattern feature corresponding to each of the at least two directional line patterns. For example, if the line patterns include five directions, after performing feature extraction on the biological part image based on the line patterns in the five directions, for each pixel in the biological part image, such as pixel A, pixel A includes a directional pattern feature corresponding to each directional line pattern, i.e., pixel A includes five directional pattern features. In a specific implementation, the server can determine a line feature extraction unit corresponding to each directional line pattern and perform feature extraction on the biological part image based on the line feature extraction unit to obtain a pattern feature for each pixel in the biological part image. As shown in Figure 5, the server can determine to use at least two of feature extraction unit A, feature extraction unit B, feature extraction unit C, and feature extraction unit D to perform feature extraction on the biological part image.

[0073] For each pixel whose directional morphological features are extracted from the biological part image, the server fuses the various directional morphological features of the pixel to obtain the directional fusion feature of the pixel. After traversing each pixel, the server obtains the directional fusion feature of each pixel. In a specific application, for each pixel, the server can average and fuse the various directional morphological features of the pixel, or perform weighted fusion to obtain the directional fusion feature of the pixel. The server can obtain the morphological features of each pixel based on the directional fusion features of each pixel. The specific server can directly use the directional fusion features of each pixel as their own morphological features; the server can also perform further feature mapping processing on the directional fusion features of each pixel to obtain the morphological features of each pixel.

[0074] In this embodiment, for the biological part image of the palm, the characteristic pattern type of the palm is a linear pattern. The server performs feature extraction on the biological part image according to the linear patterns in at least two directions to obtain the directional pattern features of each pixel in the biological part image. The directional pattern features of each pixel are fused to obtain a directional fusion feature, and the pattern features of each pixel are obtained based on the directional fusion features of each pixel. Thus, the palm image can be feature extracted according to the linear patterns in multiple directions. By fusing the extraction results corresponding to the linear patterns in various directions, the palm features can be accurately obtained, thereby improving the accuracy of identity recognition based on palm features.

[0075] In an exemplary embodiment, extracting morphological features of each pixel from the biological part image according to at least one image feature pattern that is compatible with the feature pattern type includes: determining at least one image feature pattern that is compatible with the feature pattern type at at least one image feature scale; and extracting morphological features of each pixel from the biological part image according to at least one image feature pattern at at least one image feature scale.

[0076] Among them, the image feature scale represents the range of image pixels covered during each feature extraction. Different image feature scales can cover different pixel ranges. The larger the value of the image feature scale, the more pixels are targeted each time feature extraction is performed, that is, the more pixels are covered by feature extraction each time feature extraction is performed. The numerical value of the image feature scale can be flexibly set according to actual needs, and multiple numerical values ​​of the image feature scale can also be set. The image feature scale reflects the number of pixels covered by feature extraction each time feature extraction is performed on the biological part image, and the image feature pattern reflects the distribution position of the pixels covered by feature extraction each time feature extraction is performed on the biological part image. By combining the image feature scale and the image feature pattern, the effective morphological features of the biological part image can be fully extracted.

[0077] Specifically, the server determines at least one image feature scale, and each image feature scale can determine at least one corresponding image feature pattern, with the image feature pattern being compatible with the target part's feature pattern type. For example, the image feature scales can include three scales: 4*4, 8*8, and 16*16, while the image feature pattern can include line patterns in 10 directions. The server performs feature extraction on the biological part image according to the at least one image feature pattern at the at least one determined image feature scale, obtaining morphological features for each pixel in the biological part image. In a specific implementation, for each pixel in the biological part image from which morphological features are extracted, if the number of morphological features corresponding to each pixel is one, such as if feature extraction is performed according to one image feature pattern at one image feature scale, and each pixel obtains a morphological feature corresponding to the image feature scale and image feature pattern, the server may determine this as the morphological feature corresponding to the corresponding pixel. If the number of morphological features corresponding to each pixel is more than one, such as if at least one of the image feature scales or image feature patterns is more than one, resulting in multiple morphological features for each pixel, the server may fuse the multiple morphological features to obtain the morphological feature corresponding to the corresponding pixel.

[0078] In specific applications, the server can determine a feature extraction unit corresponding to at least one image feature type at each image feature scale, and perform feature extraction on the biological part image based on the feature extraction unit, obtaining the morphological features of each pixel in the biological part image. As shown in Figure 4, the annular feature extraction units 1 and 2 correspond to different image feature scales. The server can determine at least one of the two from feature extraction units 1 and 2 to perform feature extraction on the biological part image.

[0079] In this embodiment, the server performs feature extraction on the biological part image according to at least one image feature type under at least one image feature scale to obtain the morphological features of each pixel in the biological part image. Feature extraction can be performed according to actual scale requirements, which is beneficial to enhance the targeted expression of the target part features in combination with the feature scale, and can accurately obtain the biological part features of the target part, thereby improving the accuracy of identity recognition based on the biological part features of the target part.

[0080] In an exemplary embodiment, a morphological feature of each pixel is extracted from a biological part image according to at least one image feature pattern at at least one image feature scale, including: 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 pattern at at least two image feature scales; fusing, for each pixel, the scale morphological features corresponding to the pixel and the at least two image feature scales to obtain a scale fusion feature of the pixel; and obtaining a morphological feature of each pixel based on the scale fusion feature of each pixel.

[0081] Among them, there are at least two image feature scales, and each image feature scale can correspond to at least one image feature type, thereby enabling multi-scale feature extraction. For example, if the number of image feature scales is M and the number of image feature types is N, then M*N types of feature extraction can be achieved, that is, for each pixel in the biological part image undergoing feature extraction, M*N morphological features will be extracted. Scale morphological features are morphological features extracted according to an image feature type at an image feature scale, and each scale morphological feature corresponds to an image feature type at an image feature scale. For each pixel in the biological part image undergoing feature extraction, the number of scale morphological features is the product of the number of image feature scale types and the number of image feature types. For each pixel in the biological part image undergoing feature extraction, the scale fusion feature is the feature obtained by fusing the multiple scale morphological features corresponding to the pixel. Based on the scale fusion feature, the morphological feature of the pixel can be obtained.

[0082] Specifically, the image feature scales determined by the server may include at least two, and each image feature scale may include at least one image feature type. At each image feature scale, the server performs feature extraction on the biological part image according to the at least one determined image feature type, obtaining scale-morphological features for each pixel, with each scale-morphological feature corresponding to an image feature type at an image feature scale. For example, for image feature scale 1, image feature scale 2, and image feature scale 3, at each image feature scale, corresponding scale-morphological features may be extracted for pixel A in the biological part image. Specifically, these may include scale-morphological features a1 extracted at image feature scale 1, scale-morphological features a2 extracted at image feature scale 2, and scale-morphological features a3 extracted at image feature scale 3. At each image feature scale, feature extraction of at least one image feature type may be performed. For example, if these may include image feature type 1, image feature type 2, and image feature type 3, then for pixel A in the biological part image, scale-morphological features corresponding to the three image feature types may be extracted at each image feature scale. For example, the 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 based on image feature type 1, scale-morphological feature a12 is extracted based on image feature type 2, and scale-morphological feature a13 is extracted based on image feature type 3. If there are multiple image feature scales, each pixel may have multiple scale-morphological features. For each pixel, the server may fuse the scale-morphological features of the pixel, such as by performing an average fusion, to obtain the scale-fused feature of the pixel. After traversing each pixel and obtaining the scale-fused feature of each pixel, the server may obtain the morphological feature of each pixel based on the scale-fused feature of each pixel. For example, the server may directly use the scale-fused feature of each pixel as the morphological feature, or may perform further feature mapping on the scale-fused feature of each pixel to obtain the morphological feature of each pixel.

[0083] In this embodiment, the server performs feature extraction on the biological part image according to at least one image feature type under at least two image feature scales to obtain the scale morphological features of each pixel in the biological part image. The scale fusion features are obtained by fusing the scale morphological features of each pixel, and the morphological features of each pixel are obtained according to the scale fusion features of each pixel. Thus, feature extraction can be performed on the biological part image based on multiple scales. By fusing the feature extraction results of multiple scales, the biological part features of the target part can be accurately obtained, thereby improving the accuracy of identity recognition based on the biological part features of the target part.

[0084] In an exemplary embodiment, morphological features of each pixel are extracted from an image of a biological part according to at least one image feature pattern that is compatible with the feature pattern type, including: extracting morphological features of each pixel from the image of the biological part through a convolutional network in a pre-trained feature extraction model; wherein the convolutional network is used to perform feature extraction on the image of the biological part according to at least one image feature pattern that is compatible with the feature pattern type of the target part.

[0085] The feature extraction model may be pre-trained based on sample data, and the feature extraction model may be a model trained based on various neural network algorithms, and the neural network algorithms may specifically include but are not limited to CNN (Convolutional Neural Networks) algorithms, RNN (Recurrent Neural Network) algorithms, Tansformer (Transformer Model) algorithms, MLP (Multilayer Perceptron) algorithms, ResNets (Residual Network) algorithms, etc. The feature extraction model may include a convolutional network, which is used to extract features according to at least one image feature pattern that is compatible with the characteristic pattern type of the target part, that is, the convolutional network in the feature extraction model is used to extract the morphological features of the input image, so as to extract features from the input image according to at least one image feature pattern.

[0086] For example, the server may obtain a pre-trained feature extraction model and, using a convolutional network within the pre-trained feature extraction model, extract features according to at least one image feature pattern adapted to the target part's characteristic pattern type, thereby extracting the morphological features of each pixel from the biological part image. In a specific application, each image feature pattern may correspond to a convolution kernel, and feature extraction processing for the image feature pattern may be performed using the corresponding convolution kernel. Specifically, feature extraction of the biological part image may be performed using at least one convolution kernel. The convolutional network may include at least one convolution kernel. By inputting the biological part image into the convolutional network, feature extraction processing for the corresponding image feature pattern may be performed using the at least one convolution kernel, thereby obtaining the morphological features of each pixel.

[0087] Furthermore, biological part features of the user to be identified are obtained based on the morphological features of each pixel, including: extracting the biological part features of the user to be identified based on the morphological features of each pixel through a part feature extraction network in the feature extraction model.

[0088] The feature extraction model also includes a subnetwork of a part feature extraction network, which further extracts features from the morphological characteristics of each pixel through the part feature extraction network to obtain the biological part features of the user to be identified. The part feature extraction network can be constructed using an artificial neural network algorithm based on actual needs and trained based on sample data. For example, the part feature extraction network can be constructed based on at least one of a CNN algorithm, an RNN algorithm, a Tansformer algorithm, an MLP algorithm, or a ResNet algorithm.

[0089] For example, the server may input the morphological features of each pixel into a part feature extraction network within the feature extraction model. The part feature extraction network within the feature extraction model then performs 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 may concatenate the morphological features of each input pixel. Specifically, the concatenation may be performed based on the distribution of each pixel in the biological part image to obtain the morphological features of the biological part image. Feature extraction may then be performed on the morphological features of the biological part image to obtain the biological part features of the user to be identified.

[0090] In this embodiment, through the convolutional network in the pre-trained feature extraction model, the morphological features of each pixel are extracted from the biological part image according to at least one image feature pattern that is adapted to the characteristic morphological type of the target part. Through the part feature extraction network in the feature extraction model, the morphological features of each pixel are extracted to obtain the biological part features of the user to be identified. Efficient and accurate feature extraction processing can be achieved based on the artificial neural network model, which is conducive to improving the processing efficiency and accuracy of identity recognition.

[0091] In an exemplary embodiment, a feature extraction model is obtained through a model training step; the model training step includes: obtaining a plurality of biological part image samples; extracting sample morphological features of each sample pixel from the biological part image samples through a convolutional network in the feature extraction model to be trained; extracting biological part sample features for each sample pixel through a part feature extraction network in the feature extraction model to be trained; determining a training loss based on the biological part sample features and the sample morphological features; and according to the training loss, updating the convolutional network and the part feature extraction network in the feature extraction model to be trained respectively and continuing the training until the training is completed to obtain a trained feature extraction model.

[0092] The biological part image samples are sample data for training the feature extraction model. The biological part image samples may carry identity tags so that the feature extraction performance of the feature extraction model can be determined based on the identity tags. The sample morphological features are extracted by the convolutional network in the feature extraction model to be trained based on the biological part image samples. The convolutional network can perform feature extraction on the input biological part image samples according to at least one image feature type that is compatible with the feature morphological type of the target part. The biological part sample features are biological features extracted by the part feature extraction network in the feature extraction model to be trained. The training loss is derived based on the biological part sample features and the sample morphological features. Based on the training loss, the feature extraction performance of the feature extraction model can be evaluated, thereby updating the feature extraction model to improve the feature extraction performance of the feature extraction model. The biological part sample features can reflect the feature extraction performance of the part feature extraction network, and the sample morphological features can reflect the feature extraction performance of the convolutional network.

[0093] For example, when training a feature extraction model, the server may obtain multiple biological part image samples. The biological part image samples may be acquired based on a user's target part. The server performs feature extraction on the biological part image samples using a convolutional network in the feature extraction model to be trained, thereby extracting sample morphological features for each sample pixel from the biological part image samples. The server further extracts sample morphological features for each sample pixel using the part feature extraction network in the feature extraction model to be trained, thereby obtaining biological part sample features. The server determines a training loss based on the sample morphological features and the sample morphological features, and updates model parameters of the feature extraction model to be trained based on the training loss. Specifically, the server updates network parameters of the convolutional network and the part feature extraction network in the feature extraction model to be trained, such as weight parameters within the network. Training continues using the updated feature extraction model until training is completed, thereby obtaining a trained feature extraction model. For example, when training reaches a predetermined number of training times, the feature extraction model meets convergence conditions, or the feature extraction performance of the feature extraction model meets performance requirements, the training termination conditions may be satisfied, and training is terminated. The trained feature extraction model is then obtained based on the feature extraction model at the end of training.

[0094] 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, determines the training loss according to the sample morphological characteristics reflecting the feature extraction capability of the convolutional network and the biological part sample characteristics reflecting the feature extraction capability of the part feature extraction network, and updates the feature extraction model based on the training loss. This can ensure the feature extraction performance of each of the convolutional network and the part feature extraction network, thereby improving the feature extraction performance of the feature extraction model, which is conducive to improving the accuracy of identity recognition.

[0095] In an exemplary embodiment, a training loss is determined based on biological part sample features and sample morphological features, including: obtaining a part feature extraction loss based on the biological part sample features; determining a negative sample pair, the negative sample pair including biological part image samples carrying different identity labels; obtaining a sample pair loss based on the sample morphological features of each biological part image sample in the negative sample pair; and obtaining a training loss based on the part feature extraction loss and the sample pair loss.

[0096] The part feature extraction loss is obtained based on the characteristics of the biological part samples and is used to reflect the feature extraction performance of the part feature extraction network. It can be calculated using various loss algorithms, such as the arcface (additive angular margin loss) algorithm, the mean squared error (MSE) algorithm, the cross entropy algorithm, and other loss algorithms. The negative sample pairs include biological part image samples with different identity labels, that is, the biological part image samples in the negative sample pairs correspond to different users. The sample pair loss is obtained based on the sample morphological characteristics of the biological part image samples in the negative sample pairs. 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 part feature extraction loss and the sample pair loss. It can be obtained by fusing the part feature extraction loss and the sample pair loss.

[0097] Specifically, the server can obtain the part feature extraction loss based on the biological part sample features, such as by calculating it according to the arcface algorithm based on the biological part sample features. The server determines a negative sample pair including biological part image samples carrying different identity labels, and determines the sample morphological features of each biological part image sample in the negative sample pair. The server obtains the sample pair loss based on the sample morphological features of each biological part image sample in the negative sample pair, such as by calculating the sample pair loss based on the sample morphological features of each biological part image sample in the negative sample pair based on the L1 loss algorithm. The server can obtain the training loss based on the part feature extraction loss and the sample pair loss, and specifically, the training loss can be obtained according to the sum of the part feature extraction loss and the sample pair loss, so as to update the model parameters of the feature extraction model through the training loss.

[0098] In this embodiment, the server obtains the part feature extraction loss that reflects the feature extraction capability of the part feature extraction network based on the sample morphological characteristics, obtains the sample pair loss that reflects the feature extraction capability of the convolutional network based on the sample morphological characteristics of each biological part image sample in the negative sample pair, and obtains the training loss based on the part feature extraction loss and the sample pair loss. Updating the model based on the training loss can ensure the feature extraction performance of each of the convolutional network and the part feature extraction network, thereby improving the feature extraction performance of the feature extraction model, which is conducive to improving the accuracy of identity recognition.

[0099] In an exemplary embodiment, the biological part features of the user to be identified are obtained based on the morphological features of each pixel, including: splicing 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 extracting features from the morphological features of the biological part image to obtain the biological part features of the user to be identified.

[0100] Here, the distribution position refers to the spatial position of the pixel in the biological part image. For example, the server can determine the distribution position of each pixel in the biological part image, and according to the respective distribution positions, splice the morphological features of each pixel to obtain the morphological features of the biological part image. The server further performs feature extraction based on the morphological features of the biological part image to obtain the biological part features of the user to be identified. In a specific application, the server can perform feature extraction on the morphological features of the biological part image through the part feature extraction network in the pre-trained feature extraction model. For example, the morphological features of the biological part image can be input into the part feature extraction network of the feature extraction model to obtain the biological part features of the user to be identified.

[0101] In this embodiment, the server extracts the morphological features of the obtained biological part image by splicing the morphological features of each pixel to obtain the biological part features of the user to be identified. The morphological features of each pixel can be integrated, which is conducive to improving the expressive ability of the biological part features, thereby improving the accuracy of identity recognition.

[0102] In an exemplary embodiment, morphological features of each pixel are extracted from an image of a biological part according to at least one image feature pattern that is compatible with the feature pattern type, including: determining a region of interest from the image of the biological part; determining at least one image feature pattern that is compatible with the feature pattern type; and extracting morphological features of each pixel in the region of interest according to the at least one image feature pattern.

[0103] The region of interest (ROI) is an image region determined from a biological part image for which biological part feature extraction is to be performed. For example, the server may determine the ROI from the biological part image, for example by performing biological part recognition on the biological part image to determine the region including the target part as the ROI. The server may determine at least one image feature pattern that matches the characteristic pattern type of the target part. Specifically, the server may determine the characteristic pattern type of the target part and, based on the characteristic pattern type, determine at least one corresponding image feature pattern. The server then performs feature extraction on the ROI based on the determined at least one image feature pattern to obtain morphological features for each pixel in the ROI.

[0104] In this embodiment, the server determines the region of interest from the biological part image and extracts features based on the region of interest for identity recognition, which can reduce the amount of data required for feature extraction processing and thus improve the processing efficiency of identity recognition.

[0105] In an exemplary embodiment, the target part is a palm, as shown in FIG9 , the process of determining the region of interest, i.e., determining the region of interest from the biological part image, includes steps 902 to 906 . In which:

[0106] Step 902: Detect the feature points of the interdigital spaces between different thumbs in the palm from the biological part image.

[0107] The target part is a palm, the biological part image is a palm image, and the biological part features to be extracted are palm features, which may specifically include at least one of palm print features or palm vein features. Finger joint feature points may be feature points that distinguish each thumb, and specifically may be connection points between thumbs in the palm print portion. Specifically, the server may perform palm feature point recognition on the biological part image to identify each finger joint feature point between thumbs in the palm. For example, the server may identify the connection points between adjacent thumbs on the palm of the thumb, index finger, middle finger, ring finger, and pinky finger to obtain each finger joint feature point.

[0108] Step 904 : determining the focus of interest and the area range parameters from the biological part image based on the feature point positions of the finger joint feature points and the feature point distances between the finger joint feature points.

[0109] Among them, the feature point position refers to the spatial position of the finger joint feature point in the biological part image, and the feature point distance refers to the distance between the feature point positions of each finger joint feature point. The focus of interest is the feature point of the region of interest to be determined, which can specifically be the vertex, center of the circle or center point of the region of interest to be determined. The area range parameter is used to describe the range of the region of interest to be determined, and can specifically include parameters such as side length and radius. Specifically, the server can determine the feature point positions of each finger joint feature point in the biological part image, and determine the feature point distance between each finger joint feature point based on each feature point position. The server determines the focus of interest and area range parameters in the biological part image based on the feature point positions and feature point distances. For example, the center and radius of the circle can be determined based on the feature point positions and feature point distances, so that the circular region of interest determined based on the center and radius can cover each finger joint feature point and the palm area.

[0110] Step 906 : determining a region of interest in the biological part image according to the focus of interest and the region range parameters.

[0111] Specifically, the server determines a region of interest (ROI) in the biological part image based on the focal point and the region range parameter. For example, if the focal point is the center point and the region range parameter is the side length, a polygonal ROI can be constructed with the focal point as the center and the region range parameter as the geometric side length. For another example, if the focal point is the center point and the region range parameter is the radius, a circular ROI can be constructed with the focal point as the circle point and the region range parameter as the radius.

[0112] In this embodiment, the server detects the feature points of each finger joint between the palm and the thumb in the biological part image, and determines the focus of interest and area range parameters based on the feature point position of each finger joint feature point and the feature point distance between each finger joint feature point. The area of ​​interest is determined in the biological part image according to the focus of interest and area range parameters, which can ensure that the area of ​​interest can cover the palm area. While ensuring the accuracy of biometric feature extraction, the amount of data processed by feature extraction is reduced, thereby improving the processing efficiency of identity recognition.

[0113] In an exemplary embodiment, feature matching is performed on biological part features and registered part features to obtain feature matching results, and an identity recognition result for the user to be identified is determined based on the feature matching results, including: obtaining the registered part features of each registered user; determining the feature similarity between the biological part features and each registered part feature; and determining the identity recognition result for the user to be identified based on each feature similarity.

[0114] The registered part feature is a biological part feature obtained by registering the user's target part using an image of the biological part corresponding to the target part. Feature similarity is calculated based on the biological part feature and the registered part feature, and can specifically include at least one of various metrics, such as cosine similarity, Euclidian distance, Manhattan distance, and Minkowski distance.

[0115] Specifically, the server can obtain the pre-stored registration part features of each registered user, and calculate the similarity between the biological part features and the registration part features to determine the feature similarity between the biological part features and the registration part features. The server obtains the identity recognition result of the user to be identified based on the similarity of each feature. For example, the server can determine the registered user corresponding to the registration part feature with the highest similarity as the identity recognition result of the user to be identified. In addition, a similarity threshold can be set. Only when the similarity value exceeds the similarity threshold, the features are considered to match, and the registered user corresponding to the registration part feature with the highest similarity is determined as the identity recognition result of the user to be identified. If the similarity value is less than the similarity threshold, it can be considered that the features do not match, that is, the user to be identified does not belong to the registered users who have completed registration in advance.

[0116] In this embodiment, the server determines the identity recognition result of the user to be identified based on the feature similarity between the biological part features of the user to be identified and the registered part features of each registered user. The biological part features can be used to implement effective identity recognition processing and ensure the accuracy of identity recognition.

[0117] This application also provides an application scenario, which applies the above-mentioned identity recognition method. Specifically, the application of the identity recognition method in this application scenario is as follows:

[0118] Palmprint recognition is a new generation of biometric identification technology, following facial recognition. It has already been applied in areas such as mobile payments and identity verification. Compared to facial recognition, palmprint recognition uses images of the palm print area to identify different users. Palmprints are also more discreet, better protecting user privacy and are unaffected by factors such as masks, makeup, and sunglasses.

[0119] Existing palmprint recognition technologies can generally be divided into three major categories. The first is palmprint recognition based on geometric features, which mainly performs palmprint recognition by identifying geometric shapes on the palm, such as finger spacing, finger width, palm length, etc. Specifically, when implementing a palmprint recognition solution based on geometric features, image preprocessing is performed, specifically performing operations such as denoising, enhancement, and binarization on the input palmprint image; geometric feature extraction is performed, specifically calculating palm geometric features, such as finger width, finger spacing, palm length, etc.; pattern matching is performed, specifically using the extracted geometric features to perform pattern matching with templates in the database to identify the individual's identity. Palmprint recognition methods based on geometric features are more robust to lower-resolution images and environmental interference, but the recognition accuracy is limited.

[0120] The second type is palmprint recognition based on texture features. Texture feature methods focus on the characteristics of the skin texture on the palm surface, such as Directional Field and Gabor Filter. When implementing a palmprint recognition solution based on texture features, image preprocessing is performed, specifically by performing operations such as denoising, enhancement, and filtering on the input palmprint image; texture feature extraction is performed, specifically using signal processing methods such as Directional Field and Gabor Filter to extract texture features from the preprocessed image; and feature matching is performed, specifically by comparing the input image with the texture features in the database to achieve palmprint recognition. Palmprint recognition methods based on texture features use signal processing and machine learning techniques to extract and match palmprint texture features. Texture feature methods have good robustness and recognition accuracy for complex scenarios.

[0121] The third type is deep learning-based palmprint recognition, which leverages 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 solutions involve image preprocessing, specifically denoising, enhancing, and normalizing the input palmprint image; deep learning model training, specifically using a deep convolutional neural network to train labeled palmprint images and learn the hierarchical features of palmprints; and palmprint recognition, specifically inputting the preprocessed input image into the pretrained model for palmprint recognition. Deep learning-based palmprint recognition methods can automatically learn the hierarchical feature representation of palm images, improving the accuracy and robustness of palmprint recognition.

[0122] However, in mobile payment scenarios, the user base is very large, including many highly similar samples. Existing methods based on geometric features, texture features, and deep learning all extract palm features using square convolution kernels. Since palm features are primarily linear, these methods are not very effective in feature extraction.

[0123] The characteristic information of the palm is mainly concentrated in the palm lines, so the extraction of palm line features is very important for distinguishing different palms. In response to the problems of the existing methods, this embodiment proposes a palm print feature extraction method based on linear convolution, which fully learns the linear features of the palm lines to distinguish different palm lines. In response to the form of palm line features, this embodiment proposes a palm print feature extraction method based on linear convolution. By designing a linear convolution feature extractor, the model's extraction effect on linear features is effectively improved, and the model's ability to distinguish different palms is effectively improved, thereby improving the accuracy of palm print recognition. This embodiment proposes a palm print recognition method based on linear convolution: the first step is to use the detection model to detect the positions of the three key points between the index finger, middle finger, and ring finger of the palm; the second step is to extract the palm region of interest based on the positions of the three key points between the finger joints; the third step is to use a multi-scale linear convolution kernel to extract palm print features from the region of interest; the fourth step is to fuse the linear extraction features of different scales; the fifth step is to use a pairwise loss function for feature constraint. This embodiment can more effectively extract the linear features of palm prints by introducing multi-scale linear convolution, while increasing the pixel-level directional map comparison to better distinguish highly similar samples, thereby improving the accuracy of palm print recognition.

[0124] Palmprint recognition technology has broad application prospects in commercial scenarios such as mobile payment and identity verification. This embodiment provides a palmprint feature extraction method based on linear convolution, which can identify the user's identity through feature matching based on the user's palm image. As shown in FIG10 , the palmprint recognition processing process for payment includes: step 1001, performing image acquisition through a terminal payment device; step 1002, acquiring a user's hand image; step 1003, performing finger joint key point detection, specifically, the detection model can detect the three finger joint key points of the user's hand; step 1004, extracting the palm region of interest, specifically, the palm region of interest can be extracted based on the hand image and the key point positions; step 1005, multi-scale linear feature extraction, specifically, using a multi-scale linear model to extract the palm region of interest features; step 1006, feature fusion, specifically, fusing multi-scale linear features; step 1007, extracting features using a recognition model, specifically, reasoning on multi-scale information through the recognition model; step 1008, calculating the similarity between the features and the base database, specifically, calculating the cosine similarity between the spliced ​​features and the base database features; step 1009, determining the recognition result based on the similarity, specifically, the identity information corresponding to the base database photo with the highest similarity can be determined as the target user.

[0125] Palm region detection is performed on the captured image, specifically using object detection technology to locate the finger joints and extract the palm region image from the image. Specifically, as shown in Figure 11, the palm region of interest (ROI) extraction process begins by locating key points between the fingers: index finger A, middle finger B, and ring finger C. A finger joint point detector based on the YOLOv2 (You Only Look Once) object detection algorithm is used to detect key points between the fingers: A, B, and C. The local coordinate system is then determined. The x-axis of the local coordinate system is determined based on key points A and C. The y-axis, perpendicular to the x-axis, is established based on the third point B. The palm print center point D is located along the negative y-axis at a distance AC from the origin, with the distance DE equal to 6 / 5 of the distance AC. Finally, the ROI (Region of Interest) is extracted. The distance from point A to point C is multiplied by 3 / 2 to obtain the ROI (Region of Interest) with a side length d of 1 / 2. The ROI is extracted with point D as the center and a side length of 2d as the side length, serving as the input for the recognition model. As shown in FIG11 , a square area with point D as the center and 3AC as the side length is the determined ROI area.

[0126] For model training processing, each local region of interest extracted from each sample data is resized (scaled) to the same 224*224 scale. Furthermore, according to the distribution of palm lines, linear convolution kernels of different scales and directions can be set to extract palm line features at different scales and in different directions. As shown in Figure 12, for a 9*9 linear convolution, the center pixel of the convolution kernel is taken as the origin, and linear convolution weights are constructed along the directions of 0 degrees a, 30 degrees b, 60 degrees c, 90 degrees d, 120 degrees e, and 150 degrees f with the horizontal direction. The linear convolution weight width is 1, and the weight of the rest is 0, that is, the weight of the convolution unit filled with lines is 1, and its convolution result is valid, while the weight of the unfilled blank convolution unit is 0, and its convolution result is 0. As shown in Figure 13, for a 16*16 linear convolution, the kernel center pixel is taken as the origin, and linear convolution weights are constructed along the directions of 0 degree a, 30 degree b, 60 degree c, 90 degree d, 120 degree e, and 150 degree f with the horizontal direction. The linear convolution weight width is 4, and the weights of the rest are 0.

[0127] Furthermore, the features of the region of interest are extracted by linear convolution kernels of different scales and directions. The feature extraction results can be expressed as follows:

[0128] Among them, y i,j is the linear feature extracted from the pixel in row i and column j of the input image based on the linear convolution kernel, x i,j represents the pixel value of the i-th row and j-th column of the input image, x i+m,j+n Refers to the pixel corresponding to the mth row and nth column within the range of the linear convolution kernel; m is the width of the linear convolution kernel, and n is the height of the linear convolution kernel; ω m,n is the weight parameter of the linear convolution kernel in the mth row and nth column, b is the bias term, and k is the width of the linear convolution kernel, which is 9 and 16 respectively. In this way, 12 linear convolution results can be obtained for each pixel of the input image, specifically including 2 scales and 6 directions for each scale.

[0129] Furthermore, for each pixel, the 12 linear convolution results are averaged to represent the linear convolution fusion result of the pixel, as shown in the following formula:

[0130] Among them, y i,j is the result of linear convolution fusion, is the linear convolution result of each column.

[0131] Furthermore, the linear convolution results of all pixels are spliced ​​according to the corresponding pixel positions to obtain the orientation map of the entire image, and the orientation map is used as the input of the subsequent feature extraction network. The subsequent feature extraction network can be inception resnet50, and its output is 512-dimensional feature feature_id.

[0132] Furthermore, we use ArcFace to calculate the loss function between the features of palm images of different identities, and on this basis, we add the difference of the pairwise pixel orientation map of negative samples. Specifically, we use two palm images of inconsistent identities as negative samples, calculate the L1 loss function between their orientation maps, and take the sum of the L1 losses of all negative samples as the pairwise loss, as shown in the following formula:

[0133] Where n is the number of negative samples, is the direction map of the i-th negative sample, is the direction map of the j-th negative sample.

[0134] The final loss function loss_final is obtained by adding arcface loss and pairwise loss, as shown in the following formula:

[0135] loss_final=arcface_loss+pairwise_loss

[0136] After the final loss function loss_final is calculated, gradient backpropagation is performed and training is continued until the training is completed to obtain a trained feature extraction network.

[0137] When performing identity recognition, the user's hand image can be collected through the terminal payment device camera; the detection model detects the key points between the three fingers of the user's hand; the palm area of ​​interest is extracted based on the hand image and the key point positions; multi-scale linear convolution is used to extract multi-scale linear features; the multi-scale linear features are averaged and fused; the palm print feature encoding vector is extracted using a feature extraction network; the cosine similarity between the palm print feature encoding vector and the base database feature is calculated. The cosine similarity calculation formula is as follows:

[0138] Among them, vector reg 、vector rec The sample ID with the highest similarity is taken as the final recognition result and returned to the terminal payment device as the identity recognition result.

[0139] In a specific application, the effectiveness of the high-similarity palmprint fine-grained recognition algorithm was verified. Table 1 below shows the verification results of the linear convolution method of this embodiment on the twin dataset.

[0140] Table 1

[0141] To verify the effectiveness of linear convolutional feature extraction, we used high-definition and blurred palmprint images of forty pairs of twins as highly similar palmprint patterns for testing. The results are shown in Table 1. The left and right hands of the same twin pair were used as a sample pair, for a total of 3,600 sample pairs. On the high-definition twin images, the existing ArcFace method misidentified 37 pairs of samples, while the method of this embodiment misidentified none. On the blurred twin images, ArcFace misidentified 46 pairs of samples, while the method of this embodiment misidentified none, demonstrating the effectiveness of this method in recognizing highly similar samples.

[0142] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence 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 can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0143] Based on the same inventive concept, the present application also provides an identity recognition device for implementing the aforementioned identity recognition method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more of the following identity recognition device embodiments can be found in the above-mentioned limitations of the identity recognition method and will not be further elaborated here.

[0144] In one embodiment, as shown in FIG14 , an identity recognition device 1400 is provided, which is applied to a computer device and includes: a body part image acquisition module 1402 , a morphological feature extraction module 1404 , a biological body part feature acquisition module 1406 , and a feature matching module 1408 , wherein:

[0145] The body part image acquisition module 1402 is used to acquire a biological body part image obtained for a target body part of the user to be identified;

[0146] The morphological feature extraction module 1404 is configured to determine a characteristic morphological type of the target part and extract the morphological features of each pixel from the biological part image according to at least one image characteristic pattern that matches the characteristic morphological type. The image characteristic pattern is the pattern resulting from the combination of the distribution positions of the pixels covered by the feature extraction, and the pixels covered by the feature extraction are the pixels in the biological part image targeted during each feature extraction.

[0147] A biological part feature acquisition module 1406 is configured to acquire a biological part feature of the user to be identified based on the morphological features of each pixel; and

[0148] The feature matching module 1408 is used to perform feature matching between the biological part features and the registered part features to obtain a feature matching result, and determine the identity recognition result for the user to be identified based on the feature matching result, wherein the registered part features are the biological part features obtained by performing identity registration on the biological part image corresponding to the target part of the registered user.

[0149] In one embodiment, the target part is a palm, and the image feature pattern is a linear pattern; the pattern feature extraction module 1404 is further configured to extract the pattern features of each pixel of the biological part image according to the linear pattern in at least one direction.

[0150] In one embodiment, the morphological feature extraction module 1404 is further configured to extract directional morphological features corresponding to each pixel of the biological part image and the linear patterns in at least two directions; for each pixel, fuse the directional morphological features corresponding to the pixel and the linear patterns in at least two directions to obtain a directional fusion feature of the pixel; and obtain a morphological feature of each pixel based on the directional fusion feature of each pixel.

[0151] In one embodiment, the morphological feature extraction module 1404 is further configured to determine at least one image feature pattern that matches the feature pattern type at at least one image feature scale; and to extract the morphological features of each pixel from the biological part image according to the at least one image feature pattern at at least one image feature scale.

[0152] In one embodiment, the morphological feature extraction module 1404 is further configured to extract, from the biological part image, scale morphological features corresponding to each pixel and the at least two image feature scales, according to at least one image feature pattern at the at least two image feature scales; fuse, for each pixel, the scale morphological features corresponding to the pixel and the at least two image feature scales to obtain a scale fusion feature for the pixel; and obtain a morphological feature for each pixel based on the scale fusion feature of each pixel.

[0153] In one embodiment, the morphological feature extraction module 1404 is further used to extract the morphological features of each pixel from the biological part image through the convolutional network in the pre-trained feature extraction model; wherein the convolutional network is used to perform feature extraction on the biological part image according to at least one image feature pattern adapted to the feature pattern type; the biological part feature acquisition module 1406 is further used to extract the biological part features of the user to be identified based on the morphological features of each pixel through the part feature extraction network in the feature extraction model.

[0154] In one embodiment, the system further includes a model training module for obtaining a plurality of biological part image samples; extracting sample morphological features of each sample pixel from the biological part image samples through the convolutional network in the feature extraction model to be trained; extracting biological part sample features for each sample pixel through the part feature extraction network in the feature extraction model to be trained; determining the training loss based on the biological part sample features and the sample morphological features; and, based on the training loss, updating the convolutional network and the part feature extraction network in the feature extraction model to be trained respectively and continuing the training until the training is completed to obtain a trained feature extraction model.

[0155] In one embodiment, the model training module is also used to obtain a part feature extraction loss based on the biological part sample features; determine a negative sample pair; the negative sample pair includes biological part image samples carrying different identity labels; obtain a sample pair loss based on the sample morphological features of each biological part image sample in the negative sample pair; and obtain a training loss based on the part feature extraction loss and the sample pair loss.

[0156] In one embodiment, the biological part feature acquisition module 1406 is further configured to concatenate the morphological features of each pixel according to their respective distribution positions in the biological part image to obtain the morphological features of the biological part image; and to perform feature extraction on the morphological features of the biological part image to obtain the biological part features of the user to be identified.

[0157] In one embodiment, the morphological feature extraction module 1404 is further configured to determine a region of interest from the biological part image; determine at least one image feature pattern that matches the feature pattern type; and extract morphological features of each pixel in the region of interest according to the at least one image feature pattern.

[0158] In one embodiment, the target part is a palm; the morphological feature extraction module 1404 is further used to detect the feature points of each finger joint between different thumbs in the palm from the biological part image; determine the focus of interest and area range parameters from the biological part image based on the feature point position of each finger joint feature point and the feature point distance between each finger joint feature point; and determine the area of ​​interest in the biological part image according to the focus of interest and area range parameters.

[0159] In one embodiment, the feature matching module 1408 is further used to obtain the registered part features of each registered user; determine the feature similarities between the biological part features and each registered part feature; and determine the identity recognition result for the user to be identified based on each feature similarity.

[0160] Each module in the above-mentioned identity recognition device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor of the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0161] In one embodiment, a computer device is provided. The computer device may be a server or a terminal, and its internal structure diagram may be shown in FIG15 . The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, the memory, and the 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 computing and control capabilities. 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 environment for the operation of the operating system and 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 implemented.

[0162] Those skilled in the art will understand that the structure shown in FIG15 is merely a block diagram of a portion of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0163] In one embodiment, a computer device is further provided, including a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps in the above-mentioned method embodiments when executing the computer-readable instructions.

[0164] In one embodiment, a computer-readable storage medium is provided, which stores computer-readable instructions. When the computer-readable instructions are executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0165] In one embodiment, a computer program product is provided, comprising computer-readable instructions, which implement the steps in the above-mentioned method embodiments when executed by a processor.

[0166] It should be noted that the 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.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0167] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through computer-readable instructions. The computer-readable instructions can be stored in a non-volatile computer-readable storage medium. When the computer-readable instructions are executed, they can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0168] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, 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, they should be considered to be within the scope of this specification.

[0169] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. An identity recognition method, executed by a computer device, comprising: Acquire a biological part image obtained for a target part of a user to be identified; Determine the characteristic morphological type of the target part, and extract the morphological features of each pixel from the biological part image according to at least one image characteristic morphological type that is compatible with the characteristic morphological type, wherein the image characteristic morphological type is the morphological type obtained by combining the distribution positions of the pixels covered by the feature extraction, and the pixels covered by the feature extraction are the pixels in the biological part image targeted at each feature extraction; Obtaining biological part features of the user to be identified based on the morphological features of each pixel; and The biological part feature is feature matched with the registered part feature to obtain a feature matching result, and the identity recognition result for the user to be identified is determined based on the feature matching result, wherein the registered part feature is a biological part feature obtained by identity registration of a biological part image corresponding to a target part of the registered user.

2. The method according to claim 1, wherein the target part is a palm, and the image feature pattern is a linear pattern; The step of extracting the morphological features of each pixel from the biological part image according to at least one image characteristic morphology adapted to the characteristic morphological type comprises: The shape feature of each pixel of the biological part image is extracted according to the line shape in at least one direction.

3. The method according to claim 1 or 2, wherein the step of extracting the shape features of each pixel of the biological part image according to the linear shape in at least one direction comprises: Extracting directional pattern features corresponding to each pixel and the at least two directional line patterns from the biological part image according to the at least two directional line patterns; For each pixel in the pixels, the directional pattern features corresponding to the pixel and the line patterns in the at least two directions are fused to obtain a directional fusion feature of the pixel; and According to the directional fusion features of each pixel, the morphological features of each pixel are obtained.

4. The method according to any one of claims 1 to 3, wherein the step of extracting the morphological features of each pixel from the biological part image according to at least one image characteristic pattern adapted to the characteristic pattern type comprises: Determining at least one image feature pattern that matches the feature pattern type at at least one image feature scale; and According to the at least one image feature pattern at the at least one image feature scale, the pattern feature of each pixel is extracted from the biological part image.

5. The method according to any one of claims 1 to 4, wherein extracting the morphological features of each pixel from the biological part image according to the at least one image feature morphology at the at least one image feature scale comprises: According to the at least one image feature pattern under at least two image feature scales, extracting scale pattern features corresponding to each pixel and the at least two image feature scales from the biological part image; For each pixel among the pixels, fusing the scale morphological features corresponding to the pixel and the at least two image feature scales to obtain a scale fusion feature of the pixel; and According to the scale fusion features of each pixel, the morphological features of each pixel are obtained.

6. The method according to any one of claims 1 to 5, wherein the step of extracting the morphological features of each pixel from the biological part image according to at least one image characteristic pattern adapted to the characteristic pattern type comprises: Extracting the morphological features of each pixel from the biological part image through a convolutional network in a pre-trained feature extraction model; The convolutional network is used to generate at least one image feature type that matches the feature type. Performing feature extraction on the biological part image; The obtaining of the biological part feature of the user to be identified based on the morphological features of each pixel includes: The biological part features of the user to be identified are extracted by the part feature extraction network in the feature extraction model based on the morphological features of each pixel.

7. According to the method according to any one of claims 1 to 6, the feature extraction model is obtained through a model training step; the model training step comprises: Acquire multiple biological part image samples; Extracting sample morphological features of each sample pixel from the biological part image sample through a convolutional network in the feature extraction model to be trained; Extracting biological part sample features from the sample morphological features of each sample pixel through the part feature extraction network in the feature extraction model to be trained; Determining a training loss based on the biological part sample features and the sample morphology features; and According to the training loss, the convolutional network and the part feature extraction network in the feature extraction model to be trained are updated respectively and then continued to be trained until the training is completed, thereby obtaining the trained feature extraction model.

8. The method according to any one of claims 1 to 7, wherein determining the training loss based on the biological part sample features and the sample morphology features comprises: Obtaining a part feature extraction loss based on the biological part sample feature; Determine a negative sample pair, wherein the negative sample pair includes biological part image samples carrying different identity labels; Obtaining a sample pair loss based on the sample morphological features of each biological part image sample in the negative sample pair; and The training loss is obtained according to the part feature extraction loss and the sample pair loss.

9. The method according to any one of claims 1 to 8, wherein obtaining the biological part feature of the user to be identified based on the morphological features of each pixel comprises: According to the distribution positions of the pixels in the biological part image, the morphological features of the pixels are combined to obtain the morphological features of the biological part image; and Feature extraction is performed on the morphological features of the biological part image to obtain biological part features of the user to be identified.

10. The method according to any one of claims 1 to 9, wherein the step of extracting the morphological features of each pixel from the biological part image according to at least one image characteristic pattern adapted to the characteristic pattern type comprises: determining a region of interest from the biological part image; Determining at least one image feature pattern that matches the feature pattern type; and According to the at least one image feature type, a morphological feature of each pixel in the region of interest is extracted.

11. The method according to any one of claims 1 to 10, wherein the target part is a palm; and determining the region of interest from the biological part image comprises: Detecting the feature points of the finger gaps between different thumbs in the palm from the biological part image; Determine a focus of interest and a region range parameter from the biological part image based on a feature point position of each of the finger joint feature points and a feature point distance between each of the finger joint feature points; and A region of interest is determined in the biological part image according to the focus of interest and the region range parameter.

12. The method according to any one of claims 1 to 11, wherein the feature matching of the biological part feature with the registered part feature to obtain a feature matching result, and determining the identity recognition result for the user to be recognized according to the feature matching result, comprises: Obtain the registered part characteristics of each registered user; respectively determining the feature similarities between the biological part feature and each of the registered part features; and An identity recognition result for the user to be recognized is determined based on the similarities of each of the features.

13. An identity recognition device, applied to a computer device, comprising: A part image acquisition module, used to acquire a biological part image obtained for a target part of a user to be identified; A morphological feature extraction module is used to determine the characteristic morphological type of the target part, and extract the morphological features of each pixel from the biological part image according to at least one image characteristic pattern adapted to the characteristic morphological type, wherein the image characteristic pattern is the pattern obtained by combining the distribution positions of the pixels covered by the feature extraction, and the pixels covered by the feature extraction are the pixels in the biological part image targeted at each feature extraction; A biological part feature acquisition module, used to obtain the biological part feature of the user to be identified based on the morphological features of each pixel; and The feature matching module is used to perform feature matching between the biological part feature and the registered part feature to obtain a feature matching result, and determine the identity recognition result for the user to be identified based on the feature matching result, wherein the registered part feature is the biological part feature obtained by performing identity registration on the biological part image corresponding to the target part of the registered user.

14. A computer device, comprising a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the method according to any one of claims 1 to 12 when executing the computer-readable instructions.

15. A computer-readable storage medium having computer-readable instructions stored thereon, wherein the computer-readable instructions, when executed by a processor, implement the steps of the method according to any one of claims 1 to 12.

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

Images