Image processing method and apparatus, computer device, computer-readable storage medium, and computer program product

Abstract

The present application discloses an image processing method and apparatus, a computer device, a computer-readable storage medium, and a computer program product. The method comprises: acquiring a first image to be recognized and a first embedding feature of the first image, the first image being a palmprint image; recognizing a target image type to which the first image belongs; acquiring an image feature set associated with the target image type, the image feature set comprising image embedding features of target images of a plurality of objects, the target images being palmprint images belonging to the target image type, and the objects in an object set having respective object identity information; respectively acquiring feature matching degrees between the first embedding feature and the image embedding features; and on the basis of the feature matching degrees between the first embedding feature and the image embedding features and the object identity information of the objects, determining identity recognition information for the object to which the first image belongs.

Description

Image processing method and device, computer device, computer readable storage medium and computer program product

[0001] Cross-reference to related applications

[0002] The present application is based on the Chinese patent application No. 2024109478671, filed on July 15, 2024, and claims priority to the above Chinese patent application, the entire contents of which are incorporated herein by reference. TECHNICAL FIELD

[0003] The present application relates to the technical field of image processing, and in particular to an image processing method and device, a computer device, a computer readable storage medium and a computer program product. BACKGROUND

[0004] In the field of identifying a user based on a palmprint image, a most standard and perfect palmprint image of the user can be pre-acquired and stored as a base image for the user, and subsequent identity recognition of the user can be performed based on the stored base image. However, in the specific process of identifying the user, the real-time acquired palmprint image can have great randomness, and the real-time acquired palmprint image can be different in posture or style compared with the pre-stored base image, which not only causes a large amount of invalid comparison and low recognition efficiency, but also causes the inconsistency between the real-time acquired palmprint image and the pre-stored base image, and thus cannot accurately identify the identity of the user. SUMMARY

[0005] The embodiments of the present application provide an image processing method and device, a computer device, a computer readable storage medium and a computer program product, which can improve the accuracy of identifying an object based on a palmprint image.

[0006] The embodiments of the present application provide an image processing method applied to a computer device, and the method comprises:

[0007] Obtaining a first image to be identified and a first embedding feature of the first image, the first image being a palmprint image;

[0008] Identifying a target image type to which the first image belongs;

[0009] Obtaining an image feature set associated with the target image type, the image feature set comprising image embedding features of target images of a plurality of objects, the target images being palmprint images belonging to the target image type, each object having its own object identity information;

[0010] Determining a feature matching degree between the first embedding feature and each image embedding feature;

[0011] determine the identity recognition information of the object to which the first image belongs based on the feature matching degree between the first embedding feature and each image embedding feature and the object identity information of each object.

[0012] Embodiments of the present application provide an image processing device, which comprises:

[0013] The first obtaining module is configured to obtain a first image to be recognized and a first embedding feature of the first image, the first image being a palmprint image.

[0014] The recognition module is configured to recognize a target image type to which the first image belongs.

[0015] The second obtaining module is configured to obtain a set of image features associated with the target image type, the set of image features comprising image embedding features of target image images of a plurality of objects, the target image being a palmprint image belonging to the target image type, each object having its own object identity information.

[0016] The third obtaining module is configured to obtain a feature matching degree between the first embedding feature and each image embedding feature.

[0017] The determination module is configured to determine the identity recognition information of the object to which the first image belongs based on the feature matching degree between the first embedding feature and each image embedding feature and the object identity information of each object.

[0018] Embodiments of the present application provide a computer device, which comprises a memory and a processor, the memory storing a computer program, the computer program being executed by the processor to make the processor execute the image processing method provided in the embodiments of the present application.

[0019] Embodiments of the present application provide a computer readable storage medium, which stores a computer program, the computer program being executed by a processor to make the processor execute the image processing method.

[0020] Embodiments of the present application provide a computer program product, which comprises a computer program stored in a computer readable storage medium. A processor of a computer device reads the computer program from the computer readable storage medium, and the processor executes the computer program to make the computer device execute the image processing method provided in the embodiments of the present application.

[0021] The embodiment of the application can obtain a first image to be identified and a first embedding feature of the first image, wherein the first image is a palm print image, a target image type to which the first image belongs is identified from a plurality of image types set, an image feature set associated with the target image type is obtained, the image feature set includes image embedding features of target images of a plurality of objects, the target images are palm print images belonging to the target image type, each object has its own object identity information, a feature matching degree between the first embedding feature and each image embedding feature is obtained, and identity recognition information of an object to which the first image belongs is determined based on the feature matching degree between the first embedding feature and each image embedding feature and the object identity information of each object. As can be seen, the method proposed in the embodiment of the application can identify a specific target image type to which the first image belongs, so that the identity of the object to which the first image belongs can be identified by using the specific image feature set associated with the specific target image type and the embedding feature (such as the first embedding feature) of the first image. Since the specific type of the image is considered for identification within a specific range (such as the image feature set associated with the target image type), the interference of cross-type matching can be reduced, the accuracy of the same-type feature comparison can be improved, and thus the identity recognition of the object to which the image belongs is realized, so that the accuracy of the identity recognition of the object by the palm print image is improved. Moreover, dynamically loading the image feature set associated with the target image type can reduce a large amount of invalid feature comparison, thereby significantly improving the recognition efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0023] FIG. 1 is a structural schematic diagram of a network architecture of image recognition provided by an embodiment of the application;

[0024] FIG. 2 is a scene schematic diagram of image recognition provided by an embodiment of the application;

[0025] FIG. 3 is a flow schematic diagram of an image processing method provided by an embodiment of the application;

[0026] FIG. 4 is a scene schematic diagram of image recognition by referring to a feature set provided by an embodiment of the application;

[0027] FIG. 5 is a flow schematic diagram of identity recognition of an object by object identification information provided by an embodiment of the application;

[0028] FIG. 6 is a schematic diagram of an interface for inputting object identification information according to an embodiment of the present application;

[0029] FIG. 7 is a schematic diagram of a scenario for training a quality detection network according to an embodiment of the present application;

[0030] FIG. 8 is a schematic diagram of a process for identity recognition of an object according to an embodiment of the present application;

[0031] FIG. 9 is a schematic diagram of a structure of an image processing apparatus according to an embodiment of the present application;

[0032] FIG. 10 is a schematic diagram of a structure of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the scope of protection of the embodiments of the present application.

[0034] First, it needs to be noted that all the data (such as images, object identification information, object identity information, and embedded features of images, and related data) collected by the embodiments of the present application are collected with the consent and authorization of the owner of the data (such as a user, an institution, or an enterprise), and the collection, use, and processing of the related data need to comply with the relevant laws, regulations, and standards in the relevant region.

[0035] Here, the related technical concepts involved in the embodiments of the present application are described:

[0036] 1) Color image: a color image collected by a color sensor (Sensor) from natural light imaging, which can be used for palm optimization and contrast recognition in palm payment.

[0037] 2) Infrared image: an infrared image collected by an infrared sensor from pan-infrared light imaging, which can be used for living body detection in palm payment.

[0038] 3) Support vector machine (SVM): its basic model is a linear classifier with the largest interval defined in the feature space.

[0039] 4) Palmprint image: a digital representation of the human palm region (including biological features such as palm skin texture, main lines, minutiae, and wrinkles) obtained by an image acquisition device, which is an important data form in the field of biometric recognition. The palmprint image has uniqueness, stability, and collectability.

[0040] 5) Image embedding feature, which is a numerical representation of an original image mapped to a low-dimensional vector space by a deep neural network or a feature extraction algorithm. The image embedding feature can be understood as a feature vector. This feature vector can compress and retain the key discriminative information of the image (such as the line structure of the palm print and the distribution of the minutia points), and can also eliminate redundancy (such as removing interference such as illumination and background).

[0041] 6) Association storage, which is a process of binding and persistently storing two or more data items (such as "reference embedding features" and "identity information") in a certain logical relationship. The core goal is to establish a searchable association between data for subsequent query, analysis and application.

[0042] Please refer to FIG. 1, which is a structural schematic diagram of a network architecture for image recognition provided by an embodiment of the present application. As shown in FIG. 1, the network architecture can include a palm print image collection device 100 and a server 200. The palm print image collection device 100 can be used to collect the palm print image of a user and report it to the server 200, and the server 200 can be used to identify and process the palm print image reported by the palm print image collection device 100. Therefore, the palm print image collection device 100 and the server 200 can be connected by a network, so that the palm print image collection device 100 and the server 200 can interact with each other through the network connection.

[0043] The server 200 shown in FIG. 1 can be a standalone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (content distribution network), and basic cloud computing services such as big data and artificial intelligence platforms. The palm print image collection device 100 can be any type of terminal device.

[0044] Please refer to FIG. 2, which is a scene schematic diagram of image recognition provided by an embodiment of the present application. As shown in FIG. 2, the palm print image collected and reported by the palm print image collection device 100 to the server 200 can be a first image. The server 200 can perform feature extraction processing on the first image to obtain a first embedding feature of the first image, and can also perform type identification on the first image to identify the target image type to which the first image belongs from a plurality of image types set.

[0045] Further, the server 200 can also obtain an image feature set associated with the target image type, which can include image embedding features of images of each object in the object set in the target image type. The objects in the object set can be users whose object identities and image associations are registered at the server 200, and each object in the object set has registered (also bound) object identity information.

[0046] The server 200 can calculate the feature matching degree (which can be feature similarity) between the first embedding feature of the first image and each image embedding feature in the image feature set, so that the server 200 can determine the identity recognition information of the object to which the first image belongs by the feature matching degree between the first embedding feature and each image embedding feature in the image feature set, and the object identity information of each object in the object set. The identity recognition information is the identity information recognized for the object to which the first image belongs. The specific process of identity recognition can also be referred to the related description in the embodiments below.

[0047] By using the method provided in the embodiments of the present application, since the randomness of the user when shooting the palmprint image (such as palm tilt behavior, palm posture action, palm cleanliness, clothing shielding state, etc.), the palmprint image collected by the user also has randomness, therefore, the embodiments of the present application can provide a plurality of image types set according to these random factors, so that when the object is identified by the image, the image type to which the currently collected palmprint image (such as the first image) belongs is first identified, and then the palmprint image is identified according to the identified specific image type, which greatly improves the accuracy of identifying the palmprint image of the user.

[0048] Please refer to FIG. 3, which is a flowchart of an image processing method provided by an embodiment of the present application. The execution subject in the embodiment of the present application can be a computer device, which can be a computer device or a computer device cluster composed of multiple computer devices. The computer device can be a server or other types of devices, and the present application does not limit this. As shown in FIG. 3, the method can include:

[0049] In step S101, a first image to be identified and a first embedding feature of the first image are obtained.

[0050] In some embodiments, the computer device can obtain a first image to be identified, which can be a palmprint image collected in real time for any object (such as a user), and the object identity of the object to which the first image belongs can be identified by the first image in the embodiments of the present application.

[0051] In some embodiments, the process of obtaining the first image to be recognized can include:

[0052] The computer device can obtain a second image collected by a palmprint image collection device. The palmprint image collection device can send the collected second image to the computer device. For example, the second image can be an image of a palm collected by the palmprint image collection device through a configured camera and infrared.

[0053] The computer device can perform feature extraction processing on the second image to generate image embedding features of the second image, which can be referred to as second embedding features. For example, the computer device can use a trained first feature extraction network (which can be a deep learning network) to perform feature extraction processing on the second image to generate the second embedding features of the second image.

[0054] The computer device can also obtain a reference feature set, which can include reference embedding features of reference images of objects in the object set in a reference posture. The reference images are standard and of good quality, such as palmprint images without dirt, without occlusion, and with a posture that is substantially parallel to the palmprint image collection device without a large angle.

[0055] The object set can be a user set, and the objects in the object set can include all users who have performed palmprint image registration. For users who have performed palmprint image registration, there are reference embedding features of reference images stored for the users. The reference embedding features of the reference images of the users can be understood as the registered palmprint images of the users. It should be noted that, in order to protect the security and privacy of the palmprint images of the users, the computer device generally does not directly store the palmprint images collected for the users when the users perform palmprint image registration, but stores image embedding features of the palmprint images as the base images of the users.

[0056] Each object in the object set has its own registered object identity information, which is information used to reflect the identity of the object. For example, the object identity information of an object can include the name, gender, account information, and identity number of the object.

[0057] The computer device can obtain a feature matching degree between the second embedding feature of the second image and each reference embedding feature in the reference feature set, which can reflect a feature similarity between the second embedding feature and the reference embedding feature in the reference feature set. There can be one feature matching degree between the second embedding feature and one reference embedding feature in the reference feature set. The greater the feature matching degree, the more similar the second embedding feature and the reference embedding feature in the reference feature set. Conversely, the smaller the feature matching degree, the less similar the second embedding feature and the reference embedding feature in the reference feature set.

[0058] For example, the image embedding feature of the palmprint image can be a feature vector. The computer device can calculate a cosine similarity between the second embedding feature and each reference embedding feature in the reference feature set as the feature matching degree between the second embedding feature and the reference embedding feature in the reference feature set.

[0059] In some embodiments, the Euclidean distance between the second embedding feature and each reference embedding feature can be determined as the feature matching degree between the second embedding feature and the reference embedding feature. The Manhattan distance between the second embedding feature and each reference embedding feature can also be determined as the feature matching degree between the second embedding feature and the reference embedding feature. The Hamming distance between the second embedding feature and each reference embedding feature can also be determined as the feature matching degree between the second embedding feature and the reference embedding feature. The determination manner of the feature matching degree is not limited in the embodiments of the present application.

[0060] Further, the computer device can evaluate the feature matching degree between the second embedding feature and each reference embedding feature. If the feature matching degree between the second embedding feature and the reference embedding feature is less than the first matching degree threshold, it can be indicated that the identity recognition of the object to which the second image belongs by the reference feature set fails, that is, there is no reference embedding feature in the reference feature set that matches the second embedding feature. At this time, the second image can be regarded as the first image described above, and the recognition process described in the embodiments of the present application is performed on the first image again, that is, the recognition process described in the embodiments of the present application is further performed on the second image.

[0061] The first matching degree threshold described above can be set according to an actual application scenario. The first matching degree threshold can be the minimum feature matching degree used to evaluate the matching between the second embedding feature and the reference embedding feature in the reference feature set.

[0062] If there is a reference embedding feature in the reference feature set that has a feature matching degree with the second embedding feature greater than or equal to the first matching degree threshold, the computer device can filter, from the reference feature set, the reference embedding feature that has a feature matching degree with the second embedding feature greater than or equal to the first matching degree threshold, and all of the reference embedding features are first candidate embedding features. At this time, there can be one or more first candidate embedding features.

[0063] Therefore, the computer device can take the object to which the first candidate embedding feature with the largest feature matching degree with the second embedding feature belongs as a first recognition object, and the first recognition object is the object recognized by the second image on the object to which the second image belongs. The first candidate embedding feature with the largest feature matching degree with the second embedding feature can be understood as a reference embedding feature in the reference feature set that matches the second embedding feature. At this time, the second embedding feature and the first candidate embedding feature with the largest feature matching degree with the second embedding feature can be considered as image embedding features of palm print images of the same object (such as the same user).

[0064] Therefore, the computer device can take the object identity information of the first recognition object as identity recognition information of the object to which the second image belongs. In this case, the identity of the object to which the second image belongs is successfully recognized by the reference feature set, and there is no need to further perform secondary recognition on the second image, that is, there is no need to further take the second image as the first image to perform the recognition process described in the embodiments of the present application, which can reduce the recognition steps and improve the recognition efficiency, thereby improving the business execution efficiency.

[0065] As can be seen from the above, the first image and the second image are the same image, and when the object identity of the object to which the second image belongs is not successfully recognized by the reference feature set, the second image can be taken as the first image to further accurately perform secondary recognition on the first image.

[0066] Referring to FIG. 4, FIG. 4 is a scene diagram of image recognition by a reference feature set according to an embodiment of the present application. As shown in FIG. 4, here, it is assumed that the object set can include six objects, object 1, object 2, object 3, object 4, object 5, and object 6. Therefore, the reference feature set can include reference embedding feature 1 of a reference image of object 1, reference embedding feature 2 of a reference image of object 2, reference embedding feature 3 of a reference image of object 3, reference embedding feature 4 of a reference image of object 4, reference embedding feature 5 of a reference image of object 5, and reference embedding feature 6 of a reference image of object 6.

[0067] The computer device can obtain a feature matching degree 1 between the first embedding feature and the reference embedding feature 1, a feature matching degree 2 between the first embedding feature and the reference embedding feature 2, a feature matching degree 3 between the first embedding feature and the reference embedding feature 3, a feature matching degree 4 between the first embedding feature and the reference embedding feature 4, a feature matching degree 5 between the first embedding feature and the reference embedding feature 5, and a feature matching degree 6 between the first embedding feature and the reference embedding feature 6.

[0068] If only the feature matching degree 1 and the feature matching degree 3 among the feature matching degree 1, the feature matching degree 2, the feature matching degree 3, the feature matching degree 4, the feature matching degree 5, and the feature matching degree 6 are greater than or equal to the first matching degree threshold, the reference embedding feature 1 and the reference embedding feature 3 can be taken as the first candidate embedding features.

[0069] Further, if the feature matching degree 3 is greater than the feature matching degree 1, the object identity information of the object 3 to which the reference embedding feature 3 belongs can be directly taken as the identity recognition information of the object to which the first image belongs.

[0070] If the feature matching degree 1, the feature matching degree 2, the feature matching degree 3, the feature matching degree 4, the feature matching degree 5, and the feature matching degree 6 are all less than the first matching degree threshold, the second image can be further taken as the first image for image type recognition and then for identity recognition.

[0071] The computer device can obtain the first embedding feature of the first image. In some embodiments, the first embedding feature can be the second embedding feature; or, different image embedding features of the palmprint image can be used in different image recognition stages, and therefore, the computer device can perform feature extraction processing on the first image by using the trained second feature extraction network (which can also be a deep learning network) to generate the first embedding feature of the first image. The trained second feature extraction network and the trained first feature extraction network can be different feature extraction networks, and in this case, the first embedding feature and the second embedding feature are different.

[0072] In this way, different image embedding features of the same palmprint image are generated by using different feature extraction networks in different palmprint image recognition stages (such as the stage of performing palmprint image recognition by using the reference feature set, and the stage of performing palmprint image recognition by using the image type in the embodiments of the present application), and the different image embedding features of the same palmprint image are used for recognition processing in different image recognition stages, which can enrich the style and dimension of the image embedding features of the same palmprint image, and thus more accurate and comprehensive recognition processing of the same palmprint image can be achieved.

[0073] Step S102, identify the target image type to which the first image belongs.

[0074] In some embodiments, the computer device can obtain a trained multi-classification network, which can be a trained network capable of identifying (i.e., classifying) images of a set of multiple image types. The trained multi-classification network can be obtained through supervised training using sample images of the set of multiple image types (sample images with image type labels). Exemplarily, the trained multi-classification network can be an SVM network (a generalized linear classifier using supervised learning).

[0075] Therefore, the computer device can identify the image type (also referred to as type label) to which the first image belongs through the trained multi-classification network, and the identified image type of the first image can be referred to as the target image type. In some embodiments, the computer device can input the first image into the trained multi-classification network, so that the trained multi-classification network can identify the target image type to which the first image belongs from the set of multiple image types.

[0076] Alternatively, the multiple image types can include a first image type, which can be a type in which the included angle between the image plane of the first image (which can be a plane formed by three points on the first image) and the coordinate system of the palmprint image acquisition device is greater than a certain included angle threshold. For the first image type, a specific identification method can be used, including:

[0077] The computer device can obtain a target included angle between the image plane of the first image (which can be a plane formed by three points on the first image) and the coordinate system of the palmprint image acquisition device. In some embodiments, the target included angle can be detected by the palmprint image acquisition device that acquires the first image and sent to the computer device. The position of the camera component of the palmprint image acquisition device can be the origin position of the coordinate system.

[0078] If the target included angle is greater than the set included angle threshold, the computer device can consider that the target image type to which the first image belongs is the large-angle image type. In some embodiments, the target included angle can include a pitch angle (lateral inclination angle) and a roll angle (roll angle), both of which are angles used to describe the included angle between a plane and a coordinate system in space. The pitch angle and the roll angle can each have a corresponding included angle threshold. If the pitch angle and the roll angle are each greater than the corresponding included angle threshold, it can be considered that the target included angle is greater than the set included angle threshold.

[0079] Or, in this case, the trained multi-classification network can also be trained by sample images of other image types (sample images with image type labels) in the above-mentioned multiple image types except the first image type, so that the trained multi-classification network can be used to identify (i.e., classify) the images of other image types in the set multiple image types except the first image type.

[0080] Therefore, the computer device can identify the image type to which the first image belongs by the above-mentioned target angle and the trained multi-classification network, and at this time, the first image type and other image types (i.e., image types in the above-mentioned multiple image types except the first image type) can be respectively set with a judgment priority, so that when it is detected that the target angle is greater than the angle threshold and the trained multi-classification network identifies any of the other image types, the target image type to which the first image belongs can be finally determined by the judgment priority.

[0081] For example, when it is detected that the target angle is greater than the angle threshold and the trained multi-classification network also identifies any of the other image types, if the judgment priority of the first image type is higher than that of the other image types, the first image type can be taken as the target image type finally identified for the first image.

[0082] For another example, when it is detected that the target angle is greater than the angle threshold and the trained multi-classification network also identifies any of the other image types, if the judgment priority of the first image type is lower than that of the other image types, the image type identified by the trained multi-classification network can be taken as the target image type finally identified for the first image.

[0083] Through the above process, the target image type to which the first image belongs can be identified from the set multiple image types. Or, in actual application scenarios, the image type of an image can also be identified by other identification methods, which are not limited by the present application.

[0084] For example, the plurality of image types in the above setting can include a second image type (characterizing that the photographed palm is dirty), a third image type (characterizing that the photographed palm is partially blocked), the above first image type, a fourth image type (characterizing that the light is weak when the image is photographed), a fifth image type (characterizing that the photographed image is overexposed), a seventh image type (characterizing that the photographed image is distorted), and an eighth image type (in the embodiment of the present application, the photographed image includes not only a palm print but also a vein blood vessel image photographed by infrared rays, and the eighth image type characterizes that the vein blood vessel is weak), and the like. Similarly, in actual application scenarios, more image types can also be set, and the embodiment of the present application does not limit this. The palm print image can be a color image, and the vein blood vessel image can be an infrared image.

[0085] In step S103, an image feature set associated with the target image type is obtained; the image feature set includes image embedding features of target images of a plurality of objects, and the target images are palm print images belonging to the target image type, and each object has its own object identity information.

[0086] In some embodiments, the computer device can obtain an image feature set associated with the target image type, and the image feature set includes image embedding features of target images of a plurality of objects in the above object set, and the target images are palm print images belonging to the target image type, in other words, the image feature set can include image embedding features of palm print images of the target image type collected for each object in the object set before.

[0087] It can be understood that the target image type can be any one of the plurality of image types described above, each of the plurality of image types can have its own associated image feature set, and the image feature set associated with one image type can include image embedding features of images of a plurality of objects in this image type. In some embodiments, the image feature set associated with each image type can be obtained by performing feature extraction on the images of the objects in the object set in each image type through the above trained second feature extraction network.

[0088] In step S104, a feature matching degree between the first embedding feature and each image embedding feature is obtained.

[0089] In some embodiments, the computer device can obtain a feature matching degree between the first embedding feature of the first image and each image embedding feature in the image feature set associated with the target image type, and there can be a feature matching degree between the first embedding feature and one image embedding feature in the image feature set. In some embodiments, the feature matching degree between the first embedding feature and the image embedding feature in the image feature set associated with the target image type can also be the cosine similarity between the first embedding feature and the image embedding feature in the image feature set.

[0090] Similarly, the feature matching degree between the first embedding feature and the image embedding feature in the image feature set associated with the target image type can be used to reflect the feature similarity between the first embedding feature and the image embedding feature in the image feature set. The greater the feature matching degree between the first embedding feature and the image embedding feature in the image feature set associated with the target image type, the more similar the first embedding feature and the image embedding feature in the image feature set, and vice versa. The smaller the feature matching degree between the first embedding feature and the image embedding feature in the image feature set associated with the target image type, the less similar the first embedding feature and the image embedding feature in the image feature set.

[0091] In step S105, based on the feature matching degree between the first embedding feature and each image embedding feature and the object identity information of each object, the identity recognition information of the object to which the first image belongs is determined.

[0092] In some embodiments, if there is an image embedding feature in the image feature set associated with the target image type, the feature matching degree between which and the first embedding feature is greater than or equal to the second matching degree threshold, the computer device can filter out the image embedding feature from the image feature set associated with the target image type, the feature matching degree between which and the first embedding feature is greater than or equal to the second matching degree threshold, as the second candidate embedding feature. At this time, there can be one or more second candidate embedding features. The second matching degree threshold can be set according to the actual application scenario, such as the second matching degree threshold can be equal to the first matching degree threshold described above. The second matching degree threshold can be the minimum feature matching degree for evaluating whether the first embedding feature and the image embedding feature in the image feature set associated with the target image type are matched.

[0093] Therefore, the computer device can take the object to which the second candidate embedding feature with the largest feature matching degree with the first embedding feature belongs as the second recognition object, which is the object recognized by the first image through the first image. Wherein, the second candidate embedding feature with the largest feature matching degree with the first embedding feature can be understood as the image embedding feature in the image feature set associated with the target image type that matches the first embedding feature. At this time, it can be considered that the first embedding feature and the second candidate embedding feature with the largest feature matching degree with the first embedding feature are image embedding features of images of the same object (such as the same user).

[0094] Therefore, the computer device can take the object identity information of the second recognition object as the identity recognition information of the object to which the first image belongs. In this case, the image feature set associated with the target image type successfully identifies the identity of the object to which the first image belongs, and there is no need to further identify the first image three times, such as there is no need to ask the user (such as the object to which the first image belongs) to input the object identification information in the palmprint image collection device, and then further identify the object identity of the object to which the first image belongs through the input object identification information. The specific process of further identifying the object to which the first image belongs through the object identification information input by the user can be referred to the description of the corresponding embodiment of FIG. 5 below.

[0095] If the feature matching degree between the first embedding feature and each image embedding feature in the image feature set associated with the target image type is less than the second matching degree threshold, it can be indicated that the identity of the object to which the first image belongs is not successfully identified through the image feature set associated with the target image type, that is, there is no image embedding feature in the image feature set associated with the target image type that matches the first embedding feature. At this time, the user needs to input the registered object identification information to further identify the identity of the object to which the first image belongs through the input object identification information (1v1 identification, that is, 1-to-1 identification). The specific identification process can be referred to the related description in the corresponding embodiment of FIG. 5 below.

[0096] The method provided by the embodiment of the present application can perform secondary identification on the palmprint image of the user under the corresponding image type in the case of image difference caused by specific reasons (such as different behaviors or gestures of shooting palmprint images), effectively improve the problem of palm recognition failure caused by specific reasons, finally improve the success rate of identifying the palm of the user, and reduce the experience damage caused by the user inputting the mobile phone number (that is, the object identification information) for verification, effectively improve the user experience.

[0097] The embodiment of the application can obtain a first image to be identified and a first embedding feature of the first image; identify a target image type to which the first image belongs from a plurality of image types set; obtain an image feature set associated with the target image type; the image embedding features of the target image of the image feature set contain a plurality of objects, the target image is a palmprint image belonging to the target image type, and each object in the object set has its own object identity information; obtain a feature matching degree between the first embedding feature and each image embedding feature in the image feature set; and determine the identity recognition information of the object to which the first image belongs based on the feature matching degree between the first embedding feature and each image embedding feature in the image feature set and the object identity information of each object. As can be seen, the method proposed in the embodiment of the application can first identify the specific target image type to which the first image belongs, so that the identity recognition of the object to which the first image belongs can be realized through the specific image feature set associated with the specific target image type and the embedding feature (such as the first embedding feature) of the first image. Since the specific type of the image is considered for the identification within a specific range (such as the image feature set associated with the target image type), the interference of cross-type matching can be reduced, the accuracy of the same-type feature comparison can be improved, and thus the identity recognition of the object to which the image belongs is realized, so that the accuracy of the identity recognition of the object through the palmprint image is improved. Moreover, dynamically loading the image feature set associated with the target image type can reduce a large amount of invalid feature comparison, thereby significantly improving the recognition efficiency.

[0098] Please refer to FIG. 5, which is a flowchart of the identity recognition of an object through object identification information according to an embodiment of the application. The execution subject in this embodiment can also be the computer device described above. As shown in FIG. 5, the flowchart can include the following steps:

[0099] In step S201, if the feature matching degrees between the first embedding feature and each image embedding feature are all less than the second matching degree threshold, the target object identification information of the object to which the first image belongs is obtained.

[0100] In some embodiments, if the feature matching degrees between the first embedding feature and each image embedding feature in the image feature set associated with the target image type are all less than the second matching degree threshold, the computer device can send prompt information to the palmprint image collection device that the object identification information needs to be input, so that the palmprint image collection device can display an input interface for the user to input the object identification information registered by the user, and the palmprint image collection device can obtain the object identification information input by the user (such as the object to which the first image belongs) in the input interface. For example, the object identification information can be the communication number (such as the mobile phone number) of the user.

[0101] The palmprint image acquisition device can send the obtained object identification information to the computer device, so that the computer device can also receive the object identification information input by the user for the object to which the first image belongs, and the obtained object identification information can be referred to as target object identification information.

[0102] Please refer to FIG. 6, which is an interface diagram provided by an embodiment of the present application for inputting object identification information. As shown in FIG. 6, assuming that the object identification information of the object is a registered (i.e., bound) mobile phone number of the object, the palmprint image acquisition device can have a display screen with touch points. When the user needs to input the object identification information, the palmprint image acquisition device can display interface 1a on the display screen, which is an interface for inputting a mobile phone number. The user can input his own mobile phone number in interface 1a. After the input is completed, the display screen of the palmprint image acquisition device can display interface 2a, which can prompt the user that his identity is being identified by the inputted mobile phone number (the identification process can be performed by the computer device). When the subsequent identity identification of the user by the computer device is successful, the computer device can also return a prompt message of successful identification to the palmprint image acquisition device. The palmprint image acquisition device can output the prompt message of successful identification to prompt the user that his identity identification is successful, or directly prompt the user that the relevant service (e.g., a target service described below) is executed successfully.

[0103] In step S202, feature extraction processing is performed on the first image to generate a third embedding feature of the first image.

[0104] In some embodiments, if the first embedding feature and the second embedding feature are the same, the third embedding feature can also be the same as the first embedding feature and the second embedding feature, i.e., the first embedding feature, the second embedding feature, and the third embedding feature can be the same embedding feature, which can be the second embedding feature generated first. In this way, the workload of obtaining the first embedding feature, the second embedding feature, and the third embedding feature can be saved.

[0105] If the first embedding feature and the second embedding feature are different, the third embedding feature can also be different from the first embedding feature and the second embedding feature. For example, the computer device can call a trained third feature extraction network to perform feature extraction processing on the first image to generate the third embedding feature of the first image. The trained third feature extraction network, the trained first feature extraction network, and the trained second feature extraction network can be different feature extraction networks. For example, the trained first feature extraction network, the trained second feature extraction network, and the trained third feature extraction network can be trained by using different sample images and different training methods, or / and the trained first feature extraction network, the trained second feature extraction network, and the trained third feature extraction network can have different network structures (for example, different types of models).

[0106] In this case, each different image recognition stage (for example, the stage of image recognition by referring to the feature set, the stage of image recognition by image type, and the stage of image recognition by object identification information) uses different image embedding features of the same image (that is, the second image) to perform image recognition, which enriches the style and dimension of the image embedding features of the same image, thereby achieving more accurate and comprehensive recognition of the same image, and further improving the success rate and accuracy of object identity recognition.

[0107] In step S203, identity recognition information of the object to which the first image belongs is determined based on the target object identification information and the third embedding feature.

[0108] In some embodiments, each object in the object set can have an image embedding feature of a reference image in a reference posture of the object, and the reference feature set includes the image embedding features of the reference images of the objects. The image embedding feature of the reference image of each object can be stored in association with the object identification information of the object (for example, stored in the background, such as stored in the computer device or a database accessible by the computer device), and the association storage can be understood as that the image embedding feature of the reference image of an object and the object identification information registered by the object have a mapping relationship.

[0109] Therefore, the computer device can obtain the reference embedding feature in the reference feature set of the reference image of the plurality of objects and associated with the target object identification information, and the reference embedding feature obtained and associated with the target object identification information can be referred to as a target embedding feature. The target embedding feature is the reference embedding feature of the reference image of the user to which the input target object identification information belongs.

[0110] Therefore, the computer device can obtain the feature matching degree between the target embedding feature and the third embedding feature. For example, the feature matching degree between the target embedding feature and the third embedding feature can also be the cosine similarity between the target embedding feature and the third embedding feature.

[0111] If the feature matching degree between the target embedding feature and the third embedding feature is greater than or equal to the third matching degree threshold, it indicates that the target embedding feature and the third embedding feature are matched, and the computer device can take the object identity information of the object to which the target embedding feature belongs as the identity recognition information of the object to which the first image belongs. In this way, the object identity of the object to which the first image belongs is successfully recognized. The third matching degree threshold can be the minimum feature matching degree for evaluating the matching between the target embedding feature and the third embedding feature, and the third matching degree threshold can be set according to an actual application scenario.

[0112] In this process, the requirement for identifying the user's identity by using the third embedding feature can be reduced, such as appropriately reducing the third matching degree threshold for evaluating the matching between the target embedding feature and the third embedding feature. Therefore, it can be understood that the third matching degree threshold can be less than the first matching degree threshold and less than the second matching degree threshold.

[0113] If the feature matching degree between the target embedding feature and the third embedding feature is less than the third matching degree threshold, it indicates that the target embedding feature and the third embedding feature are not matched, and at this time, the object identity recognition of the object to which the first image belongs fails. The computer device generates recognition failure information, and in some embodiments, the computer device can take the recognition failure information as the identity recognition information of the object to which the first image belongs.

[0114] The computer device can return the recognition failure information to the palmprint image acquisition device, so that the palmprint image acquisition device can output prompt information for re-acquiring an image when receiving the recognition failure information, to prompt the user (such as the object to which the first image belongs) to re-enter an image, and then re-identify the identity by using the re-entered image, that is, the currently entered second image does not meet the specification.

[0115] Alternatively, in some embodiments, the user can also only input part of the target object identification information (e.g., only the last 4 digits of the target object identification information of the user) when inputting the object identification information, so that the computer device can filter out the reference embedding features of the reference images of the objects whose object identification information at the corresponding position (e.g., the last 4 digits) is the same as the part of the object identification information input by the user, and the filtered reference embedding features can constitute a reference feature subset, so that the user can be finally identified through the reference feature subset. In this process, a matching degree threshold (e.g., a fourth matching degree threshold) can also be set to evaluate the matching between the third embedding feature and the reference embedding features in the reference feature subset, and the fourth matching degree threshold is also the minimum feature matching degree for evaluating the matching between the third embedding feature and the reference embedding features in the reference feature subset.

[0116] The principle of identifying the object identity of the object to which the first image belongs through the third embedding feature, the fourth matching degree threshold, and the reference feature subset is the same as the principle of identifying the object identity of the object to which the first image belongs through the first embedding feature, the second matching degree threshold, and the image feature set associated with the target image type. The only difference is that the identification of the object identity of the object to which the first image belongs through the third embedding feature, the fourth matching degree threshold, and the reference feature subset is performed under the condition that the part of the object identification information input by the user has been matched with the part of the object identification information registered by the user, so the fourth matching degree threshold can be smaller than the second matching degree threshold. However, since only part of the object identification information is input, the fourth matching degree threshold can be greater than the third matching degree threshold when the object identity is identified by inputting the complete target object identification information.

[0117] In some embodiments, if the feature matching degree between the target embedding feature and the third embedding feature is greater than or equal to the third matching degree threshold, it indicates that the embedding feature extracted from the current first image is matched with the reference embedding feature of the reference image of the object to which the first image belongs to a certain extent, and then the computer device can also detect the image quality of the first image.

[0118] If the image quality of the first image meets the set quality standard, the computer device can store any one of the first embedding feature, the second embedding feature and the third embedding feature as an image embedding feature of an image of the object to which the first image belongs under a target image type, for subsequent identity recognition of the object under the target image type. In some embodiments, which identity recognition stage the embedding feature is stored (i.e., which one of the first embedding feature, the second embedding feature and the third embedding feature is specifically stored) as the image embedding feature of the image of the object to which the first image belongs under the target image type can be set according to actual application scenarios.

[0119] If the image quality of the first image does not meet the set quality standard, indicating that the first image is not suitable for storing embedding features as a base image of the object, the computer device can refuse to store the embedding features of the first image, such as the first embedding feature, the second embedding feature and the third embedding feature.

[0120] In some embodiments, the manner in which the computer device detects the image quality of the first image can include:

[0121] The computer device can obtain a trained quality detection network, which can be a network (also a deep learning network) that has been trained to detect the quality of an image. The computer device can input the first image into the trained quality detection network to invoke the trained quality detection network to detect the quality of the first image and generate a quality parameter (which can be a quality score) of the first image. The quality parameter of the first image can be used to reflect the image quality of the first image. The higher the quality parameter of the first image, the better the image quality of the first image, and vice versa.

[0122] On this basis, the set quality standard can include a set quality parameter threshold, which can be the minimum quality parameter for evaluating the image quality to meet the quality standard. Therefore, if the quality parameter of the first image is greater than or equal to the quality parameter threshold, the computer device can determine that the image quality of the first image meets the set quality standard, i.e., the image quality of the first image is good. If the quality parameter of the first image is less than the quality parameter threshold, the computer device can determine that the image quality of the first image does not meet the set quality standard, i.e., the image quality of the first image is poor.

[0123] For example, in some embodiments, the manner in which the computer device trains the trained quality detection network can include:

[0124] The computer device can obtain a quality detection network to be trained and a sample data set for training the quality detection network. The sample data set can include N groups of sample data, where N is a positive integer. One group of sample data can include a sample image to be detected and a standard image of a sample palm to which the sample image belongs. One group of sample data can include a plurality of sample images to be detected, and the plurality of sample images can correspond to the same standard image, which is a canonical image with good image quality. Each sample image in one group of sample data and the standard image corresponding to each sample image are images of a palm of the same sample object (such as a sample user). Each sample image in the group of sample data can be an image of various styles (such as images of various image types) taken of the palm of the sample object, and the standard image corresponding to the sample image can be similar to the reference image, which is a reference gesture image with good image quality.

[0125] The computer device can also obtain a trained similarity recognition network. The trained similarity recognition network can be a pre-trained network capable of recognizing the similarity between images. The network structure of the trained similarity recognition network can be selected according to the actual application scenario, which is not limited in the embodiments of the present application.

[0126] The computer device can call the trained similarity recognition network to recognize the similarity between the sample image and the corresponding standard image in each group of sample data, to generate a similarity parameter (which can be a similarity score or an identification score) between the sample image and the corresponding standard image in each group of sample data. One sample image and the corresponding standard image in one group of sample data can have one similarity parameter, which is used to evaluate the similarity between the sample image and the corresponding standard image. For example, the greater the value of the similarity parameter, the higher the similarity between the sample image and the corresponding standard image, and vice versa.

[0127] The computer device can also call the quality detection network to be trained to detect the quality of each sample image in each group of sample data, to generate a quality parameter of each sample image in each group of sample data. One sample image in one group of sample data can have one quality parameter.

[0128] Therefore, the computer device can update the network parameters of the quality detection network to be trained by using the similarity parameter between the sample image and the corresponding standard image in each group of sample data and the quality parameter of each sample image in each group of sample data, to obtain a trained quality detection network.

[0129] The manner of updating the network parameters of the quality detection network to be trained can include: updating the network parameters of the quality detection network to be trained, so that the first parameter distribution approaches the second parameter distribution. The first parameter distribution can be a parameter distribution constituted by quality parameters generated by the quality detection network to be trained on sample images in the N groups of sample data, and the second parameter distribution can be a parameter distribution constituted by similarity parameters generated by the trained similarity recognition network between the sample images in the N groups of sample data and the corresponding standard images. Therefore, the loss function L of the quality detection network to be trained (which can be referred to as a quality detection deviation of the quality detection network to be trained for sample images) can be a square loss function as shown in the following formula:

[0130] Y in the above formula can refer to the similarity parameters between the sample images in the N groups of sample data and the corresponding standard images, X can be the sample images in the N groups of sample data, and f(X) can be the quality parameters generated by the quality detection network to be trained on the sample images X. The square difference between Y and f(X) can be used to reflect the difference between the first parameter distribution (i.e., the parameter distribution constituted by Y) and the second parameter distribution (i.e., the parameter distribution constituted by f(X)). Therefore, the purpose of updating the quality detection network to be trained can be to make the loss function L generated in each iteration process tend to a minimum value (such as 0).

[0131] The computer device can continuously iterate and correct (i.e., iteratively update) the network parameters of the quality detection network to be trained through the above-described principles. When the updating of the network parameters of the quality detection network to be trained is completed (such as when the training reaches a convergence state, or when the number of iteration training reaches a set number threshold), the trained quality detection network can be obtained.

[0132] Please refer to FIG. 7, which is a scene diagram of training a quality detection network provided by an embodiment of the present application. As shown in FIG. 7, the computer device can obtain N groups of sample data (including sample data 1 to sample data N here), wherein each group of sample data can include a plurality of sample images and standard images corresponding to the plurality of sample images.

[0133] The computer device can input the sample images and the standard images in the N groups of sample data into the trained similarity recognition network to call the similarity recognition network to generate similarity parameters between the sample images and the corresponding standard images in each group of sample data. The similarity parameters between the sample images and the corresponding standard images in the N groups of sample data can constitute the second parameter distribution.

[0134] The computer device can also input the sample images in the N groups of sample data into the quality detection network to be trained, so as to call the quality detection network to be trained to generate quality parameters of the sample images in each group of sample data. The quality parameters of the sample images in the N groups of sample data can constitute the first parameter distribution.

[0135] Therefore, the computer device can update the network parameters of the quality detection network to be trained through the difference between the first parameter distribution and the second parameter distribution, and the target of the update can be to make the difference between the first parameter distribution and the second parameter distribution tend to be minimum (such as tend to 0), and finally a trained quality detection network can be obtained.

[0136] It should be noted that the traditional image quality detection algorithm is set by experts for various types of images (such as different image types) through various expert experiences. Such algorithm needs to rely on expert knowledge, and can only design different quality evaluation rules for different types of images, and has insufficient robustness for complex scenes. Therefore, the training method of the above quality detection network designed by the embodiments of the present application can assist in training the quality detection network through the similarity between the image with poor quality (such as the sample image) and the image with optimal quality (such as the standard image), so that the quality detection network can also generate the quality parameters of the input image according to the similarity between the image with poor quality (such as the sample image) and the image with optimal quality (such as the standard image). Since the sample image with poor quality can be various, the method provided by the embodiments of the present application can realize accurate quality detection of various types of images through the trained quality detection network, and improve the robustness of quality detection of various types of images.

[0137] In some embodiments, the first image can be obtained in real time when triggering the execution of the target service. The target service can be any service that needs to be identified by image recognition for execution, such as a service of palm payment, a service of palm borrowing (such as palm borrowing of power banks), or a service of palm cash withdrawal, etc. If the identity recognition information determined for the object to which the first image (or the second image) belongs (which can be determined at any identity recognition stage) is the object identity information of the target object (which can be any object in the object set) in the object set, the computer device can further obtain the service authority of the target object under the target service.

[0138] The computer device can execute the target service for the target object through the service authority of the target object under the target service, that is, the target service can be executed for the target object in a manner matched with the service authority of the target object under the target service.

[0139] In some embodiments, the process of identifying the object to which the first image belongs can also be performed by the palmprint image acquisition device, in which the palmprint image acquisition device can obtain relevant information (such as the reference feature set, the image feature set associated with the target image type, and / or the image embedding feature stored in association with the input target object identification information) stored in the background from the computer device. The process of identifying the object to which the first image belongs can be performed by anyone, which can be determined according to the actual application scenario, and the embodiments of the present application do not limit this.

[0140] In the above manner, the object to which the first image belongs is accurately identified in each identity recognition stage. In this process, the object identity of the object to which the first image belongs is identified before the user needs to input the object identification information by combining the reference feature set and the image feature set associated with the target image type, thereby improving the possibility of identifying the object identity of the object to which the first image belongs before the user needs to input the object identification information, improving the accuracy of object identity recognition, thereby reducing the frequency of inputting the object identification information of the object in the object set when performing image recognition, i.e. reducing the operation complexity when performing image recognition, and further improving the efficiency of image recognition.

[0141] Please refer to FIG. 8, which is a flowchart of an embodiment of the present application for identifying the object. As shown in FIG. 8, the process can include:

[0142] Step S301, the palmprint image acquisition device acquires a palmprint image.

[0143] In some embodiments, when the user brushes the palm in the palmprint image acquisition device, the palm can be aimed at the shooting assembly of the palmprint image acquisition device, and the palmprint image acquisition device acquires the palmprint image (such as the second image described above) through the shooting assembly.

[0144] Step S302, the palmprint image acquisition device sends the palmprint image to the computer device.

[0145] Step S303, the computer device judges whether the quality of the palmprint image meets the requirements.

[0146] In some embodiments, the computer device performs quality detection on the palmprint image to determine whether the image quality of the palmprint image sent by the palmprint image acquisition device meets the requirements, and the quality detection can also be implemented by using the trained quality detection network. For example, the quality detection can be implemented by determining whether the quality parameter of the palmprint image is greater than a threshold. Since the detection requirement can be lower than the requirement for storing the base image, the threshold can be lower than the quality parameter threshold. If the quality detection determines that the image quality of the palmprint image meets the requirements, the computer device can perform step S304. If the quality detection determines that the image quality of the palmprint image does not meet the requirements, the computer device can prompt the palmprint image acquisition device to directly request the user to re-enter the palmprint image, and the computer device can discard the currently collected palmprint image.

[0147] In step S304, the computer device determines whether the recognition is successful based on the reference feature set.

[0148] In some embodiments, the computer device can identify the object identity of the object by using the second embedding feature of the palmprint image and the reference feature set. The computer device can determine whether the recognition of the currently collected image is successful based on the reference feature set, that is, the computer device can determine whether the object identity of the object to which the currently collected image (for example, the second image) belongs is successfully recognized by using the reference feature set. If the recognition is successful, the computer device can perform step S305. If the recognition fails, the computer device can perform step S306.

[0149] In step S305, the computer device returns the identity recognition information.

[0150] In some embodiments, the computer device returns the identity recognition information of the object to which the currently collected image belongs. For example, the returned identity recognition information can be the object identity information of the first identified object.

[0151] In step S306, the computer device identifies the image type to which the currently collected palmprint image belongs.

[0152] For example, the identified image type can be the target image type.

[0153] In step S307, the computer device determines whether the recognition is successful based on the target image type.

[0154] In some embodiments, when the computer device identifies through the target image type, it can be based on the image feature set associated with the target image type to identify, which includes the image embedding features of the images of each object in the object set under the target image type. Therefore, the identification process is to identify the object identity of the object through the first embedding features and the image feature set associated with the target image type. If the identification is successful, the computer device can perform the following step S308, and if the identification fails, the computer device can perform the following step S309.

[0155] Step S308, the computer device returns the identity recognition information.

[0156] Here, the returned identity recognition information can be the object identity information of the second identification object.

[0157] Step S309, the computer device prompts the palmprint image acquisition device that the identification fails.

[0158] Here, the computer device prompts the palmprint image acquisition device that the identification fails, so that the palmprint image acquisition device makes the user input the object identification information (such as the mobile phone number), and the palmprint image acquisition device can return the user input mobile phone number to the computer device.

[0159] Step S310, the computer device judges whether the current user has opened the palm service.

[0160] Here, the judgment of whether to open the palm service, that is, whether the user input mobile phone number has been registered for palm service, if it has been registered, the following step S312 can be executed, if it has not been registered, the following step S311 can be executed.

[0161] Step S311, the computer device opens the palm service based on the mobile phone number and registers the base map.

[0162] Here, the computer device can open (i.e. register) the palm service for the user's current input mobile phone number, and can register the user's base map, which can be to require the user to input the reference image of the reference posture, so that the computer device can store the reference embedding features of the reference image of the user, to realize the registration of the palmprint image of the user.

[0163] Step S312, the computer device performs 1v1 identification and returns the result.

[0164] Here, the computer device can perform 1v1 identification on the user according to the user-inputted mobile phone number and return an identity recognition result. The 1v1 identification process is the process of associating the target embedding feature stored according to the target object identification information inputted by the user with the third embedding feature to perform identity recognition on the object to which the current collected image belongs.

[0165] In step S313, the computer device determines whether the current image can be used as a sub-base image.

[0166] In some embodiments, the current image is a palmprint image collected by the palmprint image collection device. The computer device can detect the image quality of the current image by using the trained quality detection network. If the quality parameter of the current image is greater than or equal to the quality parameter threshold, it indicates that the current image can be used as a sub-base image. Otherwise, if the quality parameter of the current image is less than the quality parameter threshold, it indicates that the current image cannot be used as a sub-base image. If the current image can be used as a sub-base image, the computer device can perform step S315. If the current image cannot be used as a sub-base image, the computer device can perform step S314.

[0167] In step S314, the computer device discards the current image.

[0168] In step S315, the computer device stores the sub-base image.

[0169] Here, the current image can be stored as a sub-base image, or the image embedding feature of the current image can be stored as a sub-base image. For example, the image embedding feature of the current image can be stored as the image embedding feature of the target image of the object to which the image belongs under the target image type.

[0170] Through the above process, the user's identity is identified three times in a progressive manner, and accurate and efficient identification of the user's identity is achieved.

[0171] Please refer to FIG. 9, which is a structural schematic diagram of an image processing apparatus provided by an embodiment of the present application. As shown in FIG. 9, the image processing apparatus 90 can include a first obtaining module 901, an identification module 902, a second obtaining module 903, a third obtaining module 904, and a determination module 905.

[0172] The first acquisition module 901 is configured to acquire a first image to be identified and a first embedding feature of the first image, the first image being a palmprint image; the identification module 902 is configured to identify a target image type to which the first image belongs; the second acquisition module 903 is configured to acquire an image feature set associated with the target image type; the image feature set includes image embedding features of target images of a plurality of objects, the target images being palmprint images belonging to the target image type, and each object having object identity information of the object; the third acquisition module 904 is configured to acquire a feature matching degree between the first embedding feature and each image embedding feature; and the determination module 905 is configured to determine identity recognition information of an object to which the first image belongs based on the feature matching degree between the first embedding feature and each image embedding feature and the object identity information of each object.

[0173] In some embodiments, the first acquisition module 901 acquires the first image in the following manner: a second image collected for a palm is acquired, and a feature extraction process is performed on the second image to generate a second embedding feature of the second image; a reference feature set is acquired; the reference feature set includes reference embedding features of a plurality of reference images, the reference images being palmprint images of the object in a reference posture; a feature matching degree between the second embedding feature and each reference embedding feature is acquired; and if the feature matching degree between the second embedding feature and each reference embedding feature is less than a first matching degree threshold, the second image is determined as the first image.

[0174] In some embodiments, the first acquisition module 901 is further configured to: from the reference feature set, filter out a reference embedding feature having a feature matching degree greater than or equal to the first matching degree threshold with the second embedding feature as a first candidate embedding feature; determine an object to which the first candidate embedding feature having the greatest feature matching degree with the second embedding feature belongs as a first identified object; and determine object identity information of the first identified object as the identity recognition information of the object to which the second image belongs.

[0175] In some embodiments, the determination module 905 determines the identity recognition information of the object to which the first image belongs based on the feature matching degree between the first embedding feature and each image embedding feature and the object identity information of each object in the following manner: if the feature matching degree between the first embedding feature and each image embedding feature is less than a second matching degree threshold, target object identification information of the object to which the first image belongs is acquired; a feature extraction process is performed on the first image to generate a third embedding feature of the first image; and the identity recognition information of the object to which the first image belongs is determined based on the target object identification information and the third embedding feature.

[0176] In some embodiments, the determining module 905 is further configured to: determine a second candidate embedding feature from the image feature set, wherein the feature matching degree between the second candidate embedding feature and the first embedding feature is greater than or equal to a second matching degree threshold; determine the object to which the second candidate embedding feature with the largest feature matching degree with the second embedding feature belongs as the second identification object; and determine the object identity information of the second identification object as the identity recognition information for the object to which the first image belongs.

[0177] In some embodiments, each object has its own reference embedding feature of a reference image in a reference pose, and the reference embedding feature corresponding to each object is associated with the object identification information of each object. The determining module 905 determines the method of identifying the identity information of the object to which the first image belongs based on the target object identification information and the third embedding feature, including: determining the reference embedding feature associated with the target object identification information as the target embedding feature; obtaining the feature matching degree between the target embedding feature and the third embedding feature; if the feature matching degree between the target embedding feature and the third embedding feature is greater than or equal to the third matching degree threshold, then determining the object identity information of the object to which the target embedding feature belongs as the identity identification information of the object to which the first image belongs.

[0178] In some embodiments, the first image is acquired by a palmprint image acquisition device; the determining module 905 is further configured to: generate recognition failure information if the feature matching degree between the target embedded feature and the third embedded feature is less than the third matching degree threshold; return the recognition failure information to the palmprint image acquisition device, so that the palmprint image acquisition device outputs a prompt message that the image needs to be acquired again based on the recognition failure information.

[0179] In some embodiments, the determining module 905 is further configured to: if the feature matching degree between the target embedding feature and the third embedding feature is greater than or equal to the third matching degree threshold, then detect the image quality of the first image; if the image quality of the first image meets the set quality standard, then store the first embedding feature or the third embedding feature as the image embedding feature of the object to which the first image belongs under the target image type.

[0180] In some embodiments, the method by which the determining module 905 detects the image quality of the first image includes: inputting the first image into a trained quality detection network; calling the trained quality detection network to perform quality detection on the first image and generating quality parameters of the first image; the quality parameters of the first image are used to reflect the image quality of the first image; the quality standard includes a set quality parameter threshold; the determining module is further configured to: if the quality parameters of the first image are greater than or equal to the quality parameter threshold, then determine that the image quality of the first image meets the quality standard.

[0181] In some embodiments, the determining module 905 is further configured to: determine that the image quality of the first image does not meet the quality standard if the quality parameter of the first image is less than the quality parameter threshold; and refuse to store the first embedding feature or the third embedding feature of the first image.

[0182] In some embodiments, the image processing apparatus further includes a training module 906, which is configured to: acquire a quality detection network to be trained and a sample dataset; the sample dataset contains N sets of sample data, where N is a positive integer, and each set of sample data includes a sample image to be detected and a standard image of the sample hand to which the sample image belongs; call the trained similarity recognition network to identify the similarity between the sample image and the corresponding standard image in each set of sample data, and generate similarity parameters between the sample image and the corresponding standard image in each set of sample data; call the quality detection network to be trained to perform quality detection on the sample images in each set of sample data, and generate quality parameters of the sample images in each set of sample data; and update the network parameters of the quality detection network to be trained based on the similarity parameters between the sample images and the corresponding standard images in each set of sample data and the quality parameters of the sample images in each set of sample data, to obtain a trained quality detection network.

[0183] In some embodiments, the training module 906 updates the network parameters of the quality detection network to be trained by: updating the network parameters of the quality detection network to be trained so that the first parameter distribution approaches the second parameter distribution; wherein the first parameter distribution is a parameter distribution composed of the quality parameters of the sample images in N sets of sample data, and the second parameter distribution is a parameter distribution composed of the similarity parameters between the sample images and the corresponding standard images in N sets of sample data.

[0184] In some embodiments, the method by which the recognition module 902 identifies the target image type to which the first image belongs from a set of multiple image types includes: inputting the first image into a trained multi-classification network; and calling the trained multi-classification network to identify the target image type to which the first image belongs from the multiple image types.

[0185] In some embodiments, the first image is acquired by a palmprint image acquisition device; the multiple image types include a large-angle image type; the recognition module 902 identifies the target image type to which the first image belongs from the set multiple image types by: obtaining the target angle between the image plane corresponding to the first image and the coordinate system where the palmprint image acquisition device is located; if the target angle is greater than a set angle threshold, then the large-angle image type is determined as the target image type to which the first image belongs.

[0186] In some embodiments, the third acquisition module 904 acquires the feature matching degree between the first embedded feature and each image embedded feature in the image feature set by: calculating the cosine similarity between the first embedded feature and each image embedded feature in the image feature set; and determining the cosine similarity between the first embedded feature and the image embedded features in the image feature set as the feature matching degree between the first embedded feature and the image embedded features in the image feature set.

[0187] In some embodiments, the first image is acquired in real time when the execution of the target service is triggered; the image processing device further includes an execution module 907, which is configured to: if the identity information determined for the object to which the first image belongs is the object identity information of the target object in the object set, then obtain the business permissions of the target object under the target service; and execute the target service on the target object based on the business permissions of the target object under the target service.

[0188] According to one embodiment of this application, the steps involved in the image processing method shown in FIG3 can be executed by various modules in the image processing apparatus 90 shown in FIG9. For example, step S101 shown in FIG3 can be executed by the first acquisition module 901 in FIG9, step S102 shown in FIG3 can be executed by the recognition module 902 in FIG9; step S103 shown in FIG3 can be executed by the second acquisition module 903 in FIG9, step S104 shown in FIG3 can be executed by the third acquisition module 904 in FIG9, and step S105 shown in FIG3 can be executed by the determination module 905 in FIG9.

[0189] This application can acquire a first image to be identified and a first embedding feature of the first image; identify the target image type to which the first image belongs from a set of multiple image types; acquire an image feature set associated with the target image type; the image feature set includes image embedding features of multiple objects in the object set under the target image type, and each object in the object set has its own object identity information; acquire the feature matching degree between the first embedding feature and each image embedding feature in the image feature set; and determine the identity recognition information for the object to which the first image belongs based on the feature matching degree between the first embedding feature and each image embedding feature in the image feature set and the object identity information of each object. Therefore, the apparatus proposed in this application can first identify the specific target image type to which the first image belongs, and then, through the specific image feature set associated with the specific target image type and the embedding features of the first image (such as the first embedding feature), realize the identity recognition of the object to which the first image belongs. Because the specific type of the image is considered and the image is identified within a specific range (such as the image feature set associated with the target image type), the identity recognition of the object to which the image belongs is realized, thus improving the accuracy of object identification through images.

[0190] According to one embodiment of this application, the modules in the image processing apparatus 90 shown in FIG. 9 can be individually or entirely combined into one or more units, or some of the units can be further divided into multiple functionally smaller sub-units to achieve the same operation without affecting the technical effect of the embodiment of this application. The above modules are based on logical function division. In practical applications, the function of one module can also be implemented by multiple units, or the function of multiple modules can be implemented by one unit. In other embodiments of this application, the image processing apparatus 90 may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0191] In this application embodiment, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0192] According to one embodiment of this application, a computer program capable of executing the steps involved in the corresponding methods shown in the various embodiments of this application can be run on a general-purpose computer device (which may include processing elements and storage elements such as a central processing unit (CPU), random access storage medium (RAM), and read-only storage medium (ROM)) to construct the image processing apparatus 90 shown in FIG9. The aforementioned computer program can be recorded on a computer-readable recording medium, and can be loaded into and run in the aforementioned computer device via the computer-readable recording medium.

[0193] Please refer to Figure 10, which is a schematic diagram of the structure of a computer device provided in an embodiment of this application. As shown in Figure 10, the computer device 1000 may include: a processor 1001, a network interface 1004, and a memory 1005. Furthermore, in some embodiments, the computer device 1000 may also include: a user interface 1003, and at least one communication bus 1002. The communication bus 1002 is used to implement communication between these components. The user interface 1003 may include a display screen and a keyboard; optionally, the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may include a standard wired interface and a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device. The memory 1005 may also be at least one storage device located remotely from the aforementioned processor 1001. As shown in Figure 10, the memory 1005, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a device control application.

[0194] In the computer device 1000 shown in Figure 10, the network interface 1004 provides network communication functions; the user interface 1003 is mainly used to provide an input interface for users; and the processor 1001 can be used to call the device control application stored in the memory 1005 to achieve: acquiring a first image to be identified and a first embedded feature of the first image; identifying the target image type to which the first image belongs; acquiring an image feature set associated with the target image type; the image feature set contains image embedded features of target images of multiple objects, the target image is a palm print image belonging to the target image type, and each object has its own object identity information; acquiring the feature matching degree between the first embedded feature and each image embedded feature respectively; and determining the identity recognition information of the object to which the first image belongs based on the feature matching degree between the first embedded feature and each image embedded feature and the object identity information of each object.

[0195] In some embodiments, the processor 1001 may also be used to invoke a device control application stored in the memory 1005 to: acquire a second image captured from the palm, and perform feature extraction processing on the second image to generate a second embedded feature of the second image; acquire a reference feature set; the reference feature set includes reference embedded features of reference images of objects in the object set under a reference pose; acquire the feature matching degree between the second embedded feature and each reference embedded feature; if the feature matching degree between the second embedded feature and each reference embedded feature is less than a first matching degree threshold, then the second image is determined as the first image.

[0196] In some embodiments, the processor 1001 may also be used to call a device control application stored in the memory 1005 to: determine a first candidate embedded feature from a reference feature set, wherein the feature matching degree between the first candidate embedded feature and the second embedded feature is greater than or equal to a first matching degree threshold; determine the object to which the first candidate embedded feature with the largest feature matching degree with the second embedded feature belongs as a first identification object; and determine the object identity information of the first identification object as the identity recognition information for the object to which the second image belongs.

[0197] In some embodiments, the processor 1001 may also be used to call the device control application stored in the memory 1005 to: if the feature matching degree between the first embedded feature and each image embedded feature is less than the second matching degree threshold, then obtain the target object identification information of the object to which the first image belongs; perform feature extraction processing on the first image to generate the third embedded feature of the first image; and determine the identity recognition information of the object to which the first image belongs based on the target object identification information and the third embedded feature.

[0198] In some embodiments, the processor 1001 may also be used to call a device control application stored in the memory 1005 to: determine a second candidate embedding feature from the image feature set, wherein the feature matching degree between the second candidate embedding feature and the first embedding feature is greater than or equal to a second matching degree threshold; determine the object to which the second candidate embedding feature with the largest feature matching degree with the second embedding feature belongs as a second identification object; and determine the object identity information of the second identification object as the identity recognition information for the object to which the first image belongs.

[0199] In some embodiments, each object has its own reference embedding feature of a reference image in a reference pose, and the reference embedding feature corresponding to each object is stored in association with the object identification information of each object; the processor 1001 can also be used to call the device control application stored in the memory 1005 to: determine the reference embedding feature that is stored in association with the target object identification information as the target embedding feature; obtain the feature matching degree between the target embedding feature and the third embedding feature; if the feature matching degree between the target embedding feature and the third embedding feature is greater than or equal to the third matching degree threshold, then determine the object identity information of the object to which the target embedding feature belongs as the identity recognition information of the object to which the first image belongs.

[0200] In some embodiments, the first image is acquired by a palmprint image acquisition device; the processor 1001 can also be used to call the device control application stored in the memory 1005 to achieve: if the feature matching degree between the target embedded feature and the third embedded feature is less than the third matching degree threshold, then generate recognition failure information; return the recognition failure information to the palmprint image acquisition device, so that the palmprint image acquisition device outputs a prompt message that the image needs to be acquired again based on the recognition failure information.

[0201] In some embodiments, the processor 1001 may also be used to call the device control application stored in the memory 1005 to: detect the image quality of the first image if the feature matching degree between the target embedding feature and the third embedding feature is greater than or equal to the third matching degree threshold; and if the image quality of the first image meets the set quality standard, store the first embedding feature or the third embedding feature as the image embedding feature of the object to which the first image belongs under the target image type.

[0202] In some embodiments, the processor 1001 can also be used to call a device control application stored in the memory 1005 to: input a first image into a trained quality detection network; call the trained quality detection network to perform quality detection on the first image and generate quality parameters of the first image; the quality parameters of the first image are used to reflect the image quality of the first image; the quality standard includes a set quality parameter threshold; the processor 1001 can also be used to call a device control application stored in the memory 1005 to: if the quality parameters of the first image are greater than or equal to the quality parameter threshold, determine that the image quality of the first image meets the quality standard.

[0203] In some embodiments, the processor 1001 may also be used to call a device control application stored in the memory 1005 to: determine that the image quality of the first image does not meet the quality standard if the quality parameter of the first image is less than the quality parameter threshold; and refuse to store the first embedding feature or the third embedding feature of the first image.

[0204] In some embodiments, the processor 1001 can also be used to call the device control application stored in the memory 1005 to: acquire a quality detection network to be trained and a sample dataset; the sample dataset contains N sets of sample data, where N is a positive integer, and each set of sample data includes a sample image to be detected and a standard image of the sample hand to which the sample image belongs; call the trained similarity recognition network to identify the similarity between the sample image and the corresponding standard image in each set of sample data, and generate similarity parameters between the sample image and the corresponding standard image in each set of sample data; call the quality detection network to be trained to perform quality detection on the sample images in each set of sample data, and generate quality parameters for the sample images in each set of sample data; based on the similarity parameters between the sample images and the corresponding standard images in each set of sample data, and the quality parameters of the sample images in each set of sample data, update the network parameters of the quality detection network to be trained, and obtain the trained quality detection network.

[0205] In some embodiments, updating the network parameters of the quality detection network to be trained includes: updating the network parameters of the quality detection network to be trained so that the first parameter distribution approaches the second parameter distribution; wherein the first parameter distribution is a parameter distribution composed of the quality parameters of the sample images in N sets of sample data, and the second parameter distribution is a parameter distribution composed of the similarity parameters between the sample images and the corresponding standard images in N sets of sample data.

[0206] In some embodiments, the processor 1001 can also be used to call the device control application stored in the memory 1005 to: input the first image into the trained multi-classification network; and call the trained multi-classification network to identify the target image type to which the first image belongs from multiple image types.

[0207] In some embodiments, the first image is acquired by a palmprint image acquisition device; multiple image types include the first image type; the processor 1001 can also be used to call the device control application stored in the memory 1005 to achieve: obtaining the target angle between the image plane corresponding to the first image and the coordinate system where the palmprint image acquisition device is located; if the target angle is greater than a set angle threshold, then the first image type is determined as the target image type to which the first image belongs.

[0208] In some embodiments, the processor 1001 may also be used to call a device control application stored in the memory 1005 to: calculate the cosine similarity between the first embedded feature and each image embedded feature respectively; and determine the cosine similarity between the first embedded feature and the image embedded feature as the feature matching degree between the first embedded feature and the image embedded feature.

[0209] In some embodiments, the first image is acquired in real time when the execution of the target service is triggered; the processor 1001 can also be used to call the device control application stored in the memory 1005 to achieve: if the identity information determined for the object to which the first image belongs is the object identity information of the target object in the object set, then the business permissions of the target object under the target service are obtained; based on the business permissions of the target object under the target service, the target service is executed on the target object.

[0210] It should be understood that the computer device 1000 described in the embodiments of this application can execute the image processing methods described in the embodiments of this application, and can also execute the image processing apparatus 90 described in the embodiments corresponding to FIG9 above. In addition, the beneficial effects of using the same method are also the same.

[0211] Furthermore, it should be noted that this application also provides a computer-readable storage medium storing a computer program. When a processor executes the computer program, it can perform the image processing methods described in the various embodiments of this application. Additionally, the beneficial effects of using the same method are also the same. For technical details not disclosed in the computer storage medium embodiments of this application, please refer to the description of the method embodiments of this application.

[0212] As an example, the aforementioned computer program can be deployed and executed on a single computer device, or deployed and executed on multiple computer devices located in one location, or executed on multiple computer devices distributed across multiple locations and interconnected via a communication network. These multiple computer devices distributed across multiple locations and interconnected via a communication network can form a blockchain network.

[0213] The aforementioned computer-readable storage medium can be an internal storage unit of the computer device, such as a hard drive or memory. It can also be an external storage device, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, or flash card. Furthermore, the computer-readable storage medium can include both internal and external storage units of the computer device. This computer-readable storage medium is used to store the computer program and other programs and data required by the computer device. It can also be used to temporarily store data that has been output or will be output.

[0214] This application provides a computer program product comprising a computer program stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform the image processing methods described in the various embodiments of this application. Furthermore, the beneficial effects of using the same method are also the same. For technical details not disclosed in the computer-readable storage medium embodiments of this application, please refer to the description of the method embodiments of this application.

[0215] The terms "first," "second," etc., in the specification, claims, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the term "comprising," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or modules, but may optionally include steps or modules not listed, or may optionally include other step units inherent to these processes, methods, apparatuses, products, or devices.

[0216] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0217] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. An image processing method applied to a computer device, the method comprising: obtaining a first image to be recognized and a first embedding feature of the first image, the first image being a palmprint image; recognizing a target image type to which the first image belongs; obtaining an image feature set associated with the target image type, the image feature set containing image embedding features of target images of a plurality of objects, the target images being palmprint images belonging to the target image type, each object having respective object identity information; determining feature matching degrees between the first embedding feature and each image embedding feature; based on the feature matching degrees between the first embedding feature and each image embedding feature and the object identity information of each object, determining identity recognition information of an object to which the first image belongs.

2. The method of claim 1, wherein, obtaining the first image comprises: obtaining a second image collected for a palm and performing feature extraction processing on the second image to generate a second embedding feature of the second image; obtaining a reference feature set, the reference feature set containing reference embedding features of a plurality of reference images, the reference images being palmprint images of the object in a reference posture; obtaining feature matching degrees between the second embedding feature and each reference embedding feature; if the feature matching degrees between the second embedding feature and each reference embedding feature are all less than a first matching degree threshold, determining the second image as the first image.

3. The method of claim 1 or 2, wherein, The method further comprises: from the reference feature set, determining a first candidate embedding feature, the feature matching degree between the first candidate embedding feature and the second embedding feature being greater than or equal to the first matching degree threshold; determining an object to which a first candidate embedding feature with the greatest feature matching degree between the second embedding feature belongs as a first recognized object; determining object identity information of the first recognized object as the identity recognition information of the object to which the second image belongs.

4. The method of any one of claims 1 to 3, wherein, The method further comprises: if the feature matching degrees between the first embedding feature and each image embedding feature are all less than a second matching degree threshold, obtaining target object identification information of the object to which the first image belongs; performing feature extraction processing on the first image to generate a third embedding feature of the first image; based on the target object identification information and the third embedding feature, determining the identity recognition information of the object to which the first image belongs.

5. The method of any one of claims 1 to 4, wherein, The method further comprises: from the image feature set, determining a second candidate embedding feature, the feature matching degree between the second candidate embedding feature and the first embedding feature being greater than or equal to the second matching degree threshold; determining an object to which a second candidate embedding feature with the greatest feature matching degree between the second embedding feature belongs as a second recognized object; determining object identity information of the second recognized object as the identity recognition information of the object to which the first image belongs.

6. The method of any one of claims 1 to 5, wherein, Each of the objects has a reference embedding feature of a reference image in a reference posture, and the reference embedding feature corresponding to each of the objects is stored in association with object identification information of the object; The identity recognition information of the object to which the first image belongs is determined based on the target object identification information and the third embedding feature, including: The reference embedding feature stored in association with the target object identification information is determined as a target embedding feature; A feature matching degree between the target embedding feature and the third embedding feature is obtained; If the feature matching degree between the target embedding feature and the third embedding feature is greater than or equal to a third matching degree threshold, object identity information of the object to which the target embedding feature belongs is determined as the identity recognition information of the object to which the first image belongs.

7. The method of any one of claims 1 to 6, wherein, The first image is obtained by a palmprint image acquisition device; the method further includes: If the feature matching degree between the target embedding feature and the third embedding feature is less than the third matching degree threshold, identification failure information is generated; The identification failure information is returned to the palmprint image acquisition device, so that the palmprint image acquisition device outputs prompt information for re-performing palmprint image acquisition based on the identification failure information.

8. The method of any one of claims 1 to 7, wherein, The method further includes: If the feature matching degree between the target embedding feature and the third embedding feature is greater than or equal to the third matching degree threshold, the image quality of the first image is detected; If the image quality of the first image meets a set quality standard, the first embedding feature or the third embedding feature is determined as an image embedding feature of an image of the object to which the first image belongs in the target image type.

9. The method of any one of claims 1 to 8, wherein, The detection of the image quality of the first image includes: The first image is input into a trained quality detection network; The trained quality detection network is called to perform quality detection on the first image, to generate a quality parameter of the first image; the quality parameter of the first image is used to reflect the image quality of the first image; The quality standard includes a set quality parameter threshold; the method further includes: If the quality parameter of the first image is greater than or equal to the quality parameter threshold, it is determined that the image quality of the first image meets the quality standard.

10. The method of any one of claims 1 to 9, wherein, The method further includes: If the quality parameter of the first image is less than the quality parameter threshold, it is determined that the image quality of the first image does not meet the quality standard; and The first embedding feature or the third embedding feature of the first image is rejected.

11. The method of any one of claims 1 to 9, wherein, The method further includes: A quality detection network to be trained and a sample data set are obtained; the sample data set includes N groups of sample data, N is a positive integer, and one group of sample data includes a sample image to be detected and a standard image of a sample palm to which the sample image belongs; A trained similarity recognition network is called to recognize and process a similarity between the sample image and the corresponding standard image in each group of sample data, to generate a similarity parameter between the sample image and the corresponding standard image in each group of sample data; calling the to-be-trained quality detection network to perform quality detection on the sample images in each set of sample data respectively, to generate quality parameters of the sample images in each set of sample data; updating network parameters of the to-be-trained quality detection network based on the similarity parameters between the sample images in each set of sample data and the corresponding standard images and the quality parameters of the sample images in each set of sample data, to obtain the trained quality detection network.

12. The method of any one of claims 1 to 11, wherein, The manner of updating the network parameters of the to-be-trained quality detection network comprises: updating the network parameters of the to-be-trained quality detection network so that a first parameter distribution approaches a second parameter distribution. The first parameter distribution is a parameter distribution formed by the quality parameters of the sample images in the N sets of sample data, and the second parameter distribution is a parameter distribution formed by the similarity parameters between the sample images in the N sets of sample data and the corresponding standard images.

13. The method of any one of claims 1 to 12, wherein, The identification of the target image type to which the first image belongs comprises: inputting the first image into the trained multi-classification network; calling the trained multi-classification network to identify the target image type to which the first image belongs from the set multiple image types.

14. The method of any one of claims 1 to 13, wherein, The first image is obtained by a palmprint image acquisition device. The identification of the target image type to which the first image belongs comprises: obtaining a target angle between an image plane corresponding to the first image and a coordinate system in which the palmprint image acquisition device is located; if the target angle is greater than a set angle threshold, determining a first image type as the target image type to which the first image belongs.

15. The method of any one of claims 1 to 14, wherein, The determination of the feature matching degree between the first embedding feature and each image embedding feature comprises: determining a cosine similarity between the first embedding feature and each image embedding feature in the image feature set; determining the cosine similarity between the first embedding feature and the image embedding feature as the feature matching degree between the first embedding feature and the image embedding feature.

16. The method of any one of claims 1 to 15, wherein, The first image is obtained in real time when triggering the execution of a target service; the method further comprises: if the identity recognition information determined for the object to which the first image belongs is object identity information of a target object, obtaining a service authority of the target object under the target service; performing the target service on the target object based on the service authority of the target object under the target service.

17. An image processing apparatus, the apparatus comprising: a first obtaining module configured to obtain a first image to be identified and a first embedding feature of the first image, the first image being a palmprint image; an identification module configured to identify a target image type to which the first image belongs; a second obtaining module configured to obtain an image feature set associated with the target image type, the image feature set containing image embedding features of target images of multiple objects, the target images being palmprint images belonging to the target image type, each object having its own object identity information; The first determining module is configured to determine feature matching degrees between the first embedding feature and each image embedding feature; The second determining module is configured to determine identity recognition information of the object to which the first image belongs based on the feature matching degrees between the first embedding feature and the each image embedding feature and the object identity information of the each object.

18. A computer program product comprising a computer program which, when executed by a processor, implements the steps of the method of any one of claims 1 to 16.

19. A computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the method of any one of claims 1 to 16.

20. A computer-readable storage medium storing a computer program which is adapted to be loaded and executed by a processor to perform the steps of the method of any one of claims 1 to 16.

Images