Palm feature processing method and apparatus for identity authentication, and device and medium

Abstract

A palm feature processing method for identity authentication, the method being executed by a computer device. The method comprises: acquiring a plurality of palm images for different collection angles of the same palm, wherein the region occupied by a palm in each palm image constitutes a palm region (202); for each palm image, locating a palm key region comprised in a palm region in the palm image, and determining an auxiliary region other than the palm key region in the palm region, wherein a palm portion in the palm key region is related to the collection angle of the palm image (204); performing feature extraction on the palm image, and assigning different contribution weights to the palm key region and the auxiliary region in the palm image during the feature extraction, so as to obtain a single-angle palm feature of the palm image, wherein the contribution weight assigned to the palm key region in the palm image is higher than the contribution weight assigned to the auxiliary region in the palm image (206); and fusing respective single-angle palm features of the plurality of palm images, so as to obtain a multi-angle palm feature of the palm, wherein the multi-angle palm feature is used for identity authentication (208).

Description

Palm feature processing method, device, equipment and medium for identity authentication

[0001] Related applications

[0002] This application claims priority to Chinese patent application number 2023110357669, filed on August 17, 2023, entitled “Palm feature processing method, device, equipment and medium for identity authentication,” the entire text of which is hereby incorporated by reference. Technical Field

[0003] The present application relates to artificial intelligence technology and more particularly to the field of image processing, and in particular to a palm feature processing method, apparatus, device and medium for identity authentication. Background Art

[0004] Palm recognition is a biometric technology based on palm texture. It has the following characteristics and benefits: Uniqueness makes palm recognition a highly reliable method for personal identity verification. Difficulty in forging: Compared with other biometric recognition technologies, palm texture is not easily forged. Efficiency: Palm recognition technology has low requirements for acquisition and recognition speed, enabling rapid identity verification. Non-contact: Palm recognition does not require contact with special equipment; it only requires the palm to be captured by a camera or sensor, making palm recognition more hygienic, convenient, and comfortable. Diversity: Palm recognition technology is suitable for people of all ages and genders. Palm recognition technology can be widely used in various fields involving identity authentication, such as internet security, payment systems, access control systems, and self-service equipment. Due to its uniqueness, difficulty in forging, and efficiency, palm recognition technology has become a convenient, secure, and reliable method for personal identity authentication, and has broad application prospects in multiple application fields.

[0005] In the traditional palm-based identity authentication method, identity authentication is often performed based on a palm image from a single angle. The feature information of the palm image involved in the identity authentication process is thin, which affects the effect of identity authentication and has a low accuracy rate, resulting in a waste of hardware resources used to support the identity authentication function.

[0006] Summary of the Invention

[0007] Provided are a palm feature processing method, apparatus, device, and medium for identity authentication.

[0008] In a first aspect, the present application provides a palm feature processing method for identity authentication, which is executed by a computer device, and the method includes:

[0009] Acquire multiple palm images of the same palm at different acquisition angles, where the area occupied by the palm in each palm image constitutes a palm area;

[0010] For each of the palm images, locating a palm key area included in a palm region within the palm image, and determining an auxiliary area in the palm region other than the palm key area, wherein the palm position in the palm key area is related to an acquisition angle of the palm image;

[0011] Extracting features from the palm image, and assigning different contribution weights to a palm key region and an auxiliary region in the palm image during the feature extraction process to obtain a single-angle palm feature of the palm image, wherein the contribution weight assigned to the palm key region in the palm image is higher than the contribution weight assigned to the auxiliary region in the palm image; and

[0012] The single-angle palm features of each of the multiple palm images are fused to obtain multi-angle palm features of the palm; the multi-angle palm features are used for identity authentication.

[0013] In a second aspect, the present application provides a palm feature processing device for identity authentication, the device comprising:

[0014] An acquisition module is used to acquire multiple palm images of the same palm at different acquisition angles, where the area occupied by the palm in each palm image constitutes a palm area;

[0015] a positioning module, configured to locate, for each of the palm images, a palm key area included in the palm region of the palm image, and determine an auxiliary area in the palm region other than the palm key area, wherein the palm position in the palm key area is related to the acquisition angle of the palm image;

[0016] an extraction module, configured to extract features from the palm image, and in the process of extracting features, assign different contribution weights to the palm key area and the auxiliary area in the palm image to obtain single-angle palm features of the palm image; wherein the contribution weight assigned to the palm key area in the palm image is higher than the contribution weight assigned to the auxiliary area in the palm image; and

[0017] The fusion module is used to fuse the single-angle palm features of the multiple palm images to obtain the multi-angle palm features of the palm; the multi-angle palm features are used for identity authentication.

[0018] In a third aspect, the present application provides a computer device including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps in the method embodiments of the present application are implemented.

[0019] In a fourth aspect, the present application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the various method embodiments of the present application.

[0020] In a fifth aspect, the present application provides a computer program product, including a computer program, which implements the steps in the method embodiments of the present application when the computer program is executed by a processor.

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

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

[0023] FIG1 is a diagram illustrating an application environment of a palm feature processing method for identity authentication according to an embodiment;

[0024] FIG2 is a flow chart of a palm feature processing method for identity authentication according to one embodiment;

[0025] FIG3 is a schematic diagram of obtaining multiple captured images of the same palm at different capture angles in one embodiment;

[0026] FIG4 is a schematic diagram showing the positions of key palm areas included in the palm area in one embodiment;

[0027] FIG5 is a flowchart of super-resolution reconstruction in one embodiment;

[0028] FIG6 is a flowchart of super-resolution reconstruction based on a super-resolution model in one embodiment;

[0029] FIG7 is a diagram showing the structure of a super-resolution model according to an embodiment;

[0030] FIG8 is a schematic diagram of the training principle of a feature extraction model in one embodiment;

[0031] FIG9 is a diagram showing the overall framework of identity authentication based on palm authentication in one embodiment;

[0032] FIG10 is a flow chart of a palm feature processing method for identity authentication according to another embodiment;

[0033] FIG11 is a schematic diagram of a scenario in which a palm feature processing method for identity authentication is applied to palm authentication for logging into an instant messaging application in one embodiment;

[0034] FIG12 is a schematic diagram showing a scenario in which a palm feature processing method for identity authentication can be applied to palm authentication to achieve resource transfer in one embodiment;

[0035] FIG13 is a block diagram of a palm feature processing device for identity authentication according to an embodiment;

[0036] FIG14 is a block diagram of a palm feature processing device for identity authentication according to another embodiment;

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

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

[0039] The palm feature processing method for identity authentication provided in this application can be applied in the application environment shown in Figure 1. Terminal 102 communicates with server 104 via a network. A data storage system can be separately provided to store data that server 104 needs to process. The data storage system can be integrated with server 104 or placed in the cloud or on another server. Terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart car devices, etc. Portable wearable devices can include smart watches, smart bracelets, head-mounted devices, etc. Terminal 102 is equipped with a camera for capturing palm images. Server 104 can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides network security services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, cloud security, host security, CDN, and basic cloud computing services such as big data and artificial intelligence platforms. The terminal 102 and the server 104 may be connected directly or indirectly via wired or wireless communication, which is not limited in this application.

[0040] Server 104 can obtain multiple palm images of the same palm at different acquisition angles, where the area occupied by the palm in each palm image constitutes a palm area. For each palm image, the palm key area included in the palm area within the palm image is located, and the auxiliary area in the palm area other than the palm key area is determined. The palm position in the palm key area is related to the acquisition angle of the palm image. Server 104 can extract features from the palm image, and in the feature extraction process, assign different contribution weights to the palm key area and the auxiliary area in the palm image to obtain single-angle palm features of the palm image. The contribution weight assigned to the palm key area in the palm image is higher than the contribution weight assigned to the auxiliary area in the palm image. Server 104 can fuse the single-angle palm features of multiple palm images to obtain multi-angle palm features of the palm. The multi-angle palm features are used for identity authentication.

[0041] It is understood that terminal 102 can capture multiple palm images of the same palm at different capture angles and send the multiple palm images to server 104. Server 104 can receive the multiple palm images sent by terminal 102. It is understood that server 104 can also obtain multiple palm images of the same palm at different capture angles from a third-party storage device. This embodiment is not limited to this. It is understood that the application scenario in Figure 1 is for illustrative purposes only and is not limited to this.

[0042] It should be noted that the palm feature processing methods for identity authentication in some embodiments of this application utilize artificial intelligence technology. For example, key palm regions are located using artificial intelligence technology, and single-angle palm features from palm images are also encoded using artificial intelligence technology. To facilitate understanding of artificial intelligence, the concept of artificial intelligence will now be explained. Specifically, artificial intelligence refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. Artificial intelligence is the study of the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making capabilities. This application utilizes artificial intelligence technology to locate key palm regions and extract single-angle palm features from palm images, further improving the accuracy of identity authentication and thus further avoiding the waste of hardware resources used to support identity authentication.

[0043] In one embodiment, as shown in FIG2 , a palm feature processing method for identity authentication is provided. This embodiment uses the method applied to the server 104 in FIG1 as an example for description, including the following steps:

[0044] Step 202 : Acquire multiple palm images of the same palm at different acquisition angles, where the area occupied by the palm in each palm image constitutes a palm area.

[0045] The acquisition angle is the angle formed between the orientation of the palm in the imaging plane and the calibration direction of the image acquisition device when the image acquisition device is capturing an image of the palm. The orientation of the palm can be the direction indicated by two key points fixed relative to the palm, or the direction from the palm of the hand to the middle finger of the palm. The key point can be the connection point between two adjacent fingers. The calibration direction is the direction preset by the image acquisition device in the imaging plane, such as using the upward direction (or other arbitrarily specified direction) in the imaging plane as the calibration direction.

[0046] A palm image is an image captured by capturing the palm of the hand. It is understood that the capture angles may include frontal, oblique, and side angles. The palm region may be the area surrounding the palm in the palm image. Compared to the palm image, the palm region contains less redundancy beyond the palm.

[0047] In one embodiment, an image capture device may be deployed on the terminal. The user of a palm may place the same palm at different angles within the image capture device's field of view. The terminal may use the image capture device to capture images of the same palm placed above the image capture device at different angles, thereby obtaining multiple captured images of the same palm at different capture angles. The terminal may then send the multiple captured images to a server. The server may receive the multiple captured images sent by the terminal and directly use the multiple captured images as multiple palm images.

[0048] In one embodiment, as shown in FIG3 , the terminal can use an image capture device to capture images of the same palm placed at different angles above the image capture device, thereby obtaining multiple captured images of the same palm at different capture angles. It is understood that the same palm can be placed at eight different positions, 301 to 308, to obtain eight captured images of the same palm at different capture angles.

[0049] In one embodiment, a third-party storage device may store multiple captured images of the same palm at different capture angles. The server may directly obtain the multiple captured images of the same palm at different capture angles from the third-party storage device and use the multiple captured images directly as palm images.

[0050] In one embodiment, the server may obtain multiple captured images of the same palm at different capture angles and perform image resolution enhancement on the multiple captured images to obtain multiple palm images. It is understood that the resolution of the palm image is higher than the resolution of the captured image corresponding to the palm image.

[0051] In one embodiment, the server may obtain multiple captured images of the same palm from different capture angles and perform pixel interpolation on each of the multiple captured images to obtain multiple corresponding palm images. It is understood that the resolution of the palm image is higher than the resolution of the corresponding captured image. Interpolation may employ methods such as bilinear interpolation and bicubic interpolation. In this embodiment, interpolation calculations are performed between known pixel values ​​in the captured images to generate new pixel values, thereby achieving improved image resolution.

[0052] Step 204 , for each palm image, locate the palm key area included in the palm area of ​​the palm image, and determine the auxiliary area in the palm area except the palm key area, where the palm position in the palm key area is related to the acquisition angle of the palm image.

[0053] The palm key region is the area in the palm image that has unique biometric features. The auxiliary region is the area within the palm image that is not the palm key region. It can be understood that the palm key region in the palm image has more unique biometric features than the auxiliary region, is more resistant to forgery, and has higher recognition accuracy and security in the identity authentication field.

[0054] Specifically, for each of the multiple acquired palm images, the server may perform key area recognition on the palm image to locate the palm key area included in the palm area within the palm image. It will be appreciated that after locating the palm key area included in the palm area within the palm image, the area of ​​the palm image other than the palm key area becomes the auxiliary area included in the palm image.

[0055] In one embodiment, for each of the multiple acquired palm images, the server may extract image features of the palm image, and locate the palm key area included in the palm area within the palm image based on the extracted image features.

[0056] In one embodiment, for each of the multiple acquired palm images, the server may input the palm image into a pre-trained key region recognition model to extract image features of the palm image using the pre-trained key region recognition model. Based on the extracted image features, the server may locate and output the palm key region included in the palm region within the palm image. It will be understood that the key region recognition model is a target detection model capable of identifying the palm key region from a palm image.

[0057] In one embodiment, as shown in FIG4 , the palm key area of ​​the palm image may specifically include at least one of the palm area 402 in the palm area 401 or the area 403 between the fingers and the palm. The palm area 402 is the area within the palm area 401 in the palm image and includes the center of the palm. It can be understood that the palm is the central area of ​​the palm, and its skin has many unique lines and skin features, and the line features of the palm are relatively stable. In addition, the palm has a large number of sweat glands and sebaceous glands, and these secretions can provide additional biometric information during palm-based identity authentication. The area between the fingers and the palm includes the base connection area of ​​the thumb and the other four fingers, which is the transition area of ​​the palm biometric features. In these areas, including the skin on the back and palm sides of the fingers, there are unique lines and features.

[0058] To further understand how the palm area within the palm key region is related to the palm image acquisition angle, let's take an example. When the acquisition angle is a frontal angle, the palm key region within the palm region within the palm image is the palm center. When the acquisition angle is an oblique angle, the palm key region within the palm region within the palm image can be the area between the fingers and the palm center.

[0059] Step 206: extract features from the palm image, and in the process of feature extraction, assign different contribution weights to the palm key area and the auxiliary area in the palm image to obtain the single-angle palm feature of the palm image; wherein the contribution weight assigned to the palm key area in the palm image is higher than the contribution weight assigned to the auxiliary area in the palm image.

[0060] The single-angle palm feature is a feature extracted from a single palm image. It can be understood that a single palm image is any one of multiple palm images of the same palm captured at different angles.

[0061] Specifically, when extracting features from a palm image among multiple palm images, the server may assign a higher contribution weight to the key palm region in the palm image than to the auxiliary region, thereby obtaining a single-angle palm feature for the palm image. It is understood that when extracting features from a palm image among multiple palm images, the server will pay more attention to the key palm region in the palm image than to the auxiliary region in the palm image. It is understood that the single-angle palm feature can characterize the superficial features of the palm, specifically the basic structure of the palm in the palm image, as well as the local features of the palm at the angle at which the palm image was captured.

[0062] In one embodiment, the server may input a palm image into a pre-trained feature extraction model to extract features from the palm image. During the feature extraction process, the server assigns different contribution weights to the key palm region and auxiliary regions in the palm image to obtain single-angle palm features of the palm image. The contribution weight assigned to the key palm region is higher than the contribution weight assigned to the auxiliary regions. It can be understood that the feature extraction model is a neural network model used to extract features from an image, and it has the ability to extract single-angle palm features from a palm image.

[0063] Step 208: The single-angle palm features of the multiple palm images are fused to obtain multi-angle palm features of the palm; the multi-angle palm features are used for identity authentication.

[0064] Multi-angle palm features are derived by fusing the single-angle palm features of multiple palm images acquired at different angles. Multi-angle palm features characterize the deep features of the palm, specifically the palm structure and local features at multiple acquisition angles, thereby more comprehensively reflecting the palm's detailed characteristics. Multi-angle palm features are a combined representation of the single-angle palm features of multiple palm images, and they contain richer feature information than the single-angle palm features of any single palm image.

[0065] In one embodiment, the server may perform a weighted summation of the single-angle palm features of the multiple palm images to obtain the multi-angle palm features of the palm. It is understood that the server may assign corresponding weights to the single-angle palm features of the multiple palm images and weight the single-angle palm features of the multiple palm images based on the assigned weights to obtain the multi-angle palm features of the palm.

[0066] In one embodiment, the server may perform feature concatenation on the single-angle palm features of the multiple palm images to obtain the multi-angle palm features of the palm. It is understood that the server may concatenate the single-angle palm features of the multiple palm images to obtain the multi-angle palm features of the palm.

[0067] In one embodiment, the server may perform feature selection on the single-angle palm features of the multiple palm images to obtain the multi-angle palm features of the palm. It is understood that the server may select the single-angle palm features with the richest feature information from the single-angle palm features of the multiple palm images as the multi-angle palm features of the palm.

[0068] In the palm feature processing method for identity authentication, multiple palm images of the same palm captured at different angles are acquired. For each palm image, the palm key region included in the palm area of ​​the palm image is located, and auxiliary regions other than the palm key region are determined. The palm position in the palm key region is related to the acquisition angle of the palm image. Features are extracted from the palm image, and during the feature extraction process, different contribution weights are assigned to the palm key region and auxiliary regions in the palm image to obtain single-angle palm features of the palm image; wherein the contribution weight assigned to the palm key region in the palm image is higher than the contribution weight assigned to the auxiliary regions in the palm image. The single-angle palm features of the multiple palm images are fused to obtain multi-angle palm features of the palm, which can be used for identity authentication. Compared to traditional identity authentication methods based on single-angle palm images, the present application acquires multiple palm images of the same palm captured at different angles and locates the palm key region of the palm images. This approach places greater emphasis on the palm key region during feature extraction, thereby extracting more accurate single-angle palm features for the located palm image. Furthermore, by fusing the single-angle palm features of each palm image into richer and more accurate multi-angle palm features, and performing identity authentication based on the richer and more accurate multi-angle palm features, the accuracy of identity authentication can be improved, thereby avoiding the waste of hardware resources used to support the identity authentication function.

[0069] In one embodiment, obtaining multiple palm images of the same palm at different acquisition angles includes: obtaining multiple acquisition images of the same palm at different acquisition angles; performing super-resolution reconstruction on the multiple acquisition images respectively to obtain multiple palm images; the resolution of the palm images is higher than the resolution of the acquisition images corresponding to the palm images.

[0070] The captured image is an image obtained by capturing an image of the palm and has not been reconstructed by super-resolution.

[0071] Specifically, the server can obtain multiple captured images of the same palm from different capture angles and perform super-resolution reconstruction on each of the multiple captured images to obtain multiple palm images corresponding to the multiple captured images. It can be understood that the palm images obtained after super-resolution reconstruction have a higher resolution than the corresponding captured images before super-resolution reconstruction.

[0072] In one embodiment, for each collected image, when performing super-resolution reconstruction on the collected image, the server may assign the same contribution weight to each pixel area in the collected image, and reconstruct a palm image corresponding to the collected image.

[0073] In the above embodiment, by performing super-resolution reconstruction on multiple captured images respectively, multiple palm images are obtained accordingly, the resolution of the image is improved, and better quality images are obtained. In particular, more precise and accurate image reconstruction of complex textures and details in the image can be performed, effectively improving the image resolution, thereby further improving the accuracy of identity authentication, and further avoiding the waste of hardware resources used to support the identity authentication function.

[0074] In one embodiment, as shown in FIG5 , super-resolution reconstruction is performed on multiple collected images to obtain multiple palm images, including:

[0075] Step 502: for each captured image, determine a palm focus area included in the palm area of ​​the captured image, and determine a secondary focus area in the palm area of ​​the captured image except the palm focus area, wherein the palm position in the palm focus area is related to the capture angle of the captured image.

[0076] The palm region of interest is the area in the captured image obtained by capturing the palm that has unique biometric features. The secondary region of interest is the area in the captured image obtained by capturing the palm that does not include the palm region of interest. It can be understood that the palm region of interest in the captured image has more unique biometric features than the secondary region of interest, is more resistant to forgery, and has higher recognition accuracy and security in the identity authentication field.

[0077] Specifically, for each of the multiple acquired images, the server may perform region of interest recognition on the acquired image to locate the palm region of interest included in the palm region of the acquired image. It will be appreciated that after locating the palm region of interest included in the palm region of the acquired image, the region of interest in the acquired image other than the palm region of interest becomes the secondary region of interest included in the acquired image.

[0078] In one embodiment, for each of the multiple acquired images, the server may extract image features of the acquired image, and locate the palm focus area included in the palm area in the acquired image based on the extracted image features.

[0079] In one embodiment, for each of the multiple acquired images, the server may input the acquired image into a pre-trained region of interest recognition model. The pre-trained region of interest recognition model may be used to extract image features of the acquired image, and based on the extracted image features, the server may locate and output a palm region of interest included in the palm area within the acquired image. It will be understood that the region of interest recognition model is an object detection model capable of identifying a palm region of interest from an acquired image.

[0080] In one embodiment, the palm focus area may specifically include at least one of a palm center area, an area between fingers and a palm center, and the like.

[0081] Step 504 , when super-resolution reconstruction is performed on the captured image, different contribution weights are assigned to the palm focus area and the secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, and a palm image corresponding to the captured image is reconstructed.

[0082] Specifically, when super-resolution reconstructing multiple acquired images, the server assigns different contribution weights to the palm focus region and the secondary focus region, with the contribution weight assigned to the palm focus region being higher than the contribution weight assigned to the secondary focus region. An image corresponding to the acquired image is reconstructed and directly used as the palm image. It is understood that when super-resolution reconstructing the acquired image from multiple acquired images, the server will pay more attention to the palm focus region in the acquired image than to the secondary focus region in the acquired image.

[0083] In one embodiment, as shown in FIG6 , the server may input the captured image into a pre-trained super-resolution model to determine the palm focus region included in the palm area within the captured image through the super-resolution model. When performing super-resolution reconstruction on a captured image among multiple captured images, the server assigns a higher contribution weight to the palm focus region than to the secondary focus region within the captured image, thereby obtaining and outputting a palm image corresponding to the captured image. It will be understood that the super-resolution model is a neural network model used for super-resolution reconstruction, and it has the ability to reconstruct a high-resolution palm image from the captured image.

[0084] In one embodiment, as shown in FIG7 , the super-resolution model may include a convolution unit with 64 channels and a 9×9 kernel, an activation function, multiple residual units, a normalization unit, a convolution unit with 256 channels and a 3×3 kernel, and a pixel reconstruction unit. It will be appreciated that by inputting a captured image into the super-resolution model for super-resolution reconstruction, a palm image with a higher resolution than the input captured image can be output.

[0085] In the above embodiment, for each captured image, the palm region of interest and the secondary region of interest included in the palm area of ​​the captured image are determined. Because the palm region of interest in the captured image has richer and more unique biometric features than the secondary region of interest, different contribution weights are assigned to the palm region of interest and the secondary region of interest during super-resolution reconstruction of the captured image, with the contribution weight assigned to the palm region of interest being higher than the contribution weight assigned to the secondary region of interest. This results in a reconstructed palm image corresponding to the captured image, further improving the image resolution, thereby further improving the accuracy of identity authentication and further avoiding waste of hardware resources used to support identity authentication.

[0086] In one embodiment, when performing super-resolution reconstruction on a captured image, different contribution weights are assigned to a palm focus area and a secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, and a palm image corresponding to the captured image is reconstructed, including: when performing super-resolution reconstruction on the captured image, different contribution weights are assigned to a palm focus area and a secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, and an initial palm image corresponding to the captured image is reconstructed; an initial palm key area included in the palm area within the initial palm image is determined; the palm position in the initial palm key area is related to the acquisition angle of the initial palm image; and based on the initial palm image, local image enhancement processing is performed on the initial palm key area to obtain a palm image corresponding to the initial palm image.

[0087] The region obtained after enhancing the initial palm key region has more prominent texture details than the initial palm key region. The initial palm image is the image obtained after super-resolution reconstruction of the acquired image, before local image enhancement. The initial palm key region is the palm key region included in the palm region within the initial palm image.

[0088] Specifically, when performing super-resolution reconstruction on the captured image, the server may assign different contribution weights to the palm focus area and the secondary focus area, with the contribution weight assigned to the palm focus area being higher than the contribution weight assigned to the secondary focus area. This reconstructs an initial palm image corresponding to the captured image. It is understood that the resolution of the initial palm image is higher than the resolution of the captured image corresponding to the palm image. The server may determine an initial palm key region included in the palm area within the initial palm image and, based on the initial palm image, perform local image enhancement processing on the initial palm key region to obtain a palm image corresponding to the initial palm image.

[0089] In one embodiment, for each of the multiple initial palm images obtained, the server may extract image features of the initial palm image, and locate the initial palm key area included in the palm area within the initial palm image based on the extracted image features.

[0090] In one embodiment, for each of the multiple initial palm images obtained, the server may input the initial palm image into a pre-trained key area recognition model to extract image features of the initial palm image through the pre-trained key area recognition model, and locate and output the initial palm key area included in the palm area within the initial palm image based on the extracted image features.

[0091] In one embodiment, the local image enhancement processing may be implemented by local histogram equalization or by local adaptive contrast enhancement.

[0092] In the above embodiment, an initial palm key region is determined within the palm area of ​​an initial palm image obtained after super-resolution reconstruction, and local image enhancement processing is performed on the initial palm image to obtain a palm image corresponding to the initial palm image. Because the region obtained after enhancing the initial palm key region has more prominent texture details than the initial palm key region, the image quality of the palm image is further improved, thereby further improving the accuracy of identity authentication and further avoiding the waste of hardware resources used to support identity authentication.

[0093] In one embodiment, when performing super-resolution reconstruction on a captured image, different contribution weights are assigned to a palm focus area and a secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, and a palm image corresponding to the captured image is reconstructed, including: when performing super-resolution reconstruction on the captured image, different contribution weights are assigned to a palm focus area and a secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, and a palm image to be denoised corresponding to the captured image is reconstructed; based on the palm image to be denoised, noise distribution information of the palm image to be denoised is determined; image features of the palm image to be denoised are extracted, and according to the noise distribution information, noise features are eliminated from the image features to obtain image structural features; and image reconstruction is performed based on the image structural features to obtain a denoised palm image for the palm image to be denoised.

[0094] The palm image to be denoised is the image obtained by super-resolution reconstruction of the captured image before denoising. Noise features are the characteristics of the noise present in the palm image to be denoised. Image structural features are features that characterize the image structure of the palm image to be denoised. Noise distribution information is information indicating the distribution of noise in the palm image to be denoised. Noise distribution information can be obtained by detecting pixels where noise is present and recording their locations.

[0095] Specifically, when performing super-resolution reconstruction on a captured image, the server may assign different contribution weights to the palm focus area and the secondary focus area, with the contribution weight assigned to the palm focus area being higher than the contribution weight assigned to the secondary focus area. The reconstruction results in a palm image corresponding to the captured image to be denoised. It is understood that the resolution of the palm image to be denoised is higher than the resolution of the captured image corresponding to the palm image. The server may determine noise distribution information of the palm image to be denoised based on the palm image to be denoised. The server may extract image features of the palm image to be denoised and, based on the noise distribution information, remove noise features from the image features to obtain image structural features. Furthermore, the server may perform image reconstruction based on the image structural features to obtain a palm image after denoising the palm image to be denoised. It is understood that the denoised palm image has less noise and better image quality than the palm image to be denoised.

[0096] In one embodiment, the server can input the palm image to be denoised into a pre-trained image denoising model, so as to determine the noise distribution information of the palm image to be denoised based on the palm image to be denoised through the image denoising model, extract the image features of the palm image to be denoised, eliminate the noise features from the image features according to the noise distribution information, obtain the image structural features, and reconstruct the image based on the image structural features to obtain and output the denoised palm image of the palm image to be denoised.

[0097] In the above embodiment, the palm image to be denoised is obtained by super-resolution reconstruction, and the noise distribution information of the palm image to be denoised is determined. Then, by extracting the image features of the palm image to be denoised, the noise features are eliminated from the image features according to the noise distribution information to obtain the image structural features, and the image is reconstructed based on the image structural features to obtain the palm image after denoising the palm image to be denoised, which further improves the image quality of the palm image, thereby further improving the accuracy of identity authentication and further avoiding the waste of hardware resources used to support the identity authentication function.

[0098] In one embodiment, for each captured image, the server may determine a palm focus region included in the captured image and determine a secondary focus region within the palm region of the captured image, excluding the palm focus region. The location of the palm in the palm focus region is correlated with the acquisition angle of the captured image. When performing super-resolution reconstruction on the captured image, the server may assign different contribution weights to the palm focus region and the secondary focus region, with the contribution weight assigned to the palm focus region being higher than the contribution weight assigned to the secondary focus region. The server then reconstructs a palm image corresponding to the captured image to be denoised. Based on the palm image to be denoised, the server may determine noise distribution information for the palm image to be denoised. The server may extract image features of the palm image to be denoised and, based on the noise distribution information, remove noise features from the image features to obtain image structural features. The server may perform image reconstruction based on the image structural features to obtain an initial palm image after denoising the palm image to be denoised. The server may determine an initial palm key region included in the palm region of the initial palm image. The location of the palm in the initial palm key region is correlated with the acquisition angle of the initial palm image. The server may perform local image enhancement processing on the key areas of the initial palm based on the initial palm image to obtain a palm image corresponding to the initial palm image.

[0099] In one embodiment, for each captured image, the server may determine a palm focus region included in the palm region of the captured image and determine secondary focus regions within the palm region of the captured image, excluding the palm focus region. The location of the palm in the palm focus region is correlated with the acquisition angle of the captured image. During super-resolution reconstruction of the captured image, the server may assign different contribution weights to the palm focus region and the secondary focus region, with the contribution weight assigned to the palm focus region being higher than the contribution weight assigned to the secondary focus region, to reconstruct an initial palm image corresponding to the captured image. The server may determine an initial palm key region included in the palm region of the initial palm image. The location of the palm in the initial palm key region is correlated with the acquisition angle of the initial palm image. Based on the initial palm image, the server may perform local image enhancement processing on the initial palm key region to obtain a palm image to be denoised corresponding to the initial palm image. Based on the palm image to be denoised, the server may determine noise distribution information of the palm image to be denoised. The server may extract image features of the palm image to be denoised and, based on the noise distribution information, remove noise features from the image features to obtain image structural features. The server may perform image reconstruction based on the image structural features to obtain a denoised palm image for the palm image to be denoised.

[0100] In one embodiment, a palm image is reconstructed using a pre-trained super-resolution model; the method further includes: obtaining at least one first palm image pair; the first palm image pair includes a first sample palm image and a reference palm image; the reference palm image has a higher resolution than the corresponding first sample palm image; the first sample palm image carries a first attention tag, the first attention tag indicating a sample palm attention region included in the palm region of the first sample palm image; the palm position in the sample palm attention region is related to the acquisition angle of the first sample palm image; the palm region of the first sample palm image also includes a sample sub-attention region in addition to the sample palm attention region; the first attention tag is used to indicate that when super-resolution reconstruction is performed on the first sample palm image, different contribution weights are assigned to the sample palm attention region and the sample sub-attention region, and the contribution weight assigned to the sample palm attention region is higher than the contribution weight assigned to the sample sub-attention region; inputting the first sample palm image into the super-resolution model to be trained to obtain a reconstructed palm image; and training the super-resolution model to be trained based on the difference between the reconstructed palm image and the corresponding reference palm image to obtain a trained super-resolution model.

[0101] The first palm image pair is a palm image pair used to train the super-resolution model to be trained. The sample palm region of interest is a palm region of interest included in the palm region in the first sample palm image. The sample secondary region of interest is a secondary region of interest included in the palm region in the first sample palm image.

[0102] Specifically, the server may obtain at least one first palm image pair and input the first sample palm image from the first palm image pair into the super-resolution model to be trained, thereby obtaining a reconstructed palm image. Furthermore, the server may iteratively train the super-resolution model to be trained based on the difference between the reconstructed palm image and the corresponding reference palm image from the first palm image pair until an iteration stopping condition is met, thereby obtaining a trained super-resolution model.

[0103] In one embodiment, the iteration stopping condition may be that the number of iterations reaches a preset number of iterations, or that the difference between the reconstructed palm image and the corresponding reference palm image in the first palm image pair is less than a preset difference threshold.

[0104] In the above embodiment, by inputting a first sample palm image into a super-resolution model to be trained to obtain a reconstructed palm image, and then training the super-resolution model based on the difference between the reconstructed palm image and the corresponding reference palm image, a trained super-resolution model with high-resolution reconstruction capabilities can be obtained. Furthermore, reconstructing a palm image from the captured image based on the super-resolution model can further improve the image resolution, thereby further improving the accuracy of identity authentication and further avoiding the waste of hardware resources used to support identity authentication.

[0105] In one embodiment, the single-angle palm feature is extracted by a pre-trained feature extraction model; the method also includes: obtaining at least one set of second palm image pairs; the second palm image pair includes a label palm image and a second sample palm image; the label palm image carries a second focus label, the second focus label indicates the sample palm key area included in the palm area in the label palm image; the palm part in the sample palm key area is related to the acquisition angle of the label palm image; the second focus label is used to indicate that when extracting features from the label palm image, different contribution weights are assigned to the sample palm key area and the sample auxiliary area, and the contribution weight assigned to the sample palm key area is higher than the contribution weight assigned to the sample auxiliary area; the label palm image is input into the feature extraction model to be trained to extract the reference single-angle palm feature; the second palm image is input into the feature extraction model to be trained to extract the predicted single-angle palm feature; according to the difference between the predicted single-angle palm feature and the corresponding reference single-angle palm feature, the feature extraction model to be trained is trained to obtain a trained feature extraction model.

[0106] The second palm image pair is a palm image pair used to train the feature extraction model to be trained. The sample palm key region is a palm key region included in the palm region in the second sample palm image. The sample auxiliary region is an auxiliary region included in the palm region in the second sample palm image.

[0107] Specifically, the server may obtain at least one set of second palm image pairs and input the labeled palm images from the second palm image pairs into the feature extraction model to be trained to extract reference single-angle palm features. The server may input the second palm images into the feature extraction model to be trained to extract predicted single-angle palm features. The server may iteratively train the feature extraction model to be trained based on the difference between the predicted single-angle palm features and the corresponding reference single-angle palm features until an iteration stopping condition is met, thereby obtaining a fully trained feature extraction model.

[0108] In one embodiment, the iteration stopping condition may be that the number of iterations reaches a preset number of iterations, or that the difference between the predicted single-angle palm feature and the corresponding reference single-angle palm feature is less than a preset difference threshold.

[0109] In one embodiment, as shown in FIG8 , the feature extraction model includes a convolution unit, an activation function, and a pooling unit. The server may input the labeled palm image in the second palm image pair into the feature extraction model to be trained to extract reference single-angle palm features. The server may also input the second palm image into the feature extraction model to be trained to extract predicted single-angle palm features. Furthermore, the server may determine a loss value based on the difference between the predicted single-angle palm features and the corresponding reference single-angle palm features, and iteratively train the feature extraction model to be trained based on the loss value until an iteration stop condition is met, thereby obtaining a trained feature extraction model.

[0110] In one embodiment, the server may perform data augmentation on the labeled palm image and the second sample palm image in the second palm image pair, and use the augmented images as training data to train the feature extraction model to be trained. For example, data augmentation may include at least one of rotation, translation, or scaling. This can increase the diversity of training data, expand the training dataset, and improve the generalization capability of the model.

[0111] In one embodiment, the server may perform data preprocessing on the labeled palm image and the second sample palm image in the second palm image pair, and use the preprocessed images as training data to train the feature extraction model to be trained. For example, the data preprocessing method may specifically include at least one of grayscale conversion and histogram equalization. This can reduce the effects of lighting, noise, and other factors on image quality.

[0112] In the above embodiment, the labeled palm image is input into the feature extraction model to be trained to extract the reference single-angle palm feature, and the second palm image is input into the feature extraction model to be trained to extract the predicted single-angle palm feature. Then, according to the difference between the predicted single-angle palm feature and the corresponding reference single-angle palm feature, the feature extraction model to be trained is trained to obtain a trained feature extraction model with the ability to accurately improve image features, thereby further improving the accuracy of identity authentication and further avoiding the waste of hardware resources used to support the identity authentication function.

[0113] In one embodiment, single-angle palm features of multiple palm images are fused to obtain multi-angle palm features of the palm, including: determining the acquisition angles corresponding to the multiple palm images; assigning corresponding fusion weights to the single-angle palm features corresponding to the multiple palm images according to the acquisition angles corresponding to the multiple palm images; and fusing the single-angle palm features of the multiple palm images according to the corresponding fusion weights to obtain multi-angle palm features of the palm.

[0114] In one embodiment, the server may input the multiple palm images into a pre-trained image angle recognition model, so as to recognize and output the acquisition angles corresponding to the multiple palm images through the image angle recognition model.

[0115] In one embodiment, the server may determine the acquisition angles corresponding to the multiple palm images and assign corresponding fusion weights to the single-angle palm features corresponding to the multiple palm images based on the acquisition angles. The server may perform a weighted fusion of the single-angle palm features corresponding to the multiple palm images based on the corresponding fusion weights to obtain multi-angle palm features of the palm.

[0116] In one embodiment, the server may determine the acquisition angles corresponding to the multiple palm images and assign fusion weights to the single-angle palm features corresponding to the multiple palm images based on the acquisition angles. The server may perform feature concatenation on the single-angle palm features corresponding to the multiple palm images based on the fusion weights to obtain multi-angle palm features of the palm.

[0117] In one embodiment, the server may determine the acquisition angles corresponding to the multiple palm images, and assign fusion weights to the single-angle palm features corresponding to the multiple palm images based on the acquisition angles. The server may select multi-angle palm features of the palm from the single-angle palm features of the multiple palm images based on the fusion weights of the single-angle palm features corresponding to the multiple palm images.

[0118] In the above embodiment, since palm images collected at different collection angles will contain feature information of different richness, corresponding fusion weights are assigned to the single-angle palm features corresponding to the multiple palm images according to the collection angles corresponding to the multiple palm images, and the single-angle palm features of the multiple palm images are fused according to the corresponding fusion weights of the single-angle palm features corresponding to the multiple palm images to obtain multi-angle palm features of the palm. More accurate and richer multi-angle palm features can be obtained, thereby further improving the accuracy of identity authentication and further avoiding the waste of hardware resources used to support the identity authentication function.

[0119] In one embodiment, the multi-angle palm feature is obtained by fusing the single-angle palm features of each of the multiple palm images in a target fusion manner. The method further includes: obtaining multiple test palm images and a first reference category to which each test palm image belongs; the multiple test palm images are collected for different test palms, and each test palm is collected from a different collection angle; determining test single-angle palm features for the same test palm at different collection angles based on the test palm images collected for the same test palm at different collection angles; determining multiple candidate fusion methods; for each candidate fusion method, fusing the test single-angle palm features for the same test palm at different collection angles according to the candidate fusion method to obtain a test multi-angle palm feature corresponding to the candidate fusion method; inputting the test palm image and the test multi-angle palm feature into a pre-trained palm classification model, so that the pre-trained palm classification model predicts the category of the test palm image based on the test multi-angle palm feature, and obtains a first predicted category to which the test palm image corresponding to the candidate fusion method belongs; and determining a target fusion method from the multiple candidate fusion methods based on the difference between the first predicted category corresponding to each candidate fusion method and the first reference category.

[0120] The first reference category is the category to which the test palm image belongs. The first reference category to which each test palm image belongs may refer to the test palm corresponding to each test palm image. It is understood that a single palm can be considered a category, and multiple palm images of the same palm captured at different acquisition angles may belong to the same category.

[0121] Specifically, the server may obtain multiple test palm images and a first reference category to which each test palm image belongs. The server may perform feature extraction on each test palm image acquired at different acquisition angles for the same test palm, obtaining test single-angle palm features for the same test palm at different acquisition angles. The server may obtain multiple candidate fusion methods. For each candidate fusion method, the server may fuse the test single-angle palm features acquired at different acquisition angles for the same test palm according to the candidate fusion method to obtain a test multi-angle palm feature corresponding to the candidate fusion method. The server may input the test palm image and the test multi-angle palm feature corresponding to the candidate fusion method into a pre-trained palm classification model. The pre-trained palm classification model may then predict the category of the test palm image based on the test multi-angle palm feature, obtaining a first predicted category to which the test palm image belongs corresponding to the candidate fusion method. The server may determine a target fusion method from the multiple candidate fusion methods based on the difference between the first predicted category corresponding to each candidate fusion method and the corresponding first reference category. Furthermore, the server may fuse the single-angle palm features of the multiple palm images using the target fusion method to obtain a multi-angle palm feature.

[0122] In the above embodiment, by traversing multiple candidate fusion methods, selecting the target fusion method with the best fusion effect from the multiple candidate fusion methods, and then fusing the single-angle palm features of multiple palm images in the target fusion method, more accurate and richer multi-angle palm features can be obtained, thereby further improving the accuracy of identity authentication and further avoiding the waste of hardware resources used to support the identity authentication function.

[0123] In one embodiment, the method further includes: obtaining at least one third sample palm image and a second reference category to which the third sample palm image belongs; the third sample palm image is collected for different sample palms, and each sample palm is collected from a different collection angle; based on the third sample palm images collected at different collection angles for the same sample palm, determining sample single-angle palm features for different collection angles of the same sample palm; fusing the sample single-angle palm features for different collection angles of the same sample palm to obtain sample multi-angle palm features of the sample palm; inputting the third sample palm image and the sample multi-angle palm features into a palm classification model to be trained, so as to perform category prediction on the third sample palm image based on the sample multi-angle palm features by the palm classification model to be trained, and obtain a second predicted category to which the third sample palm image belongs; and training the palm classification model to be trained based on the difference between the second predicted category and the second reference category to obtain a trained palm classification model.

[0124] The second reference category is the category to which the third sample palm image belongs. The second reference category to which each third sample palm image belongs may refer to the sample palm corresponding to each third sample palm image.

[0125] Specifically, the server may obtain at least one third sample palm image and the second reference category to which the third sample palm image belongs. The server may perform feature extraction on each third sample palm image acquired at different acquisition angles for the same sample palm, thereby obtaining sample single-angle palm features for the same sample palm at different acquisition angles. The server may fuse the sample single-angle palm features acquired at different acquisition angles for the same sample palm to obtain sample multi-angle palm features for the sample palm. Furthermore, the server may input the third sample palm image and the sample multi-angle palm features into a palm classification model to be trained, so that the palm classification model to be trained can predict the category of the third sample palm image based on the sample multi-angle palm features, thereby obtaining the second predicted category to which the third sample palm image belongs. The server may iteratively train the palm classification model to be trained based on the difference between the second predicted category and the second reference category, until an iteration stopping condition is satisfied, thereby obtaining a trained palm classification model.

[0126] In one embodiment, the iteration stopping condition may be that the number of iterations reaches a preset number of iterations, or that the difference between the second predicted category and the second reference category is less than a preset difference threshold.

[0127] In the above embodiment, by inputting the third sample palm image and the sample multi-angle palm features into the palm classification model to be trained, the palm classification model to be trained predicts the category of the third sample palm image based on the sample multi-angle palm features, obtains the second predicted category to which the third sample palm image belongs, and trains the palm classification model to be trained based on the difference between the second predicted category and the second reference category to which the third sample palm image belongs, thereby obtaining a trained palm classification model with strong classification capabilities. Furthermore, the palm classification model can be used to determine a more suitable target fusion method from multiple candidate fusion methods, thereby further improving the accuracy of identity authentication and further avoiding the waste of hardware resources used to support identity authentication functions.

[0128] In one embodiment, the multi-angle palm features are associated with the user identity of the user to whom the palm belongs and stored in a palm feature library; the method further includes: obtaining a target palm image to be identified, and performing palm feature extraction on the target palm image to obtain a target palm feature; searching for a target multi-angle palm feature that meets a similarity condition with the target palm feature from each multi-angle palm feature stored in the palm feature library; when the target multi-angle palm feature is found, determining an identity authentication result based on the user identity associated with the target multi-angle palm feature.

[0129] The user identity is a character string used to uniquely identify the user, and the target palm image is a palm image to be identified and authenticated that is captured in an identity authentication scenario.

[0130] Specifically, in an identity authentication scenario, the terminal can capture the target palm image to be identified and authenticated using a locally deployed image acquisition device and send the target palm image to a server. The server can receive the target palm image sent by the terminal, extract palm features from the target palm image, obtain target palm features, and search for target multi-angle palm features that meet similarity conditions with the target palm features from the various multi-angle palm features stored in the palm feature library. Once the target multi-angle palm features are found, the server can generate an identity authentication result based on the user identity associated with the target multi-angle palm features. The identity authentication results include identity authentication passed and identity authentication failed.

[0131] In one embodiment, searching for a target multi-angle palm feature that satisfies a similarity condition with a target palm feature from among the multi-angle palm features stored in the palm feature library includes searching for a multi-angle palm feature with the greatest similarity with the target palm feature from among the multi-angle palm features stored in the palm feature library as the target multi-angle palm feature.

[0132] In one embodiment, as shown in Figure 9, in a palm-based identity authentication scenario, the processing process is primarily divided into two phases. The first phase is the palm feature processing phase, and the second phase is the authentication phase. Specifically, in the first phase, the server may obtain captured images of the same palm from different capture angles, sent by the terminal. For each captured image, the server may perform image quality enhancement processing on the captured image to obtain multiple high-quality palm images. It is understood that the image quality enhancement processing method may include at least one of super-resolution reconstruction, local image enhancement processing, or noise reduction processing. For each palm image, the server may locate the palm key area included in the palm region within the palm image and extract features from the palm image. During the feature extraction process, different contribution weights are assigned to the palm key area and auxiliary areas in the palm image to obtain single-angle palm features for the palm image; the contribution weight assigned to the palm key area in the palm image is higher than the contribution weight assigned to the auxiliary areas in the palm image. The server can then fuse the single-angle palm features of the multiple palm images to obtain multi-angle palm features of the palm, and associate the multi-angle palm features with the user identity of the user to whom the palm belongs and store them in the palm feature library. For the second stage, the terminal can use a locally deployed image acquisition device to capture the target palm image to be identified and authenticated, and send the target palm image to the server. The server can receive the target palm image sent by the terminal, extract the palm features of the target palm image, obtain the target palm features, and search for the target multi-angle palm features that meet the similarity conditions with the target palm features from the multi-angle palm features stored in the palm feature library. Once the target multi-angle palm feature is found, the server can generate an identity authentication result based on the user identity associated with the target multi-angle palm feature, and return the identity authentication result to the terminal.

[0133] In the above embodiment, palm features are extracted from the target palm image to be identified to obtain target palm features, and target multi-angle palm features that meet the similarity condition with the target palm features are searched from the multi-angle palm features stored in the palm feature library. When the target multi-angle palm features are found, the identity authentication result is determined based on the user identity identifier associated with the target multi-angle palm features, thereby further improving the accuracy of identity authentication and further avoiding the waste of hardware resources used to support the identity authentication function.

[0134] In one embodiment, the multi-angle palm feature is obtained by fusing the single-angle palm features of multiple palm images according to a predetermined target fusion method; the single-angle palm features of the multiple palm images are associated with the multi-angle palm features and stored in a palm feature library; the method also includes: when the target fusion method is updated, the single-angle palm features associated with the multiple palm images in the palm feature library are re-fused according to the updated fusion method to obtain the multi-angle palm features of the re-fused palm; and the multi-angle palm features in the palm feature library are updated to the re-fused multi-angle palm features.

[0135] Specifically, the server may monitor the fusion method. If it detects an update to the target fusion method, the server may re-fuse the single-angle palm features associated with multiple palm images in the palm feature library according to the updated fusion method to obtain the re-fused multi-angle palm features. Furthermore, the server may update the multi-angle palm features stored in the palm feature library to the re-fused multi-angle palm features.

[0136] It can be understood that since the multi-angle palm features are associated with the user identity of the user to whom the palm belongs and are stored in the palm feature library, and the single-angle palm features of each of the multiple palm images are associated with the multi-angle palm features and are stored in the palm feature library, it can be seen that the single-angle palm features, multi-angle palm features, and the user identity of the user to whom the palm belongs are associated and stored in the palm feature library.

[0137] In one embodiment, the server may use the user identity of the user to whom the palm belongs as a key, and the single-angle palm features and multi-angle palm features of the corresponding multiple palm images as values, and store them in the palm feature library in the form of key-value pairs.

[0138] In one embodiment, the server may use the user identity identifier of the user to whom the palm belongs as the key, and the compressed results of the single-angle palm features of the corresponding multiple palm images and the multi-angle palm features as the value, and store them in the palm feature library in the form of key-value pairs.

[0139] In one embodiment, the server can use the user identity of the user to whom the palm belongs and, through indexing technology, search the palm feature library for the compressed single-angle palm features and multi-angle palm features of the associated multiple palm images, and decompress the compressed results to obtain the single-angle palm features and multi-angle palm features of the multiple palm images.

[0140] In the above embodiment, when the target fusion method is updated, the single-angle palm features associated with multiple palm images in the palm feature library are re-fused according to the updated, more optimal fusion method to obtain re-fused multi-angle palm features of the palm. The multi-angle palm features in the palm feature library are then updated to the re-fused multi-angle palm features. Identity authentication is then performed based on the re-fused multi-angle palm features with richer characteristic information, further improving the accuracy of identity authentication and further avoiding the waste of hardware resources used to support the identity authentication function.

[0141] In one embodiment, the image acquisition device involved in this application may include a camera, optical components, an image sensor, a control system, and a data interface. The camera is responsible for capturing palm images at different acquisition angles. The camera should have a sufficient field of view and flexibility to capture clear palm images at different acquisition angles. The optical components include lenses and apertures, which are used to clearly image the palm at different acquisition angles. The image sensor is responsible for converting optical images into digital images. The sensor should have high resolution, high sensitivity, and a wide dynamic range to obtain clear palm images under various lighting conditions. The control system is responsible for controlling the operating status of the camera, the optical imaging parameters of the optical components, and the settings of the image sensor. The control system can support automatic or manual adjustment to obtain optimal image quality under different acquisition angles and lighting conditions. The data interface is responsible for transmitting the captured palm image data. The data interface should have sufficient bandwidth and low latency to transmit the palm image in a timely manner.

[0142] As shown in FIG10 , in one embodiment, a palm feature processing method for identity authentication is provided. This embodiment uses the method applied to the server 104 in FIG1 as an example for explanation. The method specifically includes the following steps:

[0143] Step 1002: Acquire multiple captured images of the same palm at different capture angles; for each captured image, input the captured image into a pre-trained super-resolution model to determine the palm focus area included in the palm area in the captured image through the pre-trained super-resolution model.

[0144] The palm position in the palm focus area is related to the acquisition angle of the acquired image; the palm area in the acquired image also includes a secondary focus area in addition to the palm focus area.

[0145] Step 1004 , when performing super-resolution reconstruction on the captured image, different contribution weights are assigned to the palm focus area and the secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, and the palm image to be denoised corresponding to the captured image is reconstructed.

[0146] Step 1006: Based on the palm image to be denoised, determine the noise distribution information of the palm image to be denoised; extract the image features of the palm image to be denoised, and remove the noise features from the image features according to the noise distribution information to obtain the image structure features.

[0147] Step 1008 : reconstructing the image based on the image structure feature to obtain an initial palm image after denoising the palm image to be denoised.

[0148] Step 1010: Determine an initial palm key area included in the palm area within the initial palm image; the palm position within the initial palm key area is related to the acquisition angle of the initial palm image.

[0149] Step 1012: Based on the initial palm image, perform local image enhancement processing on the initial palm key area to obtain a palm image corresponding to the initial palm image; the resolution of the palm image is higher than the resolution of the captured image corresponding to the palm image.

[0150] Step 1014: For each palm image, input the palm image into a pre-trained feature extraction model to locate the palm key area included in the palm area in the palm image through the pre-trained feature extraction model.

[0151] The palm position in the palm key area is related to the acquisition angle of the palm image; the palm image also includes an auxiliary area except the palm key area.

[0152] Step 1016: extract features from the palm image. During the feature extraction process, different contribution weights are assigned to the key palm area and the auxiliary palm area in the palm image to obtain single-angle palm features of the palm image. The contribution weight assigned to the key palm area in the palm image is higher than the contribution weight assigned to the auxiliary palm area in the palm image.

[0153] Step 1018 , obtaining a plurality of test palm images and the first reference category to which each test palm image belongs; the plurality of test palm images are collected for different test palms, and each test palm is collected from a different collection angle.

[0154] Step 1020 , based on the test palm images acquired at different acquisition angles for the same test palm, determine the test single-angle palm features at different acquisition angles for the same test palm.

[0155] Step 1022, determining multiple candidate fusion methods; for each candidate fusion method, according to the candidate fusion method, fusion of the test single-angle palm features of the same test palm at different acquisition angles is performed to obtain the test multi-angle palm features corresponding to the candidate fusion method.

[0156] In step 1024, the test palm image and the test multi-angle palm features are input into a pre-trained palm classification model, so that the pre-trained palm classification model can predict the category of the test palm image based on the test multi-angle palm features to obtain a first predicted category to which the test palm image corresponds to the candidate fusion method.

[0157] Step 1026 : Determine a target fusion method from the plurality of candidate fusion methods according to the difference between the first prediction category and the first reference category corresponding to each candidate fusion method.

[0158] Step 1028: Determine the acquisition angles corresponding to the multiple palm images; assign corresponding fusion weights to the single-angle palm features corresponding to the multiple palm images according to the acquisition angles corresponding to the multiple palm images.

[0159] Step 1030 , based on the corresponding fusion weights of the single-angle palm features corresponding to the multiple palm images, the single-angle palm features of the multiple palm images are fused in a target fusion manner to obtain multi-angle palm features of the palm.

[0160] The single-angle palm features, multi-angle palm features of the plurality of palm images, and the user identity identifiers of the users to whom the palms belong are associated and stored in a palm feature library.

[0161] Step 1032: Acquire a target palm image to be identified, and perform palm feature extraction on the target palm image to obtain target palm features.

[0162] Step 1034 : Search the multi-angle palm features stored in the palm feature library for a target multi-angle palm feature that satisfies a similarity condition with the target palm feature.

[0163] Step 1036: When the target multi-angle palm feature is found, the identity authentication result is determined according to the user identity identifier associated with the target multi-angle palm feature.

[0164] Step 1038, when the target fusion mode is updated, the single-angle palm features associated with the multiple palm images in the palm feature library are re-fused according to the updated fusion mode to obtain the multi-angle palm features of the re-fused palm.

[0165] Step 1040 : updating the multi-angle palm features in the palm feature library to the re-fused multi-angle palm features, and performing identity authentication based on the re-fused multi-angle palm features.

[0166] The present application also provides an application scenario that applies the above-mentioned palm feature processing method for identity authentication. Specifically, as shown in FIG11 , the palm feature processing method for identity authentication can be applied to a scenario in which palm authentication is used to log into an instant messaging application. It can be understood that the user identity identifier in the present application includes an application login identifier for logging into an instant messaging application, and the multi-angle palm features are used to perform identity authentication when logging into the instant messaging application, so that the user automatically logs into the instant messaging application after successful palm authentication, without the need to manually enter the application login identifier to log into the instant messaging application.

[0167] Specifically, the server can obtain multiple captured images of the same palm at different capture angles. For each captured image, the server inputs the captured image into a pre-trained super-resolution model to determine the palm focus region included in the palm area within the captured image, and to determine the secondary focus region within the palm area of ​​the captured image other than the palm focus region. The palm position in the palm focus region is related to the capture angle of the captured image. When super-resolution reconstruction is performed on the captured image, different contribution weights are assigned to the palm focus region and the secondary focus region, and the contribution weight assigned to the palm focus region is higher than the contribution weight assigned to the secondary focus region. The palm image to be denoised corresponding to the captured image is reconstructed.

[0168] The server can determine noise distribution information of the palm image to be denoised based on the palm image to be denoised; extract image features of the palm image to be denoised, and remove noise features from the image features according to the noise distribution information to obtain image structural features. Image reconstruction is performed based on the image structural features to obtain an initial palm image after denoising the palm image to be denoised. An initial palm key region included in the palm area within the initial palm image is determined; the palm position within the initial palm key region is related to the acquisition angle of the initial palm image. Based on the initial palm image, local image enhancement processing is performed on the initial palm key region to obtain a palm image corresponding to the initial palm image; the resolution of the palm image is higher than the resolution of the acquired image corresponding to the palm image.

[0169] For each palm image, the server may input the palm image into a pre-trained feature extraction model to locate the palm key region included in the palm area within the palm image through the pre-trained feature extraction model, and determine the auxiliary region of the palm area other than the palm key region, where the palm position in the palm key region is related to the acquisition angle of the palm image. Features are extracted from the palm image, and in the feature extraction process, different contribution weights are assigned to the palm key region and the auxiliary region in the palm image to obtain a single-angle palm feature of the palm image; wherein the contribution weight assigned to the palm key region in the palm image is higher than the contribution weight assigned to the auxiliary region in the palm image.

[0170] The server can obtain multiple test palm images and a first reference category to which each test palm image belongs; the multiple test palm images are collected for different test palms, and each test palm is collected from different collection angles. Based on the test palm images collected for the same test palm at different collection angles, test single-angle palm features for the same test palm at different collection angles are determined. Multiple candidate fusion methods are determined; for each candidate fusion method, the test single-angle palm features for the same test palm at different collection angles are fused according to the candidate fusion method to obtain test multi-angle palm features corresponding to the candidate fusion method. The test palm image and the test multi-angle palm features are input into a pre-trained palm classification model, so that the pre-trained palm classification model predicts the category of the test palm image based on the test multi-angle palm features, and obtains the first predicted category to which the test palm image corresponding to the candidate fusion method belongs. Based on the difference between the first predicted category corresponding to each candidate fusion method and the first reference category, a target fusion method is determined from the multiple candidate fusion methods.

[0171] The server can determine the acquisition angles corresponding to the multiple palm images; assign fusion weights to the single-angle palm features corresponding to the multiple palm images based on the acquisition angles corresponding to the multiple palm images; and fuse the single-angle palm features of the multiple palm images in a target fusion manner based on the fusion weights corresponding to the single-angle palm features of the multiple palm images to obtain multi-angle palm features of the palms. The single-angle palm features of the multiple palm images, the multi-angle palm features, and the application login identifier of the user to whom the palm belongs are associated and stored in a palm feature library.

[0172] Continuing with FIG11 , an instant messaging application is running on terminal 1101. A user can place a target palm 1102 above the image capture device of terminal 1101 to capture an image of the target palm to be identified. Terminal 1101 can send the target palm image to a server. The server can receive the target palm image and perform palm feature extraction on the target palm image to obtain target palm features. The server then searches for target multi-angle palm features that meet similarity requirements with the target palm features from the various multi-angle palm features stored in the palm feature library. Once the target multi-angle palm features are found, an identity authentication result is determined based on the application login identifier associated with the target multi-angle palm features. If the identity authentication result indicates successful palm authentication, the user is automatically logged into the instant messaging application, eliminating the need to manually enter the application login identifier to log into the instant messaging application.

[0173] In this way, by acquiring multiple palm images of the same palm from different acquisition angles and locating the key palm areas of the palm images, the key palm areas are given greater attention during feature extraction to extract more accurate single-angle palm features for the located palm images. Furthermore, by fusing the single-angle palm features of each palm image into richer and more accurate multi-angle palm features, and performing identity authentication when logging into an instant messaging application based on these richer and more accurate multi-angle palm features, the accuracy of identity authentication for logging into the instant messaging application can be improved, thereby improving the efficiency of palm authentication login for the instant messaging application and avoiding the waste of hardware resources used to support the identity authentication function for logging into the instant messaging application.

[0174] This application also provides an application scenario that utilizes the aforementioned palm feature processing method for identity authentication. Specifically, as shown in FIG12 , this palm feature processing method for identity authentication can be applied to a scenario where palm authentication is used to implement resource transfer. It will be understood that the user identity identifier in this application includes a resource transfer identifier for resource transfer, and the multi-angle palm features are used to perform identity authentication during resource transfer, so that resources are automatically transferred after successful palm authentication, eliminating the need to manually enter the resource transfer identifier to implement resource transfer. Specifically, by acquiring multiple palm images of the same palm at different acquisition angles and locating the key palm area of ​​the palm image, feature extraction focuses more on the key palm area to extract more accurate single-angle palm features for the located palm image. Furthermore, by fusing the single-angle palm features of each palm image into richer and more accurate multi-angle palm features, referring again to FIG12 , the terminal 1201 is provided with an image acquisition device. The user can place the target palm 1202 within the acquisition field of view of the image acquisition device of the terminal 1201 to acquire the target palm image to be identified. Terminal 1201 can send the target palm image to be identified to the server. The server can perform identity authentication during resource transfer based on the target palm image and richer and more accurate multi-angle palm features, which can improve the accuracy of identity authentication for resource transfer, thereby improving the efficiency of resource transfer and avoiding the waste of hardware resources used to support the identity authentication function for resource transfer.

[0175] It should be understood that, although the various steps in the flow chart of the above-mentioned embodiments are shown in sequence, these steps are not necessarily performed in sequence. Unless clearly stated herein, the execution of these steps does not have strict order restrictions, and these steps can be performed in other sequences. Moreover, at least a portion of the steps in the above-mentioned embodiments may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0176] In one embodiment, as shown in FIG13 , a palm feature processing device 1300 for identity authentication is provided, which specifically includes:

[0177] An acquisition module 1302 is configured to acquire multiple palm images of the same palm at different acquisition angles, wherein the area occupied by the palm in each palm image constitutes a palm area;

[0178] a positioning module 1304 configured to locate, for each palm image, a palm key region included in the palm region of the palm image, and determine auxiliary regions in the palm region other than the palm key region, wherein the palm position in the palm key region is related to the acquisition angle of the palm image;

[0179] The extraction module 1306 is configured to extract features from the palm image and assign different contribution weights to the key palm region and the auxiliary palm region in the palm image during the feature extraction process to obtain single-angle palm features of the palm image; wherein the contribution weight assigned to the key palm region is higher than the contribution weight assigned to the auxiliary palm region;

[0180] The fusion module 1308 is used to fuse the single-angle palm features of multiple palm images to obtain multi-angle palm features of the palm; the multi-angle palm features are used for identity authentication.

[0181] In one embodiment, the acquisition module 1302 is further used to acquire multiple captured images of the same palm at different acquisition angles; super-resolution reconstruction is performed on the multiple captured images respectively to obtain multiple palm images; the resolution of the palm image is higher than the resolution of the captured image corresponding to the palm image.

[0182] In one embodiment, the acquisition module 1302 is further used to determine, for each acquired image, a palm focus area included in the palm area of ​​the acquired image, and to determine a secondary focus area in the palm area of ​​the acquired image other than the palm focus area, where the palm position in the palm focus area is related to the acquisition angle of the acquired image; when performing super-resolution reconstruction on the acquired image, different contribution weights are assigned to the palm focus area and the secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, so as to reconstruct a palm image corresponding to the acquired image.

[0183] In one embodiment, the acquisition module 1302 is also used to assign different contribution weights to the palm focus area and the secondary focus area when performing super-resolution reconstruction on the acquired image, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, so as to reconstruct an initial palm image corresponding to the acquired image; determine the initial palm key area included in the palm area within the initial palm image; the palm position in the initial palm key area is related to the acquisition angle of the initial palm image; based on the initial palm image, perform local image enhancement processing on the initial palm key area to obtain a palm image corresponding to the initial palm image.

[0184] In one embodiment, the acquisition module 1302 is further used to assign different contribution weights to the palm focus area and the secondary focus area when performing super-resolution reconstruction on the acquired image, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, so as to reconstruct a palm image to be denoised corresponding to the acquired image; determine the noise distribution information of the palm image to be denoised based on the palm image to be denoised; extract image features of the palm image to be denoised, and eliminate noise features from the image features according to the noise distribution information to obtain image structural features; and reconstruct the image based on the image structural features to obtain a denoised palm image for the palm image to be denoised.

[0185] In one embodiment, the palm image is reconstructed using a pre-trained super-resolution model. As shown in FIG14 , the palm feature processing device 1300 for identity authentication further includes:

[0186] The first training module 1310 is configured to obtain at least one first palm image pair; the first palm image pair includes a first sample palm image and a reference palm image; the reference palm image has a higher resolution than the corresponding first sample palm image; the first sample palm image carries a first attention tag, the first attention tag indicating a sample palm attention region included in the palm region of the first sample palm image; the palm position in the sample palm attention region is related to the acquisition angle of the first sample palm image; the palm region of the first sample palm image also includes a sample secondary attention region in addition to the sample palm attention region; the first attention tag is configured to indicate that, when super-resolution reconstruction is performed on the first sample palm image, different contribution weights are assigned to the sample palm attention region and the sample secondary attention region, and the contribution weight assigned to the sample palm attention region is higher than the contribution weight assigned to the sample secondary attention region; the first sample palm image is input into the super-resolution model to be trained to obtain a reconstructed palm image; and the super-resolution model to be trained is trained based on the difference between the reconstructed palm image and the corresponding reference palm image to obtain a trained super-resolution model.

[0187] In one embodiment, the single-angle palm feature is extracted using a pre-trained feature extraction model. As shown in FIG14 , the palm feature processing device 1300 for identity authentication further includes:

[0188] The second training module 1312 is used to obtain at least one set of second palm image pairs; the second palm image pair includes a label palm image and a second sample palm image; the label palm image carries a second focus label, and the second focus label indicates the sample palm key area included in the palm area of ​​the label palm image; the palm part in the sample palm key area is related to the acquisition angle of the label palm image; the second focus label is used to indicate that when extracting features from the label palm image, different contribution weights are assigned to the sample palm key area and the sample auxiliary area, and the contribution weight assigned to the sample palm key area is higher than the contribution weight assigned to the sample auxiliary area; the label palm image is input into the feature extraction model to be trained to extract the reference single-angle palm feature; the second palm image is input into the feature extraction model to be trained to extract the predicted single-angle palm feature; according to the difference between the predicted single-angle palm feature and the corresponding reference single-angle palm feature, the feature extraction model to be trained is trained to obtain a trained feature extraction model.

[0189] In one embodiment, the fusion module 1308 is also used to determine the acquisition angles corresponding to the multiple palm images; assign corresponding fusion weights to the single-angle palm features corresponding to the multiple palm images according to the acquisition angles corresponding to the multiple palm images; and fuse the single-angle palm features of the multiple palm images according to the fusion weights corresponding to the single-angle palm features corresponding to the multiple palm images to obtain multi-angle palm features of the palm.

[0190] In one embodiment, the multi-angle palm feature is obtained by fusing the single-angle palm features of multiple palm images in a target fusion manner. As shown in FIG14 , the palm feature processing device 1300 for identity authentication further includes:

[0191] Determination module 1314 is used to obtain multiple test palm images and a first reference category to which each test palm image belongs; the multiple test palm images are collected for different test palms, and for each test palm, they are collected from different collection angles; based on the test palm images collected for the same test palm at different collection angles, test single-angle palm features for different collection angles of the same test palm are determined; multiple candidate fusion methods are determined; for each candidate fusion method, the test single-angle palm features for different collection angles of the same test palm are fused according to the candidate fusion method to obtain test multi-angle palm features corresponding to the candidate fusion method; the test palm image and the test multi-angle palm features are input into a pre-trained palm classification model, so that the pre-trained palm classification model predicts the category of the test palm image based on the test multi-angle palm features to obtain the first predicted category to which the test palm image corresponding to the candidate fusion method belongs; and according to the difference between the first predicted category corresponding to each candidate fusion method and the first reference category, a target fusion method is determined from the multiple candidate fusion methods.

[0192] In one embodiment, as shown in FIG14 , the palm feature processing device 1300 for identity authentication further includes:

[0193] The third training module 1316 is used to obtain at least one third sample palm image and a second reference category to which the third sample palm image belongs; the third sample palm images are collected for different sample palms, and each sample palm is collected from a different collection angle; based on the third sample palm images collected at different collection angles for the same sample palm, sample single-angle palm features for different collection angles of the same sample palm are determined; the sample single-angle palm features for different collection angles of the same sample palm are fused to obtain sample multi-angle palm features of the sample palm; the third sample palm image and the sample multi-angle palm features are input into the palm classification model to be trained, so that the palm classification model to be trained predicts the category of the third sample palm image based on the sample multi-angle palm features to obtain the second predicted category to which the third sample palm image belongs; and the palm classification model to be trained is trained according to the difference between the second predicted category and the second reference category to obtain a trained palm classification model.

[0194] In one embodiment, the multi-angle palm features are associated with the user identity identifier of the user to whom the palm belongs and stored in a palm feature library; as shown in FIG14 , the palm feature processing device 1300 for identity authentication further includes:

[0195] The authentication module 1318 is used to obtain a target palm image to be identified, perform palm feature extraction on the target palm image, and obtain target palm features; search for a target multi-angle palm feature that meets a similarity condition with the target palm feature from the various multi-angle palm features stored in the palm feature library; and when the target multi-angle palm feature is found, determine the identity authentication result based on the user identity identifier associated with the target multi-angle palm feature.

[0196] In one embodiment, the multi-angle palm feature is obtained by fusing the single-angle palm features of multiple palm images according to a predetermined target fusion method; the single-angle palm features of the multiple palm images are associated with the multi-angle palm features and stored in a palm feature library; the fusion module 1308 is also used to, when the target fusion method is updated, re-fuse the single-angle palm features associated with the multiple palm images in the palm feature library according to the updated fusion method to obtain the multi-angle palm features of the re-fused palm; and update the multi-angle palm features in the palm feature library to the re-fused multi-angle palm features.

[0197] The palm feature processing device for identity authentication acquires multiple palm images of the same palm captured at different angles. For each palm image, it locates the palm key region within the palm image and determines the auxiliary region, excluding the palm key region, of the palm area. The palm position within the palm key region is related to the angle of capture of the palm image. Features are extracted from the palm image, and during the feature extraction process, different contribution weights are assigned to the palm key region and the auxiliary region within the palm image to obtain single-angle palm features for the palm image. The contribution weight assigned to the palm key region is higher than the contribution weight assigned to the auxiliary region within the palm image. The single-angle palm features of the multiple palm images are fused to obtain multi-angle palm features of the palm. The multi-angle palm features can be used for identity authentication. Compared to traditional identity authentication methods based on single-angle palm images, the present application acquires multiple palm images of the same palm captured at different angles and locates the palm key region within the palm images. This prioritizes the palm key region during feature extraction, thereby extracting more accurate single-angle palm features for the located palm image. Furthermore, by fusing the single-angle palm features of each palm image into richer and more accurate multi-angle palm features, and performing identity authentication based on the richer and more accurate multi-angle palm features, the accuracy of identity authentication can be improved, thereby avoiding the waste of hardware resources used to support the identity authentication function.

[0198] Each module in the aforementioned palm feature processing device for identity authentication may be implemented in whole or in part through software, hardware, or a combination thereof. Each module may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a computer device's memory in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0199] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in Figure 15. The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a palm feature processing method for identity authentication is implemented.

[0200] Those skilled in the art will understand that the structure shown in FIG15 is merely a block diagram of a portion of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0201] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0202] In one embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.

[0203] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0204] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions.

[0205] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0206] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0207] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A palm feature processing method for identity authentication, executed by a computer device, the method comprising: Acquire multiple palm images of the same palm at different acquisition angles, wherein the area occupied by the palm in each palm image constitutes a palm area; For each of the palm images, locate a palm key area included in the palm area of ​​the palm image, and determine an auxiliary area in the palm area except the palm key area, wherein the palm position in the palm key area is related to the acquisition angle of the palm image; Extracting features from the palm image, and in the process of extracting features, assigning different contribution weights to the palm key area and the auxiliary area in the palm image to obtain single-angle palm features of the palm image, wherein the contribution weight assigned to the palm key area in the palm image is higher than the contribution weight assigned to the auxiliary area in the palm image; and The single-angle palm features of the multiple palm images are fused to obtain multi-angle palm features of the palm; the multi-angle palm features are used for identity authentication.

2. According to the method of claim 1, the step of acquiring multiple palm images of the same palm at different acquisition angles comprises: Acquire multiple images of the same palm at different acquisition angles; Super-resolution reconstruction is performed on the multiple collected images respectively, and a plurality of palm images are correspondingly obtained, and the resolution of the palm images is higher than the resolution of the collected images corresponding to the palm images.

3. The method according to claim 2, wherein the super-resolution reconstruction is performed on the multiple collected images respectively to obtain multiple palm images, comprising: For each of the acquired images, determine a palm focus area included in a palm area in the acquired image, and determine a secondary focus area in the palm area in the acquired image except the palm focus area, wherein the palm position in the palm focus area is related to the acquisition angle of the acquired image; When super-resolution reconstruction is performed on the acquired image, different contribution weights are assigned to the palm focus area and the secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, so as to reconstruct a palm image corresponding to the acquired image.

4. The method according to claim 3, wherein when performing super-resolution reconstruction on the acquired image, different contribution weights are assigned to the palm focus area and the secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, and the palm image corresponding to the acquired image is reconstructed, comprising: When performing super-resolution reconstruction on the acquired image, different contribution weights are assigned to the palm focus area and the secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, so as to reconstruct an initial palm image corresponding to the acquired image; Determine an initial palm key area included in the palm area in the initial palm image, where the palm position in the initial palm key area is related to the acquisition angle of the initial palm image; Based on the initial palm image, local image enhancement processing is performed on the initial palm key area to obtain a palm image corresponding to the initial palm image.

5. The method according to claim 3 or 4, wherein when performing super-resolution reconstruction on the acquired image, different contribution weights are assigned to the palm focus area and the secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, and the palm image corresponding to the acquired image is reconstructed, comprising: When super-resolution reconstruction is performed on the collected image, different contribution weights are assigned to the palm focus area and the secondary focus area, and the contribution weight assigned to the palm focus area is higher than the contribution weight assigned to the secondary focus area, so as to reconstruct a palm image to be denoised corresponding to the collected image; Based on the palm image to be denoised, determining noise distribution information of the palm image to be denoised; Extracting image features of the palm image to be denoised, and removing noise features from the image features according to the noise distribution information to obtain image structural features; Perform image reconstruction based on the image structural features to obtain a palm image after denoising the palm image to be denoised image.

6. According to the method according to any one of claims 3 to 5, the palm image is reconstructed by a pre-trained super-resolution model, and the method further comprises: Acquire at least one set of first palm image pairs, the first palm image pair comprising a first sample palm image and a reference palm image, the reference palm image having a higher resolution than the corresponding first sample palm image, the first sample palm image carrying a first attention tag, the first attention tag indicating a sample palm attention area included in a palm area in the first sample palm image, the palm position in the sample palm attention area being related to an acquisition angle of the first sample palm image, the palm area in the first sample palm image also comprising a sample secondary attention area in addition to the sample palm attention area, the first attention tag being used to indicate that when super-resolution reconstruction is performed on the first sample palm image, different contribution weights are assigned to the sample palm attention area and the sample secondary attention area, and the contribution weight assigned to the sample palm attention area is higher than the contribution weight assigned to the sample secondary attention area; Inputting the first sample palm image into the super-resolution model to be trained to obtain a reconstructed palm image; The super-resolution model to be trained is trained according to the difference between the reconstructed palm image and the corresponding reference palm image to obtain a trained super-resolution model.

7. The method according to any one of claims 1 to 6, wherein the single-angle palm feature is extracted by a pre-trained feature extraction model, and the method further comprises: Acquire at least one set of second palm image pairs, wherein the second palm image pair includes a label palm image and a second sample palm image, wherein the label palm image carries a second attention label, wherein the second attention label indicates a sample palm key area included in a palm area in the label palm image, wherein a palm position in the sample palm key area is related to an acquisition angle of the label palm image, wherein the palm area in the label palm image also includes a sample auxiliary area in addition to the sample palm key area, and wherein the second attention label is used to indicate that when extracting features from the label palm image, different contribution weights are assigned to the sample palm key area and the sample auxiliary area, and the contribution weight assigned to the sample palm key area is higher than the contribution weight assigned to the sample auxiliary area; Inputting the labeled palm image into the feature extraction model to be trained to extract reference single-angle palm features; Inputting the second palm image into the feature extraction model to be trained to extract predicted single-angle palm features; The feature extraction model to be trained is trained according to the difference between the predicted single-angle palm feature and the corresponding reference single-angle palm feature to obtain a trained feature extraction model.

8. The method according to any one of claims 1 to 7, wherein the fusing of the single-angle palm features of the plurality of palm images to obtain the multi-angle palm features of the palm comprises: Determine the acquisition angles corresponding to the multiple palm images respectively; According to the acquisition angles corresponding to the multiple palm images, assigning corresponding fusion weights to the single-angle palm features corresponding to the multiple palm images respectively; According to the fusion weights corresponding to the single-angle palm features respectively corresponding to the multiple palm images, the single-angle palm features of the multiple palm images are fused to obtain the multi-angle palm features of the palm.

9. The method according to any one of claims 1 to 8, wherein the multi-angle palm feature is obtained by fusing the single-angle palm features of the plurality of palm images in a target fusion manner, and the method further comprises: Acquire a plurality of test palm images and a first reference category to which each of the test palm images belongs, wherein the plurality of test palm images are acquired for different test palms, and each test palm is acquired from a different acquisition angle; Determine the test single-angle palm features at different acquisition angles for the same test palm based on the test palm images acquired at different acquisition angles for the same test palm; Identify multiple candidate fusion methods; For each of the candidate fusion modes, according to the candidate fusion mode, the test single-angle palm features of the same test palm at different acquisition angles are fused to obtain the test multi-angle palm features corresponding to the candidate fusion mode; The test palm image and the test multi-angle palm features are input into a pre-trained palm classification model, so that the test palm classification model can be used to classify the test palm image based on the test multi-angle palm features. Performing category prediction on the palm image to obtain a first predicted category to which the test palm image corresponding to the candidate fusion mode belongs; A target fusion mode is determined from the plurality of candidate fusion modes according to a difference between the first prediction category and the first reference category respectively corresponding to each of the candidate fusion modes.

10. The method according to claim 9, further comprising: Acquire at least one third sample palm image and a second reference category to which the third sample palm image belongs, wherein the third sample palm image is collected for different sample palms, and each sample palm is collected from a different collection angle; Determine sample single-angle palm features at different acquisition angles for the same sample palm based on the third sample palm images acquired at different acquisition angles for the same sample palm; The sample single-angle palm features collected at different angles for the same sample palm are fused to obtain the sample multi-angle palm features of the sample palm; Inputting the third sample palm image and the sample multi-angle palm features into a palm classification model to be trained, so as to perform category prediction on the third sample palm image based on the sample multi-angle palm features through the palm classification model to be trained, and obtain a second predicted category to which the third sample palm image belongs; The palm classification model to be trained is trained according to the difference between the second predicted category and the second reference category to obtain a trained palm classification model.

11. The method according to any one of claims 1 to 10, wherein the multi-angle palm features are associated with a user identity identifier of a user to which the palm belongs and stored in a palm feature library, the method further comprising: Acquire a target palm image to be identified, and perform palm feature extraction on the target palm image to obtain target palm features; Searching for a target multi-angle palm feature that satisfies a similarity condition with the target palm feature from each of the multi-angle palm features stored in the palm feature library; When the target multi-angle palm feature is found, the identity authentication result is determined according to the user identity identifier associated with the target multi-angle palm feature.

12. The method according to claim 11, wherein the multi-angle palm feature is obtained by fusing the single-angle palm features of the plurality of palm images according to a predetermined target fusion method, and the single-angle palm features of the plurality of palm images are associated with the multi-angle palm feature and stored in the palm feature library, and the method further comprises: In the case where the target fusion mode is updated, the single-angle palm features associated with each of the plurality of palm images in the palm feature library are re-fused according to the updated fusion mode to obtain the re-fused multi-angle palm features of the palm; The multi-angle palm features in the palm feature library are updated to the re-fused multi-angle palm features.

13. A palm feature processing device for identity authentication, the device comprising: An acquisition module is used to acquire multiple palm images of the same palm at different acquisition angles, wherein the area occupied by the palm in each palm image constitutes a palm area; a positioning module, for locating, for each of the palm images, a palm key area included in the palm region of the palm image, and determining an auxiliary area in the palm region other than the palm key area, wherein the palm position in the palm key area is related to the acquisition angle of the palm image; an extraction module, configured to extract features from the palm image, and in the process of extracting features, assign different contribution weights to the palm key area and the auxiliary area in the palm image to obtain single-angle palm features of the palm image, wherein the contribution weight assigned to the palm key area in the palm image is higher than the contribution weight assigned to the auxiliary area in the palm image; and The fusion module is used to fuse the single-angle palm features of the multiple palm images to obtain the multi-angle palm features of the palm, and the multi-angle palm features are used for identity authentication.

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

15. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the method according to any one of claims 1 to 12.

16. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 12 are implemented.

Images