Image processing method and device, equipment and storage medium

Abstract

The embodiment of the invention provides an image processing method and device, equipment and a storage medium, which can be applied to various scenes such as artificial intelligence, cloud technology, smart traffic and vehicle-mounted, and the method comprises the following steps: in a process of carrying out identity authentication based on a collected palm image in response to a palm brushing authentication operation, carrying out identity authentication on the basis of the palm image; meanwhile, the target age group of the target object is recognized through the palm image, the brush palm application is controlled to work in the operation mode matched with the target age group so as to meet the actual requirements of different age groups, mode switching of the brush palm application does not depend on active triggering of the target object any more, self-adaption of mode switching is achieved, and the user experience is improved. Therefore, the mode switching efficiency of the brush palm application is improved. And for the elderly user who is not familiar with the mode switching operation, the corresponding operation mode is adaptively switched on the basis of the identified age group of the elderly user, and help of an aging mechanism is provided for the elderly user, so that the use experience of the elderly user is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of artificial intelligence technology, and in particular to an image processing method, apparatus, device and storage medium. Background Art

[0002] With the development of information technology, information security and identity authentication are becoming more and more important. Compared with traditional authentication technologies, biometric recognition technology can provide more secure, reliable and fast identity authentication results, and is therefore widely used in various applications. Among them, palm scanning applications are applications that achieve identity authentication through palm vein recognition or palm print recognition.

[0003] As more and more elderly users use palm-swiping applications, palm-swiping applications have also introduced various elderly-friendly modes to adapt to the usage habits and characteristics of elderly users. Under relevant technologies, palm-swiping applications work in normal mode by default, and elderly users need to choose to switch from normal mode to elderly-friendly mode on their own.

[0004] However, in actual applications, many elderly users are not familiar with how to control the palm-brushing application to switch from the normal mode to the aging-friendly mode, and need the guidance or assistance of relevant personnel to achieve mode switching, resulting in low efficiency of mode switching of the palm-brushing application. Summary of the invention

[0005] Embodiments of the present application provide an image processing method, apparatus, device, and storage medium for improving the efficiency of mode switching of palm-swiping applications.

[0006] On the one hand, an embodiment of the present application provides an image processing method, the method comprising:

[0007] In response to a palm swiping authentication operation triggered by a target object in a palm swiping application, collecting a palm image of the target object;

[0008] Based on the palm image, identifying the target age group corresponding to the target object; and performing identity authentication on the target object based on the palm image to obtain a target authentication result;

[0009] When the current operation mode of the palm-swiping application does not match the target operation mode corresponding to the target age group, the current operation mode is switched to the target operation mode, and the target authentication result is output in the target operation mode.

[0010] On the one hand, an embodiment of the present application provides an image processing device, the device comprising:

[0011] A collection module, configured to collect a palm image of a target object in response to a palm swiping authentication operation triggered by the target object in the palm swiping application;

[0012] an identification module, configured to identify a target age group corresponding to the target object based on the palm image; and to perform identity authentication on the target object based on the palm image to obtain a target authentication result;

[0013] The judgment module is used to switch from the current operation mode to the target operation mode when the current operation mode of the palm-swiping application does not match the target operation mode corresponding to the target age group, and output the target authentication result in the target operation mode.

[0014] Optionally, the identification module is specifically used for:

[0015] Segmenting a palm portion from the palm image as a region of interest;

[0016] Extracting features of the region of interest using a trained classification model to obtain a palm vein feature map of the target object;

[0017] Based on the palm vein feature map, a target age group corresponding to the target object is predicted.

[0018] Optionally, the identification module is specifically used for:

[0019] Identify a first valley point between an index finger region and a middle finger region in the palm image, and a second valley point between a ring finger region and a little finger region in the palm image;

[0020] A connecting line between the first valley point and the second valley point is used as a reference edge, a square area including the palm portion is determined, and the square area is used as the region of interest.

[0021] Optionally, the identification module is specifically used for:

[0022] Identify the index finger root area and the little finger root area in the palm image, and determine a connecting line between the center point of the index finger root area and the center point of the little finger root area;

[0023] The connecting line is used as a reference edge, a square area including the palm portion is determined, and the square area is used as the region of interest.

[0024] Optionally, the identification module is specifically used for:

[0025] A depthwise separable convolution is performed on the region of interest through a trained classification model to obtain a palm vein feature map of the target object.

[0026] Optionally, the identification module is specifically used for:

[0027] Based on the palm vein information in the region of interest, convert the region of interest into an input feature map of M channels, where M is a positive integer;

[0028] For the input feature maps of the M channels, the following operations are performed respectively: a first convolution kernel is used to perform convolution processing on the input feature map of one channel to obtain an intermediate feature map;

[0029] For the N second convolution kernels, the following operations are performed respectively: the M intermediate feature maps obtained are fused using a second convolution kernel to obtain a palm vein feature map of one channel, where N is a positive integer.

[0030] Optionally, the identification module is further used for:

[0031] In response to a palm swiping authentication operation triggered by a target object in a palm swiping application, before collecting a palm image of the target object, sample data is acquired, wherein the sample data includes palm sample images of multiple age groups;

[0032] Iteratively pre-training the classification model to be trained based on the sample data to obtain the trained classification model, wherein each iterative pre-training process includes the following operations:

[0033] Determine the corresponding predicted age group based on the selected palm sample image through the classification model to be trained;

[0034] Based on the marked age group corresponding to the palm sample image and the predicted age group, a target loss value is determined, and the target loss value is used to adjust the parameters of the classification model to be trained.

[0035] Optionally, the identification module is further used for:

[0036] The regions of interest of multiple palm images obtained within a preset time period are used as tuning training samples, and the trained classification model is fine-tuned to obtain an updated classification model.

[0037] Optionally, the acquisition module is specifically used for:

[0038] In response to a palm swiping authentication operation triggered by a target object in a palm swiping application, near infrared light is emitted through a light emitting component in a camera module;

[0039] Receiving the reflected light of the near-infrared light through the imaging component in the camera module;

[0040] Based on the obtained reflected light, a palm image of the target object is determined.

[0041] Optionally, the judgment module is specifically used to:

[0042] When the current operating mode is the normal mode, and the target operating mode corresponding to the target age group is the aging-friendly mode, it is determined that the current operating mode does not match the target operating mode, and the normal mode is switched to the aging-friendly mode.

[0043] On the one hand, an embodiment of the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned image processing method when executing the program.

[0044] On the one hand, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program executable by a computer device. When the program is run on the computer device, the computer device executes the steps of the above-mentioned image processing method.

[0045] On the one hand, an embodiment of the present application provides a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer device, the computer device executes the steps of the above-mentioned image processing method.

[0046] In the embodiment of the present application, in response to the palm swiping authentication operation, in the process of identity authentication based on the collected palm image, the palm image is used to identify the target age group of the target object, and the palm swiping application is controlled to work in an operation mode that matches the target age group to meet the actual needs of different age groups. In addition, the mode switching of the palm swiping application no longer depends on the active triggering of the target object, and the mode switching is adaptive, thereby improving the efficiency of the mode switching of the palm swiping application. Secondly, for elderly users who are not familiar with the mode switching operation, based on the age group of the elderly user obtained by identification, the corresponding operation mode is adaptively switched to provide the elderly user with the help of the aging-friendly mechanism, thereby improving the user experience of the elderly users and avoiding the interference of fixed aging-friendly functions on young users. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0048] Figure 1 A schematic diagram of a system architecture provided in an embodiment of the present application;

[0049] Figure 2A schematic diagram of an application scenario provided in an embodiment of the present application;

[0050] Figure 3 A schematic diagram of an application scenario provided in an embodiment of the present application;

[0051] Figure 4 A flowchart of an image processing method provided in an embodiment of the present application;

[0052] Figure 5A A schematic diagram of a flow chart of a method for extracting a region of interest provided in an embodiment of the present application;

[0053] Figure 5B A schematic diagram of a region of interest provided in an embodiment of the present application;

[0054] Fig. 6A A schematic diagram of a region of interest provided in an embodiment of the present application;

[0055] Figure 6B A schematic diagram of a region of interest provided in an embodiment of the present application;

[0056] Figure 7 A schematic diagram of a depth-separable convolution process provided in an embodiment of the present application;

[0057] Figure 8 A schematic diagram of a standard convolution process provided in an embodiment of the present application;

[0058] Fig. 9 A flow chart of a classification model training method provided in an embodiment of the present application;

[0059] Fig.10 A flowchart of an image processing method provided in an embodiment of the present application;

[0060] Fig.11 A schematic diagram of the structure of an image processing device provided in an embodiment of the present application;

[0061] Fig.12 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and beneficial effects of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0063] For ease of understanding, the terms involved in the embodiments of the present invention are explained below.

[0064] Palm scanning terminal: a desktop device with palm image acquisition hardware and screen, as well as a central processing unit (CPU), memory, hard disk and operating system. The functions of the palm scanning terminal include: collecting palm images, running palm scanning applications, etc., where palm images include but are not limited to palm RGB images, palm print images, and palm vein images.

[0065] Palm-swiping application: An application that runs on the palm-swiping terminal, has an interactive interface, and interacts with the user through the screen of the palm-swiping terminal. Its functions include: supporting and guiding user operations in scenarios such as authorizing the activation of palm-swiping services and palm-swiping payments.

[0066] Palm veins: The veins on the palm of the hand, whose density, shape, size, etc. can be used as identification features.

[0067] Palm lines: The lines from the fingertips to the wrist. The main lines, wrinkles, fine textures, ridge ends, bifurcation points, etc. can be used as identification features.

[0068] Age-friendly: A product and service design strategy. When designing products and services, we should take the usage habits and characteristics of elderly users into consideration and make modifications to help them use products and services better and more conveniently.

[0069] RGB: represents the colors of the three color channels: red (R), green (G), and blue (B).

[0070] Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that machines have the functions of perception, reasoning and decision-making.

[0071] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0072] Machine Learning (ML) is a multi-disciplinary interdisciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithmic complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning generally include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and learning by teaching. In an embodiment of the present application, a classification model is trained using deep learning technology, and then the user's age group is identified based on the palm image of the user of the palm brushing application through the classification model. When it is determined that the user's age group is the elderly age group, the aging-friendly mode of the palm brushing application is triggered.

[0073] Pretrain-Finetune: Pretraining-fine-tuning mode is a transfer learning method; the specific process is: first use a large-scale data set to pre-train a deep learning model, then use the parameters of the pre-trained model as initialization, and continue to fine-tune the training on a smaller target data set to achieve better results on the target data set.

[0074] Region of interest: ROI (region of interest), in machine vision and image processing, the image area that needs to be processed is outlined from the processed image in the form of a box, circle, ellipse, irregular polygon, etc., which is called the region of interest.

[0075] The design concept of the embodiments of the present application is introduced below.

[0076] In actual applications, more and more elderly users are beginning to use palm-swiping applications, which are applications that implement identity authentication through palm vein recognition or palm print recognition. In order to adapt to the usage habits and characteristics of elderly users, palm-swiping applications have also introduced various aging-friendly modes (also known as elderly mode). Compared with the normal mode, when the palm-swiping application works in the aging-friendly mode, it uses a larger font to display prompt information, adds more prompts, and displays larger button icons.

[0077] Under the relevant technology, the palm-brushing application works in the normal mode by default, and the elderly users need to choose to switch from the normal mode to the aging-friendly mode. However, many elderly users are not familiar with how to control the palm-brushing application to switch from the normal mode to the aging-friendly mode, or are unwilling to actively trigger the palm-brushing application to switch from the normal mode to the aging-friendly mode. At this time, the mode switching needs to be achieved with the guidance or assistance of relevant personnel, which leads to the low efficiency of the mode switching of the palm-brushing application.

[0078] Considering that the palm swiping application needs to collect palm images during identity authentication through palm vein recognition or palm print recognition, and the palm vein information or palm print information in the palm image also reflects the user's age group to a certain extent. In view of this, in the embodiment of the present application, it is proposed that during the identity authentication based on the palm image, the user's age group is also recognized based on the palm image, and the palm swiping application is controlled to work in an operation mode that matches the user's age group.

[0079] Specifically, in response to a palm swiping authentication operation triggered by a target object in a palm swiping application, a palm image of the target object is collected. Then, based on the palm image, a target age group corresponding to the target object is identified; and, based on the palm image, identity authentication is performed on the target object to obtain a target authentication result. When the current operation mode of the palm swiping application does not match the target operation mode corresponding to the target age group, the current operation mode is switched to the target operation mode, and the target authentication result is output in the target operation mode.

[0080] In the embodiment of the present application, in response to the palm swiping authentication operation, in the process of identity authentication based on the collected palm image, the palm image is used to identify the target age group of the target object, and the palm swiping application is controlled to work in an operation mode that matches the target age group to meet the actual needs of different age groups. In addition, the mode switching of the palm swiping application no longer depends on the active triggering of the target object, and the mode switching is adaptive, thereby improving the efficiency of the mode switching of the palm swiping application. In addition, for elderly users who are not familiar with the mode switching operation, based on the age group of the elderly user obtained by identification, the corresponding operation mode is adaptively switched to provide the elderly user with the help of the aging-friendly mechanism, thereby improving the user experience of the elderly users and avoiding the interference of fixed aging-friendly functions on young users.

[0081] refer to Figure 1 , which is a system architecture diagram applicable to an embodiment of the present application. The system architecture at least includes a palm-swiping terminal 101 and a server 102. The number of palm-swiping terminals 101 can be one or more, and the number of servers 102 can also be one or more. The present application does not specifically limit the number of palm-swiping terminals 101 and servers 102.

[0082] The palm swiping terminal 101 has a palm swiping application pre-installed therein, wherein the palm swiping application is a client application, a web application, a mini-program application, etc. The palm swiping terminal 101 may be a palm swiping gate machine, a cash register, a tablet computer, a smart phone, a laptop computer, a desktop computer, a smart home appliance, a smart voice interaction device, a smart vehicle-mounted device, etc., but is not limited thereto.

[0083] Server 102 is the background server of the palm-swiping application. Server 102 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. Palm-swiping terminal 101 and server 102 can be directly or indirectly connected through wired or wireless communication, and this application does not limit this.

[0084] The image processing method in the embodiment of the present application can be executed by the palm-swiping terminal 101 , can be executed by the server 102 , or can be executed interactively by the palm-swiping terminal 101 and the server 102 .

[0085] In practical applications, the image processing method in the embodiments of the present application can be applied to riding scenarios, shopping scenarios, access control scenarios, check-in scenarios, etc.

[0086] For example, in the car-riding scenario, see Figure 2 The palm-swiping gate machine 201 includes an identifier 202 and a screen for displaying an application interface 203 of the palm-swiping application. The application interface 203 of the palm-swiping application displays the prompt message "Please swipe your palm". The passenger faces the palm to the identifier 202, and the palm-swiping gate machine 201 collects the palm image and identifies the target age group of the passenger based on the palm image. If the target age group of the passenger is the elderly, the palm-swiping application is controlled to switch to the aging-friendly mode, that is, the volume of the playing prompt information is increased and the font of the displayed prompt information is increased. The palm-swiping gate machine 201 authenticates the passenger based on the palm image, and when the authentication is passed, controls the gate machine to open; in the aging-friendly mode, the palm-swiping application uses a louder volume to play the prompt message "Please pass", and uses a larger font to display the prompt message "Please pass" in the application interface 203.

[0087] For example, in the shopping scenario, see Figure 3, the cash register device 301 includes an identifier 302 and a screen for displaying an application interface 303 of the palm-swiping application. When checking out for shopping, the application interface 303 of the palm-swiping application displays the prompt message "Please swipe your palm" and the corresponding consumption amount "XX yuan". The customer faces the palm of his hand to the identifier 302, and the cash register device 301 collects the palm vein image and identifies the customer's target age group based on the palm vein image. If the customer's target age group is the elderly age group, the palm-swiping application is controlled to switch to the elderly-friendly mode, that is, the volume of the prompt information played is increased and the font of the prompt information displayed is increased. The cash register device 301 authenticates the customer based on the palm vein image, and when the authentication is passed, it interacts with the background server to perform the deduction operation; the palm-swiping application uses a louder volume to play the prompt message "Payment successful" in the elderly-friendly mode, and uses a larger font to display the prompt message "Payment successful" and the consumption amount "XX yuan" in the application interface 303.

[0088] It should be noted that the above examples are merely illustrative of the application scenarios in this application and this application does not make any limitation thereto.

[0089] In addition, it can be understood that in the specific implementation of the present application, related data such as palm images and palm vein images are involved. When the embodiments involved in the present application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0090] based on Figure 1 The system architecture diagram shown in the figure, the embodiment of the present application provides a process of an image processing method, such as Figure 4 As shown, the process of the method is executed by a computer device, which may be Figure 1 The palm-swiping terminal 101 and / or server 102 shown includes the following steps:

[0091] Step S401 : in response to a palm swiping authentication operation triggered by a target object in a palm swiping application, a palm image of the target object is collected.

[0092] Specifically, the palm image may be: palm vein imaging, palm print image, palm RGB image, etc., wherein the palm vein imaging may be a palm vein image or a palm back vein image, the palm print image may be a palm palm print image or a palm back palm print image, and the palm RGB image may be a palm RGB image or a palm back RGB image.

[0093] The principle of palm vein imaging is that the near-infrared absorption rate of hemoglobin in the palm vein is stronger than that of the surrounding tissue. When the near-infrared light is absorbed by the blood vessels, the blood vessels form a shadow. Specifically, the palm skin is composed of three layers: epidermis, subcutaneous layer and dermis. The palm vein is located in the dermis. The absorption rate of its hemoglobin to near-infrared light is greater than the absorption rate of the tissue in the epidermis and subcutaneous layer to near-infrared light. Therefore, it is possible to image and capture the image of the palm vein under the irradiation of near-infrared light.

[0094] In some embodiments, when the palm image is a palm vein image, the embodiment of the present application uses at least the following methods to collect the palm image of the target object:

[0095] In response to the palm swiping authentication operation triggered by the target object in the palm swiping application, the light emitting component in the camera module emits near infrared light, and the imaging component in the camera module receives the reflected light of the near infrared light; based on the obtained reflected light, the palm image of the target object is determined.

[0096] Specifically, the camera module includes a light-emitting component and an imaging component, wherein the light-emitting component and the imaging component are located in the same plane. The camera module uses a reflective image acquisition method to acquire palm vein images, that is, when the palm is facing the camera module, the light-emitting component emits near-infrared light, and the near-infrared light is reflected when it hits the palm. The imaging component receives the reflected light and obtains the palm vein image based on the reflected light imaging.

[0097] In the embodiment of the present application, since the palm vein is a biometric feature in the body, rather than an image feature of the palm surface, there is no recognition barrier due to damage, wear, dryness or wetness of the palm surface, thereby improving the accuracy of identity authentication and age recognition. Secondly, when collecting palm vein images for age recognition and identity authentication, the hand does not need to contact the device, avoiding the unsanitary problem when the hand contacts the device and the safety problem caused by the possible duplication of the hand surface features.

[0098] Step S402, identifying the target age group corresponding to the target object based on the palm image; and authenticating the target object based on the palm image to obtain a target authentication result.

[0099] Specifically, the palm swiping authentication operation occurs during the registration phase, and is used to indicate authorization to activate the palm swiping service; after collecting the palm image, the palm image is associated with the identity information of the target object to obtain the target authentication result.

[0100] The palm swipe authentication operation occurs during the use phase and is used to indicate identity authentication. After the palm image is captured, the captured palm image is compared with the pre-recorded palm image to obtain the target authentication result. In the riding scenario, if the captured palm image matches the pre-recorded palm image, the authentication is passed and the gate is controlled to open; if the captured palm image does not match the pre-recorded palm image, the authentication fails and the gate remains closed.

[0101] After the palm image is collected, in addition to identity authentication based on the finger image, the embodiment of the present application uses at least the following methods to identify the target age group corresponding to the target object:

[0102] See also Figure 5A , the palm part is divided from the palm image as the region of interest; then the features of the region of interest are extracted through the trained classification model to obtain the palm vein feature map of the target object; based on the palm vein feature map, the target age group corresponding to the target object is predicted.

[0103] Specifically, in the palm image, the palm part is the area with the densest palm veins and the best imaging effect. Therefore, the palm part is divided from the palm image as the region of interest, and the region of interest is input into the trained classification model. The classification model can focus on extracting palm vein features and obtaining a palm vein feature map. Then, the target age group corresponding to the target object is predicted through the palm vein feature map.

[0104] In the embodiment of the present application, the region of interest is intercepted from the palm image, and then the region of interest is input into the trained classification model for age recognition, which reduces the image size of the input classification model, thereby reducing the computational complexity and improving the operation speed. Secondly, the palm image may contain a large amount of background or irrelevant information. After the region of interest is extracted from the palm image, the irrelevant information can be filtered out, so that the classification model only focuses on the features of the region of interest, thereby improving the effect of model recognition.

[0105] In practical applications, the embodiments of the present application use at least the following implementation methods to segment the palm portion from the palm image as the region of interest:

[0106] For a possible implementation, see Figure 5B , identify the first valley point 501 between the index finger area and the middle finger area in the palm image, and the second valley point 502 between the ring finger area and the little finger area in the palm image; then use the connecting line between the first valley point 501 and the second valley point 502 as a reference edge, determine the square area 503 including the palm part, and use the square area 503 as the area of ​​interest.

[0107] Specifically, the valley point between two fingers refers to the lowest point between the two fingers. A first recognition model for identifying the first valley point and the second valley point is pre-trained in a supervised learning manner, and the types of the first recognition model include but are not limited to: convolutional neural network, recurrent neural network, deep neural network. After the training is completed, the first recognition model is used to identify the first valley point and the second valley point in the palm image, and then the first valley point and the second valley point are connected to obtain a reference edge. Then, with the position of the palm as a reference, a square area is determined based on the reference edge, and the square area is the area of ​​interest in the palm image.

[0108] In the embodiment of the present application, since the distance between the valley points on the palm is less affected by the palm rotation angle and the degree of finger opening, after identifying the first valley point between the index finger area and the middle finger area in the palm image, and the second valley point between the ring finger area and the little finger area, the connecting line between the first valley point and the second valley point is used as a reference edge to determine the square area of ​​interest, which can overcome the influence of palm rotation and finger opening to a large extent, thereby improving the accuracy of locating the area of ​​interest.

[0109] For a possible implementation, see Fig. 6A , identify the index finger root area and the little finger root area in the palm image, and determine the connection line between the center point 601 of the index finger root area and the center point 602 of the little finger root area; use the connection line as a reference edge, determine the square area 603 including the palm part, and use the square area 603 as the area of ​​interest.

[0110] Specifically, a second recognition model for identifying the base area of ​​the index finger and the base area of ​​the little finger is trained in advance in a supervised learning manner, and the types of the second recognition model include but are not limited to: convolutional neural network, recurrent neural network, deep neural network. After the training is completed, the second recognition model is used to identify the base area of ​​the index finger and the base area of ​​the little finger in the palm image, and then the center point of the base area of ​​the index finger and the center point of the base area of ​​the little finger are determined. The center point of the base area of ​​the index finger and the center point of the base area of ​​the little finger are connected to obtain a reference edge. Then, with the position of the palm as a reference, a square area is determined based on the reference edge, and the square area is the area of ​​interest in the palm image.

[0111] In an embodiment of the present application, the index finger root area and the little finger root area are automatically identified through a recognition model, and then the square area of ​​interest is determined based on the connecting line between the center point of the index finger root area and the center point of the little finger root area as a reference edge, thereby improving the efficiency of locating the area of ​​interest.

[0112] For a possible implementation, see Figure 6B, identify the first valley point 501 between the index finger area and the middle finger area in the palm image, and the second valley point 502 between the ring finger area and the little finger area in the palm image; then use the connecting line between the first valley point 501 and the second valley point 502 as a reference edge to determine the square area 503 including the palm center.

[0113] The base area of ​​the index finger and the base area of ​​the little finger in the palm image are identified, and a connecting line between the center point 601 of the base area of ​​the index finger and the center point 602 of the base area of ​​the little finger is determined; the connecting line is used as a reference edge to determine a square area 603 including the palm center.

[0114] The union of the square area 503 and the square area 603 is taken as the region of interest.

[0115] It should be noted that the shape of the region of interest in the embodiment of the present application is not limited to a square, and may also be other shapes, such as a triangle, a circle, etc., and the present application does not make specific limitations on this. In addition, the embodiment of the present application may also use other methods to combine the above two methods of extracting the region of interest to locate the region of interest, so as to obtain the best region of interest location effect, and the present application does not make specific limitations on this.

[0116] In some embodiments, when the related technology recognizes and classifies images, it usually collects the convolutional neural network architecture. Although this architecture can obtain better image recognition and classification accuracy, the number of parameters and the amount of calculation of this architecture itself are large. After the model is deployed, more computing resources (such as CPU, memory, etc.) are required to support its fast operation.

[0117] However, in this application, the classification module needs to be deployed on the palm-swiping terminal, and the palm-swiping terminal itself has fewer computing resources and cannot support the operation of the traditional convolutional neural network model. In addition, this application needs to quickly identify the age group of the target object after the palm image acquisition is completed, so as to quickly determine whether to switch the running model. Therefore, if the traditional convolutional neural network model is used, it will take too long in the recognition stage, resulting in the adverse consequence of waiting for a long time. In view of this, the embodiment of the present application adopts a lightweight classification model to predict the target age group corresponding to the target object.

[0118] Specifically, through the trained classification model, a deep separable convolution is performed on the region of interest to obtain the palm vein feature map of the target object; then based on the palm vein feature map, the target age group corresponding to the target object is predicted.

[0119] Depthwise separable convolution consists of two stages: depthwise convolution and pointwise convolution. Figure 7 First, based on the palm vein information in the region of interest, the region of interest is converted into an input feature map of M channels (size DF ×D F ), M, D F is a positive integer. Then based on the M first convolution kernels (size D K ×D K ×1), perform deep convolution on the input feature maps of M channels to obtain M intermediate feature maps (size is D G ×D G ), D K , D G Specifically, for the input feature maps of the M channels, the following operations are performed respectively: a first convolution kernel is used to perform convolution processing on the input feature map of one channel to obtain an intermediate feature map.

[0120] After performing the deep convolution, the M intermediate feature maps are essentially separated from each other. In order to fuse the separated M intermediate feature maps, this application uses N second convolution kernels (size is 1×1×M) to perform point-by-point convolution on the M intermediate feature maps to obtain the palm vein feature maps of N channels (size is D G ×D G ), N is a positive integer, D G ≤D F Specifically, for the N second convolution kernels, the following operations are performed respectively: a second convolution kernel is used to fuse the obtained M intermediate feature maps to obtain a palm vein feature map of one channel.

[0121] The computational cost S1 of the above-mentioned depth-wise separable convolution process is calculated using the following formula (1):

[0122] S1=D K ×D K ×D F ×D F ×M+1×1×D G ×D G ×M×N…………(1)

[0123] In order to more intuitively compare the computational complexity of depth-wise separable convolution and standard convolution, the following describes the process of standard convolution based on input feature maps of the same size:

[0124] See also Figure 8 Based on the palm vein information in the region of interest, the region of interest is converted into an input feature map of M channels (size D F ×D F ), M is a positive integer. Then N standard convolution kernels (size D K ×D K ×M) respectively convolve the input feature maps of M channels to obtain the palm vein feature maps of N channels (size is D G ×D G).

[0125] The computational cost S2 of the above standard convolution process is calculated using the following formula (2):

[0126] S2=D K ×D K ×D F ×D F ×M×N…………(2)

[0127] Next, the following formula (3) is used to calculate the ratio Z of the computation amount S1 of the depthwise separable convolution to the computation amount S2 of the standard convolution:

[0128]

[0129] It can be seen from formula (3) that the amount of calculation of the depthwise separable convolution is significantly less than that of the standard convolution. Therefore, in this application, when the depthwise separable convolution is used to replace the standard convolution to obtain a lightweight classification model, the number of parameters and the amount of calculation of the model are greatly reduced, making the classification model smaller and more suitable for deployment on the palm-swiping terminal. Secondly, since the amount of calculation and the number of parameters of the lightweight classification model are greatly reduced, the calculation process of the lightweight classification model is more efficient, thereby improving the speed of model training and model reasoning. In addition, the lightweight classification model in this application also provides two hyperparameters, width multiplier and input image resolution, to adjust its performance and model size, so that users can flexibly adjust the classification model according to different application scenarios.

[0130] It should be noted that this application does not specifically limit the number of layers of depthwise separable convolution, nor does it specifically limit the number of layers of depthwise convolution and point-by-point convolution in each depthwise separable convolution, and can be adjusted according to actual conditions; in addition, in addition to depthwise separable convolution, the classification model can also include pooling layers, excitation layers, fully connected layers, etc., for which this application does not make specific limitations.

[0131] In some implementations, in order to allow the classification model to learn rich and generalized feature representations and adapt to the actual application environment, the present application embodiment uses pre-training and fine-tuning training to train the classification model. Fig. 9 First, a classification model is pre-trained using large-scale sample data (including palm sample images of multiple age groups), and then the palm images collected on-site by the palm-swiping terminal are used as tuning training samples to fine-tune the classification model to obtain an updated classification model.

[0132] Specifically, when collecting large-scale sample data, it is important to have a large amount of data and rich data sources so that the trained classification model has good generalization capabilities. Therefore, an open source palm image sample library can be used, combined with the palm images collected by the palm-swiping terminal in the past period of time. After the classification model is obtained through pre-training, the classification model is deployed on the palm-swiping terminal. The palm-swiping terminal identifies the target age group corresponding to the palm image collected on-site based on the classification model. At the same time, the palm-swiping terminal uploads the palm images collected on-site to the background server for storage. Subsequently, the palm images collected on-site are used as tuning training samples to fine-tune the classification model and obtain an updated classification model. The updated classification model is then used to replace the originally deployed classification model for age group recognition.

[0133] The following specifically describes the process of pre-training the classification model to be trained using sample data, including the following steps:

[0134] Obtain sample data, wherein the sample data includes sample palm images of multiple age groups. Perform iterative pre-training on the classification model to be trained based on the sample data to obtain a trained classification model, wherein each iterative pre-training process includes the following operations:

[0135] Through the classification model to be trained, the corresponding predicted age group is determined based on the selected palm sample image; then, based on the marked age group and the predicted age group corresponding to the palm sample image, the target loss value is determined, and the target loss value is used to adjust the parameters of the classification model to be trained.

[0136] Specifically, the palm sample images are labeled in advance, that is, the marked age group corresponding to the palm sample images is obtained. For example, the palm brushing application only has a normal mode and an aging-friendly mode, then the classification model only needs to identify whether the age group of the target object is the elderly age group, and does not need to identify the specific age of the target object. Based on this, the marked age group of each palm sample image is the elderly age group or the non-elderly age group. Then, the labeled sample data is used to perform binary classification training on the classification model to be trained.

[0137] Some key parameters need to be set during the training process. The quality of these parameter settings directly affects the effect of the model. Therefore, in the embodiment of the present application, the values ​​of key parameters are obtained by grid search. In addition, when training the model, the regularization term coefficient is set to prevent overfitting, thereby improving the generalization ability of the model.

[0138] During the training process, the target loss value is used to characterize the difference between the marked age group corresponding to the palm sample image and the predicted age group corresponding to the palm sample image. Model training is to minimize the difference between the marked age group and the predicted age group. Determine whether the classification model has converged based on the target loss value. When it is determined that the classification model has not converged, adjust the model parameters of the classification model according to the target loss value, and perform the next round of training with the classification model after parameter adjustment. When it is determined that the classification model has converged, end the training and output the trained classification model. Of course, in the embodiment of the present application, the training can also be ended and the trained classification model can be output when the number of iterative pre-training reaches a preset number of times. This application does not make specific limitations on this.

[0139] In the embodiment of the present application, a large-scale palm sample image set is used to pre-train the classification model, so that the classification model can learn richer and more generalized feature representations, laying a good foundation for subsequent fine-tuning training. The palm images collected on-site by the palm-swiping terminal are used to fine-tune the classification model obtained by pre-training, thereby improving the recognition performance of the classification model in the actual environment.

[0140] The process of fine-tuning training is described below. The present application embodiment adopts at least the following implementation methods to perform fine-tuning training:

[0141] In some embodiments, regions of interest of multiple palm images obtained within a preset time period are used as tuning training samples, and the trained classification model is fine-tuned to obtain an updated classification model.

[0142] Specifically, the regions of interest of multiple palm images obtained within a preset time are annotated and cleaned to obtain a tuned training sample. Then, the parameters of the trained classification model are used as initial values, and the trained classification model is iteratively pre-trained using the tuned training sample to obtain an updated classification model. The updated classification model is applied in the field environment, that is, deployed on the palm-swiping terminal.

[0143] Since the palm-swiping terminal continuously collects palm images for age group recognition, the classification model deployed in the latest time interval can be fine-tuned and trained using the region of interest of the palm image collected in the latest time interval at fixed intervals to achieve continuous updating of the classification model.

[0144] After the fine-tuning training is completed, the recognition results of the updated classification model are compared with those of the classification model before the update to evaluate the fine-tuning training effect of the classification model. Then, based on the fine-tuning training effect, the data volume of the tuning training samples and the frequency of fine-tuning training are adjusted to obtain a better fine-tuning training plan.

[0145] In the embodiment of the present application, the region of interest of the palm image used for age group recognition in the on-site environment is used as a tuning training sample to fine-tune the trained classification model, which effectively reduces the input dimension and irrelevant information, thereby improving the training efficiency. Secondly, since there may be deviations between the palm images collected on-site, such as deviations in direction, brightness, palm distance, etc., and the region of interest of the palm image reduces the differences between the palm images that are irrelevant to the palm veins themselves, therefore, when the region of interest of the palm image is used as the optimal training sample, it can play a role in balancing the training samples, thereby optimizing the training of the classification model, making the classification model more focused on the characteristics of the palm veins themselves, and improving the training effect of the classification model.

[0146] In some embodiments, in addition to using the region of interest in the palm image as a tuning training sample, multiple palm images obtained within a preset time period can also be directly used as tuning training samples to fine-tune the trained classification model to obtain an updated classification model.

[0147] Specifically, the preset duration can be set according to actual needs. Multiple palm images collected by the palm-swiping terminal within the preset duration are annotated and cleaned to obtain a tuned training sample. Then, the parameters of the trained classification model are used as initial values, and the trained classification model is fine-tuned using the tuned training sample to obtain an updated classification model. The updated classification model is applied in the field environment, i.e., deployed in the palm-swiping terminal.

[0148] When fine-tuning training is performed using palm images as tuning training samples, it also supports continuously updating the classification model over time, and evaluating the fine-tuning results to obtain a better fine-tuning training plan, which will not be described in detail here.

[0149] In the embodiment of the present application, the palm images collected on-site by the palm-swiping terminal are used to fine-tune the classification model obtained by pre-training, so that the classification model is more adapted to the image distribution and characteristics in the actual environment, making up for the shortcomings of using large-scale data for training, and improving the recognition performance and accuracy of the classification model in the actual environment. Secondly, palm images of the scene are continuously collected over time, and the classification model is re-fine-tuned using the newly collected palm images, so that the classification model always keeps up with the changes and evolution of the palm images in the on-site environment, and is updated accordingly, thereby significantly improving the performance of the classification model in the actual environment.

[0150] It should be noted that the process of pre-training and fine-tuning the classification model can be performed in the palm-swiping terminal or in the server, and this application does not make any specific limitations on this.

[0151] Step S403, when the current operation mode of the palm-swiping application does not match the target operation mode corresponding to the target age group, the current operation mode is switched to the target operation mode, and the target authentication result is output in the target operation mode.

[0152] Specifically, different age groups correspond to different operating modes. The operating modes included in the palm brushing application include but are not limited to the normal mode and the aging-friendly mode. When the target age group is the elderly age group, the corresponding target operating mode is the aging-friendly mode. When the target age group is the non-elderly age group, the corresponding target operating mode is the normal mode.

[0153] In some embodiments, when the current operating mode is the normal mode, and the target operating mode corresponding to the target age group is the aging-friendly mode, it is determined that the current operating mode does not match the target operating mode, and the normal mode is switched to the aging-friendly mode.

[0154] When the current operating mode is the aging-friendly mode, and the target operating mode corresponding to the target age group is the normal mode, it is determined that the current operating mode does not match the target operating mode, and the normal mode is switched to the aging-friendly mode.

[0155] In actual applications, compared with the normal mode, the aging-friendly mechanisms introduced in the aging-friendly mode include but are not limited to: using larger fonts and larger buttons, extending the waiting time between operations, adding more prompts, reducing the speed of prompts, and increasing the volume of prompts.

[0156] In the embodiment of the present application, in response to the palm swiping authentication operation, in the process of identity authentication based on the collected palm image, the palm image is used to identify the target age group of the target object, and the palm swiping application is controlled to work in an operation mode that matches the target age group to meet the actual needs of different age groups. In addition, the mode switching of the palm swiping application no longer depends on the active triggering of the target object, and the mode switching is adaptive, thereby improving the efficiency of the mode switching of the palm swiping application. Secondly, for elderly users who are not familiar with the mode switching operation, based on the age group of the elderly user obtained by identification, the corresponding operation mode is adaptively switched to provide the elderly user with the help of the aging-friendly mechanism, thereby improving the user experience of the elderly users and avoiding the interference of fixed aging-friendly functions on young users.

[0157] In some application scenarios, in addition to deploying classification models, palm-scanning terminals can also deploy elderly community recognition models. Palm-scanning terminals can be access control devices, health testing equipment, taxi-hailing devices, etc. in the community.

[0158] The training sample data is used in advance, and the initialized elderly community recognition model is iteratively trained by supervised learning until the iteration stop condition is met to obtain the trained elderly community recognition model, wherein the training sample data includes characteristic information and labels of multiple sample communities, and the characteristic information includes: the proportion of elderly people, the number of elderly people, resident behavior information, etc. The label of each sample community is: elderly community or non-elderly community. The iteration stop condition is that the number of iterative training reaches a preset number, or the obtained loss value represents the convergence of the elderly community recognition model. In some embodiments, the classification model and the elderly community recognition model can also be obtained by joint training.

[0159] In actual application, the palm swiping terminal responds to the palm swiping authentication operation triggered by the community residents in the palm swiping application, collects the palm image of the community residents, inputs the palm image into the classification model, identifies the target age group corresponding to the community residents, and authenticates the identity of the community residents based on the palm image to obtain the target authentication result.

[0160] If the target age group corresponding to the community residents is the elderly age group and the identity authentication is passed, the palm-swiping terminal obtains the characteristic information of the current community, where the characteristic information includes but is not limited to: the proportion of elderly people, the number of elderly people, resident behavior information, etc. The characteristic information of the current community is input into the elderly community identification model for identification to obtain the identification result. When the identification result indicates that the current community is an elderly community, the working mode of the palm-swiping application is switched to the elderly-friendly mode, and the elderly-friendly mode is set as the default working mode of the palm-swiping application.

[0161] When the palm-swiping terminal is the access control device of the community, the palm-swiping application uses a larger font to display the access control prompt information in the aging-friendly mode, and plays the access control prompt information at a louder volume and slower speed.

[0162] When the palm-swiping terminal is used as a health monitoring device in the community, the palm-swiping application displays larger operation buttons in the aging-friendly mode, extends the waiting time between each operation, adds more prompt voices, etc.

[0163] When the palm-swiping terminal is used as a taxi-hailing device in the community, the palm-swiping application displays larger operation buttons in the elderly-friendly mode, and plays vehicle information and route planning information at a louder volume and slower speed.

[0164] In the embodiment of the present application, when the target object is identified as being in the elderly age group based on the palm image of the target object through the classification model, it means that the target object currently using the palm-swiping terminal is an elderly user. When the palm-swiping terminal is currently located in an elderly community through the elderly community recognition model, it means that most of the users of the palm-swiping terminal are elderly users. Therefore, when the palm-swiping application is set to work in the elderly-friendly mode by default, the elderly-friendly mechanism can be provided for most elderly users in the elderly community, thereby improving the user experience of elderly users.

[0165] In order to better explain the embodiment of the present application, an image processing method provided by the embodiment of the present application is introduced below in combination with a specific implementation scenario. The process of the method can be as follows: Figure 1 The palm-swiping terminal 101 shown is executed. The palm-swiping terminal 101 deploys a palm image acquisition and processing module, a deep learning algorithm module, and an interactive strategy module. Among them, the palm image acquisition and processing module includes infrared camera hardware, the deep learning algorithm module includes lightweight classification models such as MobileNet, and the interactive strategy module is a part of the palm-swiping application, which is used to determine whether to switch to the aging-friendly mode.

[0166] The following is an introduction to the process of the image processing method in the embodiment of the present application based on the various modules deployed in the palm-swiping terminal 101. Fig.10 , including: asynchronous model training stage 1001, use stage 1002 and synchronous model tuning stage 1003.

[0167] First, we introduce the asynchronous training model stage 1001:

[0168] Collect palm vein image sets, and then use them as training sample data after age calibration (elderly age group or non-elderly age group). Through supervised learning, use the training sample data to pre-train the initial lightweight classification model to obtain a pre-trained classification model. The pre-trained classification model is used as a deep learning algorithm module, together with the palm image acquisition and processing module and the interaction strategy module, and is deployed on the palm scanning terminal.

[0169] Next, the use phase 1002 is introduced:

[0170] The palm image acquisition and processing module responds to the palm swiping authentication operation triggered by the target object in the palm swiping application, collects the palm vein image of the target object through the infrared camera hardware, and then authenticates the target object based on the palm vein image to obtain the target authentication result; at the same time, the palm vein image is passed to the deep learning algorithm module. The deep learning algorithm module uses a pre-trained classification model to identify the target age group corresponding to the target object based on the palm image; then the target age group corresponding to the target object is output to the interactive strategy module. The interactive strategy module determines whether to introduce an aging-friendly mechanism; if the current operating mode of the palm swiping application is the normal mode, and the target operating mode corresponding to the target age group is the aging-friendly mode, then the aging-friendly mechanism is introduced, that is, the palm swiping application is controlled to switch from the normal mode to the aging-friendly mode.

[0171] For example, in a bus scenario, assuming that the target authentication result of the palm image collection and processing module is authentication passed, the gate is controlled to open; assuming that the judgment result of the interaction strategy module is to introduce an aging-friendly mechanism, the palm-swiping application is controlled to switch from normal mode to aging-friendly mode, and in the aging-friendly mode, a prompt message "Please pass" is played at a louder volume, and the prompt message "Please pass" is displayed in a larger font in the application interface of the palm-swiping application.

[0172] Finally, we introduce the synchronous tuning model stage 1003:

[0173] In the use stage 1002, after the palm image acquisition and processing module acquires the palm vein image, the palm vein image is uploaded to the background server for storage. After a preset time interval, the palm vein images collected within the preset time are used as tuning training samples to fine-tune the currently deployed classification model to obtain an updated classification model, and then the updated classification model is used to replace the currently deployed classification model, so that the classification model is more suitable for identifying the palm vein images obtained in the actual use environment.

[0174] In the embodiment of the present application, in response to the palm swiping authentication operation, in the process of identity authentication based on the collected palm image, the palm image is used to identify the target age group of the target object, and the palm swiping application is controlled to work in an operation mode matching the target age group to meet the actual needs of different age groups, and the mode switching of the palm swiping application no longer depends on the active triggering of the target object, and the mode switching is adaptive, thereby improving the efficiency of the mode switching of the palm swiping application. Secondly, for elderly users who are not familiar with the mode switching operation, based on the age group of the elderly user obtained by identification, the corresponding operation mode is adaptively switched to provide the elderly user with the help of the aging-friendly mechanism, thereby improving the user experience of the elderly user and avoiding the interference of the fixed aging-friendly function on the young user. In addition, while the present application uses palm vein recognition technology for identity authentication, it combines palm vein recognition technology with interactive design to obtain the age group to guide the mode switching of the palm swiping application, thereby realizing the refined management of the product experience and bringing a better experience to users of different age groups.

[0175] Based on the same technical concept, the present application embodiment provides a structural schematic diagram of an image processing device, such as Fig.11 As shown, the device 1100 includes:

[0176] The acquisition module 1101 is used for acquiring a palm image of the target object in response to a palm swiping authentication operation triggered by the target object in the palm swiping application;

[0177] The identification module 1102 is used to identify the target age group corresponding to the target object based on the palm image; and to perform identity authentication on the target object based on the palm image to obtain a target authentication result;

[0178] The judgment module 1103 is used to switch from the current operation mode to the target operation mode when the current operation mode of the palm-swiping application does not match the target operation mode corresponding to the target age group, and output the target authentication result in the target operation mode.

[0179] Optionally, the identification module 1102 is specifically configured to:

[0180] Segmenting a palm portion from the palm image as a region of interest;

[0181] Extracting features of the region of interest using a trained classification model to obtain a palm vein feature map of the target object;

[0182] Based on the palm vein feature map, a target age group corresponding to the target object is predicted.

[0183] Optionally, the identification module 1102 is specifically configured to:

[0184] Identify a first valley point between an index finger region and a middle finger region in the palm image, and a second valley point between a ring finger region and a little finger region in the palm image;

[0185] A connecting line between the first valley point and the second valley point is used as a reference edge, a square area including the palm portion is determined, and the square area is used as the region of interest.

[0186] Optionally, the identification module 1102 is specifically configured to:

[0187] Identify the index finger root area and the little finger root area in the palm image, and determine a connecting line between the center point of the index finger root area and the center point of the little finger root area;

[0188] The connecting line is used as a reference edge, a square area including the palm portion is determined, and the square area is used as the region of interest.

[0189] Optionally, the identification module 1102 is specifically configured to:

[0190] A depthwise separable convolution is performed on the region of interest through a trained classification model to obtain a palm vein feature map of the target object.

[0191] Optionally, the identification module 1102 is specifically configured to:

[0192] Based on the palm vein information in the region of interest, convert the region of interest into an input feature map of M channels, where M is a positive integer;

[0193] For the input feature maps of the M channels, the following operations are performed respectively: a first convolution kernel is used to perform convolution processing on the input feature map of one channel to obtain an intermediate feature map;

[0194] For the N second convolution kernels, the following operations are performed respectively: the M intermediate feature maps obtained are fused using a second convolution kernel to obtain a palm vein feature map of one channel, where N is a positive integer.

[0195] Optionally, the identification module 1102 is further configured to:

[0196] In response to a palm swiping authentication operation triggered by a target object in a palm swiping application, before collecting a palm image of the target object, sample data is acquired, wherein the sample data includes palm sample images of multiple age groups;

[0197] Iteratively pre-training the classification model to be trained based on the sample data to obtain the trained classification model, wherein each iterative pre-training process includes the following operations:

[0198] Determine the corresponding predicted age group based on the selected palm sample image through the classification model to be trained;

[0199] Based on the marked age group corresponding to the palm sample image and the predicted age group, a target loss value is determined, and the target loss value is used to adjust the parameters of the classification model to be trained.

[0200] Optionally, the identification module 1102 is further configured to:

[0201] The regions of interest of multiple palm images obtained within a preset time period are used as tuning training samples, and the trained classification model is fine-tuned to obtain an updated classification model.

[0202] Optionally, the acquisition module 1101 is specifically used for:

[0203] In response to a palm swiping authentication operation triggered by a target object in a palm swiping application, near infrared light is emitted through a light emitting component in a camera module;

[0204] Receiving the reflected light of the near-infrared light through the imaging component in the camera module;

[0205] Based on the obtained reflected light, a palm image of the target object is determined.

[0206] Optionally, the determination module 1103 is specifically configured to:

[0207] When the current operating mode is the normal mode, and the target operating mode corresponding to the target age group is the aging-friendly mode, it is determined that the current operating mode does not match the target operating mode, and the normal mode is switched to the aging-friendly mode.

[0208] In the embodiment of the present application, in response to the palm swiping authentication operation, in the process of identity authentication based on the collected palm image, the palm image is used to identify the target age group of the target object, and the palm swiping application is controlled to work in an operation mode that matches the target age group to meet the actual needs of different age groups. In addition, the mode switching of the palm swiping application no longer depends on the active triggering of the target object, and the mode switching is adaptive, thereby improving the efficiency of the mode switching of the palm swiping application. Secondly, for elderly users who are not familiar with the mode switching operation, based on the age group of the elderly user obtained by identification, the corresponding operation mode is adaptively switched to provide the elderly user with the help of the aging-friendly mechanism, thereby improving the user experience of the elderly users and avoiding the interference of fixed aging-friendly functions on young users.

[0209] Based on the same technical concept, the embodiment of the present application provides a computer device, which can be Figure 1 The palm-swiping terminal and / or server shown, such as Fig.12 As shown, it includes at least one processor 1201 and a memory 1202 connected to the at least one processor. The specific connection medium between the processor 1201 and the memory 1202 is not limited in the embodiment of the present application. Fig.12 For example, the processor 1201 and the memory 1202 are connected via a bus. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0210] In the embodiment of the present application, the memory 1202 stores instructions that can be executed by at least one processor 1201, and the at least one processor 1201 can perform the steps of the above-mentioned image processing method by executing the instructions stored in the memory 1202.

[0211] Among them, the processor 1201 is the control center of the computer device, and can use various interfaces and lines to connect various parts of the computer device, and realize virtual production by running or executing instructions stored in the memory 1202 and calling data stored in the memory 1202. Optionally, the processor 1201 may include one or more processing units, and the processor 1201 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 1201. In some embodiments, the processor 1201 and the memory 1202 may be implemented on the same chip, and in some embodiments, they may also be implemented separately on independent chips.

[0212] Processor 1201 can be a general-purpose processor, such as a central processing unit (CPU), a digital signal processor, an application-specific integrated circuit (ASIC), a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, and can implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of the present application. A general-purpose processor can be a microprocessor or any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware processor for execution, or can be executed by a combination of hardware and software modules in the processor.

[0213] Memory 1202, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer executable programs and modules. Memory 1202 may include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (Random Access Memory, RAM), static random access memory (Static Random Access Memory, SRAM), programmable read-only memory (Programmable Read Only Memory, PROM), read-only memory (Read Only Memory, ROM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), magnetic memory, disk, optical disk, etc. Memory 1202 is any other medium that can be used to carry or store a desired program code in the form of an instruction or data structure and can be accessed by a computer device, but is not limited thereto. The memory 1202 in the embodiment of the present application can also be a circuit or any other device that can realize a storage function, for storing program instructions and / or data.

[0214] Based on the same inventive concept, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program executable by a computer device. When the program runs on the computer device, the computer device executes the steps of the above-mentioned image processing method.

[0215] Based on the same inventive concept, an embodiment of the present application provides a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer device, the computer device executes the steps of the above-mentioned image processing method.

[0216] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0217] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer device or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0218] These computer program instructions may also be stored in a computer readable memory capable of directing a computer device or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0219] These computer program instructions can also be loaded onto a computer device or other programmable data processing device so that a series of operation steps are executed on the computer device or other programmable device to produce a process implemented by the computer device, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0220] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0221] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An image processing method, characterized in that: include: In response to a palm swiping authentication operation triggered by a target object in a palm swiping application, collecting a palm image of the target object; Based on the palm image, identifying a target age group corresponding to the target object; And, performing identity authentication on the target object based on the palm image to obtain a target authentication result; When the current operation mode of the palm-swiping application does not match the target operation mode corresponding to the target age group, the current operation mode is switched to the target operation mode, and the target authentication result is output in the target operation mode.

2. The method according to claim 1, characterized in that The step of identifying a target age group corresponding to the target object based on the palm image includes: Segmenting a palm portion from the palm image as a region of interest; Extracting features of the region of interest using a trained classification model to obtain a palm vein feature map of the target object; Based on the palm vein feature map, a target age group corresponding to the target object is predicted.

3. The method according to claim 2, characterized in that The step of dividing the palm portion from the palm image as a region of interest includes: Identify a first valley point between an index finger region and a middle finger region in the palm image, and a second valley point between a ring finger region and a little finger region in the palm image; A connecting line between the first valley point and the second valley point is used as a reference edge, a square area including the palm portion is determined, and the square area is used as the region of interest.

4. The method according to claim 2, characterized in that The step of dividing the palm portion from the palm image as a region of interest includes: Identify the index finger root area and the little finger root area in the palm image, and determine a connecting line between the center point of the index finger root area and the center point of the little finger root area; The connecting line is used as a reference edge, a square area including the palm portion is determined, and the square area is used as the region of interest.

5. The method according to claim 2, characterized in that The step of extracting features from the region of interest using the trained classification model to obtain a palm vein feature map of the target object includes: A depthwise separable convolution is performed on the region of interest through a trained classification model to obtain a palm vein feature map of the target object.

6. The method according to claim 5, characterized in that The performing a depthwise separable convolution on the region of interest to obtain a palm vein feature map of the target object includes: Based on the palm vein information in the region of interest, convert the region of interest into an input feature map of M channels, where M is a positive integer; For the input feature maps of the M channels, the following operations are performed respectively: a first convolution kernel is used to perform convolution processing on the input feature map of one channel to obtain an intermediate feature map; For the N second convolution kernels, the following operations are performed respectively: the M intermediate feature maps obtained are fused using a second convolution kernel to obtain a palm vein feature map of one channel, where N is a positive integer.

7. The method according to claim 2, characterized in that In response to the palm swiping authentication operation triggered by the target object in the palm swiping application, before collecting the palm image of the target object, the method further includes: Acquire sample data, wherein the sample data includes sample palm images of multiple age groups; Iteratively pre-training the classification model to be trained based on the sample data to obtain the trained classification model, wherein each iterative pre-training process includes the following operations: Determine the corresponding predicted age group based on the selected palm sample image through the classification model to be trained; Based on the marked age group corresponding to the palm sample image and the predicted age group, a target loss value is determined, and the target loss value is used to adjust the parameters of the classification model to be trained.

8. The method according to claim 2, characterized in that Also includes: The regions of interest of multiple palm images obtained within a preset time period are used as tuning training samples, and the trained classification model is fine-tuned to obtain an updated classification model.

9. The method according to any one of claims 1 to 8, characterized in that: The step of collecting a palm image of the target object in response to a palm swiping authentication operation triggered by the target object in the palm swiping application includes: In response to a palm swiping authentication operation triggered by a target object in a palm swiping application, near infrared light is emitted through a light emitting component in a camera module; Receiving the reflected light of the near-infrared light through the imaging component in the camera module; Based on the obtained reflected light, a palm image of the target object is determined.

10. The method according to any one of claims 1 to 8, characterized in that: When the current operation mode of the palm-brushing application does not match the target operation mode corresponding to the target age group, switching from the current operation mode to the target operation mode includes: When the current operating mode is the normal mode, and the target operating mode corresponding to the target age group is the aging-friendly mode, it is determined that the current operating mode does not match the target operating mode, and the normal mode is switched to the aging-friendly mode.

11. An image processing device, characterized in that: include: A collection module, configured to collect a palm image of a target object in response to a palm swiping authentication operation triggered by the target object in the palm swiping application; A recognition module, used for identifying a target age group corresponding to the target object based on the palm image; And, performing identity authentication on the target object based on the palm image to obtain a target authentication result; The judgment module is used to switch from the current operation mode to the target operation mode when the current operation mode of the palm-swiping application does not match the target operation mode corresponding to the target age group, and output the target authentication result in the target operation mode.

12. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of any one of the methods of claims 1 to 10 are implemented.

13. A computer-readable storage medium, characterized in that: It stores a computer program executable by a computer device. When the program is run on the computer device, the computer device executes the steps of any method described in claims 1 to 10.

14. A computer program product, characterized in that The computer program product comprises a computer program stored on a computer-readable storage medium, wherein the computer program comprises program instructions, and when the program instructions are executed by a computer device, the computer device is caused to execute the steps of the method according to any one of claims 1 to 10.

Images