Method and device for determining a synthetic palmprint generation model, and storage medium thereof

Abstract

The application provides a palmprint generation method and device and a storage medium thereof. The method comprises: obtaining a first training set, the training set comprising real palmprint images and identity control curves of a plurality of users, each real palmprint image corresponding to a unique user identity; performing data enhancement processing on the real palmprint images to obtain enhanced real palmprint images; inputting the enhanced real palmprint images and the identity control curves into a deep learning model to obtain a synthetic palmprint image corresponding to the unique user identity; performing data enhancement processing on the synthetic palmprint image to obtain an enhanced synthetic palmprint image; iteratively updating parameters of the deep learning model to make the enhanced synthetic palmprint image different from the enhanced real palmprint image and the similarity degree greater than a similarity threshold, and determining a synthetic palmprint generation model. The method can effectively improve the model performance of the generative adversarial network model, strengthen the training stability, and reduce the dependence of the palmprint recognition task on real samples.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method and apparatus for determining a synthetic palm print generation model, as well as a computer storage medium. Background Technology

[0002] In the field of biometric identification, palmprint recognition technology is gradually emerging as a new identification method following fingerprint and facial recognition. Palmprint recognition has significant advantages in protecting user privacy. This technology has been widely used in mobile payments, identity verification, and other areas, playing an important role in protecting users' personal privacy and property security.

[0003] In the field of palmprint recognition, the high degree of privacy inherent in palmprints presents challenges for large-scale data collection. Therefore, the lack of publicly available palmprint datasets has become a significant factor restricting technological development. The quantity and quality of existing publicly available palmprint datasets are limited. Deep learning models are prone to overfitting and instability during training when faced with small samples of real-world data. Summary of the Invention

[0004] In view of this, embodiments of this application provide a method, apparatus, and storage medium for determining a synthetic palmprint generation model, which enables the performance of the palmprint synthesis model. The technical solution is as follows:

[0005] According to a first aspect of this application, a method for determining a synthetic palmprint generation model is provided. The method is executed by a computer device and includes: acquiring a first training set, the training set including real palmprint images and identity control curves of multiple users, wherein the real palmprint image of each user corresponds to a unique user identity, and the identity control curve of each user is a parameterized palmprint curve pre-generated for the unique user identity to simulate palmprint appearance; performing data augmentation processing on the real palmprint images to obtain augmented real palmprint images; inputting the augmented real palmprint images and the identity control curves into a deep learning model to obtain a synthetic palmprint image for the unique user identity; performing the data augmentation processing on the synthetic palmprint images to obtain augmented synthetic palmprint images; iteratively updating the parameters of the deep learning model so that the augmented synthetic palmprint image is different from the augmented real palmprint image and the similarity is greater than a similarity threshold, and determining the deep learning model with updated parameters as the synthetic palmprint generation model.

[0006] In some embodiments, the data augmentation processing of the real palm print image to obtain an enhanced real palm print image includes: performing at least one of geometric transformation, color transformation, and filtering transformation on the user's real palm print image to obtain the enhanced real palm print image.

[0007] In some embodiments, performing data augmentation processing on the real palm print image to obtain an enhanced real palm print image includes: performing data augmentation processing on the user's real palm print image according to an augmentation probability to obtain an enhanced real palm print image, wherein the augmentation probability is determined based on the following parameter: the degree of overfitting between the enhanced synthetic palm print image after the iteration update before the current iteration and the enhanced real palm print image.

[0008] In some embodiments, the enhancement probability is determined according to the following steps: determining the enhancement probability based on the degree of overfit between the enhanced synthetic palm print image and the enhanced real palm print image after the previous iteration update; wherein, when the previous iteration is the first iteration, the enhancement probability is set to an initial value, and the enhancement probability is within a first interval range.

[0009] In some embodiments, the method further includes: after iteratively updating the parameters of the synthetic palmprint generation model to a predetermined number of iterations, selecting an image from the synthetic palmprint image with an inversion probability; merging the synthetic palmprint image selected with the inversion probability into the training set for training the synthetic palmprint generation model, wherein the inversion probability is determined based on the following parameters: the degree of overfitting between the enhanced synthetic palmprint image and the enhanced real palmprint image after the iteration update before the current iteration, and the inversion probability is within a second interval range.

[0010] In some embodiments, the step of merging the synthetic palmprint images selected with the inversion probability into the training set for training the synthetic palmprint generation model includes: for the images in the synthetic palmprint images output at the first iteration number selected with the inversion probability, comparing the quality of the images in the synthetic palmprint images output at the first iteration number with different synthetic palmprint images output at the second iteration number, wherein the second iteration number is prior to the first iteration number and the second iteration number is spaced apart from the first iteration number by a predetermined number of iterations; in response to the image quality of the selected synthetic palmprint images output at the first iteration number being higher than that of the different synthetic palmprint images output at the second iteration number, adding the images in the synthetic palmprint images output at the first iteration number into the training set for training the synthetic palmprint generation model.

[0011] In some embodiments, the step of comparing the quality of the images in the synthetic palmprint images output in the first iteration number and the different synthetic palmprint images output in the second iteration number for the images selected with inversion probability in the first iteration number includes: obtaining a second training set, the second training set including multiple image pairs and their corresponding quality comparison labels, each image pair including a first image and a second image, the quality comparison label indicating the quality comparison result of the first image and the second image; encoding the first image and the second image respectively to obtain a first feature vector and a second feature vector; concatenating the first feature vector and the second feature vector to obtain a concatenated feature vector; inputting the concatenated feature vector into a first neural network to obtain a predicted quality comparison result of the first image and the second image; iteratively updating the parameters of the first neural network to minimize the difference between the predicted quality comparison result and the quality comparison result indicated by the quality comparison label, and to make the distributions of the first feature vector and the second feature vector each normally distributed; updating the parameters of the first neural network model, and determining the first neural network model with updated parameters as the trained first neural network model; inputting the images in the synthetic palmprint images output in the first iteration number and the different synthetic palmprint images output in the second iteration number into the trained first neural network model to obtain the quality comparison result.

[0012] In some embodiments, inputting the enhanced real palmprint image and the identity control curve into a deep learning model to obtain a synthetic palmprint image for the unique user identity includes: inputting the enhanced real palmprint image, the identity control curve, and random Gaussian noise into a deep learning model to obtain a synthetic palmprint image for the unique user identity.

[0013] In some embodiments, the identity control curve for each user is a parameterized palmprint curve generated in advance based on a palmprint template for simulating the appearance of a palmprint, wherein each of the palmprint templates includes a first number of first palmprint sub-curves and a second number of second palmprint sub-curves parameterized using Bézier curves.

[0014] According to a second aspect of this application, a method for determining a palmprint recognition model is provided. The method includes: acquiring a third training set, the training set including real palmprint images and synthetic palmprint images, each of the real palmprint images and synthetic palmprint images corresponding to a known user identity, wherein the synthetic palmprint images are generated using a synthetic palmprint generation model determined by the method for determining a synthetic palmprint generation model provided in the first aspect of this application; inputting both the real palmprint images and the synthetic palmprint images into the palmprint recognition model to obtain a predicted user identity; iteratively updating the parameters of the palmprint recognition model so that the similarity between the known user identity and the predicted user identity is greater than a similarity threshold, updating the parameters of the palmprint recognition model, and determining the palmprint recognition model with updated parameters as the palmprint recognition model, so that the palmprint recognition model can be used to identify the user identity corresponding to the palmprint image.

[0015] According to a third aspect of this application, an apparatus for determining a synthetic palmprint generation model is provided. The apparatus includes: an acquisition module configured to acquire a first training set, the training set including real palmprint images and identity control curves of multiple users, wherein the real palmprint image of each user corresponds to a unique user identity, and the identity control curve of each user is a parameterized palmprint curve pre-generated for the unique user identity to simulate the appearance of the palmprint; a first enhancement module configured to perform data enhancement processing on the real palmprint images to obtain enhanced real palmprint images; an input module configured to input the enhanced real palmprint images and the identity control curves into a deep learning model to obtain a synthetic palmprint image for the unique user identity; a second enhancement module configured to perform the data enhancement processing on the synthetic palmprint images to obtain enhanced synthetic palmprint images; and an iteration module configured to iteratively update the parameters of the deep learning model so that the enhanced synthetic palmprint image is different from the enhanced real palmprint image and the similarity is greater than a similarity threshold, and determine the deep learning model with updated parameters as the synthetic palmprint generation model.

[0016] According to a fourth aspect of this application, a palmprint recognition model determination apparatus is provided, the apparatus comprising: an acquisition module configured to acquire a third training set, the training set including real palmprint images and synthetic palmprint images, each of the real palmprint images and synthetic palmprint images corresponding to a known user identity, wherein the synthetic palmprint images are generated using a synthetic palmprint generation model determined by the method for determining a synthetic palmprint generation model provided in the first aspect of this application; an input module configured to input both the real palmprint images and the synthetic palmprint images into the palmprint recognition model to obtain a predicted user identity; and an update module configured to iteratively update the parameters of the palmprint recognition model so that the similarity between the known user identity and the predicted user identity is greater than a similarity threshold, update the parameters of the palmprint recognition model, and determine the palmprint recognition model with updated parameters as the palmprint recognition model, so that the palmprint recognition model can be used to identify the user identity corresponding to the palmprint image.

[0017] According to a fifth aspect of this application, a computing device is proposed, comprising: a memory and a processor, wherein the memory stores a computer program, which, when executed by the processor, causes the processor to perform a method for determining a synthetic palmprint generation model and / or a method for determining a palmprint recognition model according to some embodiments of this application.

[0018] According to a sixth aspect of this application, a computer-readable storage medium is provided that stores computer-readable instructions thereon, which, when executed, implement a method for determining a synthetic palmprint generation model and / or a method for determining a palmprint recognition model according to some embodiments of this application.

[0019] According to a seventh aspect of this application, a computer program product is proposed, comprising a computer program that, when executed by a processor, implements a method for determining a synthetic palmprint generation model and / or a method for determining a palmprint recognition model according to some embodiments of this application.

[0020] The technical solution provided in this application includes at least the following beneficial effects: This application provides a method, apparatus, and storage medium for determining a synthetic palmprint generation model. This method effectively improves training stability and palmprint synthesis quality in small-sample scenarios by performing data augmentation processing on real and synthetic palmprints, thereby enhancing the performance of the synthetic palmprint generation model. By generating palmprint images with high accuracy and controllable user identity, it provides a large amount of labeled data for deep learning model training, thus meeting the demand for large-scale datasets in deep learning-based palmprint recognition technology, significantly improving model performance, and reducing the dependence of palmprint recognition tasks on real samples. Therefore, this method solves the technical problem that synthetic palmprint generation models are prone to overfitting and unstable training processes when the quantity and quality of publicly available palmprint datasets are limited.

[0021] These and other aspects of this application will be clear from the embodiments described below, and will be elucidated with reference to the embodiments described below. Attached Figure Description

[0022] Further details, features, and advantages of this application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0023] Figure 1 A schematic diagram of an exemplary implementation environment provided in one embodiment of this application is shown;

[0024] Figure 2 The illustration shows a flowchart of palmprint recognition in an exemplary mobile payment scenario provided by one embodiment of this application;

[0025] Figure 3 This is a schematic diagram of a palmprint recognition process according to an exemplary embodiment of this application;

[0026] Figure 4 This is a schematic diagram of the structure of an optimized generative adversarial network according to an exemplary embodiment of this application;

[0027] Figure 5a -b is a schematic diagram illustrating the effect of data augmentation in a scenario of discriminator overfitting, according to an exemplary embodiment of this application;

[0028] Figure 6a -b is an example schematic diagram of a palm print template according to an exemplary embodiment of this application;

[0029] Figure 7 This is a schematic diagram illustrating a quality comparison of a synthetic palmprint image according to an exemplary embodiment of this application;

[0030] Figure 8This is a flowchart of a method for determining a synthetic palmprint generation model according to an exemplary embodiment of this application;

[0031] Figure 9 This is a flowchart of a method for determining a palmprint recognition model according to an exemplary embodiment of this application;

[0032] Figure 10 This is a schematic block diagram of a device for determining a synthetic palmprint generation model according to an exemplary embodiment of this application;

[0033] Figure 11 A schematic block diagram of a device for determining a palmprint recognition model is shown.

[0034] Figure 12 This is an example block diagram of a computing device provided in an exemplary embodiment of this application. Detailed Implementation

[0035] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. The described embodiments are merely some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0036] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, as well as machine learning / deep learning, autonomous driving, intelligent transportation, and automatic control.

[0037] Machine learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can 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 endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and active learning.

[0038] With the research and advancement of artificial intelligence (AI) technology, AI is being studied and applied in various fields, such as smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, autonomous driving, drones, robots, smart healthcare, smart customer service, vehicle networking, and intelligent transportation. It is believed that with the development of technology, AI will be applied in more fields and play an increasingly important role.

[0039] Before detailing the embodiments of this application, some related concepts will first be explained.

[0040] 1. Region of Interest (ROI): A ROI is a specific region in an image or video that is important to the task or focus of attention. In tasks such as image processing and object detection, it is usually only necessary to process and analyze the ROI to improve computational efficiency and accuracy. ROIs can be selected and labeled manually in an image or video or automatically detected. The definition and selection of ROIs vary depending on the task and application. In image processing, an ROI may be a local area of ​​the image to be enhanced, repaired, or transformed. In object detection, an ROI is usually the target object to be detected. By limiting the processing scope, computational resources can be focused on the most relevant region, improving the efficiency and performance of the algorithm.

[0041] 2. Generative Adversarial Networks (GANs): Generative models that learn by engaging in a game between two neural networks. A GAN consists of a generator and a discriminator. The generator randomly samples data from the latent space as input, and its output should closely mimic real samples from the training set. The discriminator takes either real samples or the generator's output as input, and its goal is to distinguish the generator's output from real samples as closely as possible. The generator and discriminator compete against each other and continuously learn, ultimately making it impossible for the discriminator to determine whether the generator's output is real.

[0042] In mobile payment scenarios, palmprint recognition technology directly involves users' wealth, thus requiring extremely high accuracy. Traditional palmprint recognition methods measure the similarity between different user identities through feature extraction. Deep learning-based palmprint recognition technology has become a prevalent solution in recent years. It extracts features from palmprint data by designing efficient network structures and constraining loss functions, and then uses similarity matching algorithms to achieve identity verification. Among deep learning-based palmprint recognition technologies, adversarial neural network-based methods can generate synthetic palmprint images with high accuracy and controllable user identities, thereby meeting the need for large-scale datasets in deep learning-based palmprint recognition technology.

[0043] Deep learning algorithms rely on large-scale labeled data for training to extract effective features and achieve near-realistic palmprint results. Palmprint data synthesis techniques primarily include Generative Adversarial Networks (GANs) and diffusion models. While these two methods can achieve good synthesis results on large-scale and diverse training data, the availability and quality of publicly available palmprint datasets are limited due to privacy concerns.

[0044] Existing palmprint synthesis models based on generative adversarial networks (GANs) can effectively generate a large number of palmprints with valid, identical user identities (also known as user-identity-constrained palmprints) when training samples are sufficient, thereby expanding the training dataset for palmprint recognition. However, due to limitations in the quantity and quality of publicly available palmprint datasets, existing deep learning models are prone to discriminator overfitting, loss of user identity information, and training instability under conditions of small sample real data. Due to the unique characteristics of palmprint data, existing small-sample generation models cannot be optimized for palmprint synthesis scenarios and cannot meet current needs. In scenarios with small samples (i.e., limited real data), data augmentation and transfer learning models can alleviate the need for real data. The emergence of synthetic data provides a large amount of labeled data for training deep learning models, thereby significantly improving model performance.

[0045] In the embodiments of this application, the overfitting degree of the model is effectively measured by the image quality comparison module and statistical data during the training process. Then, through dynamic data augmentation and regularization, the training stability and image synthesis quality of the model in small sample scenarios are effectively improved.

[0046] Figure 1 A schematic diagram of an exemplary computer system 100 provided in one embodiment of this application is shown. This implementation environment may include: a data acquisition terminal 110, a server 120, and a registration terminal 130.

[0047] The data acquisition terminal 110 has an application 111 installed and running. When the application 111 is running, the terminal activates its camera to capture the user's palm print image in real time and sends the image to the server 120 for payment verification. Upon receiving the payment result from the server 120, the terminal displays the result to inform the user whether the payment was successful. The application 111 can be a shopping application, a payment application, or a financial transaction application, etc.

[0048] The registration terminal 130 has application 131 installed and running. Optionally, the application installed on the acquisition terminal 110 and the registration terminal 130 may be the same, or the application installed on the two terminals may be the same type of application on different operating system platforms (Android or iOS), or the application installed on the two terminals may be different versions of the same type of application (e.g., an application that supports payment and an application that supports receiving payment). When the registration terminal 130 receives an account registration operation, it turns on its camera to capture the palm print image of the user 132, and sends the captured image and the registered account to the server 120, so that the server 120 updates the user database 123, which stores the user account and the corresponding palm print features.

[0049] Figure 1 Only one payment terminal and one account registration terminal are shown in the diagram, but in different embodiments, multiple other terminals can access the server 120. Optionally, one or more terminals may also be terminals corresponding to developers, on which an application development and editing platform is installed. Developers can edit and update applications on these terminals and transmit the updated application installation package to the server 120 via a wireless network or wired network 140. The data collection terminal 110 and the registration terminal 130 can download the application installation package from the server 120 to update the application.

[0050] The data acquisition terminal 110, the registration terminal 130, and other terminals are connected to the server 120 via a wireless network or a wired network.

[0051] Server 120 includes at least one of the following: a single server, a server cluster consisting of multiple servers, a cloud computing platform, and a virtualization center. In an illustrative example, server 120 includes a memory 121, a processor 122, a user database 123, a verification module 124, and a user-facing input / output interface (I / O interface) 125. The processor 122 is used to load instructions stored in server 120 and process data in user database 123 and verification module 124. User database 123 stores user account data used by registration terminal 130 and other terminals, including user accounts and corresponding palmprint features. Verification module 124, upon acquiring palmprint features, determines the target account for payment verification and performs payment using the target account. User-facing I / O interface 125 is used to establish communication and exchange data with acquisition terminal 110 and / or registration terminal 130 via a wireless or wired network.

[0052] Figure 2This illustration schematically shows a palmprint recognition flowchart 200 provided in an exemplary mobile payment scenario according to an embodiment of this application. As those skilled in the art will understand, this mobile payment scenario is merely an example, and the application scope of the palmprint recognition method is not limited to this scenario. First, in step 210, a user's hand image is acquired through a terminal payment device; in step 220, a region of interest (ROI) is extracted from the hand image. In step 230, the ROI is input into a palmprint recognition model to obtain an identity recognition result. Thus, the user's identity corresponding to the palmprint image is identified using the palmprint recognition model. Specifically, in one example, it can be as follows... Figure 2 As shown in the dashed box: In step 240, the feature vector of the palm corresponding to the user's identity is extracted. In step 250, the similarity between the palm feature vector and the features in the database is calculated. This similarity can be cosine similarity. The identity information corresponding to the database feature vector with the highest similarity is taken as the recognition result.

[0053] Figure 3 A schematic diagram of a palmprint recognition process 300 according to an embodiment of this application is shown. The process first generates a synthetic palmprint image using a synthetic palmprint generation model 320. Then, the filtered synthetic palmprint image is input as a pseudo-positive sample into a palmprint recognition model 340. In one example, the synthetic palmprint generation model 320 includes an adversarial generative network structure 400. For a detailed description of the adversarial generative network structure 400, please refer to... Figure 4 In block 310, the identity-controlled palmprint synthesis curve is generated as the input to the Generative Adversarial Network (GAN) 320, which constrains the identity of the synthesized palmprint generation model 320, enabling it to generate identity-controlled synthesized palmprints. In block 330, latent variables are input into the synthesized palmprint generation model 320 to make the synthesized palmprint images generated by the GAN training more robust. In one embodiment, the latent variable 330 can be random Gaussian noise. When training the palmprint recognition model 340, a portion of the real palmprint image and the aforementioned synthesized palmprint image selected according to predetermined conditions are input into the palmprint recognition model 340 for palmprint recognition. The specific methods for determining the synthesized palmprint generation model 320 and the palmprint recognition model 340 are detailed in the section on... Figure 8 and Figure 9 The description.

[0054] In the embodiments of this application, the synthetic palmprint generation model 320 was optimized for scenarios with small sample training data. A data augmentation component was added to the synthetic palmprint generation model 320. Furthermore, in another embodiment, the overfitting problem in the generative adversarial network was effectively solved based on statistical data and an image quality comparison module during training.

[0055] Figure 4A schematic diagram illustrates the structure 400 of an optimized adversarial generative network according to an embodiment of this application. The generative adversarial network includes a generator and a discriminator. In scenarios where the training set is small, the discriminator in the generative adversarial model may overfit. This can cause instability in the training of the adversarial generative network and difficulty in convergence. To address the discriminator overfitting problem, a dynamic, differentiable data augmentation module is integrated into the training of the generative adversarial network.

[0056] In one embodiment, a real palmprint image 410 is first obtained. Data augmentation module 420 performs data augmentation processing on it to obtain an enhanced real palmprint curve. The enhanced real palmprint image is input into encoder 430 to obtain vectorized real data features 440. This real data feature 440, along with random noise features 450 and a Bézier curve 460 used as an identity control curve, are input into generator 470 to obtain a synthetic palmprint curve 480 for a unique user identity. Here, the identity control curve can be a Bézier curve, which is a parameterized palmprint curve pre-generated for a unique user identity to simulate the appearance of the palmprint. Specifically, for a unique user identity, the identity control curve is pre-generated based on a palmprint template, each of which includes a first palmprint sub-curve parameterized using a Bézier curve and a second number of second palmprint sub-curves. See [link to palmprint template] for details. Figure 7 A detailed description is provided. Similarly, the synthetic palmprint image is also input into the data enhancement module 490 for the aforementioned data enhancement processing to obtain an enhanced synthetic palmprint image.

[0057] Finally, the parameters of the aforementioned generative adversarial network are iteratively updated to ensure that the enhanced synthetic palmprint image and the enhanced real palmprint image are different and their similarity exceeds a similarity threshold. The deep learning model with updated parameters is then determined as the synthetic palmprint generation model. The similarity threshold can be a value set by those skilled in the art based on experience, and this value is greater than 0.5 and less than 1. In another example, the similarity threshold can be a value greater than 0.8 and less than 1. As those skilled in the art will understand, the similarity threshold can be any other suitable value. The similarity threshold can be set such that the enhanced synthetic palmprint image and the enhanced real palmprint image are different and have the highest degree of similarity.

[0058] In one embodiment, data augmentation processing may include performing at least one of geometric transformation, color transformation, and filtering transformation on the real / synthetic palmprint image to obtain an enhanced real palmprint image.

[0059] In another embodiment, the data augmentation process performs data augmentation on the user's real palm print image according to a predetermined augmentation probability, wherein the augmentation probability depends on the degree of overfit between the augmented synthetic palm print image and the augmented real palm print image to obtain the augmented real palm print image.

[0060] The loss functions for generator 470 and discriminator 4100 are as follows:

[0061]

[0062] G is the generator, and D is the discriminator. Let G be the loss function of the generator. For transformation, It is Gaussian noise.

[0063]

[0064] G is the generator, and D is the discriminator. Let D be the loss function of the discriminator. For transformation, These are real images randomly selected with inverted probability.

[0065] here This represents a graphic transformation for random data augmentation. It performs geometric and color transformations on randomly selected images. These transformations include methods such as isotropic scattering, rotation, translation, color and brightness variations, and contrast adjustment.

[0066] In one embodiment, the data augmentation further includes filtering the sample space in which the image resides. After performing a Fourier transform on the image, the signal intensities at different frequencies are sampled, thereby achieving enhancement for a specific frequency band.

[0067] Figure 5a -b schematically illustrates the effect of data augmentation in a scenario where the discriminator is overfitting. Figure 5a Using fixed data to augment probabilities Figure 5b Dynamic data augmentation is employed to enhance probabilities. The discriminator's overfitting level varies significantly across different stages, and dynamic data augmentation can address this issue. Figure 5a In the -b parameter, the vertical axis represents the discriminator's expected values ​​for real and synthetic images, and the horizontal axis represents the number of training epochs. The expected value for real images is... The expected value of the synthesized image is .exist Approaching 1, A value approaching 0 indicates severe overfitting by the discriminator. The main reason for overfitting is that the discriminator memorizes a large portion of the real image samples, thus failing to provide effective feedback to the generator. In this case, increasing the data augmentation probability will effectively alleviate the discriminator's overfitting problem. (Overfitting degree) r It can be represented as follows:

[0068] in r To account for the degree of overfitting, For discriminator, G For generator, It is Gaussian noise. These are real photos.

[0069] In one embodiment, the augmentation probability is determined based on the following steps: setting an initial value for the augmentation probability as the initial augmentation probability of the deep learning model. For example, setting the initial data augmentation probability to... The value is equal to 0.2. After each iteration, the enhancement probability is adjusted based on the degree of overfitting between the enhanced synthetic palmprint image and the enhanced real palmprint image, and the adjusted enhancement probability is used as the enhancement probability for the next iteration. The degree of overfitting... When a certain threshold is exceeded, the probability of data augmentation is dynamically adjusted. Specifically, this dynamic adjustment is based on the following formula.

[0070]

[0071] in It is data augmentation probability, and for Expected value and Expected value Range constraints were applied. This is the lower bound of the probability of data augmentation. The upper limit of the probability of data augmentation. α It is the change in the magnitude of the single expected value. .

[0072] Due to the characteristics of palmprint images, excessively high data augmentation probabilities can lead to color collapse in the generated images, while initial augmentation probabilities that are too low cannot effectively address the overfitting problem. Therefore, an interval constraint is added to the data augmentation. In one embodiment, the augmentation probability is determined based on the following steps: determining the augmentation probability based on the degree of overfitting between the augmented synthetic palmprint image and the augmented real palmprint image after the previous iteration update; wherein, when the previous iteration is the first iteration, the augmentation probability is set to an initial value, and the augmentation probability lies within a first interval.

[0073] Figure 6a -b schematically illustrates an example of a palm print template. Here, a palm print is used as an example. Figure 6a As shown, the palmprint template can be composed of a first number of first palmprint sub-curves and a second number of second palmprint sub-curves. Figure 6a As shown, palm prints are essentially composed of a first number (e.g., 3-5) of main lines and a second number of wrinkles. To fit the geometric appearance of the palm prints, Bessel functions are used to parameterize them. For simplicity, in... Figure 6b The method uses a second-order Bézier curve with three control points on a two-dimensional plane. These three control points are: the starting point marked by a triangle, the control point marked by a pentagram, and the ending point marked by a circle. Taking the three main lines of the left hand as an example, in one example, the right hand can be considered a mirror image of the left hand.

[0074] Figure 7 A schematic diagram 700 illustrates a quality comparison of a synthetic palmprint image according to an embodiment of this application. In a generative adversarial network, the generator can effectively counteract the overfitting problem of the discriminator. When the overfitting of the discriminator exceeds a predetermined threshold, it can be dynamically (e.g., with probability) adjusted. p 2) Incorporate a portion of the synthesized palm print image into the real image during the training process.

[0075] In one embodiment, after iteratively updating the parameters of the synthetic palmprint generation model to a predetermined number of iterations, the probability is inverted. p 2. Select the image from the synthesized palm print image; then invert the probability. p 2. Selected synthetic palmprint images are merged and added to the training set for training the synthetic palmprint generation model.

[0076] In another embodiment, regarding the inversion probability p 2. The selected image from the synthetic palmprint images output in the first iteration is compared in quality with different synthetic palmprint images output in the second iteration, where the second iteration follows the first iteration and is spaced a predetermined number of iterations apart. In response to the selected image from the synthetic palmprint images output in the first iteration having higher quality than the different synthetic palmprint images output in the second iteration, the selected image from the synthetic palmprint images is added to the training set for training the synthetic palmprint generation model.

[0077] In yet another embodiment, for those with inversion probability p2. Selecting an image from the synthetic palmprint image output in the first iteration, and comparing the quality of the image from the synthetic palmprint image output in the first iteration with different synthetic palmprint images output in the second iteration, includes: obtaining a second training set, which includes multiple image pairs and their corresponding quality comparison labels. Each image pair includes a first image and a second image, and the quality comparison label indicates the quality comparison result of the first image and the second image; encoding the first image and the second image respectively to obtain a first feature vector and a second feature vector; concatenating the first feature vector and the second feature vector to obtain a concatenated feature vector; inputting the concatenated feature vector into a first neural network to obtain the predicted quality comparison result of the first image and the second image; iteratively updating the parameters of the first neural network to minimize the difference between the predicted quality comparison result and the quality comparison result indicated by the quality comparison label, updating the parameters of the first neural network model, and determining the first neural network model with updated parameters as the trained first neural network model; inputting the image from the synthetic palmprint image output in the first iteration and different synthetic palmprint images output in the second iteration into the trained first neural network model to obtain the quality comparison result.

[0078] Here, the loss function can be expressed by the following formula:

[0079]

[0080] Let D be the loss function of the discriminator. For real images randomly selected with inversion probability, the inversion probability is... p 2 is the probability extracted from the synthetic palm print image and used for inclusion in the training set.

[0081] in p The update probability of 2 can be defined as follows:

[0082]

[0083] in It is the inversion probability, and for Expected value and Expected value Range constraints were applied. It is the lower bound of the reversal probability. This is the upper limit of the reversal probability. α It is the change in the magnitude of the single expected value. .

[0084] The probability interval will be used as a hyperparameter of the model. In one embodiment, after iteratively updating the parameters of the synthetic palmprint generation model to a predetermined number of iterations, an image from the synthetic palmprint image is selected with an inversion probability; the synthetic palmprint images selected with the inversion probability are merged into the training set for training the synthetic palmprint generation model, wherein the inversion probability is determined based on the following parameters: the degree of overfitting between the enhanced synthetic palmprint image and the enhanced real palmprint image after the iteration update before the current iteration, and the inversion probability is within a second interval. In the early stages of training, the quality of the generated images is poor, and introducing positive samples too early will cause difficulties in model convergence and a decrease in the quality of generated images. In the early stages of training, the overfitting phenomenon of the discriminator is not obvious. Therefore, a constraint is proposed: the optimization method is introduced into the model training only after a certain number of rounds. This strategy helps to improve the performance and stability of the generative adversarial network.

[0085] To further ensure the image quality in the pseudo-positive sample dataset, this method designs an efficient image quality comparison module. This module compares newly generated images with other images from the previous training rounds, ensuring that only high-quality newly generated images are added to the pseudo-positive sample dataset. This method helps improve the generator's performance, thus offsetting the overfitting problem of the discriminator to some extent. The training set is derived from synthetic palmprint images generated during the training of the generative model. Figure 7 The assumption is that, within a certain range, the image quality in a generative adversarial network increases with the number of training epochs. By pairing later-synthesized palmprint images (trained in later epochs) with earlier-synthesized images (trained in earlier epochs), a large number of image pairs of varying quality can be constructed, thus forming an image quality comparison module. It is worth noting that there needs to be a certain distance between the earlier and later training epochs to produce an effective result; otherwise, their qualities will be quite similar.

[0086] Image A and image B, which is in a later training epoch, are both processed by an encoder to extract real image features. In one example, the extracted feature vector has a dimension of 16*16. The encoder can use a 4-layer convolutional structure to effectively capture features from the real palmprint image. Next, the input multidimensional tensor is flattened into a one-dimensional tensor. Constraints are then added to the feature vector. This makes the distribution of eigenvectors It follows a normal distribution, among which This is the average value. Standard deviation. Loss function. as follows:

[0087]

[0088] By concatenating the feature vectors of images A and B, and inputting them into a multi-layer neural network, the activation function... sigmoid The generated result is the final image quality comparison result.

[0089] Figure 8 This is a flowchart of a method 800 for determining a synthetic palmprint generation model according to an exemplary embodiment of this application. The method 800 for determining a synthetic palmprint generation model is executed by a computer device. Specifically, the method includes:

[0090] In step 810, a training set is obtained, which includes real palmprint images and identity control curves for multiple users. As those skilled in the art will understand, the real palmprint images here are not images of the user's hand, but images of the region of interest of the hand extracted from a hand image. Each user's real palmprint image corresponds to a unique user identity, and each user's identity control curve is a parameterized palmprint curve pre-generated for the unique user identity to simulate the appearance of the palmprint. In one embodiment, the control curve for each user identity is a parameterized palmprint curve pre-generated for the unique user identity and used to simulate the appearance of the palmprint. In one embodiment, each of the palmprint templates includes a first number of first palmprint sub-curves and a second number of second palmprint sub-curves parameterized using Bézier curves.

[0091] Palm prints, for example: Palm prints are basically composed of a first number (e.g., 3-5) of main lines and a second number of wrinkles. To fit the geometric appearance of the palm print, a Bezier function is used to parameterize it. A second-order Bezier curve with three parameter points on a two-dimensional plane is used. Here, the three control points are: the starting point marked by a triangle, the control point marked by a pentagram, and the ending point marked by a circle. Taking the three main lines of the left hand as an example, in one example, the right hand can be regarded as a mirror image of the left hand.

[0092] In step 820, data augmentation processing is performed on the real palm print image to obtain an enhanced real palm print image.

[0093] In one embodiment, at least one of geometric transformation, color transformation, and filtering transformation is performed on a user's real palm print image to obtain an enhanced real palm print image. The geometric and color transformations include methods such as isotropic scattering, rotation, translation, color brightness variation, and contrast adjustment.

[0094] In one embodiment, the data augmentation further includes filtering the sample space in which the image resides. After performing a Fourier transform on the image, the signal intensities at different frequencies are sampled, thereby achieving enhancement for a specific frequency band.

[0095] In another embodiment, data augmentation is performed on the user's real palm print image according to a predetermined augmentation probability, wherein the augmentation probability depends on the degree of overfitting between the augmented synthetic palm print image and the augmented real palm print image to obtain the augmented real palm print image. Additionally, the augmentation probability is determined based on the following steps: setting an initial value for the augmentation probability as the initial augmentation probability of the deep learning model; adjusting the augmentation probability based on the degree of overfitting between the augmented synthetic palm print image and the augmented real palm print image after each iteration, and using the adjusted augmentation probability as the augmentation probability for the next iteration.

[0096] For example, setting the initial data augmentation probability to The value is equal to 0.2. After each iteration, the enhancement probability is adjusted based on the degree of overfitting between the enhanced synthetic palmprint image and the enhanced real palmprint image, and the adjusted enhancement probability is used as the enhancement probability for the next iteration. (The last sentence appears to be incomplete and possibly refers to a setting related to overfitting.) When a certain threshold is exceeded, the probability of data augmentation is dynamically adjusted. Specifically, this dynamic adjustment is based on the following formula.

[0097]

[0098] in It is data augmentation probability, and for Expected value and Expected value Range constraints were applied. This is the lower bound of the probability of data augmentation. The upper limit of the probability of data augmentation. α It is the change in the magnitude of the single expected value. .

[0099] Due to the characteristics of palmprint images, excessively high data augmentation probabilities can lead to color collapse in the generated images, while initial augmentation probabilities that are too low cannot effectively address the overfitting problem. Therefore, interval constraints are added to the data augmentation process.

[0100] In step 830, the enhanced real palm print image and the identity control curve are input into the deep learning model to obtain a synthetic palm print image for a unique user identity.

[0101] In step 840, the synthesized palmprint image undergoes data augmentation processing to obtain an enhanced synthesized palmprint image. This data augmentation processing is the same as that in step 820. Specifically, data augmentation may include at least one of geometric transformation, color transformation, and filtering transformation to obtain an enhanced realistic palmprint image. The geometric and color transformations include methods such as isotropic scattering, rotation, translation, color brightness variation, and contrast adjustment. In one embodiment, the data augmentation further includes filtering the sample space containing the image. After performing a Fourier transform on the image, the signal intensity at different frequencies is sampled, thereby achieving enhancement for specific frequency bands.

[0102] In step 850, the parameters of the deep learning model are iteratively updated to ensure that the enhanced synthetic palmprint image is different from the enhanced real palmprint image and that the similarity is greater than a similarity threshold. The deep learning model with updated parameters is then identified as the synthetic palmprint generation model. The similarity threshold can be a value set by those skilled in the art based on experience, and this value is greater than 0.5 and less than 1. In another example, the similarity threshold can be a value greater than 0.8 and less than 1. As those skilled in the art will understand, the similarity threshold can be any other suitable value. The similarity threshold can be set such that the enhanced synthetic palmprint image and the enhanced real palmprint image are different and have the highest degree of similarity.

[0103] When the deep learning model mentioned above is a generative adversarial network, the enhanced synthetic palmprint image can be made different from the enhanced real palmprint image and the similarity is greater than a similarity threshold by constructing the following loss function:

[0104] The loss functions for the generator and discriminator are as follows:

[0105]

[0106] G is the generator, and D is the discriminator. Let G be the loss function of the generator. For transformation, It is Gaussian noise.

[0107]

[0108] G is the generator, and D is the discriminator. Let D be the loss function of the discriminator. For transformation, These are real images randomly selected with inverted probability.

[0109] In one embodiment, the method further includes: after iteratively updating the parameters of the synthetic palmprint generation model to a predetermined number of iterations, selecting an image from the synthetic palmprint images with inverted probability; merging the synthetic palmprint images selected with inverted probability into the training set for training the synthetic palmprint generation model.

[0110] Specifically, the process of merging and adding synthetic palmprint images selected with inversion probability into the training set for training the synthetic palmprint generation model includes: for images selected with inversion probability in the synthetic palmprint images output in the first iteration number, comparing the quality of the images in the synthetic palmprint images output in the first iteration number with different synthetic palmprint images output in the second iteration number, wherein the second iteration number is after the first iteration number and the interval between the second iteration number and the first iteration number is a predetermined number of iterations; in response to the selected synthetic palmprint images output in the first iteration number having higher quality than different synthetic palmprint images output in the second iteration number, adding the selected synthetic palmprint images into the training set for training the synthetic palmprint generation model.

[0111] In one embodiment, an enhanced real palmprint image, an identity control curve, and random Gaussian noise are input into a deep learning model to obtain a synthetic palmprint image for a unique user identity. The addition of random Gaussian noise enhances the robustness of the deep learning model.

[0112] This method effectively improves training stability and palmprint synthesis quality in small-sample scenarios by performing data augmentation on both real and synthetic palmprints, thus enhancing the performance of the synthetic palmprint generation model. By generating palmprint images with high accuracy and controllable user identity, it provides a large amount of labeled data for deep learning model training, meeting the demand for large-scale datasets in deep learning-based palmprint recognition technology, significantly improving model performance, and reducing the dependence of palmprint recognition tasks on real samples. Therefore, this method addresses the technical problem of overfitting and unstable training processes in synthetic palmprint generation models when the quantity and quality of publicly available palmprint datasets are limited. Furthermore, by combining an image quality comparison module with statistical data from the training process, it effectively measures the degree of overfitting and significantly improves the training stability and image synthesis quality in small-sample scenarios through dynamic data augmentation and regularization.

[0113] Figure 9 The illustration schematically depicts a method 900 for determining a palmprint recognition model according to an embodiment of this application. In step 910, a training set is obtained, comprising real palmprint images and synthetic palmprint images, each of which corresponds to a known user identity, wherein the synthetic palmprint images utilize... Figure 8This palmprint was generated using a synthetic palmprint generation model. For details on determining the synthetic palmprint generation model, please refer to the section on... Figure 8 The description will not be repeated here.

[0114] In step 920, both the real palm print image and the synthetic palm print image are input into the palm print recognition model to obtain the predicted user identity.

[0115] In step 930, the parameters of the palmprint recognition model are iteratively updated to ensure that the similarity between the known user identity and the predicted user identity is greater than the similarity threshold. The parameters of the palmprint recognition model are then updated, and the palmprint recognition model with updated parameters is determined as the palmprint recognition model.

[0116] Figure 10 A schematic block diagram of a device 1000 for determining a synthetic palm print generation model is shown. The device 1000 for determining a synthetic palmprint generation model includes: an acquisition module 1010 configured to acquire a first training set, which includes real palmprint images and identity control curves of multiple users, wherein the real palmprint image of each user corresponds to a unique user identity, and the identity control curve of each user is a parameterized palmprint curve pre-generated for the unique user identity to simulate the appearance of the palmprint; a first enhancement module 1020 configured to perform data enhancement processing on the real palmprint images to obtain enhanced real palmprint images; an input module 1030 configured to input the enhanced real palmprint images and identity control curves into a deep learning model to obtain a synthetic palmprint image for the unique user identity; a second enhancement module 1040 configured to perform data enhancement processing on the synthetic palmprint images to obtain enhanced synthetic palmprint images; and an iteration module 1050 configured to iteratively update the parameters of the deep learning model so that the enhanced synthetic palmprint images are different from the enhanced real palmprint images and the similarity is greater than a similarity threshold, and to determine the deep learning model with updated parameters as the synthetic palmprint generation model.

[0117] It should be understood that the device 1000 for determining the synthetic palm print generation model can be implemented in a software, hardware or a combination of software and hardware manner. Multiple different modules in the device can be implemented in the same software or hardware structure, or one module can be implemented by multiple different software or hardware structures.

[0118] Furthermore, the device 1000 for determining the synthetic palmprint generation model can be used to implement the method for determining the synthetic palmprint generation model described above, the details of which have been described in detail above and will not be repeated here for the sake of brevity. In addition, these devices can have the same features and advantages as those described in the corresponding methods.

[0119] Figure 11A schematic block diagram of a palmprint recognition model determination device 1100 is shown. The palmprint recognition model determination device 1100 includes: an acquisition module 1110 configured to acquire a third training set, which includes real palmprint images and synthetic palmprint images, each of which corresponds to a known user identity, wherein the synthetic palmprint images are generated using a synthetic palmprint generation model determined by the method for determining synthetic palmprint generation models provided in the embodiments of this application; an input module 1120 configured to input both the real palmprint images and the synthetic palmprint images into a palmprint recognition model to obtain a predicted user identity; and an update module 1130 configured to iteratively update the parameters of the palmprint recognition model so that the similarity between the known user identity and the predicted user identity is greater than a similarity threshold, update the parameters of the palmprint recognition model, and determine the palmprint recognition model with updated parameters as the palmprint recognition model, so that the palmprint recognition model can be used to identify the user identity corresponding to the palmprint image.

[0120] It should be understood that the palmprint recognition model determination device 1100 can be implemented in a software, hardware or a combination of software and hardware manner. Multiple different modules in the device can be implemented in the same software or hardware structure, or one module can be implemented by multiple different software or hardware structures.

[0121] Furthermore, the palmprint recognition model determining device 1100 can be used to implement the palmprint recognition model determining method described above, the details of which have been described in detail above and will not be repeated here for the sake of brevity. In addition, these devices can have the same features and advantages as those described in the corresponding methods.

[0122] Figure 12 An example system 1200 is illustrated, which includes an example computing device 1210 representing one or more systems and / or devices that can implement the various methods described herein. The computing device 1210 may be, for example, a server of a service provider, a device associated with a server, a system-on-a-chip, and / or any other suitable computing device or computing system. (Refer to above) Figure 10-11 The apparatus 1000 for determining the synthetic palmprint generation model and the apparatus 1100 for determining the palmprint recognition model described herein can take the form of a computing device 1210. Alternatively, Figure 10-11 The device 1000 for determining the synthetic palmprint generation model and the device 1100 for determining the palmprint recognition model described herein can be implemented as a computer program in the form of an application 1216.

[0123] The example computing device 1210 shown includes a processing system 1211, one or more computer-readable media 1212, and one or more I / O interfaces 1213 that are communicatively coupled to each other. Although not shown, the computing device 1210 may also include a system bus or other data and command transfer system that couples the various components to each other. The system bus may include any or a combination of different bus architectures, such as a memory bus or memory controller, a peripheral bus, a universal serial bus, and / or a processor or local bus utilizing any of the various bus architectures. Various other examples, such as control and data lines, are also conceived.

[0124] Processing system 1211 represents the functionality of performing one or more operations using hardware. Therefore, processing system 1211 is illustrated as including hardware elements 1214 that can be configured as processors, function blocks, etc. This may include other logic devices implemented in hardware as application-specific integrated circuits (ASICs) or formed using one or more semiconductors. Hardware element 1214 is not limited by the materials in which it is formed or the processing mechanism employed therein. For example, a processor may consist of semiconductors and / or transistors (e.g., integrated circuits (ICs)). In such a context, processor-executable instructions may be electronically executable instructions.

[0125] Computer-readable medium 1212 is illustrated as including memory / storage device 1215. Memory / storage device 1215 represents a memory / storage capacity associated with one or more computer-readable media. Memory / storage device 1215 may include volatile media (such as random access memory (RAM)) and / or non-volatile media (such as read-only memory (ROM), flash memory, optical disk, magnetic disk, etc.). Memory / storage device 1215 may include fixed media (e.g., RAM, ROM, fixed hard disk drive, etc.) and removable media (e.g., flash memory, removable hard disk drive, optical disk, etc.). Computer-readable medium 1212 may be configured in various other ways as further described below.

[0126] One or more I / O interfaces 1213 represent functions that allow users to input commands and information to computing device 1210 using various input devices and optionally also allow information to be presented to the user and / or other components or devices using various output devices. Examples of input devices include keyboards, cursor control devices (e.g., mice), microphones (e.g., for voice input), scanners, touch functionality (e.g., capacitive or other sensors configured to detect physical touch), cameras (e.g., capable of detecting non-touch-related motion as gestures using visible or invisible wavelengths (such as infrared frequencies), etc. Examples of output devices include display devices, speakers, printers, network interface cards, haptic-responsive devices, etc. Therefore, computing device 1210 can be configured in various ways, as further described below, to support user interaction.

[0127] The computing device 1210 also includes an application 1216. The application 1216 may, for example, be used for... Figure 10-11 The software examples of the device 1000 for determining the synthetic palmprint generation model and the device 1100 for determining the palmprint recognition model are described herein, and the techniques described herein are implemented in combination with other elements in the computing device 1210.

[0128] This application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computing device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computing device to perform the visualization data presentation methods provided in the various optional implementations described above.

[0129] This document describes various technologies within the general context of software and hardware components or program modules. Generally, these modules include routines, programs, objects, elements, components, data structures, etc., that perform specific tasks or implement specific abstract data types. As used herein, the terms "module," "function," and "component" generally refer to software, firmware, hardware, or a combination thereof. The technologies described herein are characterized as platform-independent, meaning that these technologies can be implemented on a variety of computing platforms with various processors.

[0130] Implementations of the described modules and technologies may be stored on or transmitted across some form of computer-readable medium. The computer-readable medium may include a variety of media accessible by the computing device 1210. By way of example and not limitation, the computer-readable medium may include "computer-readable storage media" and "computer-readable signal media".

[0131] In contrast to simple signal transmission, carrier waves, or signals themselves, a "computer-readable storage medium" refers to a medium and / or device capable of persistently storing information, and / or a tangible storage device. Therefore, a computer-readable storage medium refers to a non-signal-bearing medium. Computer-readable storage media include hardware such as volatile and non-volatile, removable and non-removable media and / or storage devices implemented using methods or techniques suitable for storing information (such as computer-readable instructions, data structures, program modules, logic elements / circuits, or other data). Examples of computer-readable storage media may include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical storage devices, hard disks, cassette tapes, magnetic tapes, disk storage devices or other magnetic storage devices, or other storage devices, tangible media, or articles of art suitable for storing desired information and accessible by a computer.

[0132] "Computer-readable signal medium" refers to a signal-bearing medium configured to transmit instructions, such as via a network, to hardware of computing device 1210. A signal medium typically embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave, data signal, or other transmission mechanism. Signal media also include any information transmission medium. The term "modulated data signal" refers to a signal in which one or more of its characteristics are set or altered to encode information. By way of example and not limitation, communication media include wired media such as wired networks or direct connections, and wireless media such as acoustic, RF, infrared, and other wireless media.

[0133] As previously stated, hardware element 1214 and computer-readable medium 1212 represent instructions, modules, programmable device logic, and / or fixed device logic implemented in hardware, which in some embodiments can be used to implement at least some aspects of the techniques described herein. Hardware elements may include components of integrated circuits or systems-on-a-chip, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), complex programmable logic devices (CPLDs), and other implementations or other hardware devices in silicon. In this context, hardware elements can serve as processing devices for executing program tasks defined by instructions, modules, and / or logic embodied by the hardware element, and as hardware devices for storing instructions for execution, such as the previously described computer-readable storage medium.

[0134] The foregoing combinations can also be used to implement the various techniques and modules described herein. Therefore, software, hardware, or program modules and other program modules can be implemented as one or more instructions and / or logic embodied on some form of computer-readable storage medium and / or by one or more hardware elements 1214. The computing device 1210 can be configured to implement specific instructions and / or functions corresponding to the software and / or hardware modules. Thus, for example, by using the computer-readable storage medium and / or hardware element 1214 of a processing system, modules can be implemented at least partially in hardware as modules executable as software by the computing device 1210. Instructions and / or functions can be executable / operable by one or more articles of art (e.g., one or more computing devices 1210 and / or processing systems 1211) to implement the techniques, modules, and examples described herein.

[0135] In various embodiments, the computing device 1210 can be configured in various ways. For example, the computing device 1210 can be implemented as a computer-type device, including personal computers, desktop computers, multi-screen computers, laptop computers, netbooks, etc. The computing device 1210 can also be implemented as a mobile device-type device, including mobile devices such as mobile phones, portable music players, portable gaming devices, tablet computers, multi-screen computers, etc. The computing device 1210 can also be implemented as a television-type device, including devices with or connected to a generally large screen in a leisure viewing environment. These devices include televisions, set-top boxes, game consoles, etc.

[0136] The techniques described herein can be supported by these various configurations of computing device 1210, and are not limited to specific examples of the techniques described herein. Functionality can also be implemented, wholly or partially, on “cloud” 1220 using distributed systems, such as platform 1222 as described below.

[0137] Cloud 1220 includes and / or represents platform 1222 for resource 1224. Platform 1222 is the underlying functionality of the hardware (e.g., server) and software resources of cloud 1220. Resource 1224 may include applications and / or data that can be used when performing computer processing on a server located remotely from computing device 1210. Resource 1224 may also include services provided via the Internet and / or via subscriber networks such as cellular or Wi-Fi networks.

[0138] Platform 1222 can abstract resources and functions to connect computing device 1210 to other computing devices. Platform 1222 can also be used to abstract resource hierarchy to provide a corresponding level of hierarchy for any encountered needs for resource 1224 implemented via platform 1222. Therefore, in interconnect device embodiments, the implementation of the functions described herein can be distributed throughout system 1200. For example, functions can be implemented partly on computing device 1210 and partly through platform 1222, which abstracts the functions of cloud 1220.

[0139] It should be understood that, for clarity, embodiments of this application have been described with reference to different functional units. However, it will be apparent that, without departing from this application, the functionality of each functional unit may be implemented in a single unit, in multiple units, or as part of other functional units. For example, functionality described as being performed by a single unit may be performed by multiple different units. Therefore, references to specific functional units are considered merely as references to the appropriate units used to provide the described functionality, and not as indicating a strict logical or physical structure or organization. Thus, this application may be implemented in a single unit, or may be physically and functionally distributed among different units and circuits.

[0140] In the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0141] Although this application has been described in conjunction with some embodiments, it is not intended to be limited to the specific forms set forth herein. Rather, the scope of this application is limited only by the appended claims. Additionally, although individual features may be included in different claims, these may be advantageously combined, and inclusion in different claims does not imply that such a combination of features is not feasible and / or advantageous. The order of features in the claims does not imply that the features must be in any particular order of their operation. Furthermore, in the claims, the word "comprising" does not exclude other elements, and the terms "a" or "an" do not exclude a plurality. Reference numerals in the claims are provided only by way of explicit example and should not be construed as limiting the scope of the claims in any way.

[0142] It is understood that the specific embodiments of this application involve entity-related data such as entity default information. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

Claims

1. A method for determining a synthetic palmprint generation model, characterized in that, The method is performed by a computer device, and the method includes: Obtain a first training set, which includes real palm print images and identity control curves of multiple users. The real palm print image of each user corresponds to a unique user identity, and the identity control curve of each user is a parameterized palm print curve generated in advance for the unique user identity to simulate the appearance of the palm print. The real palm print image is subjected to data augmentation processing to obtain an enhanced real palm print image; The enhanced real palm print image and the identity control curve are input into a deep learning model to obtain a synthetic palm print image for the unique user identity. The data augmentation process is applied to the synthesized palmprint image to obtain an enhanced synthesized palmprint image; The parameters of the deep learning model are iteratively updated to make the enhanced synthetic palm print image different from the enhanced real palm print image and the similarity is greater than the similarity threshold. The deep learning model with updated parameters is then determined as the synthetic palm print generation model.

2. The method according to claim 1, characterized in that, The process of performing data augmentation on the real palm print image to obtain an enhanced real palm print image includes: At least one of geometric transformation, color transformation, and filtering transformation is performed on the user's real palm print image to obtain an enhanced real palm print image.

3. The method according to claim 1, characterized in that, The process of performing data augmentation on the real palm print image to obtain an enhanced real palm print image includes: The user's real palm print image is augmented according to the augmentation probability to obtain an augmented real palm print image, wherein the augmentation probability is determined based on the following parameter: the degree of overfit between the augmented synthetic palm print image after the iteration update before the current iteration and the augmented real palm print image.

4. The method according to claim 3, characterized in that, The enhancement probability is determined according to the following steps: The enhancement probability is determined based on the degree of overfit between the enhanced synthetic palmprint image after the previous iteration and the enhanced real palmprint image; Wherein, when the previous iteration was the first iteration, the enhancement probability was set to an initial value, and the enhancement probability was within the first interval range.

5. The method according to claim 1, characterized in that, Also includes: After iteratively updating the parameters of the synthetic palmprint generation model to a predetermined number of iterations, an image from the synthetic palmprint image is selected with inverted probability. The synthetic palmprint images selected with the inversion probability are merged into the training set for training the synthetic palmprint generation model, wherein the inversion probability is determined based on the following parameters: the degree of overfitting between the enhanced synthetic palmprint image and the enhanced real palmprint image after the iteration update before the current iteration, and the inversion probability is within a second interval range.

6. The method according to claim 5, characterized in that, The step of merging the synthetic palmprint images selected with the inversion probability into the training set for training the synthetic palmprint generation model includes: For the image in the synthetic palm print image output at the first iteration number selected with the inversion probability, the quality of the image in the synthetic palm print image output at the first iteration number is compared with that of the different synthetic palm print images output at the second iteration number, wherein the second iteration number is before the first iteration number and the second iteration number is spaced apart from the first iteration number by a predetermined number of iterations; In response to the fact that the quality of the selected synthetic palmprint image output in the first iteration number is higher than that of the different synthetic palmprint images output in the second iteration number, the image of the synthetic palmprint image output in the first iteration number is added to the training set for training the synthetic palmprint generation model.

7. The method according to claim 6, characterized in that, The step of comparing the quality of the image in the synthetic palmprint image output at the first iteration number, selected with the inversion probability, with the different synthetic palmprint images output at the second iteration number includes: Obtain a second training set, which includes multiple image pairs and their corresponding quality comparison labels. Each image pair includes a first image and a second image, and the quality comparison label indicates the quality comparison result between the first image and the second image. The first image and the second image are encoded respectively to obtain the first feature vector and the second feature vector; By concatenating the first and second eigenvectors, we obtain the concatenated eigenvector. The concatenated feature vectors are input into the first neural network to obtain a comparison of the prediction quality of the first and second images. The parameters of the first neural network are iteratively updated to minimize the difference between the predicted quality comparison result and the quality comparison result indicated by the quality comparison label, and to make the distributions of the first feature vector and the second feature vector each normally distributed. The parameters of the first neural network model are updated, and the first neural network model with updated parameters is determined as the trained first neural network model. The image in the synthetic palm print image output in the first iteration and the different synthetic palm print images output in the second iteration are input into the trained first neural network model to obtain quality comparison results.

8. The method according to claim 1, characterized in that, The step of inputting the enhanced real palmprint image and the identity control curve into a deep learning model to obtain a synthetic palmprint image for the unique user identity includes: The enhanced real palm print image, the identity control curve, and random Gaussian noise are input into a deep learning model to obtain a synthetic palm print image for the unique user identity.

9. The method according to any one of claims 1-8, characterized in that, Each user's identity control curve is a parameterized palmprint curve generated in advance based on a palmprint template for the unique user identity to simulate the appearance of the palmprint. Each of the palmprint templates includes a first number of first palmprint sub-curves and a second number of second palmprint sub-curves parameterized using Bézier curves.

10. A method for determining a palmprint recognition model, characterized in that, The method includes: Obtain a third training set, which includes real palm print images and synthetic palm print images, each of which corresponds to a known user identity, wherein the synthetic palm print images are generated using the synthetic palm print generation model as described in claim 1. The real palm print image and the synthetic palm print image are input into the palm print recognition model to obtain the identified predicted user identity. The parameters of the palmprint recognition model are iteratively updated to ensure that the similarity between the known user identity and the predicted user identity is greater than a similarity threshold. The parameters of the palmprint recognition model are then updated, and the palmprint recognition model with updated parameters is determined as the palmprint recognition model, so that the palmprint recognition model can be used to identify the user identity corresponding to the palmprint image.

11. A device for determining a synthetic palmprint generation model, characterized in that, The device includes: The acquisition module is configured to acquire a first training set, which includes real palm print images and identity control curves of multiple users. The real palm print image of each user corresponds to a unique user identity, and the identity control curve of each user is a parameterized palm print curve generated in advance for the unique user identity to simulate the appearance of the palm print. The first enhancement module is configured to perform data enhancement processing on the real palm print image to obtain an enhanced real palm print image. The input module is configured to input the enhanced real palm print image and the identity control curve into a deep learning model to obtain a synthetic palm print image for the unique user identity. The second enhancement module is configured to perform the data enhancement processing on the synthesized palmprint image to obtain an enhanced synthesized palmprint image. An iterative module is configured to iteratively update the parameters of the deep learning model so that the enhanced synthetic palm print image is different from the enhanced real palm print image and the similarity is greater than a similarity threshold, and to determine the deep learning model with updated parameters as the synthetic palm print generation model.

12. A device for determining a palmprint recognition model, characterized in that, The device includes: The acquisition module is configured to acquire a third training set, the training set including real palm print images and synthetic palm print images, each of the real palm print images and synthetic palm print images corresponding to a known user identity, wherein the synthetic palm print images are generated using the synthetic palm print generation model as described in claim 1. The input module is configured to input both the real palm print image and the synthetic palm print image into the palm print recognition model to obtain the identified predicted user identity. The update module is configured to iteratively update the parameters of the palmprint recognition model so that the similarity between the known user identity and the predicted user identity is greater than a similarity threshold, update the parameters of the palmprint recognition model, and determine the palmprint recognition model with updated parameters as the palmprint recognition model so that the palmprint recognition model can be used to identify the user identity corresponding to the palmprint image.

13. A computing device, characterized in that, include: Memory, which is configured to store computer-executable instructions; A processor configured to perform the method according to any one of claims 1-10 when the computer-executable instructions are executed by the processor.

14. A computer-readable storage medium, characterized in that, It stores computer-executable instructions that, when executed, implement the method according to any one of claims 1-10.

15. A computer program product, characterized in that, Includes a computer program that, when executed, implements the steps of the method according to any one of claims 1-10.

Images