Biological characteristic matching method, device and equipment

Abstract

The invention relates to the field of machine vision, in particular to a biological feature matching method, device and equipment. The invention discloses a biological feature matching method, and the method comprises the steps: obtaining the first feature distribution of a to-be-detected biological feature image, and the second feature distribution of all preset biological feature images, and the number of the preset biological feature images is at least one; according to the first feature distribution and each second feature distribution, determining a biological feature similarity value between the to-be-detected biological feature image and each preset biological feature image; and performing biological feature matching according to the biological feature similarity value.

Description

Technical Field

[0001] The present application relates to the field of machine vision, and particularly to a biometric matching method, device, and equipment. Background Art

[0002] Biometric recognition technology refers to a technology in which an electronic device collects inherent biometric features of a human body, and then matches them with the biometric features pre-stored by a user to determine whether they belong to the same user's biometric features, thereby determining the user's identity. In related technologies, biometric features are extracted into a feature point vector, and when performing matching, the similarity between the collected biometric features and the pre-stored biometric features is determined according to the distance between two feature point vectors to determine the user's identity.

[0003] However, in some scenarios, the quality of the collected biometric features is low, resulting in errors in the extracted feature point vectors, further resulting in errors in the similarity, and incorrect verification of the user's identity. Summary of the Invention

[0004] Aiming at the problem of errors in biometric matching caused by low quality of biometric features collected in an open scenario in the prior art, the present application provides a biometric matching method, device, and equipment.

[0005] In a first aspect, an embodiment of the present invention provides a biometric matching method, including:

[0006] Obtaining a first feature distribution of a biometric feature image to be verified, and a second feature distribution of all preset biometric feature images, where the number of preset biometric feature images is at least one;

[0007] Determining a biometric similarity value between the biometric feature image to be verified and each preset biometric feature image according to the first feature distribution and each second feature distribution;

[0008] Performing biometric matching according to the biometric similarity value.

[0009] In the embodiment of the present invention, instead of using a feature vector alone to represent the biometric features of a user, the confidence of the feature vector is increased, the error of the feature vector is represented by the confidence, and the biometric features of the user are represented by a feature distribution jointly composed of the feature vector and the corresponding feature vector confidence. When calculating the similarity, comprehensive consideration is given to the feature vector and the feature vector confidence in the feature distribution.

[0010] Optionally, the first feature distribution and the second feature distribution are Gaussian distributions.

[0011] Optionally, in order to characterize the blurring degree of the biometric features in the biometric image to be tested and the preset biometric image, it is necessary to determine the biometric vector confidence of the biometric image to be tested and the preset biometric image respectively, and construct a feature distribution based on the biometric vector and the biometric vector confidence. Obtaining the first feature distribution of the biometric image to be tested and the second feature distribution of all preset biometric images includes:

[0012] Input the collected biometric image to be tested into the feature recognition model to obtain the first biometric vector of the biometric image to be tested;

[0013] Input the first biometric vector and the second biometric vector of the preset biometric image into the multi-layer perceptron respectively to obtain the first biometric vector confidence regarding the first biometric vector and the second biometric vector confidence regarding the second biometric vector;

[0014] Determine the first feature distribution according to the first biometric vector and the first biometric vector confidence;

[0015] Determine the second feature distribution according to the second biometric vector and the second biometric vector confidence.

[0016] Optionally, in order to determine the probability that there is no error in the features between the two images according to the feature distributions of the biometric image to be tested and the preset biometric image, that is, to determine the similarity value between the two images. Determining the biometric similarity value of the biometric image to be tested and each preset biometric image according to the first feature distribution and each second feature distribution includes:

[0017] Establish a third feature distribution according to the first feature distribution and the second feature distribution, and the third feature distribution is the distribution of the difference between the first feature distribution and the second feature distribution;

[0018] When the difference between the first feature distribution and the second feature distribution is zero, determine the value of the third feature distribution according to the first feature vector, the second feature vector, the first biometric vector confidence, and the second biometric vector confidence;

[0019] Determine the value of the third feature distribution as the face feature similarity value.

[0020] Optionally, for the convenience of calculation, when the image quality is not sufficient to affect biometric recognition, the biometric confidence is no longer considered when calculating the biometric similarity. Determining the biometric similarity value of the biometric image to be tested and each preset biometric image according to the first feature distribution and each second feature distribution, the method further includes:

[0021] When the confidence of the first biometric feature vector is the same as that of the second biometric feature vector, determine the Euclidean distance between the first biometric feature vector and the second biometric feature vector;

[0022] Determine the Euclidean distance as the biometric similarity value.

[0023] Optionally, in order to determine whether the user corresponding to the biometric feature image to be tested is the user who has pre-stored the preset biometric feature image in the electronic device, perform biometric matching according to the biometric similarity, including:

[0024] When the biometric similarity value between the biometric feature image to be tested and at least one preset biometric feature image is greater than the preset first threshold, determine that the user corresponding to the biometric feature image to be tested passes the matching.

[0025] Optionally, in order to determine the identity of the user corresponding to the biometric feature image to be tested, perform biometric matching according to the biometric similarity, and the method further includes:

[0026] Determine the user identity information corresponding to the preset biometric feature image with the largest biometric similarity value as the identity information of the user to be tested.

[0027] Optionally, in order to ensure that there is no error in the obtained biometric similarity value, it is necessary to ensure the accuracy of the biometric confidence output by the MLP. Therefore, it is necessary to update and train the MLP network to ensure its accuracy, and the method further includes:

[0028] Collect a set of biometric feature images as the training set;

[0029] Obtain the fourth feature distribution of each biometric feature image;

[0030] According to each fourth feature distribution, determine the target similarity value between each biometric feature image belonging to the same user identity information;

[0031] Update and train the multi-layer perceptron according to the target similarity value.

[0032] In a second aspect, an embodiment of the present invention provides a biometric matching device, including:

[0033] An acquisition module that acquires the first feature distribution of the biometric feature image to be tested and the second feature distribution of each preset biometric feature image;

[0034] A determination module that determines the biometric similarity value between the biometric feature image to be tested and each preset biometric feature image according to the first feature distribution and each second feature distribution;

[0035] A matching module that performs biometric matching according to the biometric similarity value.

[0036] In a third aspect, an embodiment of the present invention provides an electronic device, which includes a memory for storing computer program instructions and a processor for executing the computer program instructions. When the computer program instructions are executed by the processor, the electronic device is triggered to execute the method according to any one of the first aspect.

[0037] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, in which a computer program is stored. When it runs on a computer, the computer is enabled to execute the method according to any one of the first aspect. Description of the Drawings

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0039] Figure 1 Shown is a schematic diagram of the changing trend of face similarity;

[0040] Figure 2 Shown is a flowchart of a biometric matching method provided by an embodiment of the present application;

[0041] Figure 3 Shown is a schematic structural diagram of an MLP provided by an embodiment of the present application;

[0042] Figure 4 Shown is a schematic diagram of a Gaussian distribution provided by an embodiment of the present application;

[0043] Figure 5 Shown is a flowchart of a specific embodiment of a biometric matching method provided by an embodiment of the present invention;

[0044] Figure 6 Shown is a flowchart of an MLP training process provided by an embodiment of the present invention;

[0045] Figure 7 Shown is a schematic structural diagram of a biometric matching device provided by an embodiment of the present application;

[0046] Figure 8 Shown is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed Embodiments

[0047] To better understand the technical solutions of the present application, the embodiments of the present application will be described in detail below with reference to the drawings.

[0048] It should be clear that the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of this application.

[0049] Before describing the embodiments of this application, the related technologies and technical problems will be described first.

[0050] Biometric recognition technology refers to the technology that an electronic device collects the inherent biometric information of a human body, and then matches it with the biometric information pre-stored by the user to determine whether it belongs to the biometric information of the same user, so as to determine the user's identity. Biometric recognition technology generally includes face recognition, fingerprint recognition, iris recognition, etc.

[0051] Taking face recognition technology as an example, an electronic device collects a number of preset face images input by users through a sensor device such as a camera, and extracts and stores the preset face features of each preset face image correspondingly. When actually performing face matching, the face image of the user to be verified is collected, the face features are extracted from the collected face image, and are matched with the preset face features stored in advance. When it matches successfully with a certain preset face feature stored in advance, the identity of the user can be determined.

[0052] In the related technology, multiple facial feature points such as eyes, eyebrows, nose, mouth, etc. are obtained from the collected face image, and the facial feature points are convolved through a face recognition model to obtain a multi-dimensional face feature vector of a person. Generally, 128 or 256 facial feature points are collected and convolved to obtain a 128-dimensional face feature vector. The face feature vector is mapped as a point vector into the face feature space. There are several point vectors of preset face features pre-stored in the face feature space. In the face feature space, the distance between the face feature vector and each pre-stored preset face feature vector is calculated, and this distance is used as the similarity value between the two faces, thus completing the matching of face features.

[0053] However, in some open scenarios, such as when there is strong light, or when the user makes certain movements that cause the facial features to be blocked, or when the pixel of the sensor device used to collect the face image is poor, the quality of the collected face image will be poor, and there will be blurs in the face image. And the poor-quality face image will have a great impact on the accuracy of face recognition.

[0054] Among them, since the quality of the collected face images is poor, not all facial features in the face images can be extracted, or there are errors between the extracted facial features and the actual ones. When performing face recognition based on the facial features with errors, whether there are errors in the preset face features or the face features collected during the actual face recognition, it will lead to incorrect recognition.

[0055] As Figure 1 shown, it is a schematic diagram of the change trend of face similarity. Through some common methods, such as performing Gaussian blur, cropping, and adding random Gaussian noise to the collected face images or preset face images, the blurring degree of the face images is increased. After using the three methods to increase the blurring degree of the face images, the change trend of the similarity between the actually collected face and the preset face is roughly the same.

[0056] See Figure 1 , when a high-quality face image of the same user is matched with a low-quality face image, as the blurring degree of the low-quality face image further increases, the similarity with the high-quality face image will be lower, that is, the possibility of judging two photos of the same user as different users is gradually increasing; when a low-quality face image of different users is matched with a high-quality face image, as the quality of the low-quality face image further decreases, the similarity with the high-quality face image will be higher, that is, the probability of judging two photos of different users as the same user gradually increases as the quality of the face image decreases. It can be understood that if two low-quality face images are used for matching at the same time, the possibility of incorrect matching judgment will be further increased.

[0057] In summary, when performing matching on two face images with clear features, it can be clearly determined whether they belong to the same user. However, when the quality of one or both face images decreases, the extracted features may be unclear or even invalid and cannot be used as features for actual applications. But in the related technologies, the errors in the face features are not characterized, and the extracted face features with errors are still used as clear features and mapped into the face feature space for face matching. As a result, there is a deviation between its position in the face feature space and the position where it should be. Therefore, the error in the similarity between two face images will increase as the distance between the two feature points in the face feature space increases. The possibility of judging photos of different users as the same user or judging photos of the same user as different users is increasing.

[0058] Similar to face recognition technology, when performing other biometric matching such as fingerprint matching and iris matching, there will also be problems such as low quality of the collected biometric features, resulting in incorrect feature matching and further causing incorrect user identity recognition.

[0059] To solve the above problems, in the embodiments of the present invention, instead of using the feature vector alone to represent the user's biometric features, the confidence of the feature vector is increased, the error of the feature vector is characterized by the confidence, and the user's biometric features are characterized by the feature distribution jointly composed of the feature vector and the corresponding feature vector confidence. When calculating the similarity, comprehensive consideration is made through the feature vector and the feature vector confidence in the feature distribution.

[0060] As Figure 2 shown, a biometric matching method provided by an embodiment of the present invention includes the following specific steps:

[0061] S201, obtain a first feature distribution of the biometric feature image to be verified and a second feature distribution of all preset biometric feature images.

[0062] Specifically, when the electronic device is first enabled or when the biometric matching function is first used, the user will be prompted to input preset biometric feature images, such as face images, fingerprint images, etc. The number of preset biometric feature images input by the user can be one or more. For example, a user can input one fingerprint image, or a user can input three face images, or three different users can each input one of their own face images, etc., for use as a template to perform biometric matching. If the user does not input any preset biometric feature images, the biometric matching function cannot be executed.

[0063] When the user inputs preset biometric feature images, the preset biometric feature images input by the user are collected through a sensor. To facilitate subsequent biometric matching, the collected preset biometric images can be input into a feature recognition model, and the feature recognition model performs convolution on the preset biometric feature images to obtain a second biometric feature vector of the preset biometric feature images. In this way, when actually performing biometric matching each time, the second biometric feature vector can be directly used without having to extract it from the preset biometric images each time.

[0064] When actually performing biometric matching, the biometric feature image to be verified is also collected through a sensor. The collected biometric feature image to be verified is input into the feature recognition model, and the feature recognition model performs convolution on the biometric feature image to be verified to obtain a first biometric feature vector of the biometric feature image to be verified. Among them, when the first biometric feature vector and the second biometric feature vector are extracted by convolution, for the sake of simple calculation, only the diagonal covariance matrix in the first biometric feature vector and the second biometric feature vector is considered.

[0065] The first biometric vector of the biometric image to be verified and the second biometric vectors of all preset biometric images in the terminal device are respectively input into a Multiple Layer Perception (MLP). Through the MLP, the confidence of the first biometric vector corresponding to the first biometric vector and the confidence of each second biometric vector corresponding to each second biometric vector are output.

[0066] As Figure 3 shown, it is a schematic structural diagram of an MLP. Among them, to ensure the stability of the MLP, the neurons in the last layer of the MLP share the alpha parameter and the beta parameter. And for the convenience of subsequent calculations, the confidence of the biometric vector output by the MLP is logσ 2 .

[0067] The first biometric vector and the confidence of the first biometric vector jointly establish a first feature distribution to describe the features of the biometric image to be verified; each second biometric vector and the confidence of its corresponding second biometric vector jointly establish a second feature distribution to describe the features of the preset biometric image.

[0068] Among them, the first feature distribution can be expressed as:

[0069] Z i (l) ~N(μ i (l) , σ i 2(l) )

[0070] The second feature distribution can be expressed as:

[0071] Z j (l) ~N(μ j (l) , σ j 2(l) )

[0072] The established first feature distribution and second feature distribution are generally Gaussian distributions. As Figure 4 shown, it is a schematic diagram of a Gaussian distribution. μ in the Gaussian distribution is the biometric vector, and σ in the Gaussian distribution is the confidence of the feature vector. If σ in the Gaussian distribution is larger, it means that the confidence of the feature vector is worse, that is, the original image is more blurred.

[0073] S202. According to the first feature distribution and each second feature distribution, determine the biometric similarity values between the biometric image to be verified and each preset biometric image.

[0074] Specifically, calculate the biometric similarity between the biometric image to be verified and all preset biometric images respectively. The calculation method of the biometric similarity between the biometric image to be verified and each preset biometric image is the same. Taking the calculation method of the biometric similarity between any one preset biometric image as an example for illustration.

[0075] The similarity between the corresponding biometric image to be verified and the preset biometric image is represented by the similarity between the first feature distribution and the second feature distribution.

[0076] Establish a third feature distribution according to the first feature distribution and the second feature distribution. The third feature distribution is the distribution of the difference between the first feature distribution and the second feature distribution. Since both the first feature distribution and the second feature distribution are Gaussian distributions, the third feature distribution obtained by taking the difference between the first feature distribution and the second feature distribution also follows a Gaussian distribution.

[0077] The third feature distribution can be expressed as:

[0078] △Z (l) ~N(μ i (l) -μ j (l) ,σ i 2(l) +σ j 2(l) )

[0079] The third feature distribution is used to characterize the similarity between the first feature distribution and the second feature distribution. When the difference between the first feature distribution and the second feature distribution is 0, the value of the third feature distribution is the similarity value between the first feature distribution and the second feature distribution. That is, calculate the probability that there is no error between the first feature distribution and the second feature distribution as the similarity value between the first feature distribution and the second feature distribution.

[0080] The following formula is the specific calculation formula for the similarity value:

[0081]

[0082] Wherein, X i is the biometric image to be verified, X j is the preset biometric image, s(X i , X j ) represents the similarity between the biometric image to be verified and the preset biometric image. For the convenience of calculation, perform the calculation by taking the logarithm through the method of log-likelihood; Z i is the first feature distribution of the biometric image to be verified, Z j is the second feature distribution of the preset biometric image; △Z is the difference between the first feature distribution and the second feature distribution, that is, the third feature distribution; μi is the first biometric feature vector of the biometric feature image to be tested, and σ i 2 is the confidence of the first biometric feature image of the biometric feature image to be tested, and μ j is the second biometric feature vector of the preset biometric feature image, and σ j 2 is the confidence of the second biometric feature image of the preset biometric feature image; D is the number of biometric features collected, generally 128 or 256, and l is one of the biometric features.

[0083] As can be seen from the above formula, since is a constant term, it will not affect the calculated similarity value, while μ i , σ i , μ j , σ j will all affect the final similarity value.

[0084] In the first term, the smaller the difference between μ i and μ j , that is, the closer the first feature vector is to the second feature vector, the larger the similarity value, and the higher the similarity between the biometric feature image to be tested and the preset biometric feature image; the second term corrects the result of the first term through the confidence of the feature vector. In the second term, since the weight of σ is larger than that of the first term, and the logarithmic function is monotonically increasing, the larger σ is, the lower the similarity value, and the lower the similarity.

[0085] Optionally, when establishing the third feature distribution, the first feature distribution and the second feature distribution satisfy the commutative law, that is, when calculating the difference between the first feature distribution and the second feature distribution, whether subtracting the first feature distribution from the second feature distribution or subtracting the second feature distribution from the first feature distribution, the obtained results are the same and will not affect the finally calculated feature similarity.

[0086] S203. Perform biometric feature matching according to the biometric feature similarity value.

[0087] Specifically, the biometric feature matching performed includes: determining whether the biometric feature image to be tested is a pre-stored preset biometric feature image, and determining the user identity according to the biometric feature image to be tested.

[0088] When determining whether a biometric image to be verified is a pre-stored preset biometric image, when the biometric similarity value between the biometric image to be verified and at least one preset biometric image is greater than a preset first threshold, it can be determined that the biometric image to be verified is consistent with the pre-stored biometric image, that is, the user identities corresponding to the two pictures are the same, and the user to be verified passes the match. If the biometric similarities do not reach the preset threshold, it is determined that the match fails. It can generally be applied to scenarios such as face unlocking and fingerprint unlocking of electronic devices.

[0089] When determining the user identity, among the preset biometric images with a biometric similarity greater than the preset threshold to the biometric image to be verified, the user ID information corresponding to the preset biometric image with the highest biometric similarity is determined as the identity information of the user to be verified. If the biometric similarities do not reach the preset threshold, it is determined that the match fails. It can generally be applied to scenarios such as work attendance.

[0090] When performing feature matching, due to the influence of various factors such as device hardware, environmental conditions, and user status, it is not required that the face features in two face images are exactly the same. Instead, a threshold is set. When the similarity of the two face features is greater than this threshold, the two face images can be recognized as the same identity.

[0091] Among them, before performing biometric matching, the biometric image of the user is input into the terminal device as a preset biometric image, and the preset biometric features are extracted from it through a feature recognition model for performing the biometric matching of the embodiments of the present invention. Each preset biometric feature is bound with corresponding user ID information.

[0092] The embodiments of the present invention introduce feature vector confidence to characterize the biometric vectors in the biometric image, that is, use the biometric vector and the biometric vector confidence to jointly describe the biometric image, and comprehensively consider the features and the quality and fuzziness of the features. When determining the similarity value, the obtained similarity value can be corrected based on the feature vector confidence. Compared with the related art, when using only the feature vector to characterize the features of the biometric image, since the collected features may be inaccurate and the error degree is not considered and calculated, resulting in an error in the solution of the similarity value and further causing the problem of unmatched user identities, this solution can ensure the accuracy of biometric recognition in an open scenario.

[0093] As Figure 5 shown, it is a flowchart of a specific embodiment of a biometric matching method provided by the embodiments of the present invention. This embodiment is described by taking face recognition as an example.

[0094] See Figure 5, specifically, collect the face image to be tested (biometric image to be tested). Input the collected face image into the face recognition model (feature recognition model), and let the face recognition model perform convolution on the face image to be tested to extract the first face feature vector (first biometric vector). Input the first face feature vector and the second face feature vectors (second biometric vectors) of each preset face feature image (preset biometric image) into the MLP, and let the MLP output the confidence of the first face feature vector of the first face feature vector (confidence of the first biometric vector), and the confidence of the second face feature vector of the second face feature vector (confidence of the second biometric vector). Determine the first feature distribution of the face image to be tested and the second feature distributions of each preset face image.

[0095] Establish a third feature distribution of the difference between the first feature distribution and each second feature distribution respectively. And solve the face feature similarity value (biometric similarity value) between the face image to be tested and each preset face image through each third feature distribution. When the solved face feature similarity value is greater than the first threshold, it is determined that the face image to be tested corresponds to the same ID information as the preset face image, that is, the matching is successful; while when the solved similarity value is less than the first threshold, it is determined that the face image to be tested corresponds to different ID information from the preset face image, that is, the matching fails.

[0096] Optionally, in some embodiments, in S201, when obtaining the confidence of the first face feature vector of the first face feature vector and the confidence of the second face feature vector of the second face feature vector, if the confidence of the first face feature vector and the confidence of the second face feature vector obtained through the MLP are the same, it means that the image quality of the two images is the same, and there will be no matching error due to image quality problems. Therefore, there is no need to use the feature distribution to solve, and the similarity value can still be directly obtained using the feature vectors.

[0097] Specifically, determine the Euclidean distance between the first face feature vector and the second face feature vector in the face feature space, and determine the determined Euclidean distance as the face feature similarity.

[0098] Optionally, in some embodiments, in order to ensure that there is no error in the obtained biometric similarity value, it is necessary to ensure the accuracy of the biometric confidence output by the MLP in S201. Therefore, it is necessary to update and train the MLP network to ensure its accuracy.

[0099] As Figure 6 shown, it is a flowchart of an MLP training process provided by an embodiment of the present invention. Refer to Figure 6, obtain a number of biometric images as the training set. Each biometric image is set with a corresponding label, and the label is the ID of the user corresponding to the biometric image.

[0100] Obtain the biometric vectors of each biometric image through the feature recognition model. Among them, the feature recognition model is only used to obtain biometric vectors and will not participate in the subsequent MLP training process.

[0101] Input the obtained set of biometric vectors into the MLP in sequence, obtain the biometric vector confidence corresponding to each biometric vector through the MLP, and determine the fourth feature distribution of each biometric image based on the biometric vector and the biometric vector confidence.

[0102] Among this set of biometric images, calculate the target similarity value of the biometric images with the same ID, that is, calculate the target similarity value between different images belonging to the same user. The calculated target similarity value is backpropagated to the MLP model to update and train the output result of the MLP.

[0103] Figure 7 The following is a schematic structural diagram of a biometric matching device provided by an embodiment of the present invention. As Figure 7 shown, the device includes: an acquisition module 701, a determination module 702, and a matching module 703.

[0104] The acquisition module 701 acquires the first feature distribution of the biometric image to be inspected and the second feature distribution of all preset biometric images, and the number of preset biometric images is at least one.

[0105] The determination module 702 determines the biometric similarity value between the biometric image to be inspected and each preset biometric image according to the first feature distribution and each second feature distribution.

[0106] The matching module 703 performs biometric matching according to the biometric similarity value.

[0107] As Figure 8 The following is a schematic hardware structure diagram of an electronic device provided by an embodiment of the present invention. As Figure 8 shown, the electronic device 800 may include a processor 801, an internal memory 802, a camera module 803, etc.

[0108] It can be understood that the structure illustrated in the embodiments of the present application does not constitute a specific limitation on the electronic device 800. In other embodiments of the present application, the electronic device 800 may include more or fewer components than those illustrated, or combine certain components, or split certain components, or have different component arrangements. For example, in a computer device, there may be no antenna, mobile communication module, and wireless communication module. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0109] The processor 801 of the electronic device 800 may be a system-on-chip (SOC). The processor may include a central processing unit (CPU), and may further include other types of processors. For example, the processor 801 may include an application processor (AP) and / or a neural-network processing unit (NPU), etc.

[0110] The processor involved in the processor 801 may include, for example, a CPU, a DSP, a microcontroller, or a digital signal processor, and may further include a GPU, an embedded neural-network processing unit (NPU), and an image signal processor (ISP). The processor 801 may also include necessary hardware accelerators or logic processing hardware circuits, such as an ASIC, or one or more integrated circuits for controlling the execution of the technical solution of the present application. In addition, the processor 801 may have the function of operating one or more software programs, and the software programs may be stored in a storage medium.

[0111] The processor 801 may include one or more processing units. For example, the processor 801 may include an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, a video codec, a digital signal processor (DSP), a baseband processor, etc. Among them, different processing units may be independent components or integrated in one or more processors. In some embodiments, the electronic device 800 may also include one or more processors 801. Among them, the controller may generate operation control signals according to the instruction operation code and timing signals to complete the control of fetching and executing instructions.

[0112] The internal memory 802 of the electronic device 800 may be used to store one or more computer programs, and the one or more computer programs include instructions. The processor 801 may execute the above instructions stored in the internal memory 802, so that the electronic device 800 executes the methods provided in some embodiments of the present application, as well as various applications and data processing, etc. The internal memory 802 may include a code storage area and a data storage area. Among them, the code storage area may store an operating system. The data storage area may store data created during the use of the electronic device 800, etc. In addition, the internal memory 802 may include high-speed random access memory, and may also include non-volatile memory, such as one or more disk storage components, flash memory components, universal flash storage (UFS), etc.

[0113] The internal memory 802 may be a read-only memory (ROM), other types of static storage devices that can store static information and instructions, random access memory (RAM), or other types of dynamic storage devices that can store information and instructions. It may also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or it may also be any computer-readable medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer.

[0114] The processor 801 and the internal memory 802 may be integrated into a processing device. More commonly, they are independent components. The processor 801 is used to execute the program code stored in the internal memory 802 to implement the methods described in the embodiments of the present application. Specifically, in implementation, the internal memory 802 may also be integrated in the processor, or independent of the processor.

[0115] Furthermore, the devices, apparatuses, and modules described in the embodiments of the present application may specifically be implemented by computer chips or entities, or by products with certain functions.

[0116] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, apparatus, or computer program product. Therefore, the present invention can take the form of an all-hardware embodiment, an all-software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0117] In several embodiments provided by the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application.

[0118] Specifically, in one embodiment of the present application, a computer-readable storage medium is further provided. The computer-readable storage medium stores a computer program, and when it runs on a computer, it causes the computer to execute the method provided by the embodiment of the present application.

[0119] One embodiment of the present application further provides a computer program product. The computer program product includes a computer program, and when it runs on a computer, it causes the computer to execute the method provided by the embodiment of the present application.

[0120] The embodiments described in the present application are described with reference to the flowcharts and / or block diagrams of methods, devices (apparatuses), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows 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 the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0121] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements the functions in Figure 1 one process or multiple processes and / or blocksFigure 1 The functions specified in one or more boxes.

[0122] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide for implementing the steps of the functions specified in one or more processes and / or boxes Figure 1 One process or more processes and / or boxes Figure 1 The steps of the functions specified in one or more boxes.

[0123] It should also be noted that in the embodiments of the present application, "at least one" means one or more, and "a plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent the situation where A exists alone, A and B exist simultaneously, or B exists alone. Where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after. "At least one of the following" and its similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0124] In the embodiments of the present application, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, commodity or device including the said element.

[0125] The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in a distributed computing environment where tasks are executed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0126] Each embodiment in this application is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other, and the differences between each embodiment and other embodiments are emphasized. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the corresponding descriptions in the method embodiments.

[0127] Those of ordinary skill in the art will recognize that the units and algorithm steps described in the embodiments of this application can be implemented by a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0128] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described devices, devices, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0129] The above is only the specific implementation manner of this application. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application and should be covered by the protection scope of this application. The protection scope of this application shall be subject to the protection scope of the claims.

Claims

1. A biometric matching method, characterized in that, comprising: obtaining a first feature distribution of a biometric image to be verified, and a second feature distribution of all preset biometric images, the number of the preset biometric images being at least one; determining biometric similarity values between the biometric image to be verified and each preset biometric image according to the first feature distribution and each of the second feature distributions; performing biometric matching according to the biometric similarity values.

2. The method according to claim 1, characterized in that, the first feature distribution and the second feature distribution are Gaussian distributions.

3. The method according to claim 2, characterized in that, the obtaining a first feature distribution of a biometric image to be verified, and a second feature distribution of all preset biometric images, includes: inputting the collected biometric image to be verified into a feature recognition model to obtain a first biometric vector of the biometric image to be verified; inputting the first biometric vector and a second biometric vector of the preset biometric image into a multi-layer perceptron respectively to obtain a first biometric vector confidence level of the first biometric vector and a second biometric vector confidence level of the second biometric vector; determining the first feature distribution according to the first biometric vector and the first biometric vector confidence level; determining the second feature distribution according to the second biometric vector and the second biometric vector confidence level.

4. The method according to claim 3, characterized in that, the determining biometric similarity values between the biometric image to be verified and each preset biometric image according to the first feature distribution and each of the second feature distributions, includes: establishing a third feature distribution according to the first feature distribution and the second feature distribution, the third feature distribution being a distribution of the difference between the first feature distribution and the second feature distribution; when the difference between the first feature distribution and the second feature distribution is zero, determining the value of the third feature distribution according to the first feature vector, the second feature vector, the first biometric vector confidence level, and the second biometric vector confidence level; determining the value of the third feature distribution as the face feature similarity value.

5. The method according to claim 2, characterized in that, the determining biometric similarity values between the biometric image to be verified and each preset biometric image according to the first feature distribution and each of the second feature distributions, the method further includes: when the first biometric vector confidence level is the same as the second biometric vector confidence level, determining the Euclidean distance between the first biometric vector and the second biometric vector; determining the Euclidean distance as the biometric similarity value.

6. The method according to claim 1, characterized in that, the performing biometric matching according to the biometric similarity, includes: When the biometric similarity value between the biometric image to be verified and at least one of the preset biometric images is greater than a preset first threshold, it is determined that the user to be verified corresponding to the biometric image to be verified passes the matching.

7. The method according to claim 6, wherein, for the biometric matching according to the biometric similarity, the method further includes: determining the user identity information corresponding to the preset biometric image with the maximum biometric similarity value as the identity information of the user to be verified.

8. The method according to claim 1, wherein, the method further includes: collecting a set of biometric images as a training set; obtaining the fourth feature distribution of each of the biometric images; determining, according to each of the fourth feature distributions, the target similarity value between the biometric images belonging to the same user identity information; updating the trained multi-layer perceptron according to the target similarity value.

9. A biometric matching device, wherein, it includes: an obtaining module, which obtains the first feature distribution of the biometric image to be verified and the second feature distribution of each preset biometric image; a determining module, which determines the biometric similarity value between the biometric image to be verified and each preset biometric image according to the first feature distribution and each of the second feature distributions; a matching module, which performs biometric matching according to the biometric similarity value.

10. An electronic device, wherein, the electronic device includes a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the electronic device is triggered to execute the method according to any one of claims 1-8.

11. A computer-readable storage medium, wherein, a computer program is stored in the computer-readable storage medium, and when it runs on a computer, the computer is caused to execute the method according to any one of claims 1-8.

Images