Biometric method, apparatus and electronic device

Abstract

The embodiment of the present application provides a biological recognition method, device and electronic equipment. The embodiment dynamically determines an optimal biological feature vector for representing a current state of a biological object according to an existing target feature template and all reference feature templates stored in a buffer corresponding to the target feature template, so as to recognize a biological feature image corresponding to the biological object collected subsequently based on the optimal biological feature vector. This obviously avoids the problem of low biological recognition accuracy caused by the change of some biological features (such as a face) over time or environment change or the poor quality of a base library template used for biological recognition.

Description

TECHNICAL FIELD

[0001] The present application relates to biometric technology, in particular to a biometric method, device and electronic equipment. BACKGROUND

[0002] Biometrics is a technology of using human biological characteristics for identity authentication. The biological characteristics here are, for example, fingerprints, finger veins, faces, irises, palm prints, etc. Compared with the use of external objects such as identity items (such as keys, certificates, ATM cards, etc.) and identity identifiers (such as usernames and passwords, etc.), the accuracy and reliability of biometrics are greatly improved.

[0003] Biometrics determines the identity of a user by comparing an input user biological characteristic with a registered biological characteristic template (also referred to as a registered template). However, in actual applications, there are often situations that affect the accuracy of biometrics, such as some biological characteristics (such as faces) changing over time or with environmental changes, or the quality of the base template used for performing biometrics being relatively poor, etc., which affects biometrics and results in a lower accuracy of biometrics. SUMMARY

[0004] Embodiments of the present application provide a biometric method, device and electronic equipment to improve the accuracy of biometrics.

[0005] The technical solutions provided by the present application include:

[0006] A biometric method, applied to an electronic device, includes:

[0007] From all obtained feature templates, a target feature template matching a target biological feature vector is determined; the target biological feature vector is extracted from a biological feature image corresponding to a biological object to be identified; according to the matching degree of the target biological feature vector and the target feature template, the target biological feature vector is stored as a reference feature template in a buffer area corresponding to the target feature template; the target feature template and each reference feature template already stored in the buffer area have the same dimension;

[0008] If the number of reference feature templates already stored in the buffer area reaches a set storage number corresponding to the buffer area, for each dimension, according to the feature values of the target feature template and all reference feature templates already stored in the buffer area at the dimension, an optimal feature value corresponding to the dimension is determined; according to the optimal feature values corresponding to each dimension, an optimal biological feature vector representing the current state of the biological object is determined, so as to identify the biological feature image corresponding to the biological object collected thereafter according to the optimal biological feature vector and the target feature template.

[0009] A biometric device applied to an electronic device, comprising:

[0010] A determining unit configured to determine a target feature template matching a target biometric feature vector from all obtained feature templates; the target biometric feature vector is extracted from a biometric feature image corresponding to a biometric object to be identified;

[0011] A caching unit configured to store the target biometric feature vector as a reference feature template in a buffer corresponding to the target feature template according to a matching degree between the target biometric feature vector and the target feature template;

[0012] A processing unit configured to, when a number of reference feature templates stored in the buffer reaches a set storage number corresponding to the buffer, determine, for each dimension, an optimal feature value corresponding to the dimension according to feature values of the target feature template and all reference feature templates stored in the buffer in the dimension; determine an optimal biometric feature vector representing a current state of the biometric object according to the optimal feature values corresponding to the dimensions; and identify a biometric feature image corresponding to the biometric object collected subsequently according to the optimal biometric feature vector and the target feature template.

[0013] An electronic device, comprising a processor and a machine-readable storage medium;

[0014] The machine-readable storage medium stores machine-executable instructions executable by the processor;

[0015] The processor is configured to execute the machine-executable instructions to implement steps in the above method

[0016] As can be seen from the above technical solutions, the embodiment determines optimal feature values of each dimension according to feature values of all reference feature templates stored in a buffer corresponding to a target feature template in each dimension, and dynamically determines an optimal biometric feature vector representing a current state of a biometric object based on the optimal feature values of each dimension, so as to identify a biometric feature image corresponding to the biometric object collected subsequently based on the optimal biometric feature vector. This obviously avoids the problem of low biometric recognition accuracy caused by changes in some biometric features (such as human faces) over time or changes in environment or poor quality of a base template used for biometric recognition. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present disclosure and serve to explain the principles of the present disclosure together with the specification.

[0018] Figure 1A flowchart illustrating the method provided in this application embodiment;

[0019] Figure 2 A reference biometric vector extraction flowchart is provided for embodiments of this application;

[0020] Figure 3 A CNN structure diagram provided for embodiments of this application;

[0021] Figure 4a A flowchart illustrating the implementation of step 101 provided in this application embodiment;

[0022] Figure 4b Another implementation flowchart of step 101 provided in the embodiments of this application;

[0023] Figure 4c Another implementation flowchart of step 101 provided in the embodiments of this application;

[0024] Figure 5 This is a schematic diagram showing the correspondence between feature templates and buffers provided in the embodiments of this application;

[0025] Figure 6 A flowchart illustrating the implementation of step 102 provided in this application embodiment;

[0026] Figure 7 This is a schematic diagram of template fusion provided in an embodiment of this application;

[0027] Figure 8 A flowchart illustrating the template fusion implementation provided in this application embodiment;

[0028] Figure 9 This is a schematic diagram illustrating the determination of optimal feature values ​​provided in an embodiment of this application.

[0029] Figure 10 This is a structural diagram of the device provided in the embodiments of this application;

[0030] Figure 11 This is a hardware structure diagram of the device provided in the embodiments of this application. Detailed Implementation

[0031] To make the method provided in this application easier to understand, the method provided in this application will be described in detail below with reference to the accompanying drawings and embodiments:

[0032] See Figure 1 , Figure 1 This is a flowchart illustrating a method provided in an embodiment of this application. This process can be applied to electronic devices. In one example, the electronic device may be a front-end device (such as a terminal), a back-end server, etc., but this embodiment is not specifically limited to these.

[0033] like Figure 1 As shown, the process may include the following steps:

[0034] Step 101: From all the obtained feature templates, determine the target feature template that matches the target biological feature vector.

[0035] As an example, all feature templates obtained in step 101 can be registered feature templates (specifically, already registered feature templates), and / or feature templates after the registered feature templates have been updated. The description of the feature templates will be presented in step 103 below, and will not be repeated here.

[0036] As an example, the target biometric vector in step 101 is extracted from the biometric image corresponding to the biological object to be identified, such as a face or iris. Specifically, there are many ways to extract the target biometric vector from the biometric image corresponding to the biological object, such as a face or iris. For example, a reference biometric vector can be extracted using depth methods, and the target biometric vector is determined based on the reference biometric vector. Taking the iris as the biological object and the iris image as the biometric image as an example, such as... Figure 2 As shown, a reference biological feature vector is extracted using a convolutional neural network (CNN). Figure 3 Examples illustrate the structure of each convolutional layer in a CNN, applied to... Figure 3 The dimension of the reference biometric vector can be 128.

[0037] As an example, there are many ways to determine the target biometric vector based on the reference biometric vector. For instance, if the CNN further includes a normalization layer, the reference biometric vector can be directly determined as the target biometric vector. Alternatively, if the CNN does not include a normalization layer, the reference biometric vector can be input into the pre-set normalization network to obtain the target biometric vector.

[0038] Ultimately, the target biometric vector can be a normalized feature vector, with the feature value in each dimension ranging from [0, 1], in order to perform subsequent biometric identification, which will be described below and will not be repeated here.

[0039] Optionally, in this embodiment, to improve the accuracy of biometric identification, the quality of the biometric image can be scored before extracting the target biometric vector. If the quality score (denoted as S) of the biometric image is greater than or equal to a preset score (indicating that the quality of the biometric image is relatively good), then the above-mentioned step of extracting the target biometric vector is further executed. Here, there are many ways to score the quality of the biometric image, and this embodiment is not specifically limited, as long as it can ensure that the quality of the biometric image is scored.

[0040] After the target biological feature vector is obtained, a target feature template matching the target biological feature vector can be determined from all the obtained feature templates, as described in step 101. In this embodiment, there are many ways to determine the target feature template matching the target biological feature vector from all the obtained feature templates, Figure 4a 、 Figure 4b 、 Figure 4c Three of which are exemplified, which will not be described here.

[0041] In step 102, the target biological feature vector is stored as a reference feature template in the buffer area corresponding to the target feature template according to the matching degree of the target biological feature vector and the target feature template.

[0042] In this embodiment, a corresponding buffer area is introduced for each obtained feature template, and one feature template corresponds to one buffer area, as shown in Figure 5 Initially, each buffer area is empty. In this embodiment, each buffer area is set to have a corresponding set storage quantity, and the set storage quantity of each buffer area is set according to actual conditions. The set storage quantities of the buffer areas can be the same or different, and the embodiment is not specifically limited. Figure 5 For example, the set storage quantities of the buffer areas are uniformly set to 4 reference feature templates (such as 4 biological feature vectors of 128 dimensions). Alternatively, in this embodiment, the target feature template and the reference feature template have the same dimension, such as 128 dimensions.

[0043] It should be noted that in this embodiment, the obtained feature template and the corresponding buffer area can be in the same memory or hard disk, and the embodiment is not specifically limited.

[0044] As an embodiment, there are many ways to store the target biological feature vector as a reference feature template in the buffer area corresponding to the target feature template according to the matching degree of the target biological feature vector and the target feature template in this step 102, Figure 6 For example, one of the implementation ways is exemplified, which will not be described here.

[0045] In step 103, if the number of reference feature templates stored in the buffer area reaches the set storage quantity corresponding to the buffer area, the optimal feature value corresponding to each dimension is determined according to the feature values of the target feature template and all the reference feature templates stored in the buffer area in the dimension, the optimal biological feature vector representing the current state of the biological object is determined according to the optimal feature values corresponding to each dimension, and the biological feature image of the biological object collected thereafter is identified according to the optimal biological feature vector and the target feature template.

[0046] As an embodiment, in the step 103, the optimal biometric feature vector used to represent the current state of the biological object can be determined by fusing the feature values of each template in each dimension, according to the target feature template and all the reference feature templates stored in the buffer area corresponding to the target feature template. For example, as shown in Figure 7 , the stored four reference feature templates and the target feature template in the buffer area corresponding to the target feature template are fused in each dimension. For each dimension, the optimal feature value corresponding to the dimension is determined according to the feature values of the target feature template and all the reference feature templates stored in the buffer area corresponding to the target feature template in the dimension. The optimal biometric feature vector used to represent the current state of the biological object is determined according to the optimal feature values corresponding to each dimension. Figure 8 An implementation manner of template fusion is described by way of example, which is not described herein.

[0047] In the embodiment, there are many implementation manners for identifying the biological feature image of the biological object collected subsequently in the step 103 according to the optimal biometric feature vector and the target feature template. As one of the implementation manners, the target feature template in the original template database can be updated. For example, the optimal biometric feature vector (i.e., a new template) replaces the target feature template in the original template database, and each reference feature template in the buffer area corresponding to the target feature template is emptied. Then, the process shown in Figure 1 is executed based on the updated target feature template to identify the biological feature image of the biological object collected subsequently.

[0048] As another implementation form, the embodiment can also not update the target feature template in the original template database, but apply corresponding weight parameters to each dimension of the target feature template to improve the biometric recognition performance. For example, for each dimension, the weight parameter corresponding to the dimension is determined according to the feature value in the dimension of the target feature template and the optimal biometric feature vector. The weight parameters corresponding to each dimension are combined to form a weight parameter vector corresponding to the target feature template. When the target biometric feature vector in the biological feature image of the biological object is collected subsequently, the process shown in Figure 1 is executed based on the target feature template and the weight parameter vector corresponding to the target feature template.

[0049] Optionally, in the embodiment, there are many implementation manners to determine the weight parameter corresponding to each dimension. For example, for each dimension, the feature value in the target feature template and the above optimal biometric feature vector in the dimension is obtained first. For the convenience of description, the feature value in the target feature template in the dimension is denoted as a first feature value, and the feature value in the optimal biometric feature vector in the dimension is denoted as a second feature value. Then, the gap between the first feature value and the second feature value is determined. For example, the first feature value is 0.21, and the second feature value is 0.68. The gap (denoted as Δ) between the first feature value and the second feature value can be represented by the following formula: Δ = |0.68-0.21| = 0.47. The weight parameter corresponding to the dimension is determined according to the above gap.

[0050] In the embodiment, the greater the gap between the feature values in the target feature template and the above optimal biometric feature vector in the same dimension (such as the first feature value and the second feature value described above), the more unreliable the feature value in the target feature template in the dimension (the first feature value described above) is. It can be considered that the probability of the feature value being noise is relatively large. In order to reflect this degree of untrustworthiness, the embodiment can apply a corresponding weight parameter α n (n represents the number of dimensions) to the dimension. The lower the weight parameter, the more unreliable the feature value in the dimension is, that is, the less the feature value in the dimension is adopted.

[0051] Based on this, in the embodiment, for each dimension, the weight parameter α n corresponding to the dimension can be determined according to the gap between the feature values in the target feature template and the above optimal biometric feature vector in the dimension. The greater the gap, the smaller the weight parameter.

[0052] As an example, α n can be represented by the following formula:

[0053] α n = 1-Δ.

[0054] Wherein, Δ is the gap between the feature values in the target feature template and the above optimal biometric feature vector in the dimension n, and α n represents the weight parameter corresponding to the dimension n.

[0055] If the above target feature template and the above optimal biometric feature vector are both 128-dimensional vectors, based on the above description, a 128-dimensional weight parameter vector (denoted as an α vector) can be finally determined. This way of applying weights to weaken the unreliable feature values or noise values in the feature template can improve the final recognition effect.

[0056] As described above, the greater the gap between the feature value in the same dimension in the target feature template and the optimal biometric feature vector (such as the first feature value and the second feature value described above), the more unreliable the feature value (the first feature value described above) in the same dimension in the target feature template, and the greater the probability that the feature value is noise. When the gap between the feature value in the same dimension in the target feature template and the optimal biometric feature vector (such as the first feature value and the second feature value described above) is greater than or equal to a certain threshold, as an embodiment, the weight parameter corresponding to the dimension can be directly determined as a specified weight parameter such as 0. After that, when the weight parameter in any dimension in the weight parameter vector corresponding to the feature template is a specified weight parameter or less than or equal to a certain weight threshold (such as 0.01, etc., which can be set according to actual needs), the feature value in the dimension in the feature template can be excluded when the target biometric feature vector is matched with the target feature template in the subsequent process, so as to ensure that the feature value in the dimension does not participate in template matching, weaken the untrusted feature value or noise value in the feature template, and thus improve the final recognition effect. For details, see Figure 4c the flow shown in FIG. 4.

[0057] At this point, the process shown in Figure 1 FIG. 4 is completed.

[0058] As can be seen from the process shown in Figure 1 , the embodiment dynamically determines the optimal biometric feature vector representing the current state of the biological object according to the existing target feature template and all reference feature templates stored in the buffer area corresponding to the target feature template, so as to identify the biometric feature image of the biological object collected subsequently based on the optimal biometric feature vector. This obviously avoids the problem of low biometric recognition accuracy caused by the change of some biometric features (such as human face) over time or environment or the poor quality of the template used in biometric recognition.

[0059] The process shown in Figure 4a FIG. 4 will be described below.

[0060] Referring to Figure 4a , Figure 4a the step 101 flowchart provided by the embodiment of the present application. As shown in Figure 4a , the flowchart can include the following steps:

[0061] Step 401a, determining the matching degree between the target biometric feature vector and each feature template obtained.

[0062] In this embodiment, the obtained feature template does not have a corresponding weight vector parameter, such as in the scenario of updating the target feature template in the original template library, the feature template does not have a corresponding weight vector parameter, and the like, which is not specifically limited in this embodiment. Based on this, in step 401a, for each obtained feature template, there are many ways to determine the matching degree between the target biometric feature vector and the feature template, such as in the deep method, the cosine similarity between the target biometric feature vector and the feature template can be calculated to represent the matching degree, or the Euclidean distance between the target biometric feature vector and the feature template can be calculated to represent the matching degree, and the like, which is not specifically limited in this embodiment.

[0063] Step 402a, selecting the target feature template according to the matching degree between the target biometric feature vector and each feature template.

[0064] Here, the matching degree between the target biometric feature vector and the target feature template is greater than or equal to the matching degree between the target biometric feature vector and any other obtained feature template. For example, the maximum matching degree is selected from the matching degrees between the target biometric feature vector and each feature template, and it is determined whether the maximum matching degree is greater than or equal to a preset matching degree value. If yes, the feature template with the maximum matching degree to the target biometric feature vector is determined as the target feature template. Of course, if the maximum matching degree is less than the preset matching degree value, the biological object can be directly determined as an unregistered biological object, and the current process is ended.

[0065] At this point, the process shown in Figure 4a is completed.

[0066] By Figure 4a , it is realized how to determine the target feature template matched with the target biometric feature vector from all obtained feature templates.

[0067] Referring to Figure 4b , Figure 4b is another implementation flowchart of step 101 provided by the embodiments of the present application. As Figure 4b shown, the flowchart can include the following steps:

[0068] Step 401b, for each obtained feature template, determining the matching degree between the target biometric feature vector and the feature template according to the feature values of the feature template and the target biometric feature vector in each dimension, and the weight parameters of the weight parameter vector corresponding to the feature template in each dimension.

[0069] For example, for each dimension, the comparison result of the feature values ​​of the feature template and the target biofeature vector in that dimension (such as the difference or Euclidean distance) is multiplied by the weight parameter in the weight parameter vector corresponding to the feature template in that dimension to determine the matching degree between the target biofeature vector and the feature template in that dimension. Then, the matching degrees in each dimension are accumulated to finally obtain the matching degree between the target biofeature vector and the feature template.

[0070] Taking the comparison result as Euclidean distance as an example, the matching degree (represented by P) between the target biometric feature vector and the feature template can be calculated by the following formula:

[0071]

[0072] Where i is the dimension number, α i Let x be the weight parameter in the i-th dimension of the weight parameter vector. i Let X be the feature value in the i-th dimension of the target biological feature vector. i is the feature value in the i-th dimension of the feature template, and n is the total number of dimensions.

[0073] Step 402b: Select the target feature template based on the matching degree between the target biometric vector and each feature template.

[0074] Step 402b is similar to step 402a above, and will not be described again here.

[0075] This concludes the process. Figure 4b The process is shown below.

[0076] pass Figure 4b It demonstrates how to determine the target feature template that matches the target biological feature vector from all the obtained feature templates.

[0077] See Figure 4c , Figure 4c This is another implementation flowchart of step 101 provided in an embodiment of this application. For example... Figure 4c As shown, the process may include the following steps:

[0078] Step 401c: For each obtained feature template, based on the weight parameter vector corresponding to the feature template, exclude feature values ​​in at least one dimension of the feature template; wherein, the dimension of the weight parameter vector is the same as the dimension of the feature template, and when the weight parameter in any dimension of the weight parameter vector is a specified weight parameter or less than or equal to a set weight threshold, the feature value in the feature template in that dimension is excluded.

[0079] The specified weight parameters and weight thresholds are as described above, and will not be repeated here.

[0080] Step 402c, for each remaining dimension of the feature template, determining the matching degree between the target biometric feature vector and the feature template according to the feature value of the feature template in the remaining dimension, the feature value of the target biometric feature vector in the remaining dimension, and the weight parameter of the weight parameter vector corresponding to the feature template in the remaining dimension.

[0081] For example, for each remaining dimension of the feature template, multiplying the comparison result (such as the difference value or the Euclidean distance, etc.) of the feature values of the feature template and the target biometric feature vector in the dimension by the weight parameter of the weight parameter vector corresponding to the feature template in the dimension to determine the matching degree of the target biometric feature vector and the feature template in the dimension, and then accumulating the matching degrees in each dimension to finally obtain the matching degree between the target biometric feature vector and the feature template. For details, please refer to the above description, which will not be repeated here.

[0082] Step 403c, selecting the target feature template according to the matching degrees between the target biometric feature vector and each feature template.

[0083] This step 403c is similar to the above step 402b or step 402a, which will not be repeated here.

[0084] So far, the process shown in Figure 4c is completed.

[0085] By Figure 4c , it is realized how to determine the target feature template matched with the target biometric feature vector from all the obtained feature templates. And it is further realized that when performing template matching, not the entire feature template is matched, but the noise values (feature values in at least one dimension) that are not reliable or not reliable in the feature template are excluded, the feature values or noise values that are not reliable in the feature template are weakened, so as to avoid the target feature template that is not accurate due to the noise values (feature values in at least one dimension) that are not reliable or not reliable in the feature template participating in the matching, and to affect the accuracy of subsequent biometric recognition.

[0086] The process shown in Figure 6 will be described below:

[0087] Referring to Figure 6 , Figure 6 , the step 102 provided by the embodiment of the present application realizes the flowchart. As shown in Figure 6 , the process can include the following steps:

[0088] Step 601, if the image quality score of the obtained biometric feature image is greater than or equal to the preset image quality score threshold, and the matching degree between the target biometric feature vector and the target feature template is greater than or equal to the preset matching degree, then step 602 is performed.

[0089] That is, in the embodiment, before storing the target biometric feature vector as a reference feature template into the buffer area corresponding to the target feature template, it is verified whether the image quality score of the biometric feature image and the matching degree between the target biometric feature vector and the target feature template meet the requirements, such as judging whether the image quality score of the biometric feature image is greater than or equal to a preset image quality score threshold and whether the matching degree between the target biometric feature vector and the target feature template is greater than or equal to a preset matching degree. Once the image quality score of the biometric feature image and the matching degree between the target biometric feature vector and the target feature template meet the requirements, the step 602 of storing the target biometric feature vector as a reference feature template into the buffer area corresponding to the target feature template is executed. Of course, if the image quality score of the biometric feature image and / or the matching degree between the target biometric feature vector and the target feature template do not meet the requirements, the step 602 is not executed.

[0090] Step 602, storing the target biometric feature vector as a reference feature template into the buffer area corresponding to the target feature template.

[0091] At this point, the process shown in Figure 6 is completed.

[0092] Through the process shown in Figure 6 , the step 102 of how to store the target biometric feature vector as a reference feature template into the buffer area corresponding to the target feature template according to the matching degree between the target biometric feature vector and the target feature template is implemented.

[0093] The process shown in Figure 8 will be described below.

[0094] Referring to Figure 8 , Figure 8 is a flowchart of the process of determining the optimal biometric feature vector in step 103 provided by the embodiment of the application. As shown in Figure 8 , the process can include the following steps.

[0095] Step 801, obtaining, for each dimension, the feature value of the target feature template and all the reference feature templates stored in the buffer area in the dimension.

[0096] Step 802, determining the current representative value of the dimension according to the distribution of the feature values of the target feature template and all the reference feature templates stored in the buffer area in the dimension; the current representative value is used to indicate that the feature values of the target feature template and all the reference feature templates stored in the buffer area in the dimension tend to the current representative value, and the number of feature values that satisfy the set difference requirement between the current representative value and the dimension is the largest.

[0097] In this embodiment, the target feature template and the reference feature templates stored in the buffer corresponding to the target feature template have the same dimension, for example, 128 dimensions.

[0098] In this embodiment, when determining the current representative value of each dimension, it can be based on the concentrated distribution trend of the feature values ​​under that dimension, so as to successively select the most stable reliable value (i.e. the current representative value) under each dimension from the target feature template and all the reference feature templates stored in the buffer corresponding to the target feature template.

[0099] For example, such as Figure 9 As shown, for each dimension, the feature values ​​of the target feature template and all reference feature templates stored in the corresponding buffer are arranged in descending or ascending order to form a sequence. The distance between adjacent feature values ​​in the sequence is calculated. L feature values ​​in the sequence that satisfy a condition (L > 1) are counted, where the distance between any two adjacent feature values ​​is less than a set distance, for example... Figure 9 The sequence shown has three feature values: 0.66, 0.67, and 0.71. The average of these feature values ​​is then calculated to obtain the result. Based on this result, the current representative value for that dimension is determined; for example, the result can be directly used to determine the current representative value for that dimension. Let L feature values ​​be... Figure 9 Given the three feature values ​​shown: 0.66, 0.67, and 0.71, the average value of 0.68 (0.68) can be determined as the current representative value for this dimension.

[0100] For example, based on the distribution of feature values ​​under each dimension and according to the clustering algorithm, the current representative value of each dimension (which is generally a centrally distributed value) can be determined. Specifically, for each dimension, the feature values ​​of the target feature template and all reference feature templates stored in the buffer corresponding to the target feature template can be clustered under that dimension based on the clustering algorithm to obtain the clusters under that dimension. From the clusters under that dimension, the target cluster with the highest feature value density is determined. The average value of each feature value in the target cluster is calculated to obtain the calculation result. Based on the calculation result, the optimal feature value corresponding to that dimension is determined, for example, the calculation result is directly used to determine the current representative value of that dimension.

[0101] Ultimately, for each dimension, the current representative value for that dimension will be determined.

[0102] Step 803: Determine the optimal feature value corresponding to the dimension based on the current representative value of the dimension.

[0103] In the embodiment, the current representative value of the dimension can be directly determined as the optimal feature value corresponding to the dimension, or the current representative value of the dimension is multiplied by the current weight parameter of the dimension, and the obtained result is determined as the optimal feature value corresponding to the dimension. The embodiment is not specifically limited.

[0104] After the optimal feature values corresponding to the dimensions are determined, the optimal feature values corresponding to the dimensions, such as the optimal feature values corresponding to 128 dimensions respectively, can form a new feature vector. The feature vector can be the optimal biological feature vector described above.

[0105] Optionally, after the optimal biological feature vector used for representing the current state of the biological object is determined, all reference feature templates stored in the buffer area corresponding to the target feature template can be further deleted, so that subsequent biological feature images corresponding to the biological object are identified according to the optimal biological feature vector and the target feature template. Figure 1 The flow continues to circulate.

[0106] At this point, the flow shown in Figure 8 is completed.

[0107] Through the flow shown in Figure 8 , how to determine the optimal feature values corresponding to the dimensions in the step 103 described above is implemented.

[0108] The method provided by the present application is described above, and the device provided by the present application is described below:

[0109] Referring to Figure 10 , Figure 10 , a device structure diagram provided by an embodiment of the present application. The device is applied to an electronic device and includes:

[0110] A determination unit is configured to determine a target feature template matched with a target biological feature vector from all obtained feature templates; the target biological feature vector is extracted from a biological feature image corresponding to a biological object to be identified;

[0111] A cache unit is configured to store the target biological feature vector as a reference feature template in a buffer area corresponding to the target feature template according to a matching degree of the target biological feature vector and the target feature template;

[0112] A processing unit is configured to, when the number of reference feature templates stored in the buffer area reaches a set storage number corresponding to the buffer area, determine, for each dimension, an optimal feature value corresponding to the dimension according to feature values of the target feature template and all reference feature templates stored in the buffer area in the dimension; determine an optimal biological feature vector used for representing a current state of the biological object according to the optimal feature values corresponding to the dimensions; and identify biological feature images corresponding to the biological object collected subsequently according to the optimal biological feature vector and the target feature template.

[0113] Optionally, the step of storing the target biometric feature vector as a reference feature template into the buffer corresponding to the target feature template according to the matching degree between the target biometric feature vector and the target feature template comprises: if the image quality score of the obtained biometric image is greater than or equal to a preset image quality score threshold, and the matching degree between the target biometric feature vector and the target feature template is greater than or equal to a preset matching degree, then storing the target biometric feature vector as a reference feature template into the buffer corresponding to the target feature template.

[0114] Optionally, the step of determining the optimal feature value corresponding to each dimension according to the feature values of the target feature template and all the reference feature templates stored in the buffer in the dimension comprises: obtaining the feature values of the target feature template and all the reference feature templates stored in the buffer in each dimension; determining a current representative value of the dimension according to the distribution of the feature values of the target feature template and all the reference feature templates stored in the buffer in the dimension; the current representative value is used to indicate that the feature values of the target feature template and all the reference feature templates stored in the buffer in the dimension are close to the current representative value, and the number of feature values that satisfy a set gap requirement between the current representative value is the largest; determining the optimal feature value corresponding to the dimension according to the current representative value of the dimension.

[0115] The step of determining the current representative value of the dimension according to the distribution of the feature values of the target feature template and all the reference feature templates stored in the buffer in the dimension comprises: arranging the feature values of the target feature template and all the reference feature templates stored in the buffer in the dimension in order to form a sequence; calculating the interval between adjacent feature values in the sequence; counting L feature values that satisfy a condition in the sequence, L is greater than 1, and the condition means that the interval between any two adjacent feature values in the L feature values is less than a set interval; performing average value calculation on the counted L feature values to obtain a calculation result; determining the current representative value of the dimension according to the calculation result; or,

[0116] arranging the feature values of the target feature template and all the reference feature templates stored in the buffer in each dimension according to a clustering algorithm to form a target cluster; determining the current representative value of the dimension based on the feature values in the target cluster; the target cluster refers to one of the clusters corresponding to the dimension, and the density of the feature values in the target cluster is the highest compared with other clusters corresponding to the dimension.

[0117] Optionally, the step of identifying the biological feature image corresponding to the biological object collected later according to the optimal biological feature vector and the target feature template comprises: updating the target feature template to the optimal biological feature vector, so that when a target biological feature vector is collected in the biological feature image corresponding to the biological object collected later, the step of determining the target feature template matching the target biological feature vector from all the obtained feature templates is returned.

[0118] Optionally, the step of determining the target feature template matching the target biological feature vector from all the obtained feature templates comprises: determining the matching degree between the target biological feature vector and each obtained feature template; and selecting the target feature template according to the matching degree between the target biological feature vector and each feature template, wherein the matching degree between the target biological feature vector and the target feature template is greater than or equal to the matching degree between the target biological feature vector and any other obtained feature template.

[0119] Optionally, the step of identifying the biological feature image corresponding to the biological object collected later according to the optimal biological feature vector and the target feature template comprises: determining the weight parameter corresponding to each dimension according to the feature values of the target feature template and the optimal biological feature vector in the dimension; wherein the greater the difference between the feature values of the target feature template and the optimal biological feature vector in the dimension, the lower the weight parameter corresponding to the dimension; and determining the weight parameter vector corresponding to the target feature template by using the weight parameters corresponding to each dimension, so that when a target biological feature vector is collected in the biological feature image corresponding to the biological object collected later, the step of determining the target feature template matching the target biological feature vector from all the obtained feature templates is returned based on the target feature template and the weight parameter vector corresponding to the target feature template.

[0120] Optionally, the step of determining the target feature template matching the target biological feature vector from all the obtained feature templates comprises: determining the matching degree between the target biological feature vector and each obtained feature template according to the feature values of the target biological feature vector and the feature template in each dimension, and the weight parameters in each dimension of the weight parameter vector corresponding to the feature template; and selecting the target feature template according to the matching degree between the target biological feature vector and each feature template, wherein the matching degree between the target biological feature vector and the target feature template is greater than or equal to the matching degree between the target biological feature vector and any other obtained feature template.

[0121] Optionally, determining the target feature template that matches the target biometric vector from all obtained feature templates includes: for each obtained feature template, excluding feature values ​​in at least one dimension of the feature template based on the weight parameter vector corresponding to the feature template; wherein the dimension of the weight parameter vector is the same as the dimension of the feature template, and when the weight parameter in any dimension of the weight parameter vector is a specified weight parameter or less than or equal to a set weight threshold, the feature value in that dimension of the feature template is excluded; for each remaining dimension of the feature template, determining the matching degree between the target biometric vector and the feature template based on the feature value in the remaining dimension of the feature template, the feature value of the target biometric vector in the remaining dimension, and the weight parameter of the weight parameter vector corresponding to the obtained feature template in the remaining dimension; selecting a target feature template based on the matching degree between the target biometric vector and each feature template; the matching degree between the target biometric vector and each feature template is greater than or equal to the matching degree between the target biometric vector and any other obtained feature template;

[0122] Optionally, after determining the optimal biological feature vector to represent the current state of the biological object, the processing unit further deletes all reference feature templates already stored in the buffer.

[0123] This concludes the process. Figure 10 Structural description of the device shown.

[0124] Correspondingly, this application also provides Figure 10 The hardware structure of the device shown. See also Figure 11 The hardware structure may include: a processor and a machine-readable storage medium, the machine-readable storage medium storing machine-executable instructions that can be executed by the processor; the processor is used to execute the machine-executable instructions to implement the method disclosed in the above example of this application.

[0125] Based on the same application concept as the above method, this application embodiment also provides a machine-readable storage medium storing a plurality of computer instructions, which, when executed by a processor, can implement the method disclosed in the above examples of this application.

[0126] For example, the aforementioned machine-readable storage medium can be any electronic, magnetic, optical, or other physical storage device that can contain or store information such as executable instructions, data, etc. For instance, machine-readable storage media can be: RAM (Random Access Memory), volatile memory, non-volatile memory, flash memory, storage drives (such as hard disk drives), solid-state drives, any type of storage disk (such as optical discs, DVDs, etc.), or similar storage media, or combinations thereof.

[0127] The systems, apparatuses, modules, or units illustrated by the above embodiments can be implemented by entities or products with certain functions. A typical implementation device is a computer, and the specific form of the computer can be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0128] For the convenience of description, the above apparatuses are described as various units by functions respectively in the description. Of course, the functions of the units can be implemented in a software and / or hardware manner in the same or multiple software and / or hardware in the implementation of the present application.

[0129] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0130] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce the functions described in the flowcharts and / or block diagrams for implementing the flows and / or blocks in the flowcharts and / or block diagrams. Figure 1 The functions specified in the flow or the flows and / or the block or the blocks. Figure 1 The apparatus that implements the functions specified in the flow or the flows and / or the block or the blocks.

[0131] Moreover, these computer program instructions can also be stored in a computer-readable memory capable of directing the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction apparatus, which implements the functions described in the flowcharts and / or block diagrams for implementing the flows and / or blocks in the flowcharts and / or block diagrams. Figure 1 The functions specified in the flow or the flows and / or the block or the blocks. Figure 1 The apparatus that implements the functions specified in the flow or the flows and / or the block or the blocks.

[0132] These computer program instructions can also be loaded into computer or other programmable data processing devices, so that a series of operation steps are performed on the computer or other programmable data processing devices to generate computer-implemented processes, thus the instructions executed on the computer or other programmable data processing devices provide processes for implementing the functions specified in the flowcharts Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0133] The above only describes the embodiments of the present application and is not intended to limit the present application. The present application can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. A biometric identification method, characterized in that, This method is applied to electronic devices, including: From all the obtained feature templates, a target feature template that matches the target biometric vector is determined; the target biometric vector is extracted from the biometric image corresponding to the biological object to be identified; based on the matching degree between the target biometric vector and the target feature template, the target biometric vector is stored as a reference feature template in the buffer corresponding to the target feature template; the target feature template and each reference feature template already stored in the buffer have the same dimension; If the number of reference feature templates stored in the buffer reaches the set storage limit corresponding to the buffer, then for each dimension, the optimal feature value corresponding to that dimension is determined based on the feature values ​​of the target feature template and all reference feature templates stored in the buffer. Based on the optimal feature values ​​corresponding to each dimension, the optimal biological feature vector representing the current state of the biological object is determined, so as to identify the biological feature image corresponding to the biological object collected subsequently based on the optimal biological feature vector and the target feature template.

2. The method according to claim 1, characterized in that, The step of storing the target biometric vector as a reference feature template in the buffer corresponding to the target feature template based on the matching degree between the target biometric vector and the target feature template includes: If the image quality score of the obtained biometric image is greater than or equal to a preset image quality score threshold, and the matching degree between the target biometric vector and the target feature template is greater than or equal to a preset matching degree, then the target biometric vector is stored as a reference feature template in the buffer corresponding to the target feature template.

3. The method according to claim 1, characterized in that, The step of determining the optimal feature value for each dimension based on the target feature template and the feature values ​​of all reference feature templates stored in the buffer for that dimension includes: For each dimension, obtain the feature values ​​of the target feature template and all reference feature templates stored in the buffer under that dimension; Based on the distribution of feature values ​​of the target feature template and all reference feature templates stored in the buffer in this dimension, the current representative value of this dimension is determined; the current representative value is used to indicate that the feature values ​​of the target feature template and all reference feature templates stored in the buffer in this dimension should move closer to the current representative value, and the number of feature values ​​in this dimension that meet the set gap requirement is the largest. Based on the current representative value of this dimension, determine the optimal feature value corresponding to this dimension.

4. The method according to claim 3, characterized in that, The step of determining the current representative value of a dimension based on the distribution of feature values ​​of the target feature template and all reference feature templates stored in the buffer in that dimension includes: For each dimension, the feature values ​​of the target feature template and all reference feature templates stored in the buffer under that dimension are arranged in order to form a sequence; the distance between adjacent feature values ​​in the sequence is calculated; L feature values ​​in the sequence that meet the conditions are counted, and the average of the counted L feature values ​​is calculated to obtain the calculation result; the current representative value of that dimension is determined based on the calculation result; wherein, L is greater than 1, and the condition means that the distance between any two adjacent feature values ​​among the L feature values ​​is less than a set distance; or, For each dimension, the feature values ​​of the target feature template and all reference feature templates stored in the buffer are clustered according to the clustering algorithm. The current representative value of the dimension is determined based on the feature values ​​in the target cluster. The target cluster refers to one of the clusters corresponding to the dimension, which has the highest density of feature values ​​compared to other clusters corresponding to the dimension.

5. The method according to claim 1, characterized in that, Based on the optimal biometric vector and the target feature template, the identification of the biometric image corresponding to the subsequently acquired biological object includes: For each dimension, a weight parameter corresponding to that dimension is determined based on the feature value in the target feature template and the feature value in the optimal biofeature vector that falls within that dimension; wherein, for each dimension, the greater the difference between the feature value in the target feature template and the feature value in the optimal biofeature vector that falls within that dimension, the lower the weight parameter corresponding to that dimension. By utilizing the weight parameters corresponding to each dimension, the weight parameter vector corresponding to the target feature template is determined. When the target biological feature vector in the biological feature image corresponding to the biological object is subsequently collected, the step of determining the target feature template that matches the target biological feature vector from all the obtained feature templates is returned based on the target feature template and the weight parameter vector corresponding to the target feature template.

6. The method according to claim 1 or 5, characterized in that, The step of determining the target feature template that matches the target biometric feature vector from all the obtained feature templates includes: For each feature template that has been obtained, the matching degree between the target biofeature vector and the feature template is determined based on the feature values ​​of the feature template and the target biofeature vector in each dimension, as well as the weight parameters of the weight parameter vector corresponding to the obtained feature template in each dimension. Based on the matching degree between the target biometric vector and each feature template, a target feature template is selected; the matching degree between the target biometric vector and each feature template is greater than or equal to the matching degree between the target biometric vector and any other obtained feature template.

7. The method according to claim 1 or 5, characterized in that, The step of determining the target feature template that matches the target biometric feature vector from all the obtained feature templates includes: For each obtained feature template, based on the weight parameter vector corresponding to the feature template, feature values ​​in at least one dimension of the feature template are excluded; wherein, the dimension of the weight parameter vector is the same as the dimension of the feature template, and when the weight parameter in any dimension of the weight parameter vector is a specified weight parameter or less than or equal to a set weight threshold, the feature values ​​in the feature template in that dimension are excluded. For each remaining dimension in the feature template, the matching degree between the target biological feature vector and the feature template is determined based on the feature value of the feature template in the remaining dimension, the feature value of the target biological feature vector in the remaining dimension, and the weight parameter of the weight parameter vector corresponding to the feature template in the remaining dimension. Based on the matching degree between the target biometric vector and each feature template, a target feature template is selected; the matching degree between the target biometric vector and each feature template is greater than or equal to the matching degree between the target biometric vector and any other obtained feature template.

8. The method according to claim 1, characterized in that, After determining the optimal biological feature vector to represent the current state of the biological object, the method further includes deleting all reference feature templates stored in the buffer.

9. A biometric identification device, characterized in that, This device is used in electronic devices, including: The determining unit is used to determine the target feature template that matches the target biometric vector from all the obtained feature templates; the target biometric vector is extracted from the biometric image corresponding to the biological object to be identified. A caching unit is used to store the target biometric vector as a reference feature template in the buffer corresponding to the target feature template based on the matching degree between the target biometric vector and the target feature template. The processing unit is configured to, when the number of reference feature templates stored in the buffer reaches the set storage quantity corresponding to the buffer, determine the optimal feature value corresponding to each dimension based on the target feature template and the feature values ​​of all reference feature templates stored in the buffer for that dimension; and determine the optimal biofeature vector representing the current state of the biological object based on the optimal feature values ​​corresponding to each dimension, so as to identify the biofeature image corresponding to the biological object acquired subsequently based on the optimal biofeature vector and the target feature template.

10. The apparatus according to claim 9, characterized in that, The step of storing the target biometric vector as a reference feature template in the buffer corresponding to the target feature template based on the matching degree between the target biometric vector and the target feature template includes: if the image quality score of the obtained biometric image is greater than or equal to a preset image quality score threshold, and the matching degree between the target biometric vector and the target feature template is greater than or equal to a preset matching degree, then the target biometric vector is stored as a reference feature template in the buffer corresponding to the target feature template; The step of determining the optimal feature value for each dimension based on the feature values ​​of the target feature template and all reference feature templates stored in the buffer includes: obtaining the feature values ​​of the target feature template and all reference feature templates stored in the buffer for each dimension; determining the current representative value for the dimension based on the distribution of the feature values ​​of the target feature template and all reference feature templates stored in the buffer; the current representative value is used to indicate that the feature values ​​of the target feature template and all reference feature templates stored in the buffer in the dimension tend to converge towards the current representative value, and the number of feature values ​​in the dimension whose difference from the current representative value meets the set difference requirement is the largest; and determining the optimal feature value for the dimension based on the current representative value. The step of determining the current representative value of a dimension based on the distribution of feature values ​​of the target feature template and all reference feature templates stored in the buffer in that dimension includes: for each dimension, arranging the feature values ​​of the target feature template and all reference feature templates stored in the buffer in that dimension in order to form a sequence; calculating the distance between adjacent feature values ​​in the sequence; counting L feature values ​​in the sequence that satisfy the condition that L is greater than 1, where the condition is that the distance between any two adjacent feature values ​​among the L feature values ​​is less than a set distance; calculating the average of the L counted feature values ​​to obtain the calculation result; and determining the current representative value of the dimension based on the calculation result; or... For each dimension, the feature values ​​of the target feature template and all reference feature templates stored in the buffer under that dimension are clustered according to a clustering algorithm, and the current representative value of that dimension is determined based on the feature values ​​in the target cluster; the target cluster refers to one of the clusters corresponding to that dimension, and the target cluster has the highest density of feature values ​​compared with other clusters corresponding to that dimension. The process of identifying the biometric images corresponding to the biological object acquired subsequently, based on the optimal biometric vector and the target feature template, includes: for each dimension, determining the weight parameter corresponding to that dimension based on the feature values ​​in the target feature template and the optimal biometric vector within that dimension; wherein, for each dimension, the greater the difference between the feature values ​​in the target feature template and the optimal biometric vector within that dimension, the lower the weight parameter corresponding to that dimension; using the weight parameters corresponding to each dimension, determining the weight parameter vector corresponding to the target feature template, so that when the target biometric vector in the biometric image corresponding to the biological object is acquired subsequently, based on the target feature template and the weight parameter vector corresponding to the target feature template, the step of determining the target feature template matching the target biometric vector from all obtained feature templates is returned; The step of determining the target feature template that matches the target biometric vector from all the obtained feature templates includes: for each obtained feature template, determining the matching degree between the target biometric vector and the feature template based on the feature values ​​of the feature template and the target biometric vector in each dimension, and the weight parameters of the weight parameter vector corresponding to the feature template in each dimension; selecting a target feature template based on the matching degree between the target biometric vector and each feature template; the matching degree between the target biometric vector and each feature template is greater than or equal to the matching degree between the target biometric vector and any other obtained feature template; Alternatively, determining the target feature template that matches the target biometric vector from all obtained feature templates includes: for each obtained feature template, excluding feature values ​​in at least one dimension of the feature template based on the weight parameter vector corresponding to the feature template; wherein the dimension of the weight parameter vector is the same as the dimension of the feature template, and when the weight parameter in any dimension of the weight parameter vector is a specified weight parameter or less than or equal to a set weight threshold, the feature value in that dimension of the feature template is excluded; for each remaining dimension of the feature template, determining the matching degree between the target biometric vector and the feature template based on the feature value in the remaining dimension of the feature template, the feature value of the target biometric vector in the remaining dimension, and the weight parameter of the weight parameter vector corresponding to the obtained feature template in the remaining dimension; selecting a target feature template based on the matching degree between the target biometric vector and each feature template; the matching degree between the target biometric vector and each feature template is greater than or equal to the matching degree between the target biometric vector and any other obtained feature template; After determining the optimal biological feature vector to represent the current state of the biological object, the processing unit further deletes all reference feature templates stored in the buffer.

11. An electronic device, characterized in that, The electronic device includes: a processor and a machine-readable storage medium; The machine-readable storage medium stores machine-executable instructions that can be executed by the processor; The processor is configured to execute machine-executable instructions to implement the method steps of any one of claims 1-8.

Images