Adaptive vein template registration method and apparatus

Abstract

The application discloses an adaptive vein template registration method and device, and belongs to the technical field of biometric recognition. The adaptive vein template registration method comprises the following steps: performing feature extraction and image quality analysis on a plurality of acquired initial sample vein images, and acquiring a first sample feature and a first quality score respectively; performing deduplication processing on the plurality of initial sample vein images based on the first sample feature and the first quality score, and acquiring a plurality of first sample vein images; and acquiring a target template vector from the plurality of first sample vein images based on a first feature similarity between a target sample vein image and remaining sample vein images other than the target sample vein image and a first threshold. The adaptive vein template registration method can select the most representative vein samples at different angles as templates, reduces the number of templates, and can reduce the consumption of storage space while improving the vein recognition performance.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of biometric recognition, and particularly relates to a self-adaptive vein template registration method and device. BACKGROUND

[0002] Vein recognition is a new biometric recognition technology. When performing vein recognition, a vein image registered by a user needs to be collected first. However, due to the limitation of a collection device, the collected vein image often has a certain offset and rotation, so a plurality of templates of different angles need to be registered, which is large in workload and cumbersome in work. In addition, when the collection device is unstable, it is difficult to find a template with the best matching degree with the vein image registered by the user during vein recognition, and the performance of vein recognition is poor. SUMMARY

[0003] The present application aims to at least solve one of the technical problems existing in the prior art. To this end, the present application provides a self-adaptive vein template registration method and device, which can select the most representative vein samples of different angles as templates, while reducing the number of templates, and can reduce the consumption of storage space while improving the performance of vein recognition.

[0004] In a first aspect, the present application provides a self-adaptive vein template registration method, which comprises:

[0005] performing feature extraction and image quality analysis on the obtained plurality of initial sample vein images to obtain first sample features and first quality scores, respectively;

[0006] performing deduplication processing on the plurality of initial sample vein images based on the first sample features and the first quality scores to obtain a plurality of first sample vein images;

[0007] obtaining a target template vector from the plurality of first sample vein images based on a first feature similarity between a target sample vein image and remaining sample vein images other than the target sample vein image and a first threshold;

[0008] The method for obtaining a target template vector from the plurality of first sample vein images based on a first feature similarity between a target sample vein image and remaining sample vein images other than the target sample vein image and a first threshold comprises:

[0009] randomly obtaining a second target number of first sample vein images from the plurality of first sample vein images, determining the target sample vein image, and generating a plurality of selections based on the second target number and the number of the plurality of first sample vein images;

[0010] In each selection, the total hit number of the remaining sample vein images is calculated based on the following formula:

[0011]

[0012] wherein, is the total hit number of the remaining sample vein images, is the remaining sample vein image, is the target sample vein image, is the first threshold value, is the number of the plurality of first sample vein images, is the number of the target sample vein images, is the number of the remaining sample vein images, is the index of the remaining sample vein image, is the index of the target sample vein image, is the first feature similarity between the ith remaining sample vein image and the jth target sample vein image, the first feature similarity being a cosine similarity, is the highest first feature similarity among the first feature similarities between the ith remaining sample vein image and each of the target sample vein images, is used to represent the hit condition of the remaining sample vein image;

[0013] In the selection in which the total hit number of the remaining sample vein images is the highest among the plurality of selections, the selected target sample vein image is determined as the target template vector;

[0014] wherein, when the highest first feature similarity among the plurality of first feature similarities corresponding to the ith remaining sample vein image is greater than the first threshold value, the corresponding is 1, i.e., it is determined that the ith remaining sample vein image hits; when the highest first feature similarity among the plurality of first feature similarities corresponding to the ith remaining sample vein image is less than or equal to the first threshold value, the corresponding is 0, i.e., it is determined that the ith remaining sample vein image does not hit.

[0015] According to the adaptive vein template registration method provided in the embodiments of the present application, the first sample features and the first quality scores of a plurality of initial sample vein images are obtained, and then the plurality of initial sample vein images are deduplicated to obtain a plurality of first sample vein images. Then, based on the first feature similarity between a target sample vein image and the remaining sample vein images other than the target sample vein image in the plurality of first sample vein images and a first threshold, a target template vector is obtained from the plurality of first sample vein images. The most representative vein samples at different angles can be selected as templates, the number of templates is reduced, the storage space consumption can be reduced while improving the vein recognition performance, and in actual operation, only the number of initial sample vein images and target sample vein images needs to be set to adaptively select a suitable target sample vein image as a target template vector, and the steps are simple and easy to implement.

[0016] In an embodiment of the present application, the adaptive vein template registration method based on the first sample features and the first quality scores, the plurality of initial sample vein images are deduplicated to obtain a plurality of first sample vein images, comprising:

[0017] In the case that the number of the plurality of initial sample vein images is not zero, the initial sample vein image corresponding to the highest first quality score in the plurality of initial sample vein images is determined as the first sample vein image, and the first sample vein image is deleted from the plurality of initial sample vein images.

[0018] In the case that the second feature similarity between the first sample feature corresponding to the initial sample vein image and the first sample feature corresponding to the first sample vein image exceeds a second threshold, the initial sample vein image is deleted from the plurality of initial sample vein images.

[0019] In an embodiment of the present application, the adaptive vein template registration method based on the first sample features and the first quality scores, the plurality of initial sample vein images are deduplicated to obtain a plurality of first sample vein images, and then based on the first feature similarity between a target sample vein image and the remaining sample vein images other than the target sample vein image in the plurality of first sample vein images and a first threshold, a target template vector is obtained from the plurality of first sample vein images. Before the above steps, the method further comprises:

[0020] In the case that the number of the plurality of first sample vein images is less than a third threshold, a first target number of sample vein images are obtained as new initial sample vein images, and the first target number of sample vein images are different from the obtained plurality of initial sample vein images.

[0021] The adaptive vein template registration method of one embodiment of the application comprises feature extraction and image quality analysis on a plurality of acquired initial sample vein images, and acquiring a first sample feature and a first quality score, respectively, which comprises:

[0022] inputting the initial sample vein image into a first model to acquire a first sample feature corresponding to the initial sample vein image output by the first model; wherein the first model is trained by taking a sample finger vein image as a sample and taking a sample feature corresponding to the sample finger vein image as a sample label;

[0023] inputting the initial sample vein image into a second model to acquire a first quality score corresponding to the initial sample vein image output by the second model; wherein the second model is trained by taking a sample finger vein image as a sample and taking a sample quality corresponding to the sample finger vein image as a sample label.

[0024] In a second aspect, the application provides an adaptive vein template registration device, which comprises:

[0025] a first processing module configured to perform feature extraction and image quality analysis on a plurality of acquired initial sample vein images, and acquire a first sample feature and a first quality score, respectively;

[0026] a second processing module configured to perform deduplication processing on the plurality of initial sample vein images based on the first sample feature and the first quality score, and acquire a plurality of first sample vein images;

[0027] a third processing module configured to acquire a target template vector from the plurality of first sample vein images based on a first feature similarity between a target sample vein image and remaining sample vein images other than the target sample vein image and a first threshold;

[0028] The third processing module is further configured to:

[0029] randomly acquire a second target number of first sample vein images from the plurality of first sample vein images, determine the target sample vein image, and generate a plurality of selections based on the second target number and a number of the plurality of first sample vein images;

[0030] In each selection, the total hit number of the remaining sample vein images is calculated based on the following formula:

[0031]

[0032] wherein, the total hit number of the remaining sample vein images, the remaining sample vein image, a target sample vein image, a first threshold value, a number of the plurality of first sample vein images, a number of the target sample vein images, a number of remaining sample vein images, an index of a remaining sample vein image, an index of a target sample vein image, a first feature similarity between an ith remaining sample vein image and a jth target sample vein image, the first feature similarity being a cosine similarity, a highest first feature similarity among the first feature similarities between the ith remaining sample vein image and each of the target sample vein images, a hit condition for characterizing the remaining sample vein image;

[0033] from among the plurality of selections, the target sample vein image selected from among the selection with the highest total number of hits of the remaining sample vein images is determined as the target template vector;

[0034] wherein, in a case where the highest first feature similarity among the plurality of first feature similarities corresponding to the ith remaining sample vein image is greater than the first threshold value, the corresponding is 1, i.e., the ith remaining sample vein image is determined to be a hit; in a case where the highest first feature similarity among the plurality of first feature similarities corresponding to the ith remaining sample vein image is less than or equal to the first threshold value, the corresponding is 0, i.e., the ith remaining sample vein image is determined to be a miss.

[0035] The adaptive vein template registration device provided in the embodiments of the present application can select the most representative vein samples at different angles as templates while reducing the number of templates, can reduce the consumption of storage space while improving the vein recognition performance, and only needs to set the number of initial sample vein images and target sample vein images to adaptively select a suitable target sample vein image as a target template vector, which is simple and easy to implement.

[0036] In a third aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the adaptive vein template registration method according to the first aspect.

[0037] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium, having a computer program stored thereon, wherein the computer program is executable by a processor to implement the adaptive vein template registration method according to the first aspect.

[0038] In a fifth aspect, the present application provides a chip, comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is configured to execute a program or an instruction to implement the adaptive vein template registration method according to the first aspect.

[0039] In a sixth aspect, the present application provides a computer program product, comprising a computer program executable by a processor to implement the adaptive vein template registration method according to the first aspect.

[0040] The one or more technical solutions in the embodiments of the present application have at least one of the following technical effects:

[0041] By obtaining the first sample features and the first quality scores of the plurality of initial sample vein images, and then deduplicating the plurality of initial sample vein images to obtain a plurality of first sample vein images, and then obtaining the target template vector from the plurality of first sample vein images based on the first feature similarity between the target sample vein image and the remaining sample vein images other than the target sample vein image and the first threshold, the most representative vein samples at different angles can be selected as the template, the number of templates is reduced, the storage space consumption can be reduced while improving the vein recognition performance, and in actual operation, only the number of initial sample vein images and target sample vein images needs to be set to adaptively select a suitable target sample vein image as the target template vector, which is simple and easy to implement.

[0042] Further, by determining the initial sample vein image corresponding to the highest first quality score in the plurality of initial sample vein images as the first sample vein image and deleting the first sample vein image from the plurality of initial sample vein images in a case where the number of the plurality of initial sample vein images is not zero, the best quality sample vein image can be screened out, so as to improve the accuracy of the final template matching; in a case where the second feature similarity between the first sample feature corresponding to the initial sample vein image and the first sample feature corresponding to the first sample vein image exceeds the second threshold, the initial sample vein image is deleted from the plurality of initial sample vein images, so as to delete the vein image with high similarity to the first sample vein image, and then the plurality of initial sample vein images are de-duplicated, so as to reduce the storage space of the final template and improve the accuracy of the template matching.

[0043] Still further, in a case where the number of the plurality of first sample vein images is less than the third threshold, the first target number of sample vein images are acquired as new initial sample vein images, so as to obtain images with sufficient feature types on the basis of reducing the number of the overall sample vein images, thereby enriching the diversity of the template and improving the accuracy of the template matching.

[0044] Still further, by randomly acquiring the second target number of first sample vein images from the plurality of first sample vein images and determining the first sample vein images as target sample vein images, in a case where the first feature similarity exceeds the first threshold, it is determined that the remaining sample vein images hit, and then the target sample vein image corresponding to the remaining sample vein image with the maximum hit number is determined as the target template vector, so as to further perform the de-redundancy processing on the de-duplicated sample pool, thereby obtaining the most representative registration template feature vector and improving the accuracy of the template matching.

[0045] Additional aspects and advantages of the application will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following and / or by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0046] The above and / or additional aspects and advantages of the application will become apparent and be readily appreciated from the description of the embodiments, which follows, including the accompanying drawings.

[0047] Figure 1 is one of the flow diagrams of the adaptive vein template registration method provided by the embodiments of the application;

[0048] Figure 2 is the second flow diagram of the adaptive vein template registration method provided by the embodiments of the application;

[0049] Figure 3 is the structural diagram of the adaptive vein template registration device provided by the embodiments of the application;

[0050] Figure 4 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0051] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0052] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0053] The following is combined with Figure 1 and Figure 2 This application describes an adaptive vein template registration method according to an embodiment.

[0054] It should be noted that the entity executing the adaptive vein template registration method can be a server, an adaptive vein template registration device, or a user's terminal, including but not limited to mobile terminals and non-mobile terminals.

[0055] For example, mobile terminals include, but are not limited to, mobile phones, PDA smart terminals, tablets, and in-vehicle smart terminals; non-mobile terminals include, but are not limited to, PCs.

[0056] like Figure 1 As shown, the adaptive vein template registration method includes steps 110, 120 and 130.

[0057] Step 110: Perform feature extraction and image quality analysis on the acquired multiple initial sample vein images to obtain the first sample features and the first quality score respectively;

[0058] In this step, the initial sample vein images are vein images registered by the user from different angles. The initial sample vein images can be the user's finger vein images, which include at least one finger feature. It is understood that different types of fingers have different corresponding features.

[0059] The first sample feature is obtained by performing feature extraction on the plurality of initial sample vein images, and is used to represent the features of the fingers in the initial sample vein images. The first sample feature can be multiple.

[0060] The first quality score is obtained by performing image quality analysis on the plurality of initial sample vein images, and is used to represent the quality of the initial sample vein images. The first quality score is in one-to-one correspondence with the initial sample vein images.

[0061] It can be understood that, in actual execution, a score can be set to represent the first quality score. For example, the first quality score can be any score between 0 and 1. The closer the first quality score is to 0, the poorer the quality of the initial sample vein image corresponding to the first quality score. The closer the first quality score is to 1, the better the quality of the initial sample vein image corresponding to the first quality score. For example, if the first quality score is greater than 0.5, it is considered that the quality of the initial sample vein image corresponding to the first quality score is good.

[0062] Alternatively, a percentage can also be set to represent the first quality score, which can be based on user customization, and the present application does not make any limitation.

[0063] In some embodiments, step 110 can include:

[0064] inputting the initial sample vein image into the first model to obtain the first sample feature corresponding to the initial sample vein image output by the first model;

[0065] inputting the initial sample vein image into the second model to obtain the first quality score corresponding to the initial sample vein image output by the second model.

[0066] In this embodiment, the first model is used to perform feature extraction on the initial sample vein image. The first model is a pre-trained model, and the first model can use a sample finger vein image as a sample and use the sample feature corresponding to the sample finger vein image as a sample label to obtain.

[0067] The sample finger vein image is a pre-acquired finger vein image of the user.

[0068] Each finger of the user is a sample feature, and each sample finger vein image can correspond to multiple sample features.

[0069] The second model is used to perform quality analysis on the initial sample vein image. The second model is a pre-trained model, and the second model can use a sample finger vein image as a sample and use the sample quality corresponding to the sample finger vein image as a sample label to obtain.

[0070] Among them, sample quality is used to characterize the quality of the finger vein images of the samples.

[0071] In actual implementation, the Arcface loss function can be used to train the first model.

[0072] The Arcface loss function can be expressed by the following formula:

[0073]

[0074] in, The number of finger vein images in a batch of samples in the training set. The number of features in the training set. For the current batch The sample features corresponding to each sample finger vein image This is the scaling factor. For sample features The parameters, The set interval size.

[0075] During the testing phase, the last fully connected layer of the first model can be removed to output 512-dimensional sample features.

[0076] A second model can be trained using the sigmoid loss function.

[0077] The sigmoid loss function can be expressed by the following formula:

[0078]

[0079] in, The number of finger vein images in a batch of samples in the training set. These are the weights of the last fully connected layer in the second model.

[0080] exist In the case of high quality samples, the finger vein images of the samples corresponding to the sample quality are classified as high-quality samples.

[0081] exist In the case of poor quality, the finger vein images of the samples corresponding to the sample quality are classified as poor quality samples.

[0082] During the testing phase, the second model can output located at Sample quality within the interval.

[0083] According to the adaptive vein template registration method provided in the embodiments of the present application, the first model and the second model are pre-trained, then the first sample features corresponding to the initial sample vein images are obtained based on the first model, and the first quality scores corresponding to the initial sample vein images are obtained based on the second model. In actual application, only pre-training is required before, and then data can be directly obtained, which is high in calculation efficiency and good in accuracy. In addition, the first model and the second model have strong learning ability, and the data in each application process can be used as training data in the next training process, so that the precision and accuracy of the model are improved, the model is convenient for users to use, has high universality, and the precision of the final vein recognition is improved.

[0084] In step 120, the plurality of initial sample vein images are de-duplicated based on the first sample features and the first quality scores to obtain a plurality of first sample vein images.

[0085] In this step, the first sample vein image is a vein image obtained after de-duplicating the plurality of initial sample vein images. The first sample vein image can include vein images of multiple types, and each type of vein image can include vein images of multiple angles.

[0086] For example, the de-duplication process can include removing images of the same angle, or removing duplicate images, or removing images with large noise, or removing images with poor quality, or removing similar images. For example, the initial sample vein image with the highest first quality score can be selected as the first sample vein image, and then based on the similarity between the remaining initial sample vein images and the first sample vein image, images with high similarity to the first sample vein image are deleted, and images with low similarity are retained. Finally, the images with low similarity are further sorted by image quality, and the vein image with the highest first quality score is selected. The above steps are repeated until the number of initial sample vein images is zero.

[0087] In some embodiments, step 120 can include:

[0088] In the case where the number of the plurality of initial sample vein images is not zero, the initial sample vein image corresponding to the highest first quality score in the plurality of initial sample vein images is determined as the first sample vein image, and the first sample vein image is deleted from the plurality of initial sample vein images.

[0089] In the case where the second feature similarity between the first sample feature corresponding to the initial sample vein image and the first sample feature corresponding to the first sample vein image exceeds a second threshold, the initial sample vein image is deleted from the plurality of initial sample vein images.

[0090] In this embodiment, the second similarity is used to represent the similarity between the initial sample vein image and the first sample vein image.

[0091] The second threshold is used to determine whether the second similarity meets the requirement, and the second threshold can be 95%, 96%, or 97%, etc., which can be customized based on the user, and the present application does not make any limitation.

[0092] For example, the plurality of initial sample vein images can include images A, B, C, D, E, F, G, H, I, and J, the initial sample vein image A with the highest first quality score is selected from the plurality of initial sample vein images as the first sample vein image, and then the second similarity between the initial sample vein images B, C, D, E, F, G, H, I, and J and A is calculated respectively, and the initial sample vein images B, C, and D corresponding to the second similarity exceeding the second threshold are deleted; the remaining images E, F, G, H, I, and J are obtained.

[0093] Then the initial sample vein image E with the highest first quality score is selected again from the remaining E, F, G, H, I, and J, and the second similarity between the initial sample vein images F, G, H, I, and J and E is calculated respectively, and the initial sample vein images F, G, and H corresponding to the second similarity exceeding the second threshold are deleted; the remaining images I and J are obtained.

[0094] The above operation is repeated until the number of initial sample vein images is zero.

[0095] In actual execution process, the non-maximum suppression method can be used to perform the de-duplication operation on the plurality of initial sample vein images, and the specific operation is as follows:

[0096] The first sample features corresponding to the plurality of initial sample vein images can be sorted in descending order of confidence, and the plurality of initial sample vein images are saved in list B.

[0097] The initial sample vein image corresponding to the highest first quality score is selected from the list B, which is determined as the first sample vein image, and is recorded as Then is put into list D, and is deleted from list B.

[0098] Each initial sample vein image in list B is traversed , and the second similarity between and is calculated, for example, the second similarity can be the cosine similarity. In the case that the second similarity exceeds the second threshold, the initial sample vein image corresponding to the second similarity is deleted from list B.

[0099]

[0100] ​​In the case that the number of the plurality of initial sample vein images is not zero, the above steps are repeated until the number of the plurality of initial sample vein images is zero, i.e., the list B is empty.

[0101] For example, in the case that the list B is not empty, the initial sample vein image corresponding to the highest first quality score is continuously selected from the list B, which is determined as the first sample vein image, denoted as Then, the initial sample vein image is put into the list D, and the initial sample vein image is deleted from the list B.

[0102] Each initial sample vein image in the list B is traversed , and the second similarity between and is calculated.

[0103] In the case that the second similarity exceeds the second threshold, the initial sample vein image corresponding to the second similarity is deleted from the list B.

[0104] According to the adaptive vein template registration method provided in the embodiments of the present application, in the case that the number of the plurality of initial sample vein images is not zero, the initial sample vein image corresponding to the highest first quality score in the plurality of initial sample vein images is determined as the first sample vein image, and the first sample vein image is deleted from the plurality of initial sample vein images, which can filter out the sample vein image with the best quality, so as to improve the precision of the final template matching; in the case that the second feature similarity between the first sample feature corresponding to the initial sample vein image and the first sample feature corresponding to the first sample vein image exceeds the second threshold, the initial sample vein image is deleted from the plurality of initial sample vein images, which can delete the vein image with high similarity to the first sample vein image, and then the plurality of initial sample vein images are de-duplicated, so as to reduce the storage space of the final template and improve the precision of the template matching.

[0105] In some embodiments, after step 120 and before step 130, the adaptive vein template registration method can further include:

[0106] In the case that the number of the plurality of first sample vein images is less than a third threshold, a first target number of sample vein images are obtained as new initial sample vein images, and the first target number of sample vein images are different from the plurality of initial sample vein images obtained.

[0107] In this embodiment, the third threshold is used to determine whether the number of the first sample vein images meets the requirement, and the first threshold can be 1000, 1500 or 2000, etc., which can be customized by a user, and the present application does not make any limitation.

[0108] ​​​​The first target number can be customized by a user, for example, the first target number can be 100, 200 or 300.

[0109] The new initial sample vein image is a sample vein image of a finger vein of a different angle for the user to continue registration.

[0110] In actual implementation, in a case where the number of the plurality of first sample vein images is less than the third threshold, the user is guided to register a sample vein image of a different angle, and a de-duplication operation is performed on the first target number of sample vein images based on a non-maximum suppression method.

[0111] According to the adaptive vein template registration method provided in the embodiments of the present application, in a case where the number of the plurality of first sample vein images is less than the third threshold, the first target number of sample vein images are acquired as new initial sample vein images, so that in actual operation, a template with a sufficient number of feature types can be acquired on the basis of reducing the overall number of templates, the diversity of the templates is enriched, and the accuracy of template matching is improved.

[0112] In step 130, a target template vector is acquired from the plurality of first sample vein images based on a first feature similarity between a target sample vein image and a remaining sample vein image other than the target sample vein image and a first threshold.

[0113] In this step, the target sample vein image is a vein image randomly selected from the plurality of first sample vein images.

[0114] The remaining sample vein image is a sample vein image remaining in the plurality of first sample vein images after the target sample vein image is selected from the plurality of first sample vein images.

[0115] The first feature similarity is used to represent the similarity between the target sample vein image and the remaining sample vein image.

[0116] The first threshold is used to determine whether the first similarity meets the requirement, and the first threshold can be 95%, 96% or 97%, etc., and can be customized by a user, which is not limited in the present application.

[0117] The target template vector is a template vector for subsequent vein recognition.

[0118] The inventor found in the research and development process that in the related art, a template is registered based on a rotation amount calculation unit, and this method depends on rotation amount calculation. In a case where the rotation amount calculation unit is not accurate or the acquisition device is not stable enough, it is difficult to select the best template.

[0119] In the present application, feature extraction and image quality analysis are first performed on the plurality of initial sample vein images to obtain first sample features and first quality scores, and then the plurality of initial sample vein images are de-duplicated to obtain a plurality of first sample vein images, which can reduce the storage space of the final template and improve the accuracy of the final template matching.

[0120] Finally, based on the first feature similarity between the target sample vein image and the remaining sample vein images other than the target sample vein image in the plurality of first sample vein images and the first threshold, a target template vector is obtained from the plurality of first sample vein images, without calculating the rotation amount, avoiding the possibility of failure in calculating the rotation amount, and being able to select the most representative vein samples at different angles as templates, thereby improving the vein recognition performance.

[0121] According to the adaptive vein template registration method provided in the embodiments of the present application, by obtaining the first sample features and the first quality scores of the plurality of initial sample vein images, then de-duplicating the plurality of initial sample vein images to obtain a plurality of first sample vein images, and then obtaining a target template vector from the plurality of first sample vein images based on the first feature similarity between the target sample vein image and the remaining sample vein images other than the target sample vein image in the plurality of first sample vein images and the first threshold, the most representative vein samples at different angles can be selected as templates, while the number of templates is reduced, the consumption of storage space can be reduced while improving the vein recognition performance, and in actual operation, only the number of initial sample vein images and target sample vein images needs to be set to adaptively select a suitable target sample vein image as a target template vector, which is simple and easy to implement.

[0122] In some embodiments, step 130 can include:

[0123] randomly obtaining a second target number of first sample vein images from the plurality of first sample vein images, determining as target sample vein images, and generating a plurality of selections based on the second target number and the number of the plurality of first sample vein images;

[0124] In each selection, the total hit number of the remaining sample vein images is calculated based on the following formula:

[0125]

[0126] wherein, is the total hit number of the remaining sample vein images, is the remaining sample vein image, is the target sample vein image, is the first threshold, is the number of the plurality of first sample vein images, a number of target sample vein images, a number of remaining sample vein images, an index of the remaining sample vein image, an index of the target sample vein image, a first feature similarity between the ith remaining sample vein image and the jth target sample vein image, the first feature similarity being a cosine similarity, a highest first feature similarity among the first feature similarities between the ith remaining sample vein image and each target sample vein image, to represent a hit situation of the remaining sample vein image;

[0127] the selected target sample vein image is determined as the target template vector in the selection with the highest total number of hits of the remaining sample vein images among the multiple selections;

[0128] wherein, in a case where the highest first feature similarity among the multiple first feature similarities corresponding to the ith remaining sample vein image is greater than a first threshold value, the corresponding is 1, that is, the ith remaining sample vein image is determined to be a hit; in a case where the highest first feature similarity among the multiple first feature similarities corresponding to the ith remaining sample vein image is less than or equal to the first threshold value, the corresponding is 0, that is, the ith remaining sample vein image is determined to be a miss.

[0129] In this embodiment, the second target number can be customized by a user, for example, the second target number can be 100, 200, or 300.

[0130] The target sample vein image is a vein image randomly selected from the multiple first sample vein images.

[0131] The target template vector is a template vector used for subsequent vein recognition.

[0132] In actual execution, as shown in FIG. 6, a number of first sample vein images are randomly obtained from the multiple first sample vein images, and a number of target sample vein images are determined, and a number of selections are generated, wherein Figure 2 the number of target sample vein images can be customized by a user, and in actual operation, can be adjusted according to the stability and accuracy requirements of the vein recognition device. In this embodiment, a number of first sample vein images are randomly obtained from the multiple first sample vein images, and a number of target sample vein images are determined, and a number of selections are generated, wherein the number of target sample vein images can be customized by a user, and in actual operation, can be adjusted according to the stability and accuracy requirements of the vein recognition device. In this embodiment, a number of first sample vein images are randomly obtained from the multiple first sample vein images, and a number of target sample vein images are determined, and a number of selections are generated, wherein

[0133] In this embodiment, a number of first sample vein images are randomly obtained from the multiple first sample vein images, and a number of target sample vein images are determined, and a number of selections are generated, wherein the number of target sample vein images can be customized by a user, and in actual operation, can be adjusted according to the stability and accuracy requirements of the vein recognition device. ​​The first sample vein image is identified as the target sample vein image, and a generation is performed. The next selection, in each of which can be calculated using a formula. The first sample vein image and The first feature similarity between the remaining sample vein images, wherein the first feature similarity can be cosine similarity.

[0134] The highest similarity score is selected. If the highest similarity score is greater than a first threshold, it indicates that the vein image of the remaining samples corresponding to that first feature similarity score has been matched. The value is 1;

[0135] In the case where the remaining vein images are not matched The value is 0.

[0136] According to the adaptive vein template registration method provided in the embodiments of this application, the remaining sample vein images are determined by the above formula, which can further de-redundancy processing be performed on the deduplicated sample pool to obtain the most representative registration template feature vector. At the same time, the number of the first sample vein image and the remaining sample vein images can be customized and adjusted based on the stability and accuracy requirements of the vein recognition device, thereby improving the performance of vein recognition.

[0137] In actual implementation, after determining that the remaining sample vein images in this selection are hit, the hit counts of the remaining sample vein images in this selection are summed, and then the selection with the highest hit count is taken as the final selection based on the following formula:

[0138]

[0139] in, For the first The total number of hits corresponding to the remaining vein images in the selection. This represents the total number of selections.

[0140] at this time, The corresponding target sample vein image is determined as the target template vector.

[0141] According to the adaptive vein template registration method provided in the embodiments of this application, a second target number of first sample vein images are randomly obtained from multiple first sample vein images and determined as target sample vein images. Then, if the first feature similarity exceeds a first threshold, the remaining sample vein images are determined to be hit. The target sample vein image corresponding to the remaining sample vein image with the most hits is then determined as the target template vector. This method can further de-redundancy processing be performed on the deduplicated sample pool to obtain the most representative registration template feature vector, thereby improving the accuracy of template matching.

[0142] The adaptive vein template registration device provided in the present application is described below, and the adaptive vein template registration device described below can be correspondingly referred to the adaptive vein template registration method described above.

[0143] The adaptive vein template registration method provided in the embodiments of the present application can be executed by the adaptive vein template registration device. In the embodiments of the present application, the adaptive vein template registration method is executed by the adaptive vein template registration device as an example to illustrate the adaptive vein template registration device provided in the embodiments of the present application.

[0144] The embodiments of the present application further provide an adaptive vein template registration device.

[0145] As shown in the adaptive vein template registration device, the adaptive vein template registration device comprises a first processing module 310, a second processing module 320 and a third processing module 330. Figure 3

[0146] The first processing module 310 is configured to perform feature extraction and image quality analysis on the obtained multiple initial sample vein images, and obtain a first sample feature and a first quality score, respectively.

[0147] The second processing module 320 is configured to perform deduplication processing on the multiple initial sample vein images based on the first sample feature and the first quality score, and obtain multiple first sample vein images.

[0148] The third processing module 330 is configured to obtain a target template vector from the multiple first sample vein images based on a first feature similarity between a target sample vein image and a remaining sample vein image other than the target sample vein image and a first threshold.

[0149] The third processing module 330 is further configured to:

[0150] randomly obtain a second target number of first sample vein images from the multiple first sample vein images, determine the first sample vein images as target sample vein images, and generate multiple selections based on the second target number and the number of the multiple first sample vein images.

[0151] In each selection, the total hit number of the remaining sample vein images is calculated based on the following formula:

[0152]

[0153] wherein, the total hit number of the remaining sample vein images is, the remaining sample vein image is, the target sample vein image is, the first threshold is,​ is a number of the first sample vein images, is a number of the target sample vein images, is a number of the remaining sample vein images, is an index of the remaining sample vein image, is an index of the target sample vein image, is a first feature similarity between the ith remaining sample vein image and the jth target sample vein image, the first feature similarity being a cosine similarity, is a highest first feature similarity among the first feature similarities between the ith remaining sample vein image and the target sample vein images, is used to represent a hit situation of the remaining sample vein image;

[0154] The selected target sample vein image is determined as the target template vector in the selection with the highest total number of hits of the remaining sample vein images in the plurality of selections;

[0155] wherein, in a case where the highest first feature similarity among the plurality of first feature similarities corresponding to the ith remaining sample vein image is greater than a first threshold value, the corresponding is 1, that is, the ith remaining sample vein image is determined to be a hit; in a case where the highest first feature similarity among the plurality of first feature similarities corresponding to the ith remaining sample vein image is less than or equal to the first threshold value, the corresponding is 0, that is, the ith remaining sample vein image is determined to be a miss.

[0156] The adaptive vein template registration device provided by the embodiments of the present application can select the most representative vein samples of different angles as templates while reducing the number of templates, can reduce the consumption of storage space while improving the vein recognition performance, and only needs to set the number of initial sample vein images and target sample vein images to adaptively select a suitable target sample vein image as a target template vector, which is simple and easy to implement.

[0157] In some embodiments, the second processing module 320 can also be configured to, in a case where the number of the plurality of initial sample vein images is not zero, determine an initial sample vein image corresponding to the highest first quality score in the plurality of initial sample vein images as the first sample vein image, and delete the first sample vein image from the plurality of initial sample vein images.

[0158] In a case where the second feature similarity between the first sample feature corresponding to the initial sample vein image and the first sample feature corresponding to the first sample vein image exceeds a second threshold, the initial sample vein image is deleted from the plurality of initial sample vein images.

[0159] According to the adaptive vein template registration device provided in the embodiments of the present application, in a case where the number of the plurality of initial sample vein images is not zero, the initial sample vein image corresponding to the highest first quality score in the plurality of initial sample vein images is determined as the first sample vein image, and the first sample vein image is deleted from the plurality of initial sample vein images, so that the best sample vein image in quality can be screened out to improve the accuracy of the final template matching; in a case where the second feature similarity between the first sample feature corresponding to the initial sample vein image and the first sample feature corresponding to the first sample vein image exceeds a second threshold, the initial sample vein image is deleted from the plurality of initial sample vein images, so that the vein image with high similarity to the first sample vein image can be deleted, and then the plurality of initial sample vein images are de-duplicated to reduce the storage space of the final template and improve the accuracy of the template matching.

[0160] In some embodiments, the adaptive vein template registration device can further include a fourth processing module configured to, in a case where the number of the plurality of first sample vein images is less than a third threshold, obtain a first target number of sample vein images as new initial sample vein images, and the first target number of sample vein images is different from the plurality of initial sample vein images obtained.

[0161] According to the adaptive vein template registration device provided in the embodiments of the present application, in a case where the number of the plurality of first sample vein images is less than a third threshold, a first target number of sample vein images are obtained as new initial sample vein images, so that in actual operation, the templates with sufficient feature types can be obtained on the basis of reducing the overall number of templates, the diversity of the templates is enriched, and the accuracy of the template matching is improved.

[0162] In some embodiments, the first processing module 310 can also be configured to input the initial sample vein image into a first model to obtain a first sample feature corresponding to the initial sample vein image output by the first model; wherein the first model is trained by taking a sample finger vein image as a sample and taking a sample feature corresponding to the sample finger vein image as a sample label.

[0163] input the initial sample vein image into the second model to obtain a first quality score corresponding to the initial sample vein image output by the second model; wherein the second model is trained by taking sample finger vein images as samples and taking sample qualities corresponding to the sample finger vein images as sample labels.

[0164] According to the adaptive vein template registration device provided in the embodiments of the present application, the first model and the second model are pre-trained, then the first sample feature corresponding to the initial sample vein image is obtained based on the first model, and the first quality score corresponding to the initial sample vein image is obtained based on the second model. In actual application, only pre-training is required before, and then data can be directly obtained, which is high in calculation efficiency and good in accuracy. In addition, the first model and the second model have strong learning ability, and the data in each application process can be used as training data in the next training process, so as to improve the precision and accuracy of the model, facilitate user use, have high universality, and improve the precision of the final vein recognition.

[0165] The adaptive vein template registration device in the embodiments of the present application can be an electronic device or a component in the electronic device, for example, an integrated circuit or a chip. The electronic device can be a terminal or other devices except the terminal. For example, the electronic device can be a mobile phone, a tablet computer, a notebook computer, a palm computer, a vehicle-mounted electronic device, a mobile Internet device (MID), an augmented reality (AR) / virtual reality (VR) device, a robot, a wearable device, an ultra-mobile personal computer (UMPC), a netbook, or a personal digital assistant (PDA), and can also be a server, a Network Attached Storage (NAS), a personal computer (PC), a television (TV), a teller machine, or a self-service machine, and the embodiments of the present application are not limited in this regard.

[0166] The adaptive vein template registration device in the embodiments of the present application can be a device with an operating system. The operating system can be an Android operating system, an IOS operating system, or other possible operating systems, and the embodiments of the present application are not limited in this regard.

[0167] The adaptive vein template registration device provided in the embodiments of the present application can achieve Figures 1 to 2The various processes implemented by the method embodiments of the adaptive vein template registration method are not repeated here to avoid repetition.

[0168] In some embodiments, as shown in Figure 4 The processor 401 is configured to execute the computer program stored in the memory 402, and implement various processes of the adaptive vein template registration method embodiments described above, and achieve the same technical effects. To avoid repetition, the various processes are not repeated here.

[0169] It should be noted that the electronic device in the embodiments of the present application includes the mobile electronic device and the non-mobile electronic device described above.

[0170] On the other hand, the present application also provides a computer program product, which comprises a computer program stored on a non-transitory computer readable storage medium, and the computer program comprises program instructions. When the program instructions are executed by a computer, the computer can execute the various processes of the adaptive vein template registration method embodiments described above, and achieve the same technical effects. To avoid repetition, the various processes are not repeated here.

[0171] In yet another aspect, the present application also provides a non-transitory computer readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the computer program implements the various processes of the adaptive vein template registration method embodiments described above, and achieves the same technical effects. To avoid repetition, the various processes are not repeated here.

[0172] In yet another aspect, the present application further provides a chip, which comprises a processor and a communication interface. The communication interface is coupled to the processor. The processor is configured to run a program or instructions, and implement the various processes of the adaptive vein template registration method embodiments described above, and achieve the same technical effects. To avoid repetition, the various processes are not repeated here.

[0173] It should be understood that the chip mentioned in the embodiments of the present application can also be referred to as a system-level chip, a system chip, a chip system, or a system-on-chip chip, etc.

[0174] The device embodiments described above are only schematic, and the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiments according to actual needs. Those skilled in the art can understand and implement it without creative labor.

[0175] Those skilled in the art can clearly understand the implementation of the various embodiments by means of software and the necessary general hardware platform from the above description of the embodiments, and of course, the embodiments can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that contributes to the technical solutions can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the methods.

[0176] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method of adaptive vein template registration, characterized by, Comprise: Feature extraction and image quality analysis are performed on the obtained multiple initial sample vein images to obtain first sample features and first quality scores respectively; Based on the first sample features and the first quality scores, the multiple initial sample vein images are processed to obtain multiple first sample vein images; Based on the first feature similarity between the target sample vein image and the remaining sample vein images other than the target sample vein image and the first threshold, the target template vector is obtained from the multiple first sample vein images; The target template vector is obtained from the multiple first sample vein images based on the first feature similarity between the target sample vein image and the remaining sample vein images other than the target sample vein image and the first threshold, comprising: Randomly obtaining a second target number of first sample vein images from the multiple first sample vein images as the target sample vein image, and generating multiple choices based on the second target number and the number of the multiple first sample vein images; In each choice, the total hit number of the remaining sample vein images is calculated based on the following formula: wherein, is a total number of hits for the remaining sample vein images, is the remaining sample vein image, is the target sample vein image, is the first threshold value, is a number of the plurality of first sample vein images, is a number of the target sample vein images, is a number of remaining sample vein images, is an index of the remaining sample vein image, is an index of the target sample vein image, is a first feature similarity between the i-th remaining sample vein image and the j-th target sample vein image, the first feature similarity being a cosine similarity, is a highest first feature similarity among the first feature similarities between the i-th remaining sample vein image and each of the target sample vein images, is used to represent a hit condition of the remaining sample vein image; The selected target sample vein image in the choice with the highest total hit number of the remaining sample vein images in the multiple choices is determined as the target template vector; wherein, in the case that the highest first feature similarity in the first feature similarities corresponding to the plurality of first feature similarities corresponding to the i-th remaining sample vein image is greater than the first threshold value, the i-th remaining sample vein image is determined to be a hit in the case that the highest first feature similarity in the first feature similarities corresponding to the plurality of first feature similarities corresponding to the i-th remaining sample vein image is less than or equal to the first threshold value, the i-th remaining sample vein image is determined to be a miss. is 1, i.e. the i-th remaining sample vein image is determined to be a hit; in the case that the highest first feature similarity in the first feature similarities corresponding to the plurality of first feature similarities corresponding to the i-th remaining sample vein image is less than or equal to the first threshold value, the i-th remaining sample vein image is determined to be a miss. is 0, i.e. the i-th remaining sample vein image is determined to be a miss.

2. The adaptive vein template registration method of claim 1, wherein, The multiple initial sample vein images are processed based on the first sample features and the first quality scores to obtain multiple first sample vein images, comprising: In the case that the number of the multiple initial sample vein images is not zero, the initial sample vein image corresponding to the highest first quality score in the multiple initial sample vein images is determined as the first sample vein image, and the first sample vein image is deleted from the multiple initial sample vein images; In the case that the second feature similarity between the first sample feature corresponding to the initial sample vein image and the first sample feature corresponding to the first sample vein image exceeds the second threshold, the initial sample vein image is deleted from the multiple initial sample vein images.

3. The adaptive vein template registration method of claim 1, wherein, After the multiple initial sample vein images are processed based on the first sample features and the first quality scores to obtain multiple first sample vein images, and before the target template vector is obtained from the multiple first sample vein images based on the first feature similarity between the target sample vein image and the remaining sample vein images other than the target sample vein image and the first threshold, the method further comprises: In the case that the number of the multiple first sample vein images is less than a third threshold, a first target number of sample vein images are obtained as new initial sample vein images, and the first target number of sample vein images are different from the obtained multiple initial sample vein images.

4. The adaptive vein template registration method according to any one of claims 1-3, wherein, The feature extraction and image quality analysis of the obtained multiple initial sample vein images to obtain first sample features and first quality scores respectively, comprising: inputting the initial sample vein image into a first model to obtain first sample features corresponding to the initial sample vein image output by the first model, wherein the first model is trained by taking sample finger vein images as samples and taking sample features corresponding to the sample finger vein images as sample labels; inputting the initial sample vein image into a second model to obtain a first quality score corresponding to the initial sample vein image output by the second model, wherein the second model is trained by taking sample finger vein images as samples and taking sample qualities corresponding to the sample finger vein images as sample labels.

5. An adaptive vein template registration device, characterized by The method comprises: a first processing module configured to perform feature extraction and image quality analysis on the obtained multiple initial sample vein images to obtain first sample features and first quality scores respectively; a second processing module configured to perform deduplication processing on the multiple initial sample vein images based on the first sample features and the first quality scores to obtain multiple first sample vein images; a third processing module configured to obtain a target template vector from the multiple first sample vein images based on a first feature similarity between a target sample vein image and remaining sample vein images other than the target sample vein image and a first threshold value; the third processing module is further configured to: randomly obtain a second target number of first sample vein images from the multiple first sample vein images to determine the target sample vein image, and generate multiple selections based on the second target number and a number of the multiple first sample vein images; in each selection, calculate a total hit number of the remaining sample vein images based on the following formula: wherein, is a total number of hits for the remaining sample vein images, is the remaining sample vein image, is the target sample vein image, is the first threshold value, is a number of the plurality of first sample vein images, is a number of the target sample vein images, is a number of remaining sample vein images, is an index of the remaining sample vein image, is an index of the target sample vein image, is a first feature similarity between the i-th remaining sample vein image and the j-th target sample vein image, the first feature similarity being a cosine similarity, is a highest first feature similarity among the first feature similarities between the i-th remaining sample vein image and each of the target sample vein images, is used to characterize a hit condition of the remaining sample vein image; in the multiple selections, the target sample vein image selected in a selection in which the total hit number of the remaining sample vein images is the highest is determined as the target template vector. wherein, in the case that the highest first feature similarity in the first feature similarities corresponding to the plurality of first feature similarities corresponding to the i-th remaining sample vein image is greater than the first threshold value, the i-th remaining sample vein image is determined to be a hit in the case that the highest first feature similarity in the first feature similarities corresponding to the plurality of first feature similarities corresponding to the i-th remaining sample vein image is less than or equal to the first threshold value, the i-th remaining sample vein image is determined to be a miss. is 1, i.e. the i-th remaining sample vein image is determined to be a hit; in the case that the highest first feature similarity in the first feature similarities corresponding to the plurality of first feature similarities corresponding to the i-th remaining sample vein image is less than or equal to the first threshold value, the i-th remaining sample vein image is determined to be a miss. is 0, i.e. the i-th remaining sample vein image is determined to be a miss.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the adaptive vein template registration method of any one of claims 1-4.

7. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the adaptive vein template registration method of any one of claims 1-4.

8. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the adaptive vein template registration method of any one of claims 1-4.

Images