Palm vein feature retrieval and matching method and system based on identity clustering

Abstract

The present invention relates to the technical field of biometric feature recognition, and proposes a palm vein feature retrieval and matching method and system based on identity clustering, comprising the following steps: obtaining a palm vein feature database, wherein the palm vein features in the database are divided into a plurality of clusters, wherein one cluster corresponds to one identity category, and one cluster corresponds to a cluster center feature; obtaining features of a palm vein image to be queried, and obtaining a query feature; calculating the similarity between the cluster center feature and the query feature, and considering a cluster with the similarity greater than a preset similarity threshold as a candidate cluster; calculating the similarity between each palm vein feature in each candidate cluster and the query feature, and obtaining a plurality of feature similarities, and sorting the palm vein features in descending order of the feature similarities, and selecting the first plurality of palm vein features as similar feature retrieval results corresponding to the palm vein image to be queried.

Description

Technical Field

[0001] The present invention relates to the technical field of biometric identification, and more specifically, to a palm vein feature retrieval and matching method and system based on identity clustering. Background Art

[0002] In today's digital age, identity authentication technology is the key to ensuring information security, preventing unauthorized access and protecting personal privacy. With the development of technology and the expansion of application areas, traditional identity authentication methods, such as passwords and PIN codes, are increasingly unable to meet people's demand for high security because they are easy to crack and forge. Therefore, biometric recognition technology has become a research hotspot in the field of identity authentication due to its unique and difficult-to-copy characteristics.

[0003] Palm vein recognition technology is currently an advanced biometric identification technology that uses the vein pattern inside an individual's palm for identity verification. The principle of this technology is based on a simple yet profound observation: everyone's palm vein pattern is unique. The palm vein pattern depends on the size and shape of the veins in the palm and their distribution in the palm. Palm vein recognition technology verifies personal identity by analyzing these patterns. Compared with other biometric identification technologies, palm vein recognition technology performs well in accuracy and stability. This is due to the complexity and uniqueness of the palm vein pattern, as well as its relatively stable characteristics, which are not easily susceptible to external damage or forgery. At the same time, because the palm vein pattern is located inside the body and is difficult to be observed by outsiders, palm vein recognition technology also has high security.

[0004] At present, the palm vein recognition process includes four key steps: image acquisition, image preprocessing, feature extraction, feature retrieval and matching. In the image acquisition step, near-infrared light is usually used to illuminate the palm, and a near-infrared camera is used to capture the vein image; in the image preprocessing step, image enhancement methods and region of interest (ROI) are usually used to ensure the accuracy of subsequent feature extraction; the feature extraction stage aims to extract recognizable feature vectors; finally, the identity authentication process is completed through feature retrieval and matching algorithms.

[0005] Although the existing palm vein recognition technology performs well in many application scenarios, the efficiency and accuracy of the recognition system on large-scale data sets still face challenges as application requirements increase. Especially in the massive data environment, relying solely on traditional feature matching methods may lead to excessive computational complexity and affect real-time performance. In addition, due to the diversity and similarity of palm vein features of different individuals, how to effectively retrieve and match features has become an urgent problem to be solved.

[0006] In summary, the existing technology has the problems of high computational complexity, poor real-time performance and low accuracy of similar feature retrieval and feature matching authentication. Summary of the invention

[0007] In order to overcome the defects of the above-mentioned prior art that the real-time performance is poor and similar feature retrieval and feature matching authentication cannot be accurately performed, the present invention provides a palm vein feature retrieval and matching method and system based on identity clustering that has strong real-time performance and can accurately perform similar feature retrieval and feature matching authentication.

[0008] To solve the above technical problems, the palm vein feature retrieval method based on identity clustering of the present invention is as follows:

[0009] Acquire a palm vein feature database, wherein the palm vein features in the database are divided into a plurality of clusters, wherein one cluster corresponds to one identity category, and one cluster corresponds to one cluster center feature;

[0010] Acquire the features of the palm vein image to be queried to obtain the query features;

[0011] Calculate the similarity between the cluster center feature and the query feature, and regard the clusters whose similarity is greater than a preset similarity threshold as candidate clusters;

[0012] The similarity between each palm vein feature in each candidate cluster and the query feature is calculated to obtain a plurality of feature similarities, the palm vein features are sorted in descending order of the feature similarities, and the first plurality of palm vein features are selected as similar feature retrieval results corresponding to the palm vein image to be queried.

[0013] The present invention also proposes a palm vein feature matching method based on identity clustering. The palm vein feature retrieval method based on identity clustering is applied. The method comprises the following steps:

[0014] Perform feature matching of the query feature based on the feature similarities corresponding to the similar feature retrieval results, and during matching, determine whether the maximum value of the feature similarities corresponding to the similar feature retrieval results is greater than a preset matching threshold;

[0015] If so, the match is successful, and the identity corresponding to the query feature is authenticated as the identity of the palm vein feature in the similar feature retrieval result corresponding to the maximum feature similarity; otherwise, the match fails.

[0016] The present invention also proposes a palm vein feature retrieval and matching system based on identity clustering for implementing the above-mentioned palm vein feature retrieval and matching method based on identity clustering, the system comprising:

[0017] A database acquisition module, used to acquire a palm vein feature database, wherein the palm vein features in the database are divided into a plurality of clusters, wherein one cluster corresponds to one identity category, and one cluster corresponds to one cluster center feature;

[0018] A query feature acquisition module is used to acquire the features of the palm vein image to be queried to obtain the query features;

[0019] A candidate feature selection module is used to calculate the similarity between the cluster center feature and the query feature, and regard the clusters with the similarity greater than a preset similarity threshold as candidate clusters;

[0020] A similar feature retrieval module is used to calculate the similarity between each palm vein feature in each candidate cluster and the query feature to obtain a plurality of feature similarities, sort the palm vein features in descending order of the feature similarities, and select the first plurality of palm vein features as similar feature retrieval results corresponding to the palm vein image to be queried;

[0021] The matching module is used to perform feature matching of the query feature based on the feature similarity corresponding to the similar feature retrieval result. During matching, it is determined whether the maximum value of the feature similarity corresponding to the similar feature retrieval result is greater than a preset matching threshold; if so, the match is successful, and the identity corresponding to the query feature is authenticated as the identity of the palm vein feature in the similar feature retrieval result corresponding to the maximum feature similarity; otherwise, the match fails.

[0022] The present invention also proposes a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the above-mentioned palm vein feature matching method based on identity clustering.

[0023] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0024] The palm vein feature database of the present invention divides the palm vein features corresponding to the same identity category into the same cluster, thereby screening out several candidate clusters of query features corresponding to the palm vein image to be queried, and then searching for several palm vein features that are most similar to the query features in the palm vein feature database from the several candidate clusters, avoiding the high computational overhead caused by global search, and significantly improving the response speed, retrieval accuracy and matching accuracy of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a schematic diagram of the process of the palm vein feature retrieval method based on identity clustering proposed in Example 1;

[0026] Figure 2 This is a schematic diagram of the first process flow of the palm vein feature matching method based on identity clustering proposed in Example 1;

[0027] Figure 3This is a second flow chart of the palm vein feature matching method based on identity clustering proposed in Example 2. DETAILED DESCRIPTION

[0028] The accompanying drawings are only used for illustrative purposes and should not be construed as limiting the present embodiment;

[0029] In order to better illustrate the present embodiment, some parts in the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product;

[0030] It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0031] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0032] Example 1

[0033] This embodiment proposes a palm vein feature retrieval method based on identity clustering. Figure 1 This is a schematic diagram of the process of the palm vein feature retrieval method based on identity clustering proposed in this embodiment.

[0034] A palm vein feature retrieval method based on identity clustering proposed in this embodiment includes the following steps:

[0035] S1: obtaining a palm vein feature database, wherein the palm vein features in the database are divided into a plurality of clusters, wherein one cluster corresponds to one identity category, and one cluster corresponds to one cluster center feature;

[0036] S2: Acquire the features of the palm vein image to be queried to obtain the query features;

[0037] S3: Calculate the similarity between the cluster center feature and the query feature, and regard the clusters whose similarity is greater than a preset similarity threshold as candidate clusters;

[0038] S4: Calculate the similarity between each palm vein feature in each candidate cluster and the query feature to obtain a plurality of feature similarities, sort the palm vein features in descending order of the feature similarities, and select the first plurality of palm vein features as similar feature retrieval results corresponding to the palm vein image to be queried.

[0039] During the specific implementation process, the palm vein feature database divides the palm vein features corresponding to the same identity category into the same cluster, thereby screening out several candidate clusters of query features corresponding to the palm vein image to be queried, and then retrieves several palm vein features that are most similar to the query features in the palm vein feature database from the several candidate clusters, avoiding the high computational overhead caused by global search, and can significantly improve the system's response speed, retrieval accuracy and matching accuracy.

[0040] In an optional embodiment, the palm vein feature database is obtained by extracting features from a number of palm vein images of a number of individuals, wherein one person corresponds to one identity category.

[0041] In an optional embodiment, the cluster center feature corresponding to the cluster includes: a mean value of all palm vein features in the cluster.

[0042] In an optional embodiment, before calculating the similarity between each palm vein feature in each candidate cluster and the query feature, the variance of each cluster is calculated. If the variance of the cluster is greater than a preset variance threshold, the cluster corresponding to the variance is added to the candidate cluster; otherwise, the cluster corresponding to the variance is not added to the candidate cluster.

[0043] In an optional embodiment, when calculating the similarity between the cluster center feature and the query feature or calculating the similarity between each palm vein feature in each candidate cluster and the query feature, the similarity includes cosine similarity;

[0044] When calculating the cosine similarity between the cluster center feature and the query feature, the cluster center feature and the query feature are normalized, and the cosine similarity between the cluster center feature and the query feature is obtained by calculating the vector inner product between the normalized cluster center feature and the query feature;

[0045] When calculating the cosine similarity between each palm vein feature in each candidate cluster and the query feature, each palm vein feature in each candidate cluster and the query feature are normalized, and the cosine similarity between each palm vein feature in each candidate cluster and the query feature is obtained by calculating the vector inner product between the normalized palm vein feature and the query feature.

[0046] As an exemplary illustration, the similarity is not limited to cosine similarity, that is, the similarity metric used is not limited to cosine distance, and other similarity metrics such as Euclidean distance may be used.

[0047] This embodiment also proposes a palm vein feature matching method based on identity clustering, which includes the following steps:

[0048] Perform feature matching of the query feature based on the feature similarities corresponding to the similar feature retrieval results, and during matching, determine whether the maximum value of the feature similarities corresponding to the similar feature retrieval results is greater than a preset matching threshold;

[0049] If so, the match is successful, and the identity corresponding to the query feature is authenticated as the identity category to which the palm vein feature in the similar feature retrieval result corresponding to the maximum feature similarity belongs; otherwise, the match fails.

[0050] As an example, Figure 2 This is a schematic diagram of the first process of the palm vein feature matching method based on identity clustering proposed in this embodiment; Figure 2 The second representation of the palm vein feature matching method based on identity clustering is shown.

[0051] In an optional embodiment, the preset matching threshold includes a first preset matching threshold and a second preset matching threshold;

[0052] When determining whether the maximum value of the feature similarities corresponding to the similar feature retrieval results is greater than a preset matching threshold, if all of the similar feature retrieval results belong to the same cluster, determining whether the maximum feature similarity corresponding to the similar feature retrieval results is greater than a preset first matching threshold, if so, the match is successful, and the identity corresponding to the query feature is authenticated as the identity category to which the palm vein feature corresponding to the maximum feature similarity belongs, otherwise, the match fails;

[0053] If the similar feature retrieval results do not all belong to the same cluster, determine whether the maximum feature similarity corresponding to the similar feature retrieval results is greater than a preset second matching threshold. If so, the match is successful, and the identity corresponding to the query feature is authenticated as the identity category to which the palm vein feature corresponding to the maximum feature similarity belongs. Otherwise, the match fails.

[0054] In an optional embodiment, the first matching threshold is lower than the second matching threshold.

[0055] Example 2

[0056] This embodiment is based on the palm vein feature retrieval method based on identity clustering and the palm vein feature matching method based on identity clustering proposed in Example 1, and proposes the following specific implementation examples:

[0057] Figure 3 This is a second flow chart of the palm vein feature matching method based on identity clustering proposed in this embodiment. Figure 3 The second process representation of the palm vein feature matching method based on identity clustering is presented.

[0058] S1. Build an index based on the palm vein feature database. For each identity category in the feature database, calculate the mean of all sample features in this category as the cluster center feature, and calculate the variance of the cluster.

[0059] S2. In the feature retrieval and matching stage of the actual recognition process, a two-stage retrieval method is adopted for the query features extracted from the acquired palm vein images: in the first stage, the query features are compared with the features of each identity cluster center, and the cluster variance information is used to screen out candidate categories. Then, the query features are compared with all samples in the candidate categories for similarity, and the k sample identities with the highest similarity and their corresponding similarities are counted.

[0060] S3. For the k most similar sample identities and corresponding similarities obtained in S2, count whether the identities of these samples are the same. If they are the same, use a lower matching threshold, otherwise use a higher matching threshold. Then compare the similarity of the most recent sample with the matching threshold and output the authentication result.

[0061] As an exemplary description, k is a positive integer. In a specific implementation process, k=5 is selected for calculation.

[0062] Furthermore, the palm vein feature database in S1 is obtained by extracting features from all samples in the palm vein image data set. In addition, since the present invention adopts cosine similarity as the feature similarity measure, each vector in the feature database will be normalized in advance. At the same time, S2 will also be normalized before the query feature is compared for similarity. Therefore, when retrieving and calculating cosine similarity, only the inner product of the vector needs to be calculated, so as to reduce the amount of calculation of similarity comparison during retrieval.

[0063] Furthermore, in the first retrieval stage of S2, the query feature will be compared with the feature of each identity cluster center for similarity. If the similarity with a category cluster center feature is greater than a preset similarity threshold or the cluster variance of the category is greater than a preset variance threshold, then the category will be placed in the candidate category. The cluster variance is used to avoid the loss of important samples due to the comparison scores of minority category samples.

[0064] Furthermore, in the feature matching process of S3, two similarity matching threshold parameters are set to obtain the identities of the k most similar samples and their corresponding similarities. When the identity labels of all neighboring samples are the same, a lower threshold parameter is used, indicating that the authentication conditions are appropriately relaxed; when there are multiple identity labels of neighboring samples, a higher threshold parameter is used, indicating that the authentication conditions are more stringent. Finally, the similarity between the query sample and the nearest sample is compared with the matching threshold. If it is greater than the threshold, the query sample passes the authentication and its category is the category of the nearest sample; otherwise, the authentication is rejected.

[0065] As an exemplary illustration, the similarity threshold, the variance threshold, the lower threshold parameter and the upper threshold parameter are set based on experience.

[0066] The present invention utilizes the identity clustering characteristics of the palm vein feature database and uses the natural identity clustering to replace clustering algorithms such as k-means to reduce the time of constructing the feature database index while achieving better clustering effect. In addition, by utilizing the clustering variance, optimization is performed for the situation where the minority category samples are relatively scattered and cause the loss of important samples. Finally, in the stage of matching the retrieval results, different matching similarity thresholds are used according to the uniqueness of the identity labels of the retrieved neighboring samples to improve the security of the authentication results.

[0067] Compared with the prior art, the present invention has the following positive effects:

[0068] 1) The present invention utilizes the identity clustering characteristics of the palm vein feature database and replaces traditional clustering algorithms such as k-means with natural identity clustering, which significantly reduces the time for constructing the feature database index, improves the real-time performance of the algorithm, and makes it more efficient in practical applications.

[0069] 2) By calculating the mean of all sample features of each identity category as the cluster center feature and using the cluster variance to screen candidate categories, it is possible to avoid the loss of important samples caused by the dispersion of samples in a few categories while maintaining a high clustering effect, thus ensuring the accuracy and reliability of feature matching.

[0070] 3) Cosine similarity is used as the feature similarity measure, and the feature vector is normalized, so that only the vector inner product needs to be calculated when comparing similarity, which simplifies the calculation process, reduces the amount of calculation, and improves the retrieval efficiency.

[0071] 4) In the matching stage of retrieval results, different matching similarity thresholds are set according to the uniqueness of the identity labels of the retrieved neighboring samples, which improves the security of the authentication results and effectively prevents potential security risks.

[0072] Example 3

[0073] This embodiment proposes a palm vein feature retrieval and matching system based on identity clustering, which is used to implement the palm vein feature retrieval and matching method based on identity clustering proposed in Example 1.

[0074] The palm vein feature retrieval and matching system based on identity clustering includes:

[0075] A database acquisition module, used to acquire a palm vein feature database, wherein the palm vein features in the database are divided into a plurality of clusters, wherein one cluster corresponds to one identity category, and one cluster corresponds to one cluster center feature;

[0076] A query feature acquisition module is used to acquire the features of the palm vein image to be queried to obtain the query features;

[0077] A candidate feature selection module is used to calculate the similarity between the cluster center feature and the query feature, and regard the clusters with the similarity greater than a preset similarity threshold as candidate clusters;

[0078] A similar feature retrieval module is used to calculate the similarity between each palm vein feature in each candidate cluster and the query feature to obtain a plurality of feature similarities, sort the palm vein features in descending order of the feature similarities, and select the first plurality of palm vein features as similar feature retrieval results corresponding to the palm vein image to be queried;

[0079] The matching module is used to perform feature matching of the query feature based on the feature similarity corresponding to the similar feature retrieval result. During matching, it is determined whether the maximum value of the feature similarity corresponding to the similar feature retrieval result is greater than a preset matching threshold; if so, the match is successful, and the identity corresponding to the query feature is authenticated as the identity of the palm vein feature in the similar feature retrieval result corresponding to the maximum feature similarity; otherwise, the match fails.

[0080] This embodiment proposes a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the palm vein feature matching method based on identity clustering described in Example 1.

[0081] It can be understood that the palm vein feature retrieval and matching system based on identity clustering and the computer device of this embodiment are used to implement the method of embodiment 1. The options in the above embodiment 1 are also applicable to this embodiment, so they will not be repeated here.

[0082] The same or similar reference numerals correspond to the same or similar components;

[0083] The terms used to describe the positional relationship in the drawings are only used for illustrative purposes and should not be construed as limiting the present embodiment;

[0084] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A palm vein feature retrieval method based on identity clustering, characterized in that: The following steps are involved: Acquire a palm vein feature database, wherein the palm vein features in the database are divided into a plurality of clusters, wherein one cluster corresponds to one identity category, and one cluster corresponds to one cluster center feature; Acquire the features of the palm vein image to be queried to obtain the query features; Calculate the similarity between the cluster center feature and the query feature, and regard the clusters whose similarity is greater than a preset similarity threshold as candidate clusters; The similarity between each palm vein feature in each candidate cluster and the query feature is calculated to obtain a plurality of feature similarities, the palm vein features are sorted in descending order of the feature similarities, and the first plurality of palm vein features are selected as similar feature retrieval results corresponding to the palm vein image to be queried.

2. The palm vein feature retrieval and matching method based on identity clustering according to claim 1 is characterized in that: The palm vein feature database is obtained by extracting features from a number of palm vein images of a number of individuals, wherein one person corresponds to one identity category.

3. The palm vein feature retrieval and matching method based on identity clustering according to claim 1, characterized in that: The cluster center feature corresponding to the cluster includes: a mean value of all palm vein features in the cluster.

4. The palm vein feature retrieval and matching method based on identity clustering according to claim 1, characterized in that: Before calculating the similarity between each palm vein feature in each candidate cluster and the query feature, the variance of each cluster is calculated. If the variance of the cluster is greater than a preset variance threshold, the cluster corresponding to the variance is added to the candidate cluster; otherwise, the cluster corresponding to the variance is not added to the candidate cluster.

5. The palm vein feature retrieval and matching method based on identity clustering according to any one of claims 1 to 4, characterized in that: When calculating the similarity between the cluster center feature and the query feature or calculating the similarity between each palm vein feature in each candidate cluster and the query feature, the similarity includes cosine similarity; When calculating the cosine similarity between the cluster center feature and the query feature, the cluster center feature and the query feature are normalized, and the cosine similarity between the cluster center feature and the query feature is obtained by calculating the vector inner product between the normalized cluster center feature and the query feature; When calculating the cosine similarity between each palm vein feature in each candidate cluster and the query feature, each palm vein feature in each candidate cluster and the query feature are normalized, and the cosine similarity between each palm vein feature in each candidate cluster and the query feature is obtained by calculating the vector inner product between the normalized palm vein feature and the query feature.

6. A palm vein feature matching method based on identity clustering, using the palm vein feature retrieval method based on identity clustering according to any one of claims 1 to 5, characterized in that: The following steps are involved: Perform feature matching of the query feature based on the feature similarities corresponding to the similar feature retrieval results, and during matching, determine whether the maximum value of the feature similarities corresponding to the similar feature retrieval results is greater than a preset matching threshold; If so, the match is successful, and the identity corresponding to the query feature is authenticated as the identity category to which the palm vein feature in the similar feature retrieval result corresponding to the maximum feature similarity belongs; otherwise, the match fails.

7. The palm vein feature retrieval and matching method based on identity clustering according to claim 6, characterized in that: The preset matching threshold includes a first preset matching threshold and a second preset matching threshold; When determining whether the maximum value of the feature similarities corresponding to the similar feature retrieval results is greater than a preset matching threshold, if all of the similar feature retrieval results belong to the same cluster, determining whether the maximum feature similarity corresponding to the similar feature retrieval results is greater than a preset first matching threshold, if so, the match is successful, and the identity corresponding to the query feature is authenticated as the identity category to which the palm vein feature corresponding to the maximum feature similarity belongs, otherwise, the match fails; If the similar feature retrieval results do not all belong to the same cluster, determine whether the maximum feature similarity corresponding to the similar feature retrieval results is greater than a preset second matching threshold. If so, the match is successful, and the identity corresponding to the query feature is authenticated as the identity category to which the palm vein feature corresponding to the maximum feature similarity belongs. Otherwise, the match fails.

8. The palm vein feature retrieval and matching method based on identity clustering according to claim 7, characterized in that: The first matching threshold is lower than the second matching threshold.

9. A palm vein feature retrieval and matching system based on identity clustering, used to implement the palm vein feature matching method based on identity clustering according to any one of claims 6 to 8, characterized in that: include: A database acquisition module, used to acquire a palm vein feature database, wherein the palm vein features in the database are divided into a plurality of clusters, wherein one cluster corresponds to one identity category, and one cluster corresponds to one cluster center feature; A query feature acquisition module is used to acquire the features of the palm vein image to be queried to obtain the query features; A candidate feature selection module is used to calculate the similarity between the cluster center feature and the query feature, and regard the clusters with the similarity greater than a preset similarity threshold as candidate clusters; A similar feature retrieval module is used to calculate the similarity between each palm vein feature in each candidate cluster and the query feature to obtain a plurality of feature similarities, sort the palm vein features in descending order of the feature similarities, and select the first plurality of palm vein features as similar feature retrieval results corresponding to the palm vein image to be queried; The matching module is used to perform feature matching of the query feature based on the feature similarity corresponding to the similar feature retrieval result. During matching, it is determined whether the maximum value of the feature similarity corresponding to the similar feature retrieval result is greater than a preset matching threshold; if so, the match is successful, and the identity corresponding to the query feature is authenticated as the identity of the palm vein feature in the similar feature retrieval result corresponding to the maximum feature similarity; otherwise, the match fails.

10. A computer device comprising a memory and a processor, wherein the memory stores computer-readable instructions, characterized in that: When the computer-readable instructions are executed by the processor, the processor performs the steps of the palm vein feature matching method based on identity clustering as claimed in any one of claims 6 to 8.

Images