Multi-device synchronous attendance checking method and device based on biological recognition

Abstract

The invention relates to the technical field of intelligent attendance checking, in particular to a multi-device synchronous attendance checking method based on biological recognition, which comprises the following steps: collecting biological characteristic data of an attendance checked person, preprocessing the biological characteristic data, extracting biological characteristic vectors with distinction degree, fusing the biological characteristic vectors with different modalities, and obtaining a multi-device synchronous attendance checking result. Generating a comprehensive biological characteristic template; calling a pre-stored employee biological characteristic database, verifying the comprehensive biological characteristic template and the employee biological characteristic database, carrying out preliminary classification on the comprehensive biological characteristic template, and obtaining a corresponding classification result; and determining a target type database in the employee biological characteristic database based on a classification result, comparing the comprehensive biological characteristic template with the target type database to verify identity data corresponding to the checked personnel, and generating a corresponding latest attendance record after verification succeeds. According to the invention, the accuracy of attendance checking can be improved by integrating multiple biological recognition technologies.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent attendance, and in particular to a multi-device synchronous attendance method and apparatus based on biometrics. Background Art

[0002] With the development of information technology, traditional paper or card-based attendance methods have gradually been eliminated due to problems such as low efficiency and easy forgery. In recent years, attendance systems based on biometric technology have been widely used due to their high efficiency, accuracy, and difficulty in forging. However, most existing biometric attendance systems mainly rely on a single biometric feature (such as fingerprint or facial recognition). This single-modal biometric method may cause recognition failure in certain special circumstances (such as finger injuries, wearing masks, etc.), affecting the accuracy of attendance.

[0003] In addition, with the expansion of enterprise scale and the diversification of work locations, the synchronization of attendance data between multiple devices has become an urgent problem to be solved. At present, although there are some multi-device attendance systems on the market, they generally have shortcomings such as data synchronization delay and information inconsistency, and cannot meet the requirements of real-time and accuracy.

[0004] From the above, we can see that how to improve the accuracy of attendance by integrating multiple biometric technologies remains to be solved. Summary of the invention

[0005] In order to improve the accuracy of attendance by integrating multiple biometric technologies, the present application provides a multi-device synchronous attendance method and apparatus based on biometrics.

[0006] In the first aspect, the present application provides a multi-device synchronous attendance method based on biometrics, which adopts the following technical solution:

[0007] A multi-device synchronous attendance method based on biometrics, comprising:

[0008] Collecting biometric data of the personnel being checked, preprocessing the biometric data, extracting a discriminative biometric vector using a deep learning algorithm, fusing biometric vectors of different modalities, and generating a comprehensive biometric template, wherein the biometric data includes fingerprints, facial images, iris images, and vein patterns;

[0009] Retrieving a pre-stored employee biometric database, verifying the comprehensive biometric template with the employee biometric database, and using a hierarchical clustering algorithm to preliminarily classify the comprehensive biometric template during the verification process and obtain corresponding classification results;

[0010] Based on the classification result, a target type database in the employee biometric database is determined, and the comprehensive biometric template is compared with the target type database to verify the identity data corresponding to the person being checked, and after successful verification, the corresponding latest attendance record is generated.

[0011] By adopting the above technical solution, by integrating multiple biometric technologies (such as fingerprints, facial images, iris images and vein patterns), combined with deep learning algorithms and hierarchical clustering algorithms, high-precision identity authentication of the attendance personnel is achieved. This method not only improves the accuracy and reliability of attendance, but also ensures stability and efficiency under different environmental conditions through multimodal biometric fusion and adaptive environmental adjustment.

[0012] Optionally, the method also includes: converting the integrated biometric template into an input format suitable for a clustering algorithm; using a K-means clustering algorithm to preliminarily classify the integrated biometric template whose input format has been converted to obtain a corresponding first preliminary classification result, and simultaneously using a DBSCAN clustering algorithm to preliminarily classify the integrated biometric template whose input format has been converted to obtain a corresponding second preliminary classification result; performing a first classification based on the first preliminary classification result, performing a second classification based on the result of the first classification using the second preliminary classification result within each cluster, integrating the results of the first classification and the results of the second classification to generate a final preliminary classification result.

[0013] By adopting the above technical solution, the comprehensive biometric template is converted into an input format suitable for the clustering algorithm, and the K-means and DBSCAN clustering algorithms are combined to perform multi-level preliminary classification. First, the K-means algorithm is used to obtain the first preliminary classification result, and then the DBSCAN algorithm is used in each cluster to obtain the second preliminary classification result. Finally, the two classification results are integrated to generate the final preliminary classification result. This method not only improves the comprehensiveness and accuracy of classification.

[0014] Optionally, the method also includes: converting the preliminary classification results into an input format; optimizing the preliminary classification results in the converted input format using a graph cut optimization algorithm to obtain a corresponding first optimization result; simultaneously optimizing the preliminary classification results in the converted input format using a probability optimization algorithm to obtain a corresponding second optimization result; integrating the first optimization result and the second optimization result to generate a final optimized classification result.

[0015] By adopting the above technical solution, the preliminary classification results are converted into an input format suitable for the optimization algorithm, and the preliminary classification results are optimized using the graph cut optimization algorithm and the probability optimization algorithm respectively to obtain the first optimization result and the second optimization result. Finally, the two optimization results are integrated to generate the final optimized classification result. This method not only improves the accuracy and consistency of the classification results.

[0016] Optionally, in the process of verifying the integrated biometric template with the employee biometric database, the method also includes: calculating the matching degree between the integrated biometric template and each candidate template in the employee biometric database, and obtaining a corresponding matching degree data set; sorting all candidate templates according to the matching degree data set, and determining the target template based on the highest matching degree data; and determining the corresponding comparison order based on the target template.

[0017] By adopting the above technical solution, the matching degree between the comprehensive biometric template and each candidate template in the employee biometric database is calculated to generate a matching degree data set, and all candidate templates are sorted according to the set. Finally, the target template is determined based on the highest matching degree data, and the corresponding comparison order is determined for verification, thereby improving the efficiency and accuracy of the comparison.

[0018] Optionally, in the process of collecting the biometric data, the method also includes: acquiring environmental data of the attendance device in real time, acquiring corresponding data of the person being attended, the environmental data including light intensity data, temperature data, and humidity data; performing light correction on the data of the person being attended according to the light intensity data to acquire first correction data, performing temperature correction on the data of the person being attended according to the temperature data to acquire second correction data, and performing humidity correction on the data of the person being attended according to the humidity data to acquire third correction data; and determining the corresponding biometric data based on the first correction data, the second correction data, and the third correction data.

[0019] By adopting the above technical solution, in the process of collecting biometric data, the environmental data of the attendance equipment (including light intensity, temperature and humidity) is obtained in real time, and the data of the person being clocked is corrected for light, temperature and humidity based on the environmental data. Through these correction steps, accurate biometric data is finally determined, thereby improving the reliability and consistency of the data.

[0020] Optionally, after generating the corresponding latest attendance record, the method also includes: retrieving the historical attendance record corresponding to the identity data and triggering incremental detection; determining the corresponding incremental data based on the latest attendance record and the historical attendance record, wherein the incremental data includes a newly added biometric template, a modified biometric template and a deleted biometric template; and adjusting the frequency of incremental detection based on the incremental data.

[0021] By adopting the above technical solution, after the latest attendance record is generated, the historical attendance record of the corresponding identity data is retrieved and the incremental detection is triggered. By comparing the latest attendance record with the historical attendance record, the incremental data (including newly added, modified and deleted biometric templates) is determined, and the frequency of incremental detection is adjusted according to the incremental data, thereby improving the efficiency of the system and the accuracy of data management.

[0022] In the second aspect, the present application provides a multi-device synchronous attendance device based on biometrics, which adopts the following technical solution:

[0023] A multi-device synchronous attendance device method based on biometrics, comprising:

[0024] A comprehensive biometric template generation module collects biometric data of the personnel being checked, pre-processes the biometric data, extracts a biometric vector with discrimination using a deep learning algorithm, and fuses biometric vectors of different modalities to generate a comprehensive biometric template, wherein the biometric data includes fingerprints, facial images, iris images, and vein patterns;

[0025] The classification result acquisition module retrieves a pre-stored employee biometric database, verifies the comprehensive biometric template with the employee biometric database, and uses a hierarchical clustering algorithm to preliminarily classify the comprehensive biometric template during the verification process and obtain the corresponding classification result;

[0026] The latest attendance record generation module determines the target type database in the employee biometric database based on the classification result, compares the comprehensive biometric template with the target type database to verify the identity data corresponding to the person being checked, and after successful verification, generates the corresponding latest attendance record.

[0027] In the third aspect, the present application provides a multi-device synchronous attendance method based on biometrics, which adopts the following technical solution:

[0028] A multi-device synchronous attendance method based on biometrics comprises a processor, wherein the processor runs a program of any one of the multi-device synchronous attendance methods based on biometrics described above.

[0029] In a fourth aspect, the present application provides a storage medium, which adopts the following technical solution:

[0030] A storage medium stores a program of any one of the above-mentioned multi-device synchronous attendance methods based on biometrics.

[0031] In summary, the present application includes at least one of the following beneficial technical effects:

[0032] 1. The multi-device synchronous attendance method based on biometrics integrates multiple biometric technologies (such as fingerprints, facial images, iris images and vein patterns) and combines deep learning algorithms and hierarchical clustering algorithms to achieve high-precision identity authentication of the personnel being clocked. This method not only improves the accuracy and reliability of attendance, but also ensures stability and efficiency under different environmental conditions through multimodal biometric fusion and adaptive environmental adjustment. In addition, through multi-level preliminary classification and optimization technology, the comprehensiveness and accuracy of classification are further improved, ensuring the robustness and intelligence level of the system.

[0033] 2. After generating the latest attendance record, this method retrieves the historical attendance record and triggers incremental detection to determine the incremental data (including newly added, modified and deleted biometric templates), and dynamically adjusts the frequency of incremental detection according to the incremental data, thereby improving the efficiency of the system and the accuracy of data management. At the same time, in the process of collecting biometric data, the environmental data of the attendance device is obtained in real time and the light, temperature and humidity are corrected, ensuring the reliability and consistency of the data, further improving the overall system performance and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 The present invention is a flowchart of a multi-device synchronous attendance method based on biometrics according to an exemplary embodiment.

[0035] Figure 2 The present invention is a structural block diagram of a multi-device synchronous attendance apparatus based on biometrics according to an exemplary embodiment. DETAILED DESCRIPTION

[0036] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings.

[0037] In the description of this specification, the description with reference to the terms "certain embodiments", "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0038] The present application embodiment discloses a multi-device synchronous attendance method based on biometrics, referring to Figure 1 ,include:

[0039] S100, collect biometric data of the personnel being checked, pre-process the biometric data, use a deep learning algorithm to extract a discriminative biometric vector, fuse biometric vectors of different modalities, and generate a comprehensive biometric template.

[0040] Among them, a variety of biometric collection devices are installed on the attendance equipment, such as fingerprint recognition, camera (for facial image and iris image collection), vein recognition, etc. When the person being checked uses the attendance equipment, his / her fingerprint, facial image, iris image, vein pattern and other biometric data are collected; when the person being checked uses the attendance equipment, he / she is guided through the user interface to correctly place his / her fingers, face the camera, etc., to ensure that high-quality biometric data is collected, and the biometric data such as fingerprint, facial image, iris image and vein pattern of the person being checked are collected, and metadata such as collection time and device ID are recorded.

[0041] Secondly, use filters or other denoising algorithms to remove noise from the collected biometric data, improve the clarity of the data, and standardize the biometric data, such as normalization and scaling, to ensure the consistency and comparability of the data.

[0042] It should be pointed out here that in the process of using deep learning algorithms to extract biometric feature vectors with discriminativeness, the following are performed: using the CNN model to extract features from fingerprint images, extracting features such as the texture and ridges of the fingerprint; using the CNN model to extract features from facial images, extracting features such as key points and contours of the face; using the CNN model to extract features from iris images, extracting features such as the texture and pattern of the iris; using the CNN model to extract features from vein images, extracting features such as the distribution and shape of the veins; and converting the extracted features into feature vectors, which should have high discrimination and be able to effectively distinguish different individuals.

[0043] In the process of fusing biometric feature vectors of different modalities, when feature alignment is required, it is first necessary to ensure that the biometric feature vectors of different modalities have the same dimension, and at the same time align the biometric feature vectors of different modalities to ensure synchronization and consistency between different modal data. In the corresponding feature fusion process, it is necessary to assign different weights according to the importance of biometric feature vectors of different modalities for weighted fusion. Multimodal fusion algorithms, such as multi-layer perceptron (MLP) and attention mechanism, can be used to fuse biometric feature vectors of different modalities into a comprehensive biometric feature vector. Finally, the fused comprehensive biometric feature vector generates a comprehensive biometric template. The template should have high discrimination and robustness and can be effectively used for identity authentication.

[0044] S110, retrieve a pre-stored employee biometric database, verify the comprehensive biometric template with the employee biometric database, and use a hierarchical clustering algorithm to preliminarily classify the comprehensive biometric template during the verification process and obtain corresponding classification results.

[0045] Among them, a connection is established between the attendance device and the central database through a network or local connection, and identity authentication is performed before connecting to the database to ensure that only authorized devices and users can access the database; then a query request is sent to the central database to request the retrieval of pre-stored employee biometric data. The central database responds to the query request and transmits the pre-stored employee biometric data to the attendance device. These data include employee fingerprints, facial images, iris images, and vein patterns, etc.; the received employee biometric data is parsed into a usable format, such as feature vectors or image data, and the parsed data is cached in the memory of the attendance device for fast access and processing.

[0046] In addition, during the verification process, the classification results generated by the hierarchical clustering algorithm are analyzed to determine the category to which the comprehensive biometric template belongs. Based on the classification results, the target type database in the employee biometric database is determined, that is, the sub-database containing the employee biometric data that matches the preliminary classification results.

[0047] Through the above steps, employee biometric data can be efficiently retrieved and processed, and preliminary classification can be performed through a hierarchical clustering algorithm to improve the accuracy and efficiency of verification.

[0048] S120, determining a target type database in the employee biometric database based on the classification result, comparing the comprehensive biometric template with the target type database to verify the identity data corresponding to the person being checked, and generating the corresponding latest attendance record after successful verification.

[0049] First, the category to which the comprehensive biometric template belongs is parsed from the classification results obtained in the preliminary classification process, where the classification results are usually one or more category labels, indicating the group to which the comprehensive biometric template most likely belongs; then, based on the parsed category labels, the specific category to which the comprehensive biometric template belongs is determined. At the same time, the employee biometric database needs to be segmented according to categories to generate multiple sub-databases, each of which contains employee biometric data under a specific category; the corresponding sub-database can be selected as the target type database based on the category determined by the classification results. It should be pointed out here that the target type database contains employee biometric data that is most likely to match the comprehensive biometric template.

[0050] In addition, the comprehensive biometric template and the employee biometric data in the target type database are converted into an input format suitable for the comparison algorithm, and a suitable similarity measurement method (such as Euclidean distance, cosine similarity) is selected to calculate the similarity between the comprehensive biometric template and each employee biometric data in the target type database. The calculated similarity value is recorded to generate a matching data set, which includes the matching value of each employee biometric data, and the employee biometric data with the highest matching degree is selected as the target template. A matching degree threshold can be set. If the highest matching degree exceeds the threshold, the match is considered to be successful and a verification success mark is generated; otherwise, the match fails and a verification failure mark is generated.

[0051] After successful verification, the verification information needs to be recorded, including the name, work number, verification time, verification location, etc. of the person being checked. By establishing a connection with the central database, the latest attendance record generated will be inserted into the central database to ensure data integrity and consistency.

[0052] Through the execution steps of S100 to S120 above, by integrating multiple biometric technologies (such as fingerprints, facial images, iris images and vein patterns), combined with deep learning algorithms and hierarchical clustering algorithms, high-precision identity authentication of the personnel being clocked is achieved; not only the accuracy and reliability of attendance are improved, but also the stability and efficiency under different environmental conditions are ensured through multimodal biometric fusion and adaptive environmental adjustment. Specifically: by integrating multiple biometric technologies, the accuracy and robustness of identity authentication are improved; environmental conditions are monitored in real time and adjusted dynamically to ensure data quality and equipment performance under different environments; hierarchical clustering algorithms are used for preliminary classification to improve the comprehensiveness and accuracy of classification; by calculating the matching degree, the templates that are most likely to match are prioritized to improve the efficiency and accuracy of the comparison.

[0053] It should be pointed out here that the generated preliminary classification results also include the following processes:

[0054] S111, converting the comprehensive biometric template into an input format suitable for a clustering algorithm.

[0055] Among them, the comprehensive biometric template is converted into a feature vector form, each feature vector represents a comprehensive biometric template, and the dimension of the feature vector depends on the number of extracted features; the feature vector is then standardized to ensure that the value of each feature is within the same range. In the embodiment of the present application, Z-score standardization or Min-Max standardization can be used; finally, the standardized feature vector is adjusted to an input format suitable for the clustering algorithm, for example, the feature vector is organized into a two-dimensional array, each row represents a sample, and each column represents a feature.

[0056] S112, use the K-means clustering algorithm to preliminarily classify the comprehensive biometric templates that have been converted into an input format, and obtain a corresponding first preliminary classification result. At the same time, use the DBSCAN clustering algorithm to preliminarily classify the comprehensive biometric templates that have been converted into an input format, and obtain a corresponding second preliminary classification result.

[0057] Among them, K samples can be randomly selected as the initial cluster centers. The value of K can be predetermined based on experience and data distribution. A suitable distance measurement method, such as Euclidean distance, is selected to calculate the distance from each sample to each cluster center. Each sample is assigned to the cluster where the nearest cluster center is located, and the center of each cluster needs to be recalculated. The mean of all samples in the cluster is usually taken as the new cluster center. Then, the process of cluster assignment and cluster center update is repeated until the cluster center no longer changes or the maximum number of iterations is reached. At this time, the final clustering result is the first preliminary classification result, and each sample is assigned to a cluster.

[0058] In addition, two main parameters of the DBSCAN algorithm can be set: ε (neighborhood radius) and MinPts (minimum number of samples). These two parameters can be adjusted according to the distribution of the data and experimental results. At the same time, the ε neighborhood of each sample needs to be defined, that is, all samples whose distance to the sample does not exceed ε.

[0059] It should be pointed out here that if the number of samples in the ε neighborhood of a sample is greater than or equal to MinPts, the sample is a core point; if the number of samples in the ε neighborhood of a sample is less than MinPts, the sample is a boundary point or a noise point.

[0060] Starting from an unvisited core point, all samples within its ε neighborhood are added to the same cluster, and the cluster is expanded until no new core points can be added. Samples not included in any cluster are marked as noise points. The final clustering result is the second preliminary classification result. Each sample is assigned to a cluster or marked as a noise point.

[0061] S113, performing a first classification based on the first preliminary classification result, performing a second classification in each cluster based on the result of the first classification using the second preliminary classification result, integrating the result of the first classification and the result of the second classification to generate a final preliminary classification result.

[0062] Among them, according to the first preliminary classification result generated by the K-means clustering algorithm, the comprehensive biometric template is divided into several clusters, and the sample data in each cluster is extracted to prepare for the second classification. In each cluster, the second preliminary classification result generated by the DBSCAN clustering algorithm is used to further classify the samples in the cluster; in each cluster, according to the classification result of the DBSCAN algorithm, the samples are further subdivided into smaller subclusters or marked as noise points. Finally, the results of the first classification and the second classification are integrated to generate the final preliminary classification result. The specific method is to determine the final classification label of each sample as its comprehensive result in the first classification and the second classification.

[0063] Through the above steps, the K-means algorithm can quickly divide the data into multiple clusters and provide preliminary classification results, while the DBSCAN algorithm can identify clusters of any shape and process noise points to provide more detailed classification results. Finally, through the dual clustering of K-means and DBSCAN, the advantages of both are combined to improve the comprehensiveness and accuracy of classification.

[0064] In addition, in the process of generating the final optimized classification result, the method further includes:

[0065] S114, converting the preliminary classification results into an input format.

[0066] Among them, as in the above format conversion process, corresponding standardization and normalization processing are required. First, from the classification results obtained in the preliminary classification process, the classification label and corresponding feature vector of each sample are extracted, and then the feature vector is standardized to ensure that the value of each feature is in the same range. Z-score standardization or Min-Max standardization is usually used. Finally, the standardized feature vector and classification label are adjusted to an input format suitable for the optimization algorithm. For example, the data is organized into a two-dimensional array, each row represents a sample, each column represents a feature, and the last column is the classification label.

[0067] S115, optimizing the preliminary classification result of the converted input format using a graph cut-based optimization algorithm to obtain a corresponding first optimization result.

[0068] First, each sample is defined as a node in the graph. The attributes of the node include the feature vector and the preliminary classification label. According to the similarity between samples (such as Euclidean distance, cosine similarity, etc.), edges are established between nodes. The weight of the edge represents the similarity between samples. At the same time, the data term energy function is defined to measure the matching degree between the feature vector of each node and its preliminary classification label. The smoothing term energy function is also defined to measure the consistency of classification labels between adjacent nodes.

[0069] A graph cut algorithm (such as the maximum flow minimum cut algorithm) is used to optimize the graph, minimize the energy function, and generate an optimized classification result. The classification label of each sample is reallocated to minimize the energy function, and the final classification label of each sample is extracted from the optimized graph to generate a first optimization result.

[0070] S116, simultaneously optimizing the preliminary classification result of the converted input format using a probability-based optimization algorithm to obtain a corresponding second optimization result.

[0071] Among them, it is necessary to select a suitable probability model, such as Gaussian mixture model (GMM), hidden Markov model (HMM), etc., and estimate the model parameters by using maximum likelihood estimation (MLE) or expectation maximization (EM) algorithm; then calculate the likelihood value of each sample under different classification labels, that is, the probability that the sample belongs to a certain category, and calculate the posterior probability of each sample according to Bayes' theorem, that is, the probability that the sample belongs to a certain category under a given feature vector. Then use a probability optimization algorithm (such as the EM algorithm) to optimize the preliminary classification results, maximize the posterior probability, and thus generate an optimized classification result. The classification label of each sample is reallocated to maximize the posterior probability. Finally, the final classification label of each sample is extracted from the optimized probability model to generate a second optimization result.

[0072] S117, integrating the first optimization result and the second optimization result to generate a final optimization classification result.

[0073] Among them, the classification label of each sample is voted, and the classification label with the most votes is selected as the final classification result. According to the confidence of the first optimization result and the second optimization result, a weight is assigned to each result, and a weighted fusion is performed to generate the final classification result. The integrated classification labels are used to generate the final optimized classification result to ensure that each sample has a unique classification label. The generated final optimized classification result can also be verified to ensure the accuracy and consistency of the classification result.

[0074] Based on the above steps S114 to S117, by converting the preliminary classification results into an input format suitable for the optimization algorithm, the preliminary classification results are optimized using a graph cut-based optimization algorithm and a probability-based optimization algorithm to obtain a first optimization result and a second optimization result, respectively. Finally, the two optimization results are integrated through a voting mechanism or weight fusion to generate a final optimized classification result. This method not only improves the accuracy and consistency of the classification results, but also enhances the robustness and adaptability of the system, ensuring high efficiency and reliability in complex data environments.

[0075] In verifying the combined biometric template against the employee biometric database, the method further comprises:

[0076] S1201, calculating the matching degree between the comprehensive biometric template and each candidate template in the employee biometric database, and obtaining a corresponding matching degree data set.

[0077] It is necessary to ensure that each candidate template in the comprehensive biometric template and employee biometric database has been converted into feature vector form, and the feature vector needs to be standardized. Then select a suitable similarity measurement method, such as Euclidean distance, cosine similarity, Hamming distance, etc.

[0078] For the comprehensive biometric template and each candidate template, calculate the similarity or distance between them. For example, if the Euclidean distance is used, the calculation formula is: Among them, x i and i are the i-th eigenvalues ​​of the comprehensive biometric template and the candidate template, respectively, and n is the dimension of the eigenvector. The calculated matching value of each candidate template can be recorded to generate a matching data set. The matching data set is usually a list or array, and each element contains the ID of the candidate template and the corresponding matching value.

[0079] S1202: sort all candidate templates according to the matching degree data set, and determine the target template based on the highest matching degree data.

[0080] Among them, it is necessary to select a suitable sorting algorithm, such as quick sort, merge sort, etc., to sort the matching data set. According to different similarity measurement methods, ascending or descending sorting can be selected. For example, if Euclidean distance is used, the lower the matching degree, the higher the similarity, so ascending sorting is selected; if cosine similarity is used, the higher the matching degree, the higher the similarity, so descending sorting is selected; finally, from the sorted matching data set, the candidate template with the highest matching degree is selected as the target template.

[0081] Threshold judgment: Set a matching threshold. If the highest matching degree exceeds the threshold, the match is considered successful; otherwise, the match fails.

[0082] S1203: Determine a corresponding comparison order based on the target template.

[0083] The comparison order is defined according to actual needs. For example, the comparison can be performed in order from high to low matching degree, or the comparison order can be determined according to other priority rules. According to the defined comparison order, the target template is verified in turn, and the verification process may include further feature comparison, identity confirmation and other steps. If a matching template is found during the comparison process, the verification is successful and the corresponding latest attendance record is generated; if no matching template is found, the verification fails and a corresponding prompt message is generated.

[0084] Based on the execution of the above steps S1201 to S1203, by calculating the matching degree between the comprehensive biometric template and each candidate template in the employee biometric database, a matching degree data set is generated, and all candidate templates are sorted according to the matching degree data set, and the template with the highest matching degree is selected as the target template. Based on the target template, the corresponding comparison order is determined for further verification. This method not only improves the efficiency and accuracy of the comparison, but also ensures the reliability of the system, and can quickly and accurately identify the identity of the attendance personnel in a large amount of data.

[0085] In the process of collecting biometric data, the solution of the present application also takes into account the impact of the real-time environment, so the method also includes:

[0086] S101, obtaining the environment data of the attendance equipment in real time, and obtaining the corresponding attendance personnel data.

[0087] As mentioned above, environmental sensors such as light sensors, temperature sensors, humidity sensors, etc. are installed on the attendance equipment to monitor the environmental conditions around the equipment in real time, and then the data of the environmental sensors can be collected regularly or in real time, including light intensity, temperature, humidity, etc. When the attendance personnel use the attendance equipment, they are guided by the user interface to correctly place their fingers and face the camera, etc., to ensure the collection of high-quality biometric data. The biometric data of the attendance personnel such as fingerprints, facial images, iris images and vein patterns can be collected, and metadata such as the collection time and device ID can be recorded.

[0088] S102, performing illumination correction on the attendance data of the person to be checked according to the illumination intensity data to obtain first correction data, performing temperature correction on the attendance data of the person to be checked according to the temperature data to obtain second correction data, and performing humidity correction on the attendance data of the person to be checked according to the humidity data to obtain third correction data.

[0089] Among them, for the light intensity data, a suitable light correction algorithm can be selected, such as histogram equalization, gamma correction, white balance correction, etc.; according to the light intensity data, the parameters of the light correction algorithm are adjusted to adapt to different lighting conditions, and the collected facial images and iris images are subjected to light correction to ensure the clarity and consistency of the images under different lighting conditions, and then the corrected image data is recorded to generate the first correction data.

[0090] For temperature data, a suitable temperature correction algorithm can be selected, such as a temperature compensation algorithm, a thermal noise correction algorithm, etc.; according to the temperature data, the parameters of the temperature correction algorithm are adjusted to adapt to different temperature conditions; then the collected biometric data is temperature corrected to eliminate the impact of temperature changes on the data, and the corrected biometric data is recorded to generate second correction data.

[0091] For humidity data, a suitable humidity correction algorithm can be selected, such as a humidity compensation algorithm, a humidity filtering algorithm, etc.; according to the humidity data, the parameters of the humidity correction algorithm are adjusted to adapt to different humidity conditions, and the collected biometric data is humidity corrected to eliminate the impact of humidity changes on the data; then the corrected biometric data is recorded to generate third correction data.

[0092] S103, determining corresponding biometric data based on the first correction data, the second correction data, and the third correction data.

[0093] Among them, first of all, it is necessary to ensure that the timestamps and sample orders of the first correction data, the second correction data and the third correction data are consistent to facilitate subsequent fusion processing; then select a suitable fusion algorithm, such as weighted average, multimodal fusion, etc., to fuse the three correction data into the final biometric data; finally, the fused data is generated into the final biometric data, and the generated final biometric data needs to be quality checked to ensure the integrity and accuracy of the data.

[0094] Based on executing the above steps S101 to S103, the environmental data (including light intensity, temperature, humidity) of the attendance equipment is acquired in real time, and the collected biometric data is corrected for light, temperature and humidity to generate the first correction data, the second correction data and the third correction data. Finally, the final biometric data is generated by data fusion. This method not only improves the accuracy and consistency of the biometric data, but also enhances the adaptability and robustness of the system, ensuring stability and efficiency under different environmental conditions. Through these steps, the system can better cope with environmental changes and improve the accuracy and reliability of attendance.

[0095] After generating the corresponding latest attendance record, the method further includes:

[0096] S121, retrieve the historical attendance records corresponding to the identity data and trigger incremental detection.

[0097] As mentioned above, it is necessary to establish a connection between the attendance device and the central database through a network or local connection to ensure the stability and security of data transmission. Then, before connecting to the database, identity authentication is performed to ensure that only authorized devices and users can access the database.

[0098] Send a query request to the central database to retrieve the historical attendance records of the person being checked. The central database responds to the query request and transmits the historical attendance records of the person being checked to the attendance device; and parses the received historical attendance records into a usable format, such as feature vectors or image data, and caches the parsed data in the memory of the attendance device for fast access and processing. After the latest attendance record is generated, the incremental detection is immediately triggered to start the incremental data detection process.

[0099] S122, determining corresponding incremental data based on the latest attendance record and the historical attendance record, wherein the incremental data includes a newly added biometric template, a modified biometric template, and a deleted biometric template.

[0100] Among them, data preparation is first required to ensure that the timestamps and sample sequences of the latest attendance records and historical attendance records are consistent, and the latest attendance records and historical attendance records are converted into a format suitable for comparison, such as a feature vector or a structured data table.

[0101] Then, the incremental data is identified by identifying the newly added biometric templates in the latest attendance record, that is, the templates that do not exist in the historical attendance records but exist in the latest attendance records; identifying the modified biometric templates in the latest attendance records, that is, the templates that exist in the historical attendance records but have changed in the latest attendance records; and identifying the biometric templates deleted in the latest attendance records, that is, the templates that exist in the historical attendance records but do not exist in the latest attendance records.

[0102] Finally, incremental data recording is performed, which requires recording the newly added, modified, and deleted biometric templates that have been identified, generating incremental data sets, and storing the incremental data sets in a central database for subsequent management and query.

[0103] S123, adjusting the frequency of incremental detection based on the incremental data.

[0104] Among them, statistical analysis is performed on the incremental data set to calculate the number and frequency of newly added, modified and deleted biometric templates, which can analyze the changing trend of the incremental data and identify whether there are abnormal or regular changes.

[0105] The frequency of incremental detection is dynamically adjusted according to the changing trend of incremental data. For example, if the frequency of incremental data changes is high, the detection frequency can be increased; if the frequency of incremental data changes is low, the detection frequency can be reduced.

[0106] Here, we need to set a threshold for incremental data changes. When the incremental data changes exceed the threshold, the detection frequency is automatically adjusted. Then, according to the adjusted frequency, the incremental detection task is performed to ensure efficient operation of the system. Accordingly, it is necessary to continuously monitor the effect of incremental detection and make further adjustments and optimizations based on the actual results.

[0107] Based on executing the above steps S121 to S123, after generating the latest attendance record, retrieve the historical attendance record of the person being checked, and trigger incremental detection. By comparing the latest attendance record with the historical attendance record, determine the incremental data (including newly added, modified, and deleted biometric templates), and dynamically adjust the frequency of incremental detection according to the changing trend of the incremental data. This method not only improves the efficiency of the system and the accuracy of data management, but also enhances the adaptability and intelligence level of the system, ensuring that attendance data can be managed efficiently and accurately under different circumstances. Through these steps, the system can detect and process data changes in a timely manner, and improve the reliability and real-time performance of attendance management.

[0108] The present application embodiment discloses a multi-device synchronous attendance device based on biometrics, referring to Figure 2 , the devices include but are not limited to:

[0109] The comprehensive biometric template generation module 200 collects the biometric data of the personnel to be checked, pre-processes the biometric data, extracts the biometric vector with discrimination using a deep learning algorithm, and fuses the biometric vectors of different modalities to generate a comprehensive biometric template, wherein the biometric data includes fingerprints, facial images, iris images and vein patterns;

[0110] The classification result acquisition module 210 retrieves the pre-stored employee biometric database, verifies the comprehensive biometric template with the employee biometric database, and uses a hierarchical clustering algorithm to preliminarily classify the comprehensive biometric template during the verification process and obtain the corresponding classification result;

[0111] The latest attendance record generation module 220 determines the target type database in the employee biometric database based on the classification result, compares the comprehensive biometric template with the target type database to verify the identity data corresponding to the person being checked, and after successful verification, generates the corresponding latest attendance record.

[0112] Furthermore, the device includes but is not limited to:

[0113] A format conversion module, used for converting the comprehensive biometric template into an input format suitable for a clustering algorithm;

[0114] A preliminary classification module, using a K-means clustering algorithm to perform preliminary classification on the integrated biometric template after the input format conversion, so as to obtain a corresponding first preliminary classification result, and simultaneously using a DBSCAN clustering algorithm to perform preliminary classification on the integrated biometric template after the input format conversion, so as to obtain a corresponding second preliminary classification result;

[0115] The preliminary classification result acquisition module performs a first classification based on the first preliminary classification result, performs a second classification in each cluster based on the result of the first classification using the second preliminary classification result, and integrates the results of the first classification and the second classification to generate a final preliminary classification result.

[0116] Furthermore, the device includes but is not limited to:

[0117] A format conversion module, used to convert the preliminary classification results into an input format;

[0118] A first optimization result acquisition module, which optimizes the preliminary classification result of the converted input format using a graph cut optimization algorithm to obtain a corresponding first optimization result;

[0119] A second optimization result acquisition module, which simultaneously optimizes the preliminary classification result of the converted input format using a probability-based optimization algorithm to obtain a corresponding second optimization result;

[0120] The integration module is used to integrate the first optimization result and the second optimization result to generate a final optimization classification result.

[0121] Furthermore, the device includes but is not limited to:

[0122] A matching degree calculation module is used to calculate the matching degree between the comprehensive biometric template and each candidate template in the employee biometric database, and obtain a corresponding matching degree data set;

[0123] The target template determination module sorts all candidate templates according to the matching degree data set, and determines the target template based on the highest matching degree data;

[0124] The comparison order determination module is used to determine the corresponding comparison order based on the target template.

[0125] Furthermore, the device includes but is not limited to:

[0126] The environmental data acquisition module is used to obtain the environmental data of the attendance equipment in real time and the corresponding attendance personnel data. The environmental data includes light intensity data, temperature data, and humidity data;

[0127] A correction module, used for performing illumination correction on the attendance data of the personnel to be checked according to the illumination intensity data to obtain first correction data, used for performing temperature correction on the attendance data of the personnel to be checked according to the temperature data to obtain second correction data, and used for performing humidity correction on the attendance data of the personnel to be checked according to the humidity data to obtain third correction data;

[0128] The biometric data determination module is used to determine corresponding biometric data based on the first correction data, the second correction data and the third correction data.

[0129] Furthermore, the device includes but is not limited to:

[0130] The historical attendance record retrieval module is used to retrieve the historical attendance records corresponding to the identity data and trigger incremental detection;

[0131] An incremental data determination module, which is used to determine corresponding incremental data based on the latest attendance record and the historical attendance record, wherein the incremental data includes a newly added biometric template, a modified biometric template, and a deleted biometric template;

[0132] The frequency adjustment module is used to adjust the frequency of incremental detection based on the incremental data.

[0133] An embodiment of the present application also discloses a multi-device synchronous attendance method based on biometrics, including a processor, in which a program of any one of the multi-device synchronous attendance methods based on biometrics described above is run.

[0134] The embodiment of the present application also discloses a storage medium storing a program of any one of the above-mentioned multi-device synchronous attendance methods based on biometrics.

[0135] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A multi-device synchronous attendance method based on biometrics, characterized in that: include: Collecting biometric data of the personnel being checked, preprocessing the biometric data, extracting a discriminative biometric vector using a deep learning algorithm, fusing biometric vectors of different modalities, and generating a comprehensive biometric template, wherein the biometric data includes fingerprints, facial images, iris images, and vein patterns; Retrieving a pre-stored employee biometric database, verifying the comprehensive biometric template with the employee biometric database, and using a hierarchical clustering algorithm to preliminarily classify the comprehensive biometric template during the verification process and obtain corresponding classification results; Based on the classification result, a target type database in the employee biometric database is determined, and the comprehensive biometric template is compared with the target type database to verify the identity data corresponding to the person being checked, and after successful verification, the corresponding latest attendance record is generated.

2. The multi-device synchronous attendance method based on biometrics according to claim 1 is characterized in that: The method also includes: Converting the synthetic biometric template into an input format suitable for a clustering algorithm; Using a K-means clustering algorithm to preliminarily classify the integrated biometric templates that have been converted into an input format, and obtaining a corresponding first preliminary classification result, and simultaneously using a DBSCAN clustering algorithm to preliminarily classify the integrated biometric templates that have been converted into an input format, and obtaining a corresponding second preliminary classification result; A first classification is performed based on the first preliminary classification result, a second classification is performed in each cluster based on the result of the first classification using the second preliminary classification result, and the result of the first classification and the result of the second classification are integrated to generate a final preliminary classification result.

3. The multi-device synchronous attendance method based on biometrics according to claim 2 is characterized in that: The method also includes: Convert the preliminary classification results into input format; Optimizing the preliminary classification result of the converted input format using a graph cut optimization algorithm to obtain a corresponding first optimization result; At the same time, a probability-based optimization algorithm is used to optimize the preliminary classification result of the converted input format to obtain a corresponding second optimization result; The first optimization result and the second optimization result are integrated to generate a final optimization classification result.

4. The multi-device synchronous attendance method based on biometrics according to claim 2 is characterized in that: In the process of verifying the integrated biometric template with the employee biometric database, the method further includes: Calculate the matching degree between the comprehensive biometric template and each candidate template in the employee biometric database, and obtain a corresponding matching degree data set; Sorting all candidate templates according to the matching degree data set, and determining the target template based on the highest matching degree data; A corresponding alignment order is determined based on the target template.

5. The multi-device synchronous attendance method based on biometrics according to claim 3 is characterized in that: In the process of collecting the biometric data, the method further includes: Obtain the environmental data of the attendance equipment in real time and the corresponding attendance personnel data. The environmental data includes light intensity data, temperature data, and humidity data; Performing illumination correction on the attendance data of the personnel to be checked according to the illumination intensity data to obtain first correction data, performing temperature correction on the attendance data of the personnel to be checked according to the temperature data to obtain second correction data, and performing humidity correction on the attendance data of the personnel to be checked according to the humidity data to obtain third correction data; Corresponding biometric data is determined based on the first correction data, the second correction data, and the third correction data.

6. The multi-device synchronous attendance method based on biometrics according to claim 5 is characterized in that: After generating the corresponding latest attendance record, the method further includes: Retrieve the historical attendance records corresponding to the identity data and trigger incremental detection; Determining corresponding incremental data based on the latest attendance record and the historical attendance record, wherein the incremental data includes a newly added biometric template, a modified biometric template, and a deleted biometric template; The frequency of delta detection is adjusted based on the delta data.

7. A multi-device synchronous attendance device based on biometrics, characterized in that: include: A comprehensive biometric template generation module collects biometric data of the personnel being checked, pre-processes the biometric data, extracts a biometric vector with discrimination using a deep learning algorithm, and fuses biometric vectors of different modalities to generate a comprehensive biometric template, wherein the biometric data includes fingerprints, facial images, iris images, and vein patterns; The classification result acquisition module retrieves a pre-stored employee biometric database, verifies the comprehensive biometric template with the employee biometric database, and uses a hierarchical clustering algorithm to preliminarily classify the comprehensive biometric template during the verification process and obtain the corresponding classification result; The latest attendance record generation module determines the target type database in the employee biometric database based on the classification result, compares the comprehensive biometric template with the target type database to verify the identity data corresponding to the person being checked, and after successful verification, generates the corresponding latest attendance record.

8. The multi-device synchronous attendance device based on biometrics according to claim 7 is characterized in that: The device also includes: A format conversion module, used for converting the comprehensive biometric template into an input format suitable for a clustering algorithm; A preliminary classification module, using a K-means clustering algorithm to perform preliminary classification on the integrated biometric template after the input format conversion, so as to obtain a corresponding first preliminary classification result, and simultaneously using a DBSCAN clustering algorithm to perform preliminary classification on the integrated biometric template after the input format conversion, so as to obtain a corresponding second preliminary classification result; The preliminary classification result acquisition module performs a first classification based on the first preliminary classification result, performs a second classification in each cluster based on the result of the first classification using the second preliminary classification result, and integrates the results of the first classification and the second classification to generate a final preliminary classification result.

9. A multi-device synchronous attendance method based on biometrics, characterized in that: It includes a processor, in which runs a program of a multi-device synchronous attendance method based on biometrics as described in any one of claims 1-6.

10. A storage medium, characterized in that: A program for a multi-device synchronous attendance method based on biometrics as described in any one of claims 1 to 6 is stored.

Images