Attendance management system based on face recognition

Abstract

The invention relates to the technical field of attendance management, in particular to an attendance management system based on face recognition. The system comprises a user registration and information management module, a cross-age sharing data network module, a dynamic personalized learning algorithm module, an intelligent template updating strategy module and a data security and user interface management module. According to the system, firstly, facial images of students are captured from multiple angles through a high-resolution camera, basic information input of the students is combined, and after being encrypted, the data are used for constructing personalized growth archives and recording facial feature changes of the students in different growth stages. The dynamic personalized learning algorithm module performs analysis according to historical facial image data, extracts and continuously optimizes facial feature parameters of each student, and adapts to changes of individual facial features along with time. According to the invention, an efficient, accurate and highly adaptive student attendance solution is provided, and campus management efficiency and data reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of attendance management, and more specifically, to an attendance management system based on face recognition. Background Art

[0002] In a campus environment, teachers and administrators face the challenge of ensuring the accuracy of student attendance records. Traditional attendance methods are not only time-consuming and laborious but also prone to problems such as proxy signing or missed records, which affect management efficiency and data reliability. The introduction of face recognition technology has automated and made the attendance process more reliable, significantly improving efficiency and reducing human error. However, although this technology performs well in the adult workplace, it has encountered bottlenecks in dealing with the rapid changes in the facial features of children and adolescents.

[0003] Due to the significant changes in the facial features of young students during their growth, existing face recognition systems often struggle to maintain long-term accuracy. A key shortcoming of existing recognition systems is the lack of an effective mechanism for tracking and managing the long-term changes in individual facial features. Traditional systems rely on fixed facial templates that are created at the initial registration and rarely updated. This static management mode cannot adapt to the significant changes in the facial features of children and adolescents over time, resulting in a decline in the recognition accuracy of the system during long-term use. The lack of archives for continuously recording and analyzing the development of individual facial features makes it difficult for existing systems to automatically adapt to the growth process of each student, limiting their application effect and reliability in an educational scenario. Summary of the Invention

[0004] The object of the present invention is to provide an attendance management system based on face recognition to solve the problem that existing systems are difficult to adapt to the long-term changes in the facial features of children and adolescents, resulting in a decline in recognition accuracy. Specifically, the cross-age data sharing network module creates personalized growth archives by integrating facial feature data at different growth stages, which details the evolution process of students' facial features from enrollment to graduation. The dynamic personalized learning algorithm module, based on these personalized growth archives, not only continuously optimizes and adjusts facial feature parameters but also uses deep learning technology to monitor and adapt to the subtle change trends of individual facial features in real time. This module uses historical data to optimize the model at the individual level and automatically adjusts parameters when new data is input each time, ensuring that the system can accurately identify the changes in students' facial features at different ages. In addition, when it is found that the change trends of the facial features of two or more students are similar, the module will also apply transfer learning technology to transfer the knowledge learned from one student to the model of another student, accelerating the model learning process and improving recognition accuracy.

[0005] To achieve the above object, an attendance management system based on face recognition is provided, including a user registration and information management module. The user registration and information management module collects and encrypts the initial facial images of users, and transfers the initial facial images to the cross-age shared data network module. The cross-age shared data network module constructs a personalized growth profile based on the initial facial images, and the personalized growth profile is used to record the changes in facial features of students at different growth stages;

[0006] It also includes a dynamic personalized learning algorithm module. The dynamic personalized learning algorithm module receives the personalized growth profile from the cross-age shared data network module, analyzes the historical facial image data in the personalized growth profile, extracts the facial feature parameters of each student, and based on the facial feature parameters, real-time monitors the input of new facial images. When the dynamic personalized learning algorithm module receives new data, it dynamically adjusts and optimizes the facial feature parameters to adapt to the changes in students' facial features. When it is found that the facial feature change trends of two users are similar, the transfer learning technology transfers the knowledge learned from one student to the model of another user to accelerate the learning process of the model. The dynamic personalized learning algorithm module transports the updated facial feature parameters to the intelligent template update strategy module to successfully perform face recognition on each student at different ages.

[0007] As a further improvement of this technical solution, the user registration and information management module captures the facial images of users from multiple angles through a high-resolution camera.

[0008] As a further improvement of this technical solution, the cross-age shared data network module realizes the anonymized sharing and updating of data through a secure transmission mechanism, identifies and adapts to the long-term changes in students' facial features, and at the same time supports data synchronization and model optimization in multiple environments.

[0009] As a further improvement of this technical solution, the dynamic personalized learning algorithm module optimizes the model at the individual level based on the data in the personalized growth profile.

[0010] As a further improvement of this technical solution, the dynamic personalized learning algorithm module also regularly updates the global model parameters.

[0011] As a further improvement of this technical solution, the user registration and information management module also collects and enters the basic information of students.

[0012] As a further improvement of this technical solution, the intelligent template update strategy module receives the optimized facial feature parameters from the dynamic personalized learning algorithm module, and evaluates the similarity between the existing face template and the latest facial image. When the similarity is lower than the preset threshold, it automatically triggers a progressive template update process.

[0013] As a further improvement of this technical solution, the intelligent template update strategy module calculates the weighted average of facial feature vectors in the recent period as the new face template to ensure the stability of the template.

[0014] As a further improvement of this technical solution, the data security and user interface management module receives the face template output by the intelligent template update strategy module and encrypts the face template using the Advanced Encryption Standard.

[0015] As a further improvement of this technical solution, the data security and user interface management module compares the updated face template with the real-time captured facial image to confirm the student's identity and record the attendance.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0017] 1. In this attendance management system based on face recognition, by introducing the dynamic personalized learning algorithm module and creating personalized growth profiles, the recognition accuracy and adaptability of the attendance management system are significantly improved. First, in the face of the challenge of rapid changes in the facial features of children and adolescents, the system can continuously optimize and adjust the facial feature parameters of each student. The dynamic personalized learning algorithm module uses deep learning technology to monitor and adapt to the changing trends of individual facial features in real time and continuously updates the model parameters according to the latest data. This mechanism not only ensures high recognition accuracy but also accelerates the learning process of the model through transfer learning technology. When similar facial feature change patterns are found, the knowledge learned from one student can be effectively transferred to the model of another student, further improving the recognition accuracy. This enables the system to maintain high performance during long-term use and greatly enhances its adaptability to students of different ages.

[0018] 2. In this attendance management system based on face recognition, the personalized growth profiles constructed by the cross-age data sharing network module provide comprehensive data support for each student. This module integrates the facial feature data of students at different growth stages and records the development process of their facial features, thus providing a solid foundation for the recognition system. These detailed profiles not only help improve the accuracy and reliability of the recognition system but also support more application scenarios, such as education management and security verification. At the same time, to ensure data security and privacy protection, all information is processed using the Advanced Encryption Standard and uploaded to the central server or cloud storage through a secure transmission mechanism. Such all-round data management and protection measures ensure that unauthorized access can be effectively prevented even during data transmission and storage, further enhancing the overall security of the system and user trust. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1It is a schematic diagram of the process structure of the attendance management system of the present invention;

[0020] Figure 2 It is a schematic diagram of the process structure of the cross - age shared data network module of the present invention;

[0021] Figure 3 It is a schematic diagram of the process structure of the dynamic personalized learning algorithm module of the present invention;

[0022] Figure 4 It is a schematic diagram of the process structure of the data security and user interface management module of the present invention.

[0023] The meanings of each label in the figure are as follows:

[0024] Among them: 100, user registration and information management module; 200, cross - age shared data network module; 300, dynamic personalized learning algorithm module; 400, intelligent template update strategy module; 500, data security and user interface management module. Specific embodiments

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0026] The following explanations are given for technical terms well - known in the art:

[0027] Robustness refers to the ability of a system, model or algorithm to maintain stable performance in the face of noise, outliers, incomplete information in the data, or deviations between the model assumptions and the actual situation. Simply put, it refers to the anti - interference ability and stability of the system.

[0028] Please refer to Figure 1 As shown, the purpose of this embodiment is to provide an attendance management system based on face recognition, aiming to provide an efficient, accurate and highly adaptable student attendance solution for users by integrating the user registration and information management module 100, cross - age shared data network module 200, dynamic personalized learning algorithm module 300, intelligent template update strategy module 400, and data security and user interface management module 500.

[0029] First, the user registration and information management module 100 is responsible for the collection of initial data and the entry of basic information. This module captures the facial images of students from multiple angles through a high-resolution camera (it should be noted here that all personal information collected in this application has been obtained with full consent and authorization, and the collection, use, and processing of relevant information need to comply with the relevant laws, regulations, and standards of the relevant countries and regions), and records relevant metadata (such as shooting time, environmental conditions, etc.). At the same time, the basic information of students (name, student number, class, etc.) is collected and entered. These data will serve as the basis for subsequent processing to ensure that the facial features of each student can be accurately identified and tracked by the system.

[0030] During the data collection process, the system first needs to obtain facial images from multiple angles. To ensure the quality and consistency of the data, the system will perform a quality assessment on each facial image, and the formula is as follows:

[0031] Q(I) = w1·S sharpness (I) + w2·S illu min ation (I) + w3·S pose (I);

[0032] In the formula, Q(I) represents the quality score of image I, which is a comprehensive score used to evaluate the overall quality of the image;

[0033] w1, w2, and w3 represent weight coefficients. These weights correspond to the clarity, lighting conditions, and pose scores of the image respectively, and their sum is 1. The weight coefficients are used to adjust the influence degree of different factors on the final quality score;

[0034] S sharpness (I) represents the clarity score of the image, which indicates the clarity of the image, that is, whether the details in the image are clear enough to be distinguishable;

[0035] S illu min ation (I) represents the lighting condition score of the image, which indicates the lighting condition of the image, that is, whether the image is taken under good lighting conditions to avoid over-bright or over-dark situations;

[0036] S pose (I) represents the pose (angle) score of the image, which indicates the angle and pose of the face in the image to ensure that the facial features can be accurately captured and identified.

[0037] Only when Q(I) reaches a certain threshold will the image be included in the personalized growth profile. This quality assessment mechanism ensures that the system can obtain high-quality facial images, thereby improving the accuracy of subsequent recognition. In addition, to ensure the comprehensiveness of facial images, the system usually captures images from multiple angles (such as the front, left, and right), providing richer facial feature information, which helps improve the adaptability of the system.

[0038] In the user registration and information management module 100, the system not only needs to capture multi-angle facial images of students, but also collect and input the basic information of students, which can be manually entered through the user interface or automatically imported from the school's existing database. All this information, together with the facial images and their metadata, forms a complete data record. For example, for each student, the system will create a dataset containing name, student ID, class affiliation, shooting time and environmental conditions, as well as multi-angle facial images. This process ensures that the personalized growth profile of each student can comprehensively reflect the development and changes of their facial features, providing a solid data foundation for subsequent dynamic learning algorithms.

[0039] Subsequently, all the collected data will be encrypted and stored in the local database, and uploaded to the cross-age shared data network module 200 as needed. Here, the cross-age shared data network module 200 receives the initial facial images and basic information from the user registration and information management module 100. The encrypted data is sent to the central server or cloud storage through a secure transmission mechanism to ensure the security of the data during transmission and storage. Through this process, the basic information of the user is successfully transformed into the initial facial image, laying the foundation for the next steps.

[0040] In practical applications, we found that relying solely on data collected from a specific age group or a specific time period is difficult to meet the high recognition accuracy and wide adaptability required for the long-term operation of the system. Over time, the facial features of students will change, resulting in a decrease in the accuracy of the model trained based on early data, affecting the overall performance and reliability of the system. For this reason, the present invention introduces the cross-age shared data network module 200, aiming to make up for this deficiency by integrating data from different age groups, so as to ensure that the system can maintain a high level of recognition ability and adaptability in various environments. This module not only enhances the system performance, but also constructs a detailed record of the evolution of facial features for each student, and then creates a personalized growth profile.

[0041] See Figure 2As shown, the cross - age shared data network module 200 enhances system performance by integrating data resources of different age groups, while building a detailed record of the facial feature evolution for each student. First, each student is assigned a unique identifier (UID) at the time of registration, which is used to track all stages of their facial image records. The entire database can be represented as:

[0042]

[0043] In the formula, UID i represents the unique identifier of the i - th student;

[0044] F i,t represents the facial image data of the student at time point t;

[0045] t1, t2,..., t n represent different time points respectively, and n is the subscript indicating the n - th time point.

[0046] [[ID=2l]]After these data are processed by deep - learning algorithms, feature vectors V i,t are generated to represent the facial features at specific time points. Next, using the cross - age shared data network module 200, the system conducts long - term tracking and analysis of the facial feature changes of each student. To capture the long - term change patterns of individual facial features, the system adopts a method based on time - series analysis to construct a development trajectory model. This model aims to extract key information from a series of feature vectors that can reflect the evolution law of individual facial features, thus forming the core part of the personalized growth profile. The formula is as follows:

[0047]

[0048] In the formula, M i represents the development trajectory model of the i - th student;

[0049] f() is a function or algorithm used to extract key information from the feature vectors V i,t obtained at each time point, showing the change law of individual facial features over time;

[0050] t1, t2,..., t n represent different time points respectively.

[0051] This personalized growth profile can not only help improve the recognition accuracy, but also support more application scenarios, such as education management, security verification, etc. In addition, to ensure data security and privacy protection, all collected information is encrypted using the Advanced Encryption Standard (AES) and uploaded to the central server or cloud storage through a secure transmission mechanism. This ensures that even during data transmission and storage, unauthorized access can be effectively prevented.

[0052] In summary, by introducing the cross - age - group shared data network module 200 and creating personalized growth profiles, the present invention greatly enhances the recognition ability and adaptability of the system, providing solid support for various application scenarios. However, in the face of the continuous change of users' facial features over time and the increasing amount of data, the static model gradually shows its limitations and is difficult to meet the ever - increasing demand for accurate recognition. This reveals the deficiencies of the existing system in dealing with long - term data accumulation and dynamic changes. To solve this problem, we further propose a dynamic personalized learning algorithm module 300. This module continuously optimizes and adjusts the model using the latest data to ensure that the system can effectively adapt to the long - term evolution of individual facial features while maintaining high accuracy and reliability. Through this dynamic adjustment mechanism, not only is the intelligence level and response speed of the system improved, but also the system becomes more flexible and efficient, enabling accurate recognition in different scenarios.

[0053] See Figure 3 As shown, once the dynamic personalized learning algorithm module 300 receives the personalized growth profile of each student from the cross - age - group shared data network module 200, the dynamic personalized learning algorithm module 300 begins to play its core optimization role. The personalized growth profile contains all the facial feature vector sequences of the student from the time of enrollment to the current time point and their corresponding unique identifiers (UIDs), and these data record the changes in the student's facial features over time.

[0054] Once the personalized growth profile is received, the system first performs model optimization at the individual level based on these input data. For each student i, the system uses an update formula to continuously adjust the facial feature parameters to adapt to the changes in individual facial features over time. Specifically, the following update formula is adopted:

[0055]

[0056] In the formula, represents the model parameters of the i - th student at time point t n+1 ;

[0057] represents the model parameters of the i - th student at time point t n ;

[0058] α represents the learning rate, which controls the step size of each update. The learning rate determines the speed and amplitude of the model parameter update;

[0059] is the loss function, which is used to measure the gap between the current model prediction and the actual feature vector ;

[0060] is the gradient of the loss function with respect to the model parameters , representing the partial derivative of the loss function with respect to the model parameters, guiding how the model parameters should be adjusted to reduce the error.

[0061] Through this method, the system can continuously optimize the facial feature parameters according to the latest facial feature data, thereby more precisely adapting to the long-term changes in individual facial features. In addition, given that there may be similar facial feature evolution patterns among different users, the dynamic personalized learning algorithm module 300 further applies transfer learning technology, which means that when it is found that the facial feature change trends of two users are similar, the system effectively transfers the knowledge learned from one user to the model of another user to accelerate the model learning process and improve the recognition accuracy. This mechanism not only enhances the adaptability and flexibility of the system, but also promotes data sharing and model optimization among different users, improving the overall performance.

[0062] In addition to individual-level optimization, this module also focuses on data accumulation and analysis at the group level. Over time, the system automatically analyzes the facial feature evolution trends of all users and extracts generally applicable rules from them, integrating them into the global model. Specifically, the system regularly updates the global model parameter G, and the training based on the datasets of all users is as follows:

[0063]

[0064] where G new represents the updated global model parameter;

[0065] G old represents the current global model parameter;

[0066] β represents the global learning rate, controlling the step size of each update. The global learning rate determines the speed and amplitude of the global model parameter update;

[0067] N represents the total number of users;

[0068] V i,t represents the facial feature vector of the i-th user at time point t;

[0069] L(V i,t ,G old ) is the loss function, used to measure the prediction error of the current global model for the facial feature vector V i,t of the i-th student;

[0070] represents the gradient of the loss function with respect to the global model parameter G oldThe gradient, which represents the partial derivative of the loss function with respect to the global model parameters, guides how the global model parameters should be adjusted to reduce the error.

[0071] This strategy not only improves the recognition accuracy when new users first use it, but also enhances the robustness and generalization ability of the entire system. By receiving personalized growth profiles and processing them using the dynamic personalized learning algorithm module 300, the present invention solves the problem that traditional static models are difficult to adapt to long-term data accumulation and dynamic changes, significantly improving the recognition ability and adaptability of the system. This enables the system to provide accurate and reliable recognition services in a variety of application scenarios while maintaining high efficiency and high reliability. This continuously evolving ability is the key innovation point of the present invention, laying a solid foundation for the realization of a more intelligent and flexible recognition system.

[0072] Through the dynamic personalized learning algorithm module 300, the system can use deep learning technology to continuously train and optimize the data in the personalized growth profile, and adjust the model parameters in real time to adapt to the changing trends of individual facial features. However, although these measures greatly enhance the adaptive ability of the system, as students grow older and their facial features continue to change, fixed facial templates may gradually lose accuracy, affecting the stability of long-term use. To solve this problem and ensure the continuity and accuracy of recognition, the present invention proposes an intelligent template update strategy module 400. This module receives the optimized facial feature parameters and update strategy suggestions transmitted from the dynamic personalized learning algorithm module 300, and based on this, automatically determines when and how to update an individual's face template.

[0073] Based on the received optimized facial feature parameters and update strategy suggestions, the intelligent template update strategy module 400 first evaluates the similarity between the existing face template and the latest facial image. If the similarity is found to be lower than the preset threshold, the template update process is automatically triggered. To prevent the problem of inconsistent recognition caused by suddenly changing the template, the module adopts a progressive update method, gradually integrating the newly collected facial image information into the existing face template during the update process. Specifically, for each user, the system calculates a weighted average of the facial feature vectors over a recent period of time as the new template:

[0074]

[0075] In the formula, T i,new represents the updated new face template;

[0076] T i,old represents the currently used old face template;

[0077] γ represents the update rate, which determines the weight distribution between new data and old templates. It is usually a value between 0 and 1;

[0078] m represents the number of new facial feature vectors used to update the template;

[0079] represents the facial feature vector of student i at the j-th time point.

[0080] This method not only helps with smooth transitions but also effectively improves the stability and accuracy of the template. In addition to adopting the progressive update method, the intelligent template update strategy module 400 also has the ability to automatically determine when to update the template. The system determines whether to update the template by analyzing the similarity score between the newly collected facial feature vectors and the existing template. If the similarity is lower than a preset threshold, the template update process is triggered. This mechanism ensures that the template is updated only when there are significant changes in the user's facial features, avoiding unnecessary frequent update operations.

[0081] After the above steps, the intelligent template update strategy module 400 will output the updated face templates. These updated templates not only contain the latest facial feature information but also retain the key features in the historical data, thus ensuring the accuracy and coherence of recognition. Finally, these updated face templates will be passed to the data security and user interface management module 500 for encrypted transmission, storage, and used for subsequent functions such as authentication and attendance record query. In this way, the intelligent template update strategy module 400 not only solves the problem of fixed templates becoming invalid over time but also further improves the recognition performance and user experience of the entire system.

[0082] After the intelligent template update strategy module 400 completes the automatic update and optimization of the face templates, the data security and user interface management module 500 further ensures that these updated templates and all related data can be accurately applied to the successful recognition process of each student. See Figure 4As shown, this module first ensures that all data is encrypted using the Advanced Encryption Standard (AES) during transmission and storage to ensure data security and privacy protection. At the same time, by implementing strict access control policies, only authorized personnel can access sensitive information, which further strengthens the security of the system. In addition, in order to achieve the final accurate identification of each student, this module integrates an efficient verification system that compares the updated face template with the real-time captured facial image to confirm the student's identity. Once the verification is successful, the system will immediately record the attendance and allow users to view information such as personal attendance records and identification results through an intuitive and user-friendly interface. Administrators can also use this interface for system configuration, monitoring the system status, and handling any exceptions to ensure the stable operation and efficient management of the system.

[0083] In summary, through the introduction of the dynamic personalized learning algorithm module 300, the present invention effectively improves the management of personalized growth profiles. By dynamically adjusting and learning the facial feature changes of each student, the characteristics of different growth stages are accurately captured to ensure the accuracy and reliability of identification. This not only supports the continuous identification and attendance record of students during their school years, but also provides strong data support for the establishment of personalized growth profiles, enabling parents and teachers to more conveniently track and support the all-round development of students.

[0084] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An attendance management system based on face recognition, including a user registration and information management module (100). The user registration and information management module (100) collects and encrypts the initial facial images of users, and the user registration and information management module (100) transfers the initial facial images to the cross-age shared data network module (200), characterized in that: The cross - age - group shared data network module (200) constructs a personalized growth profile based on the initial facial images, and the personalized growth profile is used to record the changes in facial features of students at different growth stages; It also includes a dynamic personalized learning algorithm module (300). The dynamic personalized learning algorithm module (300) receives the personalized growth profile from the cross - age - group shared data network module (200), analyzes the historical facial image data in the personalized growth profile, extracts the facial feature parameters of each student, and based on the facial feature parameters, real - time monitors the input of new facial images. When the dynamic personalized learning algorithm module (300) receives new data, it dynamically adjusts and optimizes the facial feature parameters to adapt to the changes in students' facial features. When it is found that the facial feature change trends of two users are similar, transfer learning technology transfers the knowledge learned from one student to the model of another user to accelerate the model learning process. The dynamic personalized learning algorithm module (300) sends the updated facial feature parameters to the intelligent template update strategy module (400) for face recognition of each student at different ages.

2. The attendance management system based on face recognition according to claim 1, characterized in that: The user registration and information management module (100) captures the facial images of users from multiple angles through a high - resolution camera and uses the following formula to evaluate the quality of each image: Q(I) = w1·S sharpness (I) + w2·S illumination (I) + w3·S pose (I); In the formula, Q(I) represents the quality score of the image; w1, w2, and w3 represent weight coefficients, which respectively correspond to the clarity, lighting condition, and pose score of the image, and their sum is 1; S sharpness (I) represents the clarity score of the image; S illumination (I) represents the illumination condition score of the image; S pose (I) represents the pose (angle) score of the image.

3. The attendance management system based on face recognition according to claim 1, wherein: The cross - age - group shared data network module (200) realizes the anonymized sharing and update of data through a secure transmission mechanism, identifies and adapts to the long - term changes in students' facial features, and at the same time supports data synchronization and model optimization in multiple environments.

4. The attendance management system based on face recognition according to claim 1, wherein: The dynamic personalized learning algorithm module (300) performs model optimization at the individual level based on the data in the personalized growth profile and continuously adjusts the facial feature parameters of each student using the following formula: In the formula, represents the model parameters of the i-th student at time point t n+1 ; denote the model parameters of the i-th student at time point t n ; α represents the learning rate, which controls the step size of each update. The learning rate determines the speed and amplitude of model parameter updates; is a loss function used to measure the gap between the current model prediction and the actual feature vector ; is the gradient of the loss function with respect to the model parameters , representing the partial derivative of the loss function with respect to the model parameters, guiding how the model parameters should be adjusted to reduce the error.

5. The attendance management system based on face recognition according to claim 1, characterized in that: The dynamic personalized learning algorithm module (300) also regularly updates the global model parameters through the following formula: where G new represents the updated global model parameters; G old represents the current global model parameters; β represents the global learning rate, which controls the step size of each update; N represents the total number of users; V i,t represents the facial feature vector of the i-th user at time point t; L(V i,t ,G old ) is the loss function used to measure the prediction error of the current global model for the i-th student's facial feature vector V i,t ; Denotes the gradient of the loss function with respect to the global model parameter G old .

6. The attendance management system based on face recognition according to claim 2, wherein: The user registration and information management module (100) also collects and enters the basic information of students.

7. The attendance management system based on face recognition according to claim 5, characterized in that: The intelligent template update strategy module (400) receives the optimized facial feature parameters from the dynamic personalized learning algorithm module (300) and evaluates the similarity between the existing face template and the latest facial image. When the similarity is lower than the preset threshold, it automatically triggers a progressive template update process.

8. The attendance management system based on face recognition according to claim 7, wherein: The intelligent template update strategy module (400) calculates the weighted average of the facial feature vectors in the recent period as the new face template to ensure the stability of the template.

9. The attendance management system based on face recognition according to claim 8, characterized in that: The data security and user interface management module (500) receives the face template output by the intelligent template update strategy module (400) and encrypts the face template using the Advanced Encryption Standard.

10. The attendance management system based on face recognition according to claim 9, characterized in that: The data security and user interface management module (500) compares the updated face template with the real-time captured facial image to confirm the student's identity and record the attendance situation.

Images