Portable mobile attendance design method and system based on face recognition

Abstract

The invention provides a portable mobile attendance design method and system based on face recognition, and belongs to the technical field of attendance equipment management. According to the method, a face recognition model of a lightweight algorithm framework is constructed, and after anchor frame generation, pruning optimization and transfer learning training, the face recognition model is deployed in a portable terminal of a main control chip, an embedded visual processing chip and a wireless communication module; the terminal supports battery power supply and offline work, has a fixed and mobile double deployment mode switching capability, and can be used in a scene without fixed installation conditions and wired network coverage; and by establishing communication with a remote rear-end platform, attendance data sharing and remote one-key upgrading of the model and system software are realized. The system is compact in structure, flexible in deployment and suitable for multi-scene attendance checking requirements, and the technical scheme is high in feasibility.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of attendance equipment management, in particular to a portable mobile attendance design method and system based on face recognition. BACKGROUND

[0002] With the development of face recognition technology, fixed face recognition access control equipment has been widely used in campus, office and other scenarios. It realizes the matching attendance of face and name through local integration of camera and recognition algorithm. In the prior art, the mainstream fixed equipment relies on wall installation, wired power supply and fixed network wiring, the core adopts a multi-modal collection scheme of visible light and depth camera, and combines an edge computing architecture to reduce cloud dependence, but there are still significant limitations: first, the application scenario is highly fixed, and it cannot cover outdoor or temporary scenes without power supply and network such as playgrounds and gymnasiums, resulting in incomplete coverage of attendance needs; second, software upgrade and maintenance are inconvenient, and need to be updated and maintained on site, which is high in maintenance cost. The defects of the prior art make it impossible to meet the needs of mobile and convenient attendance, and there is an urgent need for a mobile attendance solution that takes into account the adaptability of the scene and the convenience of maintenance. SUMMARY

[0003] The present application provides a portable mobile attendance design method and system based on face recognition, which is characterized by constructing a face recognition model based on a YOLO series lightweight algorithm framework, generating anchor frames adapted to different face scales through a K-means clustering algorithm, and deploying the portable terminal carrying a main control chip, an embedded visual processing chip and a wireless communication module after pruning optimization and migration learning training. The terminal is configured with a rechargeable battery, supports face image collection, recognition and attendance data caching in offline state, and has a fixed and mobile dual deployment mode switching function. The portable terminal establishes a communication connection with the remote back-end platform, on the one hand realizes the uploading and sharing of attendance data, and on the other hand receives the update instructions pushed by the platform, and the main control chip completes the decompression, deployment update and automatic restart of the model file to realize the remote one-key upgrade of the system software. The remote back-end platform synchronously stores the attendance data, face feature database and model version files to ensure data backup and sharing. The overall technical solution breaks through the scene limitations and maintenance bottlenecks of existing attendance equipment, forms a mobile attendance solution with software and hardware cooperation, flexible deployment and efficient maintenance.

[0004] The present application provides a portable mobile attendance design method based on face recognition, comprising the following steps: Construct a lightweight face recognition model, train and optimize the model based on a target detection algorithm framework, and obtain an optimized face recognition model adapted to the hardware of the portable terminal; The optimized face recognition model can be deployed on an embedded portable terminal, supporting battery power and wireless communication, and can be deployed in mobile locations without fixed installation conditions or wired network coverage. The system acquires facial images of the attendance subjects through the image acquisition module, and uses the deployed optimized facial recognition model to complete facial detection and recognition, generating attendance data. Based on the communication connection established between the portable terminal and the remote backend platform, the transmission path and receiving channel are obtained; Attendance data sharing is completed based on the transmission path, and model update instructions are received through the receiving channel to complete remote upgrades and data backups of the face recognition model and system software.

[0005] In the portable mobile attendance design method based on face recognition described in this application, the step of constructing a lightweight face recognition model, training and optimizing the model based on an object detection algorithm framework to obtain an optimized face recognition model adapted to portable terminal hardware includes: K-means clustering algorithm was used to generate anchor boxes for face datasets in attendance scenarios, resulting in multiple anchor boxes adapted to different face scales; A lightweight face recognition model is obtained by combining the YOLO series lightweight algorithm framework with the aforementioned multiple anchor boxes. Based on the deployment requirements of the lightweight face recognition model after transfer learning training, the lightweight face recognition model is deployed to an embedded vision processing chip.

[0006] The portable mobile attendance design method based on face recognition described in this application also includes: Based on the optimization requirements of the lightweight face recognition model, pruning is performed on the model to filter and remove convolutional kernels whose absolute weight values ​​are lower than a preset threshold. Based on the hardware adaptation requirements for deploying the lightweight face recognition model, the number of parameters of the lightweight face recognition model is compressed to a preset range through the pruning process. At the same time, a standard for controlling the loss of recognition accuracy is set to meet the preset first recognition accuracy.

[0007] In the portable mobile attendance design method based on face recognition described in this application, the portable terminal includes: Based on the mobile attendance function requirements of the portable terminal, a hardware architecture combining a main control chip, an embedded vision processing chip, and a wireless communication module is adopted. The main control chip is a microcontroller, the wireless communication module is a wireless network communication module, and a rechargeable battery is configured as the power supply module. Based on the usage requirements of offline scenarios, through the cooperation of the hardware architecture and power supply module, the portable terminal can complete face image acquisition, recognition, and attendance data caching in offline mode.

[0008] In the portable mobile attendance design method based on face recognition described in this application, the optimization of the lightweight face recognition model further includes: The non-maximum suppression threshold is dynamically adjusted based on the density of people in the attendance scenario. When the attendance scenario is a densely populated scenario, the non-maximum suppression threshold is adjusted to the first preset range; When the attendance scenario is a sparsely populated scenario, the non-maximum suppression threshold is adjusted to the second preset range.

[0009] In the portable mobile attendance design method based on face recognition described in this application, the training of the lightweight face recognition model further includes: Based on the training requirements of the lightweight face recognition model, data augmentation processing is performed on the collected face images; The data enhancement employs image rotation, scaling, random cropping, and color tone adjustment. Based on the functional division requirements of training, verification and testing of the lightweight face recognition model, the face images after the data augmentation process are divided into training set, verification set and test set. The facial images include different angles, lighting conditions, and facial expressions.

[0010] The portable mobile attendance design method based on face recognition described in this application also includes: Based on the multi-scenario deployment requirements of the portable terminal, a function for switching between fixed deployment mode and mobile deployment mode is configured; In the fixed deployment mode, attendance data is uploaded in real time via the network without the need for manual supervision. In the mobile deployment mode, the cached attendance data is automatically synchronized to the remote backend platform based on the network status recovery.

[0011] The portable mobile attendance design method based on face recognition described in this application also includes: Based on the model version management requirements of the remote backend platform, store the historical versions and the latest version of the face recognition model; Based on the model update requirements, the remote backend platform pushes update instructions to the portable terminal, which then downloads the latest version of the model file via the wireless communication module. The main control chip completes the decompression and deployment update of the model files. After the update is completed, the portable terminal automatically restarts to make the new version effective.

[0012] Secondly, this application provides a portable mobile attendance system based on face recognition, characterized in that it includes: A portable time and attendance terminal includes a main control module, a vision processing module, an image acquisition module, a wireless communication module, a power supply module, and a display module; The main control module is electrically connected to the vision processing module, image acquisition module, wireless communication module, power supply module, and display module, respectively. The visual processing module deploys the lightweight face recognition model; The power supply module is a lithium battery; The wireless communication module is used to communicate with a remote backend platform; The remote backend platform includes a data management module and an update and maintenance module; The data management module is used to store attendance data, facial feature database, and attendance records; The update and maintenance module is used to store the latest version of the face recognition model and system software, push update instructions to portable attendance terminals, and realize data backup and sharing.

[0013] The system also includes a memory and a processor. The memory contains a program for a portable mobile attendance design method based on face recognition. When the program for the portable mobile attendance design method based on face recognition is executed by the processor, it performs the following steps: A lightweight face recognition model is constructed, and the model is trained and optimized based on the object detection algorithm framework to obtain an optimized face recognition model adapted to portable terminal hardware. The optimized face recognition model can be deployed on an embedded portable terminal, supporting battery power and wireless communication, and can be deployed in mobile locations without fixed installation conditions or wired network coverage. The system acquires facial images of the attendance subjects through the image acquisition module, and uses the deployed optimized facial recognition model to complete facial detection and recognition, generating attendance data. Based on the communication connection established between the portable terminal and the remote backend platform, the transmission path and receiving channel are obtained; Attendance data sharing is completed based on the transmission path, and model update instructions are received through the receiving channel to complete remote upgrades and data backups of the face recognition model and system software.

[0014] As can be seen from the above, this invention provides a portable mobile attendance design method and system based on face recognition. The core lies in constructing a lightweight face recognition model adapted to portable terminals. This model is based on the YOLO series of lightweight algorithm frameworks. It performs image rotation, scaling, and other data augmentation processing on the face dataset of attendance scenarios and divides it into training, validation, and test sets. Combined with the K-means clustering algorithm, it generates anchor boxes adapted to different face scales. After transfer learning training, it further compresses parameters through pruning and dynamically adjusts the non-maximum suppression threshold according to the density of people in the attendance scenario to complete model optimization. The optimized model is deployed on a portable terminal with a hardware architecture of a main control chip, embedded vision processing chip, and wireless communication module. The terminal is equipped with a rechargeable battery as a power supply module, supporting offline face image acquisition, recognition, and attendance data caching. It also has a dual deployment mode switching function (fixed and mobile), allowing flexible use in scenarios without fixed installation conditions or wired network coverage. The portable terminal establishes a communication connection with a remote backend platform through the wireless communication module. It enables real-time uploading of attendance data or synchronous sharing of data after network recovery. On the other hand, it receives model update instructions pushed by the platform, and the main control chip completes the decompression, deployment, and automatic restart of the model files to take effect, realizing remote one-click upgrade of system software. The remote backend platform stores attendance data, facial feature database and attendance records through the data management module, and stores model version files through the update and maintenance module to ensure data backup and sharing. The overall technical solution optimizes the entire process from model construction and terminal design to platform collaboration, breaking through the bottlenecks of limited scenario coverage and inconvenient upgrade and maintenance of existing attendance devices, forming a mobile attendance solution with efficient hardware and software collaboration, flexible and convenient deployment, and adaptability to multiple scenario needs.

[0015] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing embodiments of this application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This application provides a high-level flowchart of a portable mobile attendance design method based on face recognition, used for attendance processing via face recognition.

[0018] Figure 2 A flowchart illustrating a portable mobile attendance design method based on face recognition, provided for an embodiment of this application; Figure 3 A flowchart illustrating the construction of a lightweight face recognition model for a portable mobile attendance design method based on face recognition, provided in this application embodiment; Figure 4 A flowchart illustrating the optimization of a lightweight face recognition model for a portable mobile attendance design method based on face recognition, provided in an embodiment of this application. Figure 5 This is a structural block diagram of a portable mobile attendance system based on face recognition, provided as an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0020] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first" and "second" are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0021] Please refer to Figure 1 , Figure 1This is a high-level flowchart of a portable mobile attendance design method based on face recognition, as described in some embodiments of this application. The high-level flowchart can be summarized as follows: First, through data augmentation and dataset partitioning, combined with the YOLO series of lightweight algorithm frameworks and K-means clustering anchor box generation, and through transfer learning, pruning, and dynamic adjustment of the non-maximum suppression threshold, a lightweight face recognition model adapted to portable terminals is constructed. Next, the model is deployed to a portable terminal with a specific hardware architecture, a rechargeable battery, and support for dual deployment mode switching. Subsequently, the terminal selects the deployment mode as needed, collects face images, and generates attendance data through model detection and recognition, which is then cached offline and uploaded online in real time. Finally, the terminal wirelessly communicates with a remote backend platform to achieve data sharing and synchronization, while simultaneously receiving model update instructions to complete remote upgrades, and the platform synchronously stores and backs up relevant data and model versions.

[0022] Please refer to Figure 2 , Figure 2 This is a flowchart of a portable mobile attendance design method based on face recognition, which is one of the embodiments of this application.

[0023] The first aspect of this invention discloses a portable mobile attendance design method based on face recognition for use in terminal devices, such as computers and mobile phones. This portable mobile attendance design method based on face recognition includes the following steps: S201. Construct a lightweight face recognition model, train and optimize the model based on the object detection algorithm framework, and obtain an optimized face recognition model adapted to portable terminal hardware. S202. Deploy the optimized face recognition model to an embedded portable terminal, which supports battery power and wireless communication, and can be deployed in scenarios without fixed installation conditions or wired network coverage. S203. Acquire the facial image of the attendance subject through the image acquisition module, and use the deployed optimized facial recognition model to complete facial detection and recognition, and generate attendance data; S204. Based on the communication connection established between the portable terminal and the remote backend platform, obtain the transmission path and receiving channel; S205. Based on the transmission path, attendance data sharing is completed, and model update instructions are received through the receiving channel to complete remote upgrades and data backups of the face recognition model and system software.

[0024] The core technical process of this invention begins with the construction of a lightweight face recognition model. Specifically, based on the YOLO series of lightweight object detection algorithm frameworks, it first performs data augmentation processing such as rotation and scaling on the collected face images of attendance scenes containing different angles, lighting conditions, and facial expressions, and divides them into training sets, validation sets, and test sets. Then, it uses K-means... Clustering algorithms generate multiple anchor boxes adapted to different face scales. Transfer learning is used to accelerate model training. Subsequently, pruning is performed to remove convolutional kernels with absolute weights below a preset threshold to compress the number of parameters. Simultaneously, the non-maximum suppression threshold is dynamically adjusted based on the density of people in the attendance scenario, ultimately resulting in an optimized face recognition model adapted to portable terminal hardware. This optimized model is then deployed to an embedded portable terminal. The terminal employs a hardware architecture consisting of a main control chip, an embedded vision processing chip, and a wireless communication module. The main control chip is a microcontroller, and the wireless communication module is a wireless network communication module. A rechargeable battery serves as the power supply, supporting both battery power and wireless communication. It requires no fixed installation or wired network coverage, allowing for mobile deployment in various scenarios and switching between fixed and mobile deployment modes. The terminal then acquires facial images of the attendance subjects through an image acquisition module, uses the deployed optimized face recognition model to perform face detection and recognition, and generates attendance data. (Fixed deployment mode) No manual intervention is required. In online mode, attendance data can be uploaded in real time. In mobile deployment mode, it can work in areas with weak network signals. In offline mode, attendance data is cached and automatically synchronized when the network is restored. Finally, the portable terminal establishes a communication connection with the remote backend platform. Based on this connection, a transmission path for uploading and sharing attendance data and a receiving channel for model update instructions are obtained. The transmission path enables the sharing and synchronization of attendance data to the remote backend platform. The remote backend platform stores attendance data, facial feature database, and attendance records through the data management module. At the same time, it receives model update instructions pushed by the remote backend platform through the receiving channel. The main control chip completes the decompression and deployment update of the model file. After the update is completed, the terminal automatically restarts to take effect. The update and maintenance module of the remote backend platform stores the historical and latest versions of the facial recognition model, realizing remote upgrades and data backups of the facial recognition model and system software. The entire process forms a complete mobile attendance technology solution with hardware and software collaboration, flexible deployment, and efficient maintenance.

[0025] Please refer to Figure 3 , Figure 3 This is a flowchart illustrating the construction of a lightweight face recognition model in a portable mobile attendance design method based on face recognition, as described in some embodiments of this application. According to an embodiment of the present invention, constructing the lightweight face recognition model includes: K-means clustering algorithm was used to generate anchor boxes for face datasets in attendance scenarios, resulting in multiple anchor boxes adapted to different face scales; A lightweight face recognition model is obtained by combining the YOLO series lightweight algorithm framework with the aforementioned multiple anchor boxes. Based on the deployment requirements of the lightweight face recognition model after transfer learning training, the lightweight face recognition model is deployed to an embedded vision processing chip.

[0026] First, for face datasets collected in attendance scenarios, including those from different angles, under different lighting conditions, and with varying facial expressions, K-means clustering algorithm is used to cluster the facial targets in the dataset, generating multiple anchor boxes adapted to different face scales (including large faces at close range and small faces at long range), providing accurate prior box parameters for subsequent object detection; then, YOLO... Based on a series of lightweight algorithm frameworks, multiple anchor boxes generated are integrated into the network structure design of the algorithm framework, and corresponding network parameters are configured to obtain an initial lightweight face recognition model. Subsequently, the initial lightweight face recognition model is trained by transfer learning, and pre-trained weights are used to accelerate model convergence, improving the model's adaptability and performance in face detection and recognition tasks in attendance scenarios. After the model training is completed, according to the hardware configuration and operating requirements of the portable terminal, combined with the hardware acceleration capabilities and computing power characteristics of the embedded vision processing chip, the lightweight face recognition model trained by transfer learning is deployed to the embedded vision processing chip, realizing efficient adaptation between the model and the terminal hardware. This provides core technical support for the portable terminal to quickly and accurately complete face detection and recognition in various attendance scenarios. Moreover, the entire construction process is adapted to the hardware architecture, deployment mode, and collaborative requirements of the portable terminal and the remote backend platform, ensuring the consistency and integrity of the technical solution.

[0027] According to an embodiment of the present invention, it further includes: Based on the optimization requirements of the lightweight face recognition model, pruning is performed on the model to filter and remove convolutional kernels whose absolute weight values ​​are lower than a preset threshold. Based on the hardware adaptation requirements for deploying the lightweight face recognition model, the number of parameters of the lightweight face recognition model is compressed to a preset range through the pruning process. At the same time, a standard for controlling the loss of recognition accuracy is set to meet the preset first recognition accuracy.

[0028] In this process, based on the optimization requirements for adapting the model to portable terminal hardware, model pruning technology is used to prune the lightweight face recognition model trained through transfer learning. Redundant convolutional kernels with absolute weight values ​​below a preset weight threshold are filtered out and removed. Considering the hardware adaptation requirements of the embedded vision processing chip in the portable terminal, such as computing power and storage resources, the above pruning process compresses the number of parameters of the lightweight face recognition model to a preset range that matches the terminal hardware, ensuring that the model can run efficiently on the terminal. At the same time, a clear standard for controlling the loss of recognition accuracy is set during the pruning optimization process, so that the loss of recognition accuracy after pruning does not exceed a preset first recognition accuracy threshold. This ensures that while the number of parameters is reduced, the model can still meet the accuracy requirements of face detection and recognition in attendance scenarios. This optimization step is coordinated with the model's anchor box design, transfer learning training, and terminal deployment requirements to ensure the consistency and practicality of the overall technical solution.

[0029] According to an embodiment of the present invention, the portable terminal includes: Based on the mobile attendance function requirements of the portable terminal, a hardware architecture combining a main control chip, an embedded vision processing chip, and a wireless communication module is adopted. The main control chip is a microcontroller, the wireless communication module is a wireless network communication module, and a rechargeable battery is configured as the power supply module. Based on the usage requirements of offline scenarios, through the cooperation of the hardware architecture and power supply module, the portable terminal can complete face image acquisition, recognition, and attendance data caching in offline mode.

[0030] Based on the core requirements of mobile and scenario-based attendance, the terminal adopts a hardware architecture that combines a main control chip, an embedded vision processing chip, and a wireless communication module. The main control chip is a microcontroller with high-efficiency control and data processing capabilities, responsible for coordinating the collaborative work of various modules, data transmission scheduling, and model update deployment. The wireless communication module is a wireless network communication module used to realize wireless communication connection between the terminal and the remote backend platform. A rechargeable battery is also configured as a power supply module to provide stable battery life support for the terminal and meet the usage needs in scenarios without a fixed power source. Considering the offline usage needs in attendance scenarios where there may be no network coverage or weak network signal, through the collaborative work of the above hardware architecture and the continuous power supply guarantee of the power supply module, the portable terminal can independently complete face image acquisition, face detection and recognition based on the deployed lightweight face recognition model, and local caching of attendance data in offline mode. After the terminal restores network connection, it can automatically synchronize the cached attendance data to the remote backend platform to ensure the integrity and continuity of attendance data. The terminal design is compatible with model deployment requirements and remote collaboration needs, and fully supports the efficient implementation of the entire mobile attendance process.

[0031] Please refer to Figure 4 , Figure 4 This is a flowchart illustrating the optimization of a lightweight face recognition model in a portable mobile attendance design method based on face recognition, as described in some embodiments of this application. According to embodiments of the present invention, the optimization of the lightweight face recognition model further includes: S401. Dynamically adjust the non-maximum suppression threshold according to the density of people in the attendance scenario; S402. When the attendance scenario is a densely populated scenario, adjust the non-maximum suppression threshold to the first preset range. S403. When the attendance scenario is a sparsely populated scenario, adjust the non-maximum suppression threshold to the second preset range.

[0032] Specifically, a dynamic adjustment mechanism for the non-maximum suppression threshold is established based on the differences in personnel density in attendance scenarios, enabling the model to adapt to attendance scenarios with different personnel distribution densities. When the attendance scenario is identified as a densely populated scenario, the non-maximum suppression threshold is adjusted to a first preset range to adapt to the high overlap of facial targets in dense scenarios and reduce missed detections. When the attendance scenario is identified as a sparsely populated scenario, the non-maximum suppression threshold is adjusted to a second preset range to strictly filter redundant detection boxes and improve recognition accuracy. This adjustment step, in conjunction with other optimization measures such as model pruning optimization and anchor box adaptation, further improves the adaptability and detection and recognition performance of the lightweight face recognition model in diverse attendance scenarios, ensuring that the model can stably output reliable results in different personnel distribution scenarios, and maintaining consistency with the multi-scenario deployment requirements of portable terminals and the overall technical solution.

[0033] According to an embodiment of the present invention, the training of the lightweight face recognition model further includes: Based on the training requirements of the lightweight face recognition model, data augmentation processing is performed on the collected face images; The data enhancement employs image rotation, scaling, random cropping, and color tone adjustment. Based on the functional division requirements of training, verification and testing of the lightweight face recognition model, the face images after the data augmentation process are divided into training set, verification set and test set. The facial images include different angles, lighting conditions, and facial expressions.

[0034] In this process, to meet the training requirements of adapting the model to diverse attendance scenarios and improving generalization ability, data augmentation processing is performed on the collected facial images. This data augmentation employs various methods such as image rotation, scaling, random cropping, and tone adjustment to enrich the diversity of training data and simulate different shooting angles, distances, and lighting changes that may occur in attendance scenarios. The collected facial images themselves already cover different angles, lighting conditions, and facial expressions; after the above data augmentation processing, the coverage of the data samples is further expanded. Subsequently, according to the functional division requirements of model training, validation, and testing, the data-augmented facial images are divided into training, validation, and testing sets according to a preset ratio. The training set is used for model parameter learning and iterative optimization; the validation set is used to monitor overfitting during model training and adjust hyperparameters; and the testing set is used to finally evaluate the model's detection and recognition performance. This data preprocessing step is connected to the model's anchor box generation, transfer learning training, and subsequent optimization steps, providing high-quality and diverse data sources for model training, ensuring the performance stability and scenario adaptability of the trained model, and maintaining consistency with the overall technical solution design logic.

[0035] According to an embodiment of the present invention, it further includes: Based on the multi-scenario deployment requirements of the portable terminal, a function for switching between fixed deployment mode and mobile deployment mode is configured; In the fixed deployment mode, attendance data is uploaded in real time via the network without the need for manual supervision. In the mobile deployment mode, the cached attendance data is automatically synchronized to the remote backend platform based on the network status recovery.

[0036] The portable terminal of this invention is also equipped with a dual deployment mode switching function. Specifically, based on the need for the terminal to adapt to attendance usage in various scenarios such as fixed locations and temporary outdoor venues, a switchable mechanism between fixed deployment mode and mobile deployment mode is designed to meet the attendance application needs in different scenarios. In fixed deployment mode, the portable terminal can be fixedly placed through a preset installation method and upload attendance data in real time relying on a stable network connection, without the need for manual operation. This is suitable for fixed attendance scenarios such as office areas and campus access control. In mobile deployment mode, thanks to the portability of the battery-powered terminal, it can be used flexibly in outdoor areas or areas with weak network signals where there are no fixed installation conditions or wired network coverage. In this case, the terminal first caches the generated attendance data locally. After detecting that the network status has been restored, it automatically triggers the cached data synchronization process and uploads the attendance data to the remote backend platform to ensure that the data is not lost. This dual deployment mode design is compatible with the terminal's hardware architecture, offline working capabilities, and the communication and collaboration requirements of the remote backend platform, further expanding the scenario coverage of mobile attendance, ensuring the smoothness of the attendance process and the integrity of the data in different usage scenarios, and forming an organic unity with the overall technical solution.

[0037] According to an embodiment of the present invention, it further includes: Based on the model version management requirements of the remote backend platform, store the historical versions and the latest version of the face recognition model; Based on the model update requirements, the remote backend platform pushes update instructions to the portable terminal, which then downloads the latest version of the model file via the wireless communication module. The main control chip completes the decompression and deployment update of the model files. After the update is completed, the portable terminal automatically restarts to make the new version effective.

[0038] In this system, based on the remote backend platform's requirements for face recognition model version management, the platform specifically stores historical and latest optimized versions of the model for easy version tracking and rollback, while also monitoring model updates and iterations in real time. When there are update requirements such as model performance optimization or functional expansion, the remote backend platform pushes a model update command to the associated portable terminal. After receiving the command, the portable terminal establishes a data transmission link with the platform through its configured wireless communication module and downloads the latest version of the model file. After the model file is downloaded, the terminal's main control chip coordinates the file decompression, verification, and deployment update operations to ensure the stability and accuracy of the update process. Once the deployment update is complete, the portable terminal automatically triggers a restart process, making the new version of the model and its supporting system software officially effective. This remote upgrade mechanism works in conjunction with the terminal's hardware control logic, wireless communication capabilities, and the platform's data management functions, enabling full-scale model iteration for all terminals without manual on-site operation, significantly reducing maintenance costs while ensuring the overall attendance system's technological advancement and functional stability, consistent with the overall technical solution's design philosophy.

[0039] According to an embodiment of the present invention, it further includes: Based on the cross-system collaboration requirements of the remote backend platform, establish a standardized data interface with the external management system to synchronize basic personnel information, attendance rule configuration, and access management data. After the remote backend platform aggregates the attendance data uploaded by all portable terminals, it performs consistency verification and multi-dimensional statistical analysis in conjunction with the synchronized external system data, generates standardized attendance reports, and synchronizes them to the external management system in real time. Meanwhile, based on the statistical analysis results, we identify the model adaptation shortcomings in high-frequency attendance scenarios, dynamically adjust the model update priority, and push targeted optimized model versions to portable terminals in the corresponding regions.

[0040] Specifically, based on the collaborative integration requirements between the remote backend platform and external management systems (such as enterprise HR systems, campus OA systems, and government attendance systems), a standardized data interface is established. This interface enables bidirectional synchronization of basic personnel information (related to attendance object identification), attendance rule configuration (matching attendance data generation logic), and access control data between the remote backend platform and external management systems, ensuring consistency and relevance of cross-system data. After aggregating attendance data uploaded from all portable terminals (including online real-time uploads and offline cached data synchronized), the remote backend platform, combined with the data synchronized from external management systems, performs consistency checks (including personnel identity verification, attendance time period and rule compliance verification, etc.) and multi-dimensional statistical analysis (including departmental attendance rate, personnel absence frequency, attendance popularity in different regions, etc.). Based on the verification and analysis results, standardized attendance reports are generated, and then transmitted through the aforementioned standardized data interface. Reports are synchronized to the corresponding external management system in real time. Simultaneously, based on multi-dimensional statistical analysis results, the remote backend platform accurately identifies the adaptation shortcomings of lightweight face recognition models in high-frequency attendance scenarios (such as insufficient detection accuracy in specific scenarios, response speed that needs optimization, etc.), dynamically adjusts the priority of model update iterations, and pushes targeted optimized model versions to portable terminals in corresponding areas with shortcomings. This process works in conjunction with the remote backend platform's model version storage management function, the portable terminal's wireless communication reception capability, and the main control chip deployment and update function. Ultimately, it achieves cross-system connectivity of attendance data and precise model iteration through super-system collaborative management. Moreover, the entire cross-system collaborative process does not conflict with the original model building, terminal deployment, attendance execution, remote upgrades, and other core technical aspects, further improving the overall attendance system's functional coverage and ensuring the integrity and synergy of the technical solution.

[0041] Please refer to Figure 5 , Figure 5 This is a structural block diagram of a portable mobile attendance system based on face recognition, provided in an embodiment of this application.

[0042] A second aspect of the present invention also discloses a portable mobile attendance system based on face recognition, comprising: The portable attendance terminal 501 includes a main control module 502, a vision processing module 503, an image acquisition module 504, a wireless communication module 505, a power supply module 506, and a display module 507. The main control module is electrically connected to the vision processing module, image acquisition module, wireless communication module, power supply module, and display module, respectively. The visual processing module deploys the lightweight face recognition model; The power supply module is a lithium battery; The wireless communication module is used to communicate with a remote backend platform; The remote backend platform 511 includes a data management module 512 and an update and maintenance module 513; The data management module is used to store attendance data, facial feature database, and attendance records; The update and maintenance module is used to store the latest version of the face recognition model and system software, push update instructions to portable attendance terminals, and realize data backup and sharing.

[0043] The system also includes a memory and a processor. The memory includes a program for a portable mobile attendance design method based on face recognition. When the program for the portable mobile attendance design method based on face recognition is executed by the processor, it implements the steps of the portable mobile attendance design method based on face recognition as described in any one of the first aspects.

[0044] This invention provides a portable mobile attendance design method and system based on face recognition. It utilizes the construction and optimization of a lightweight face recognition model as its core technology, the flexible deployment of portable terminals as its scenario adaptation basis, and the collaborative management of a remote backend platform as its data and iteration guarantee, forming a closed-loop mobile attendance solution. In the model construction stage, face images with different angles, lighting conditions, and facial expressions are first collected for the attendance scenario. Data enhancement processing, such as image rotation, scaling, random cropping, and tone adjustment, enriches the sample diversity, and the images are divided into training, validation, and test sets. Then, based on the YOLO series of lightweight algorithm frameworks, multiple anchor boxes adapted to different face scales are generated using the K-means clustering algorithm and integrated into the network structure to form the initial model. Subsequently, transfer learning is used to accelerate model training convergence, and redundant convolutional kernels are removed through pruning to compress the number of parameters. Simultaneously, a recognition accuracy loss control standard is set to ensure that the model meets accuracy requirements while remaining lightweight. Furthermore, the non-maximum suppression threshold can be dynamically adjusted according to the density of people in the attendance scenario, further improving the model's detection and recognition performance in diverse scenarios. In the terminal deployment phase, the portable terminal adopts a hardware architecture consisting of a main control chip, an embedded vision processing chip, and a wireless communication module. The main control chip uses a microcontroller for overall scheduling, the wireless communication module ensures network connectivity, and a rechargeable battery enables battery life in scenarios without a fixed power source. The terminal supports switching between fixed and mobile deployment modes. In fixed deployment mode, no manual intervention is required, and attendance data is uploaded in real time. In mobile deployment mode, it can operate in areas without wired network coverage or with weak signals. In offline mode, it independently completes face image acquisition, recognition, and attendance data caching, and automatically synchronizes after the network is restored. In the platform collaboration phase, the portable terminal establishes a connection with the remote backend platform through the wireless communication module, completes attendance data sharing through the transmission path, and receives model update instructions through the receiving channel. The main control chip completes the decompression, deployment update, and automatic restart of the model files. The remote backend platform stores attendance data, a face feature database, and attendance records through a data management module, and stores historical and latest versions of the model through an update and maintenance module, enabling data backup and remote upgrades. Furthermore, to expand the system's application boundaries, the remote backend platform establishes standardized data interfaces with external management systems. This synchronizes basic personnel information, attendance rule configurations, and access control data. The platform performs consistency checks and multi-dimensional statistical analysis on the aggregated attendance data, generating standardized reports that are synchronized to external systems. Simultaneously, based on the analysis results, it identifies model adaptation shortcomings, dynamically adjusts update priorities, and pushes targeted optimized versions, achieving cross-system data integration and precise model iteration. This entire technical solution optimizes the entire chain from models, terminals, and platforms to cross-system collaboration, completely overcoming the pain points of existing attendance devices, such as limited scenario coverage, inconvenient upgrades and maintenance, and data silos. It adapts to various attendance needs in offices, campuses, and temporary outdoor venues, offering significant advantages such as flexible deployment, efficient operation, convenient maintenance, and strong scalability.

[0045] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0046] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0047] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0048] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory, random access memory, magnetic disks, or optical disks.

[0049] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A portable mobile attendance design method based on face recognition, characterized in that, Includes the following steps: A lightweight face recognition model is constructed, and the model is trained and optimized based on the object detection algorithm framework to obtain an optimized face recognition model adapted to portable terminal hardware. The optimized face recognition model can be deployed on an embedded portable terminal, supporting battery power and wireless communication, and can be deployed in mobile locations without fixed installation conditions or wired network coverage. The system acquires facial images of the attendance subjects through the image acquisition module, and uses the deployed optimized facial recognition model to complete facial detection and recognition, generating attendance data. Based on the communication connection established between the portable terminal and the remote backend platform, the transmission path and receiving channel are obtained; Attendance data sharing is completed based on the transmission path, and model update instructions are received through the receiving channel to complete remote upgrades and data backups of the face recognition model and system software.

2. The portable mobile attendance design method based on face recognition according to claim 1, characterized in that, The construction of a lightweight face recognition model involves training and optimizing the model based on an object detection algorithm framework to obtain an optimized face recognition model adapted to portable terminal hardware, including: K-means clustering algorithm was used to generate anchor boxes for face datasets in attendance scenarios, resulting in multiple anchor boxes adapted to different face scales; A lightweight face recognition model is obtained by combining the YOLO series lightweight algorithm framework with the aforementioned multiple anchor boxes. Based on the deployment requirements of the lightweight face recognition model after transfer learning training, the lightweight face recognition model is deployed to an embedded vision processing chip.

3. The portable mobile attendance design method based on face recognition according to claim 2, characterized in that, Also includes: Based on the optimization requirements of the lightweight face recognition model, pruning is performed on the model to filter and remove convolutional kernels whose absolute weight values ​​are lower than a preset threshold. Based on the hardware adaptation requirements for deploying the lightweight face recognition model, the number of parameters of the lightweight face recognition model is compressed to a preset range through the pruning process. At the same time, a standard for controlling the loss of recognition accuracy is set to meet the preset first recognition accuracy.

4. The portable mobile attendance design method based on face recognition according to claim 1, characterized in that, The portable terminal includes: Based on the mobile attendance function requirements of the portable terminal, a hardware architecture combining a main control chip, an embedded vision processing chip, and a wireless communication module is adopted. The main control chip is a microcontroller, the wireless communication module is a wireless network communication module, and a rechargeable battery is configured as the power supply module. Based on the usage requirements of offline scenarios, through the cooperation of the hardware architecture and power supply module, the portable terminal can complete face image acquisition, recognition, and attendance data caching in offline mode.

5. The portable mobile attendance design method based on face recognition according to claim 1, characterized in that, The optimization of the lightweight face recognition model also includes: The non-maximum suppression threshold is dynamically adjusted based on the density of people in the attendance scenario. When the attendance scenario is a densely populated scenario, the non-maximum suppression threshold is adjusted to the first preset range; When the attendance scenario is a sparsely populated scenario, the non-maximum suppression threshold is adjusted to the second preset range.

6. The portable mobile attendance design method based on face recognition according to claim 1, characterized in that, Training the lightweight face recognition model also includes: Based on the training requirements of the lightweight face recognition model, data augmentation processing is performed on the collected face images; The data enhancement employs image rotation, scaling, random cropping, and color tone adjustment. Based on the functional division requirements of training, verification and testing of the lightweight face recognition model, the face images after the data augmentation process are divided into training set, verification set and test set. The facial images include different angles, lighting conditions, and facial expressions.

7. The portable mobile attendance design method based on face recognition according to claim 1, characterized in that, Also includes: Based on the multi-scenario deployment requirements of the portable terminal, a function for switching between fixed deployment mode and mobile deployment mode is configured; In the fixed deployment mode, attendance data is uploaded in real time via the network without the need for manual supervision. In the mobile deployment mode, the cached attendance data is automatically synchronized to the remote backend platform based on the network status recovery.

8. The portable mobile attendance design method based on face recognition according to claim 1, characterized in that, Also includes: Based on the model version management requirements of the remote backend platform, store the historical versions and the latest version of the face recognition model; Based on the model update requirements, the remote backend platform pushes update instructions to the portable terminal, which then downloads the latest version of the model file via the wireless communication module. The main control chip completes the decompression and deployment update of the model files. After the update is completed, the portable terminal automatically restarts to make the new version effective.

9. A portable mobile attendance system based on facial recognition, characterized in that, include: A portable time and attendance terminal includes a main control module, a vision processing module, an image acquisition module, a wireless communication module, a power supply module, and a display module; The main control module is electrically connected to the vision processing module, image acquisition module, wireless communication module, power supply module, and display module, respectively. The visual processing module deploys the lightweight face recognition model; The power supply module is a lithium battery; The wireless communication module is used to communicate with a remote backend platform; The remote backend platform includes a data management module and an update and maintenance module; The data management module is used to store attendance data, facial feature database, and attendance records; The update and maintenance module is used to store the latest version of the face recognition model and system software, push update instructions to portable attendance terminals, and realize data backup and sharing.

10. A portable mobile attendance system based on facial recognition, characterized in that, The system also includes a memory and a processor. The memory contains a program for a portable mobile attendance design method based on face recognition. When the program for the portable mobile attendance design method based on face recognition is executed by the processor, it performs the following steps: A lightweight face recognition model is constructed, and the model is trained and optimized based on the object detection algorithm framework to obtain an optimized face recognition model adapted to portable terminal hardware. The optimized face recognition model can be deployed on an embedded portable terminal, supporting battery power and wireless communication, and can be deployed in mobile locations without fixed installation conditions or wired network coverage. The system acquires facial images of the attendance subjects through the image acquisition module, and uses the deployed optimized facial recognition model to complete facial detection and recognition, generating attendance data. Based on the communication connection established between the portable terminal and the remote backend platform, the transmission path and receiving channel are obtained; Attendance data sharing is completed based on the transmission path, and model update instructions are received through the receiving channel to complete remote upgrades and data backups of the face recognition model and system software.

Images