Deep learning driven mutual certification type non-inductive attendance record generation method and system

Abstract

This application provides a deep learning-driven, interconnected, seamless attendance record generation method and system, relating to the field of deep learning technology. The method includes: acquiring multimodal attendance perception data; performing multimodal identity verification analysis based on the multimodal attendance perception data, configuring the degrees of freedom for identity verification, and determining the single-modal confidence level corresponding to each modality; configuring attendance optimization instructions to drive conflict resolution based on the single-modal confidence level, adjusting the underlying configuration parameters of the dual-channel identity verification model, and determining reliable attendance information; based on the reliable attendance information, determining valid attendance events in conjunction with the current attendance scenario type, and generating seamless attendance records using the valid attendance events. This application addresses the technical problem in existing technologies where attendance systems cannot accurately generate employee attendance records under modal anomalies or signal interference, further affecting the accuracy of attendance management, thus achieving the technical effect of accurately generating employee attendance records.

Description

Technical Field

[0001] This application relates to the field of deep learning technology, and in particular to a deep learning-driven method and system for generating interconnected, seamless attendance records. Background Technology

[0002] With the rapid development of enterprise digital management and intelligent office systems, employee attendance management is facing increasingly higher demands for accuracy and real-time performance. Modern enterprises commonly employ various attendance methods in their daily operations, including access control card swiping, fingerprint recognition, facial recognition, and mobile device check-in, aiming to achieve comprehensive monitoring of employee attendance status. However, existing attendance technologies still face the problem of severe reliance on single-modal data in practical applications. The lack of an effective integration mechanism between different attendance methods leads to errors or delays in overall attendance judgment when a particular modality malfunctions or fails.

[0003] Currently, while existing multimodal attendance systems can collect data from multiple sources simultaneously, they typically employ simple linear weighting or rule-based matching methods to fuse attendance information from different modalities. They lack systematic analysis and adaptive handling mechanisms for data consistency, time synchronization, and intermodal conflicts. Therefore, when modal conflicts, data delays, or signal interference occur, the attendance system struggles to accurately determine employee identity and attendance status, leading to omissions, errors, or frequent manual reviews. This problem is particularly pronounced in multi-terminal, multi-floor, or high-traffic scenarios, significantly increasing attendance anomaly rates and manual correction costs, while also reducing the system's real-time performance and scalability.

[0004] In summary, existing technologies lack technical solutions for dynamic consistency analysis and conflict adaptive handling of multimodal attendance information. This results in attendance systems being unable to accurately generate employee attendance records under conditions of modal anomalies or signal interference, further affecting the accuracy, real-time performance, and overall reliability of attendance management, and failing to meet the technical requirements for high-precision, multi-scenario attendance. Summary of the Invention

[0005] The purpose of this application is to provide a deep learning-driven, interconnected, and seamless attendance record generation method and system to solve the technical problem that existing technologies lack technical solutions for dynamic consistency analysis and conflict adaptive handling of multimodal attendance information. This results in attendance systems being unable to accurately generate employee attendance records under conditions of modal anomalies or signal interference, further affecting the accuracy, real-time performance, and overall reliability of attendance management, and failing to meet the technical requirements of high-precision, multi-scenario attendance.

[0006] In view of the above problems, this application provides a deep learning-driven method and system for generating seamless attendance records with mutual verification.

[0007] Firstly, this application provides a deep learning-driven method for generating seamless attendance records through mutual verification, implemented through a deep learning-driven system for generating seamless attendance records through mutual verification. The method includes: acquiring multimodal attendance perception data, including facial capture images, WiFi probe signal strength, Bluetooth beacon connection records, and access control card swipe logs; performing multimodal identity verification analysis based on the multimodal attendance perception data, configuring the degree of freedom for identity verification, and determining the single-modal confidence level corresponding to each modality by combining the employee's historical attendance profile database and attendance rules and strategies; configuring attendance optimization instructions to drive conflict resolution based on the degree of freedom for identity verification and the single-modal confidence levels corresponding to each modality, adjusting the underlying configuration parameters of the dual-channel identity verification model using the attendance optimization instructions, and determining reliable attendance information; and determining valid attendance events based on the reliable attendance information and the current attendance scenario type, and generating structured seamless attendance records using the employee identifier, attendance time, and attendance location from the valid attendance events.

[0008] Preferably, the deep learning-driven mutual authentication seamless attendance record generation method further includes: collecting attendance record call data and corresponding review and correction snapshots within the historical attendance period; extracting spatiotemporal consistency features and identity verification latency indicators corresponding to parallel processing tasks in the parallel inference path based on the attendance record call data and corresponding review and correction snapshots; and optimizing the decision nodes of the dual-channel identity mutual authentication model according to the spatiotemporal consistency features and identity verification latency indicators.

[0009] Preferably, the deep learning-driven mutual authentication seamless attendance record generation method further includes: obtaining the spatiotemporal consistency deviation during the synchronous parsing of multimodal attendance perception data, using the task backlog depth and priority inversion number of the identity verification queue as auxiliary indicators to measure attendance response efficiency; and adaptively adjusting the parallel inference scheduling strategy of the dual-channel identity mutual authentication model through the spatiotemporal consistency deviation, task backlog depth and priority inversion number.

[0010] Preferably, the deep learning-driven mutual authentication seamless attendance record generation method further includes: dynamically adjusting the feature encoding dimension of the explicit channel and the behavior trajectory mining depth of the potential channel based on an attendance optimization instruction that includes an explicit identity verification scheme, a potential behavior trajectory mining strategy, and attendance evidence complexity adaptation parameters; the dual-channel identity mutual authentication model includes an explicit channel and a potential channel.

[0011] Preferably, the deep learning-driven mutual verification seamless attendance record generation method further includes: optimizing the attendance evidence complexity layering threshold based on the seamless interference index under normal user activity; dynamically suppressing the feature encoding dimension of high interference modalities and reducing the mining time window of potential behavioral trajectories according to the attendance evidence complexity layering threshold.

[0012] Preferably, the deep learning-driven mutual verification seamless attendance record generation method further includes: recording model state snapshots and utility difference data, locating key configuration items that cause the attendance record review correction rate to increase; generating compensation adjustment instructions based on the key configuration items; and using the compensation adjustment instructions to update nodes and correct weights in the attendance evidence chain dependency graph.

[0013] Preferably, the deep learning-driven mutual verification seamless attendance record generation method further includes: binding the attendance evidence chain dependency graph with the temporal correspondence of multimodal attendance perception data, the constraint conditions of the identity mutual verification degree of freedom, and the decision priority of the single-modal confidence degree corresponding to each modality.

[0014] Preferably, the deep learning-driven mutual authentication seamless attendance record generation method further includes: analyzing the modal differences of each sensing terminal in a multi-terminal collaborative attendance scenario; dynamically allocating authentication tasks and synchronously optimizing the attendance parameter configuration between each sensing terminal based on the modal differences of each sensing terminal; and enabling an authentication load migration mechanism when the attendance response time difference between each sensing terminal exceeds the time difference threshold.

[0015] Preferably, the deep learning-driven mutual verification seamless attendance record generation method further includes: configuring an attendance credibility influence factor map by referring to the attendance evidence chain dependency graph; when receiving a manual correction request, predicting the attendance utility change trend based on the attendance credibility influence factor map, issuing an optimization suggestion alarm, and identifying redundant evidence items through a map backtracking mechanism; based on the redundant evidence items, mining potential interference factors in conjunction with the attendance credibility influence factor map, the potential interference factors include WiFi signal drift, sudden changes in illumination, and Bluetooth beacon obstruction; and iteratively correcting the attendance optimization instructions based on the potential interference factors until the attendance record utility verification passes.

[0016] Secondly, this application also provides a deep learning-driven, interconnected, contactless attendance record generation system for executing the deep learning-driven, interconnected, contactless attendance record generation method described in the first aspect, including: a multimodal attendance perception data acquisition module for acquiring multimodal attendance perception data including face capture images, WiFi probe signal strength, Bluetooth beacon connection records, and access control card swipe logs; and a single-modal confidence determination module for performing multimodal identity verification analysis based on the multimodal attendance perception data, configuring identity verification degrees of freedom, and combining employee historical attendance profiles and attendance rules. The strategy determines the single-modal confidence level corresponding to each modality; the trusted attendance information determination module is used to configure attendance optimization instructions that drive conflict resolution based on the identity mutual authentication degree of freedom and the single-modal confidence level corresponding to each modality, and use the attendance optimization instructions to adjust the underlying configuration parameters of the dual-channel identity mutual authentication model to determine trusted attendance information; the seamless attendance record generation module is used to determine valid attendance events based on the trusted attendance information and the current attendance scenario type, and generate structured seamless attendance records with the employee identifier, attendance time, and attendance location in the valid attendance events.

[0017] The technical solution provided in this application has at least the following technical effects or advantages: by achieving the technical goal of generating seamless attendance records based on dynamic identity verification and conflict adaptive processing of multimodal perception data, it can still accurately and in real time generate employee attendance records under the condition of modal abnormality or environmental interference, and improve the stability of attendance data, reduce the need for manual correction, and optimize the efficiency of attendance management and the reliability of the system.

[0018] The above description is merely an overview of the technical solution of this application. To enable a clearer understanding of the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the deep learning-driven, mutual-verification, seamless attendance record generation method proposed in this application.

[0021] Figure 2 This is a schematic diagram of the structure of the deep learning-driven mutual verification-based contactless attendance record generation system of this application.

[0022] Figure labeling: Module 1 for multimodal attendance perception data acquisition, Module 2 for single-modal confidence determination, Module 3 for reliable attendance information determination, and Module 4 for seamless attendance record generation. Detailed Implementation

[0023] This application provides a deep learning-driven, interconnected, contactless attendance record generation method and system. It addresses the technical problem in existing technologies where the lack of dynamic consistency analysis and conflict adaptive handling for multimodal attendance information leads to inaccurate employee attendance record generation under modal anomalies or signal interference, further impacting the accuracy, real-time performance, and overall reliability of attendance management, and failing to meet the demands for high-precision, multi-scenario attendance. The application achieves the technical goal of contactless attendance record generation based on dynamic identity verification and conflict adaptive handling of multimodal perception data. This results in accurate and real-time generation of employee attendance records even under modal anomalies or environmental interference, while improving attendance data stability, reducing the need for manual corrections, and optimizing attendance management efficiency and system reliability.

[0024] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0025] Example 1, please refer to the appendix. Figure 1 This application provides a deep learning-driven method for generating seamless attendance records based on mutual verification, applicable to a deep learning-driven system for generating seamless attendance records based on mutual verification. The method includes the following steps: Acquire multimodal attendance perception data, including face capture images, WiFi probe signal strength, Bluetooth beacon connection records, and access control card swipe logs.

[0026] Specifically, acquiring multimodal attendance perception data, including facial capture images, WiFi probe signal strength, Bluetooth beacon connection records, and access control card swipe logs, refers to the synchronous collection and structured storage of multi-source behavioral data generated by personnel within a designated area through various types of sensing terminals and data acquisition modules deployed in the attendance scenario. Facial capture images are image data containing facial feature information automatically captured by video capture devices when personnel enter or pass through the monitored area, used to extract facial feature vectors and perform identity recognition; WiFi probe signal strength refers to the received signal strength indication value obtained by listening to the wireless signal emitted by the mobile terminal through a wireless probe device, used to characterize the relative distribution characteristics of the terminals carried by personnel in space; Bluetooth beacon connection records refer to the connection timestamp, signal strength, and device identification information formed between the mobile terminal and the preset Bluetooth beacon during broadcasting and scanning, used to reflect the personnel's dwell behavior at specific spatial nodes; access control card swipe logs refer to records such as card swipe time, device number, and user identification generated when personnel verify their identity through access control devices, used to provide explicit identity verification evidence. Multimodal attendance perception data refers to a collection of data from different sources, with different forms of expression and time-series attributes. After being aligned with a unified time benchmark, it forms a basic dataset that can be used for subsequent identity verification analysis.

[0027] Based on the multimodal attendance perception data, multimodal identity verification analysis is performed, the degree of freedom for identity verification is configured, and the confidence level of each modality is determined by combining the employee historical attendance profile database and attendance rules and strategies.

[0028] Specifically, multimodal identity verification analysis based on multimodal attendance perception data refers to the process of feature extraction, vector representation, and association matching of data from different perception sources within a unified time axis and spatial identification framework. By constructing a cross-modal feature alignment mechanism, consistency verification and conflict identification are performed on visual features, wireless signal features, and access control behavior features within the same time window. Multimodal identity verification analysis includes intermodal correlation calculation, behavioral continuity testing, and identity matching probability assessment, used to determine whether the personnel identities reflected by different modalities point to the same entity, thereby forming a mutual verification relationship matrix.

[0029] Furthermore, configuring the degree of freedom for identity verification refers to setting adjustable parameters for the mutual verification strength between different modalities. By quantitatively controlling the tightness of the constraints between modalities, the dynamic adjustment of the conflict tolerance range and the mutual verification weight ratio can be achieved. The degree of freedom for identity verification can be set according to the scenario type, modal integrity level, and historical stability indicators, and is used to switch between strong consistency verification mode and relaxed fault-tolerant verification mode, thereby adapting to attendance environments of varying complexity.

[0030] Simultaneously, combining the employee historical attendance profile database involves retrieving long-term accumulated data on employee attendance time distribution characteristics, common entrance / exit trajectory patterns, device connection stability indicators, and behavioral pattern statistics. This data is used to construct individualized behavioral models, providing prior reference for current identity determination. The employee historical attendance profile database includes structured attendance record data, behavioral frequency statistical vectors, and anomaly correction historical tags, reflecting employees' typical attendance characteristics. Furthermore, combining attendance rule strategies involves logically encoding legal attendance time periods, allowable error ranges, cross-regional access rules, and multi-device concurrent verification rules based on preset attendance management systems and business constraints. This transforms institutional constraints into a computable set of rules, constraining the decision boundaries in the identity determination process. Determining the single-modal confidence level for each modality involves calculating the identity matching probability value for each type of perception modality. By comprehensively considering modality quality indicators, historical stability coefficients, signal integrity, and rule matching degree, a numerical confidence score is generated. Unimodal confidence is used to represent the degree of credibility of a single data source in supporting a specific identity conclusion within the current time window, and to provide a weighting basis for subsequent multimodal fusion and conflict resolution.

[0031] Based on the degrees of freedom of identity verification and the single-modal confidence level corresponding to each modality, configure attendance optimization instructions to drive conflict resolution, and use the attendance optimization instructions to adjust the underlying configuration parameters of the dual-channel identity verification model to determine reliable attendance information.

[0032] Specifically, the attendance optimization instructions that drive conflict resolution based on the degrees of freedom for identity verification and the confidence scores of each modality refer to the quantitative assessment of differences in identity judgments between different modalities after completing multimodal identity consistency analysis and obtaining confidence scores for each modality. This is done by combining the modal constraint strength reflected by the degrees of freedom for identity verification and generating control instructions to adjust the model's decision path. The degrees of freedom for identity verification describe the allowable range of differences between modalities and the adjustable range of verification weights; the confidence score of a single modality represents the probability level of a single modality supporting a certain identity conclusion. The attendance optimization instructions that drive conflict resolution are a set of parameters including weight redistribution parameters, threshold adjustment coefficients, and priority correction factors. These parameters guide the decision-making process when confidence scores are inconsistent between modalities, thereby reducing the probability of misjudgment and improving overall consistency.

[0033] Furthermore, adjusting the underlying configuration parameters of the dual-channel identity verification model using attendance optimization instructions refers to mapping control instructions to the model's internal structure, dynamically modifying the feature extraction dimensions, fusion weight ratios, inference path selection thresholds, and conflict determination boundaries for both the explicit and implicit channels. The dual-channel identity verification model includes an explicit channel that takes explicit identity verification data as input, and a implicit channel that takes behavioral trajectories and latent features as input. The underlying configuration parameters include feature encoding layer parameters, fusion layer weight matrix, decision node thresholds, and parallel inference scheduling strategies. Through parameter updates, the model is made more consistent with data distribution characteristics and institutional constraints in the current scenario.

[0034] After adjusting the model parameters, determining reliable attendance information involves re-fusing and calculating multimodal data under the optimized model structure to output identity verification results that meet pre-set confidence thresholds and rule constraints. Reliable attendance information includes the final employee identifier, corresponding timestamp, and spatial location identifier formed after conflict resolution, supported by a traceable chain of evidence. The degree of reliability is verified through a comprehensive confidence score and rule consistency test, ensuring the stability and interpretability of the generated results.

[0035] Based on the reliable attendance information, valid attendance events are determined in conjunction with the current attendance scenario type, and structured, seamless attendance records are generated using the employee identifier, attendance time, and attendance location in the valid attendance events.

[0036] Specifically, determining valid attendance events based on trusted attendance information and the current attendance scenario type involves matching and analyzing the identity verification results (after multimodal identity verification and conflict resolution) with actual business scenario parameters to identify event types that conform to the attendance rules. Trusted attendance information includes employee identification, timestamp data, and spatial location identifiers that have passed confidence threshold verification, along with supporting evidence chains. The current attendance scenario type refers to the business environment category defined based on location function and time period attributes, such as attendance check-in, attendance check-out, temporary absence registration, or cross-regional transfer. Through scenario rule matching algorithms, the trusted identity judgment results are logically compared with the scenario judgment conditions to filter data combinations that meet the event establishment conditions, thereby determining valid attendance events. A valid attendance event is an attendance behavior record unit that, within the rule constraints, possesses complete identity information, a legal time interval, and valid spatial location.

[0037] Furthermore, generating structured, seamless attendance records based on employee identification, attendance time, and attendance location from valid attendance events refers to mapping and formatting the core elements of valid attendance events according to a preset data model to form a standardized storage structure. Employee identification is a unique coded information identifying an individual; attendance time is a timestamp of the event occurrence calibrated using a unified time benchmark; and attendance location is a location identifier obtained based on spatial coordinate mapping or device number parsing. Structured, seamless attendance records are data entries conforming to database specifications, automatically generated and stored without manual intervention. These entries include field definitions, data type constraints, and index identifiers, supporting subsequent statistical analysis, report generation, and management review.

[0038] Furthermore, this application also includes: collecting attendance record retrieval data and corresponding review and correction snapshots within historical attendance periods; extracting spatiotemporal consistency features and identity verification latency indicators corresponding to parallel processing tasks in the parallel inference path based on the attendance record retrieval data and corresponding review and correction snapshots; and optimizing the decision nodes of the dual-channel identity mutual authentication model according to the spatiotemporal consistency features and identity verification latency indicators.

[0039] Furthermore, this application also includes: obtaining the spatiotemporal consistency deviation during the synchronous parsing of multimodal attendance perception data, using the task backlog depth and priority inversion number of the identity verification queue as auxiliary indicators to measure attendance response efficiency; and adaptively adjusting the parallel inference scheduling strategy of the dual-channel identity mutual authentication model through the spatiotemporal consistency deviation, task backlog depth and priority inversion number.

[0040] Specifically, collecting attendance record retrieval data and corresponding review and correction snapshots within historical attendance periods refers to summarizing access behavior data for queried, modified, or reviewed attendance records over multiple past attendance statistical periods, and simultaneously acquiring version change records generated during manual review or rule review processes. Historical attendance periods refer to statistical time intervals divided by natural days, weeks, or months; attendance record retrieval data includes access timestamps, access subject identifiers, call frequency, and operation type, reflecting the usage of attendance results in business processes; review and correction snapshots are data copies formed when manual intervention or rule correction occurs, preserving versions of the original attendance records and the corrected results, recording the differences before and after the correction. By structuring and aligning the above data with time series data, a foundational sample is provided for model performance evaluation and path analysis.

[0041] Furthermore, based on attendance record data and corresponding review and correction snapshots, the spatiotemporal consistency features and identity verification latency indicators corresponding to parallel processing tasks in the parallel inference path are extracted. This refers to the trajectory backtracking analysis of task branches formed by the dual-channel identity verification model during the parallel inference stage, combined with historical operation logs. The parallel inference path refers to the multi-task processing flow formed when feature calculation and identity judgment are performed simultaneously in the explicit and potential channels; parallel processing tasks include feature extraction tasks, confidence calculation tasks, and conflict determination tasks. Spatiotemporal consistency features refer to the deviation statistics of the degree of synchronization of different modal data on the time axis and the degree of matching of spatial positioning, used to reflect the degree of coordination of identity judgment results in the time and space dimensions; identity verification latency indicators refer to the processing time consumed from data input to identity judgment result output, used to characterize the model response efficiency. By performing correlation analysis on historical correction records and parallel task logs, potential performance bottlenecks and consistency anomalies that cause correction behavior are identified.

[0042] Subsequently, the spatiotemporal consistency deviation during the synchronous parsing of multimodal attendance perception data is obtained. The task backlog depth and priority inversion count of the identity verification queue are used as auxiliary indicators to measure attendance response efficiency. Spatiotemporal consistency deviation refers to the quantitative calculation of the differences in timestamp alignment and spatial location matching between different modalities during the synchronous feature parsing and identity determination of multimodal data entering the model, and the calculation results are recorded as spatiotemporal consistency deviation parameters. Synchronous parsing of multimodal attendance perception data refers to the parallel processing and fusion calculation of visual data, wireless signal data, and access control log data within a unified time window; spatiotemporal consistency deviation is used to represent the degree of difference between the time series location and spatial positioning reflected by different modalities. Meanwhile, the identity verification queue refers to the scheduling structure within the system used to queue identity determination tasks; task backlog depth represents the number of tasks that have not been processed within a unit of time; and priority inversion count represents the frequency at which low-priority tasks occupy resources during execution, causing delays in the execution of high-priority tasks. By using task backlog depth and priority inversion count as auxiliary indicators, attendance response performance can be evaluated from the perspective of system operating efficiency.

[0043] Furthermore, by adaptively adjusting the parallel inference scheduling strategy of the dual-channel identity verification model based on spatiotemporal consistency deviation, task backlog depth, and priority inversion count, this refers to dynamically optimizing the task allocation order, resource consumption ratio, and thread scheduling priority of the explicit and potential channels during the parallel execution phase, according to the real-time changes of the aforementioned indicators. The parallel inference scheduling strategy includes control parameters such as task distribution rules, computational resource allocation ratio, and result fusion waiting threshold. Adaptive adjustment means automatically modifying scheduling parameters based on changes in indicators; for example, increasing the priority of data alignment processing when spatiotemporal consistency deviation increases, compressing the computational complexity of low-confidence paths when task backlog depth increases, and resetting the task priority queue structure when the number of priority inversions increases. Through dynamic optimization of the scheduling strategy, a balance between accuracy and response efficiency is achieved in the identity verification process.

[0044] Furthermore, this application also includes: dynamically adjusting the feature encoding dimension of the explicit channel and the behavior trajectory mining depth of the potential channel based on attendance optimization instructions that include explicit identity verification schemes, potential behavior trajectory mining strategies and attendance evidence complexity adaptation parameters; the dual-channel identity mutual verification model includes explicit channels and potential channels.

[0045] Specifically, based on attendance optimization instructions that include explicit identity verification schemes, potential behavior trajectory mining strategies, and attendance evidence complexity adaptation parameters, dynamically adjusting the feature encoding dimension of explicit channels and the behavior trajectory mining depth of potential channels refers to mapping the various adjustment elements contained in the instructions to the internal structure of the dual-channel identity mutual authentication model for parameter updates after generating control instructions for conflict resolution and model optimization. Explicit identity verification schemes refer to a set of verification rules centered on direct identity recognition data, including facial feature comparison methods, access control identity matching rules, and device identification verification logic; potential behavior trajectory mining strategies refer to a set of methods for modeling and analyzing the spatial movement paths, device connection change sequences, and behavioral continuity patterns of personnel within a certain time window; attendance evidence complexity adaptation parameters are quantitative indicators used to measure the evidence chain length, modality quantity, and conflict frequency in the current attendance scenario, and are used to determine the allocation ratio of model computing resources. The feature encoding dimension of explicit channels refers to the length of the feature vector or the dimension of the representation space generated during the feature extraction stage. Increasing or compressing the dimension adjusts the recognition accuracy and computational load. The behavioral trajectory mining depth of latent channels refers to the range of historical data covered or the number of recursive levels during time series analysis, used to control the time span and granularity of trajectory inference. By dynamically adjusting these parameters, adaptive optimization of the model can be achieved under different scenario complexities.

[0046] Furthermore, the dual-channel identity verification model includes an explicit channel and a potential channel, meaning the model structure consists of two inference paths with clearly defined functions and operating in parallel. The explicit channel processes directly identifiable identity information, such as facial feature vector matching results compared with access control card swipe records, achieving rapid judgment through explicit identity credentials. The potential channel processes indirect behavioral data, inferring identity attribution by analyzing wireless signal change trends, dwell time distribution, and spatial movement trajectories. The two channels are fused and calculated at the decision layer, and the final identity judgment result is generated based on confidence weights and rule constraints. By constructing an explicit and potential dual-path structure, collaborative verification of direct evidence and behavioral evidence is achieved.

[0047] Furthermore, this application also includes: optimizing the attendance evidence complexity stratification threshold based on the non-obtrusive interference index under normal user activity; dynamically suppressing the feature encoding dimension of high-interference modalities and reducing the mining time window of potential behavioral trajectories based on the attendance evidence complexity stratification threshold.

[0048] Specifically, based on the imperceptible interference index under normal user activity, the stratified threshold for attendance evidence complexity is optimized. This refers to the intensity level of environmental disturbances and random fluctuations experienced by various sensory modalities when statisticians are in a normal attendance state during long-term operation, and the statistical results are quantified as the imperceptible interference index. The imperceptible interference index describes the amplitude of data fluctuations caused by factors such as changes in lighting, wireless signal drift, and changes in personnel density without human intervention. The stratified threshold for attendance evidence complexity refers to the boundary parameters for classifying indicators such as the number of modalities, number of conflicts, confidence dispersion, and time span in the evidence chain, used to distinguish between low-complexity, medium-complexity, and high-complexity scenarios. By using the imperceptible interference index as a reference benchmark, the stratified threshold is recalibrated to make the complexity classification more consistent with the actual operating environment, thereby avoiding excessive consumption of computing resources in low-interference scenarios or overly lenient threshold settings in high-interference scenarios.

[0049] Furthermore, based on the hierarchical threshold of attendance evidence complexity, the feature encoding dimension of high-interference modalities is dynamically suppressed, and the time window for mining potential behavioral trajectories is shortened. This means that when the current scene is determined to be at a high interference level, the feature representation dimension or feature extraction level is reduced for modalities that are more affected by the environment, thereby weakening the impact of noise features on the decision-making results. High-interference modalities refer to perception sources whose signal fluctuation amplitude exceeds the normal range within the current time period, such as wireless modalities with frequent signal strength jumps or visual modalities with degraded image quality. Feature encoding dimension refers to the dimension of the vector space used to express modal features; compressing the dimension can reduce the proportion of noise components. At the same time, the time window for mining potential behavioral trajectories refers to the time span used to analyze the historical behavior sequences of personnel; shortening the time window can reduce the interference of abnormal historical data on the current judgment. Through the above dual adjustment, the model can operate stably in high-interference environments.

[0050] Furthermore, this application also includes: recording model state snapshots and utility difference data to locate key configuration items that cause the attendance record review correction rate to increase; generating compensation adjustment instructions based on the key configuration items; and using the compensation adjustment instructions to update nodes and correct weights in the attendance evidence chain dependency graph.

[0051] Specifically, recording model state snapshots and utility difference data to identify key configuration items leading to an increase in the attendance record review correction rate involves periodically archiving the parameter configurations, structural weights, and scheduling strategies of the dual-channel identity verification model during system operation, and comparing and analyzing the utility performance of attendance results at each time point. Model state snapshots include feature encoding dimension parameters, fusion weight matrix, decision threshold parameters, and scheduling queue priority configurations; utility difference data refers to the statistical differences between changes in attendance record pass rates, manual correction ratios, conflict frequency, and confidence distribution over different time periods. By correlating the model state with the trend of review correction rate changes, the set of configuration parameters that significantly impact the increase in the correction rate can be identified, thereby determining the key configuration items. Key configuration items are parameter variables whose impact on the stability of attendance results exceeds a preset threshold during parameter perturbation experiments or historical comparisons.

[0052] Furthermore, generating compensation adjustment instructions based on key configuration items involves constructing a set of corrective control parameters for the identified key parameter variables, used to directionally correct the model's internal decision-making process. These compensation adjustment instructions include threshold callback coefficients, weight redistribution ratios, path priority correction parameters, and complexity suppression factors, used to restore the model's decision balance deviated from its optimal state under abnormal conditions. The generation process can use historically optimal state parameters as a reference, determining the correction magnitude through difference calculation and gradient estimation, thereby allowing the model's operating state to return to a low correction rate range.

[0053] Subsequently, the compensation adjustment instruction will update the nodes and correct the weights of the attendance evidence chain dependency graph. This refers to reconstructing the dependency strength and priority relationships between modal nodes at the evidence chain structure level. The attendance evidence chain dependency graph is a directed graph structure used to describe the mutual influence relationships of various types of perceived evidence in the identity judgment process. Nodes represent modal evidence or decision-making units, and edges represent dependencies and weight values. Node updates include adjusting node activation conditions or adding redundant verification paths, while weight correction includes redistributing the contribution ratios of different pieces of evidence in the fusion stage. By updating the graph structure, the evidence chain becomes more stable and reduces the probability of conflicts during subsequent operation.

[0054] Furthermore, this application also includes: the temporal correspondence between the attendance evidence chain dependency graph and the multimodal attendance perception data, the constraints on the degree of freedom of identity mutual verification, and the decision priority of the single-modal confidence degree corresponding to each modality.

[0055] Specifically, the binding of the attendance evidence chain dependency graph with the temporal correspondence of multimodal attendance perception data, the constraints of the degree of freedom for identity verification, and the decision priority of the single-modal confidence degree corresponding to each modality refers to associating the nodes and edges in the evidence chain structure with the temporal sequence relationship of the actual perception data, ensuring that each piece of attendance evidence corresponds to its collection source in chronological order. The temporal correspondence of multimodal attendance perception data refers to the alignment and synchronization order of attendance data from different sources, such as facial capture images, wireless signal strength, and access control card swipe records, within a unified time window, used to ensure the temporal consistency of evidence fusion. The constraints of the degree of freedom for identity verification refer to the adjustment parameters used to limit the allowable difference range and weight allocation range between different modalities during the multimodal identity judgment process, achieving conflict tolerance and balanced decision-making by setting upper and lower limit constraints. The decision priority of the single-modal confidence score for each modality refers to the order in which the single-modal confidence score is determined during multimodal fusion, prioritizing its role in evidence chain node updates, decision weight allocation, and conflict resolution. This ensures that the influence of high-confidence modalities on the final identity judgment result is reflected first. By binding the evidence chain dependency graph with the aforementioned temporal correspondence, degree-of-freedom constraints, and confidence priority, unified control of attendance decisions is achieved at the structural, temporal, and weight layers, ensuring the model remains stable and consistent during multimodal fusion and conflict resolution.

[0056] Furthermore, this application also includes: in a multi-terminal collaborative attendance scenario, analyzing the modal differences of each sensing terminal; dynamically allocating authentication tasks and synchronously optimizing the attendance parameter configuration between each sensing terminal based on the modal differences of each sensing terminal; and enabling an authentication load migration mechanism when the attendance response time difference between each sensing terminal exceeds the time difference threshold.

[0057] Specifically, in multi-terminal collaborative attendance scenarios, analyzing the modal differences among various sensing terminals refers to comparing and analyzing the multimodal data collected by different terminals in terms of type, quality, time synchronization, and spatial coverage when multiple attendance sensing devices are operating simultaneously. A multi-terminal collaborative attendance scenario refers to a comprehensive attendance environment including multiple personnel entry / exit monitoring points, wireless signal access points, and video acquisition devices, where each terminal independently collects sensing data and participates in a unified identity verification process. Modal differences refer to the differences in acquisition accuracy, noise level, timestamp consistency, and spatial positioning deviation between the same or different types of data. By analyzing modal differences, high-noise terminals or low-confidence data sources can be identified, thus providing a basis for task allocation.

[0058] Furthermore, based on the modal differences of each sensing terminal, dynamically allocating identity verification tasks and synchronously optimizing attendance parameter configurations among the sensing terminals refers to the real-time scheduling and allocation of identity verification tasks among the terminals according to terminal modal differences and current load conditions, and adjusting parameters such as feature extraction dimensions, fusion weights, and decision thresholds within each terminal. Dynamically allocating identity verification tasks means automatically allocating the workload for identity judgment based on terminal performance indicators, data quality, and spatiotemporal consistency during system operation; attendance parameter configuration refers to the set of adjustable controllable parameters for each terminal in feature encoding, modal fusion, conflict determination, and confidence calculation. By synchronously optimizing parameter configurations, a balance between data processing efficiency and identity judgment accuracy can be achieved among multiple terminals.

[0059] When the time difference in attendance response between various sensing terminals exceeds a time difference threshold, the authentication load migration mechanism is activated. This means that, based on real-time monitoring of the task completion time of each terminal, when the response latency exceeds a preset threshold, some authentication tasks are migrated from the delayed terminal to the faster terminal to maintain the real-time performance of the overall system. Attendance response time difference refers to the time difference that occurs when different terminals complete the same authentication task; the time difference threshold is an allowable latency range set according to business needs and system performance; the authentication load migration mechanism refers to transferring the computing tasks of high-latency terminals to other available terminals through task redistribution and scheduling priority adjustment, thereby achieving load balancing and response optimization.

[0060] Furthermore, this application also includes: configuring an attendance credibility influence factor map based on the attendance evidence chain dependency graph; upon receiving a manual correction request, predicting the attendance utility change trend based on the attendance credibility influence factor map, issuing an optimization suggestion alarm, and identifying redundant evidence items through a map backtracking mechanism; based on the redundant evidence items, mining potential interference factors in conjunction with the attendance credibility influence factor map, the potential interference factors including WiFi signal drift, sudden changes in illumination, and Bluetooth beacon obstruction; iteratively correcting the attendance optimization instructions based on the potential interference factors until the attendance record utility test passes.

[0061] Specifically, configuring an attendance credibility influence factor graph by referencing the attendance evidence chain dependency graph involves quantifying the impact of each type of evidence on the final result in identity determination based on the nodes and edges of the attendance evidence chain, and establishing a mapping graph. The attendance evidence chain dependency graph is a directed graph structure describing the dependency order, confidence weight, and mutual corroboration relationship between multimodal evidence; the attendance credibility influence factor graph is a weighted graph that integrates information such as the confidence level, modal importance, conflict frequency, and historical correction records of each node, used to guide evidence fusion and corrective judgments. By mapping the evidence chain to the influence factor graph, a visual or computable evaluation model can be formed, providing a basis for subsequent utility analysis.

[0062] Upon receiving a manual correction request, the system predicts the trend of attendance utility changes based on the attendance credibility influencing factor graph, issues optimization suggestion alerts, and identifies redundant evidence items through the graph backtracking mechanism. This means that when manual intervention is triggered, the system simulates and predicts the potential utility fluctuations of attendance records using the node weights and dependencies in the graph, and generates optimization prompts after identifying evidence nodes that may lead to conflicts or inefficiencies. A manual correction request refers to the intervention operation initiated by attendance reviewers for abnormal or disputed records; the attendance utility change trend refers to the predicted change curves of attendance record accuracy, conflict probability, and manual correction rate over a future period; the graph backtracking mechanism refers to tracing the node contribution values ​​and weights upwards along the attendance evidence chain dependency path to identify redundant or duplicated evidence items in the fusion decision. This mechanism can identify factors that may reduce attendance utility in advance and issue optimization alerts to guide adjustments.

[0063] Based on redundant evidence items, potential interference factors are identified using an attendance credibility impact factor map. These factors include WiFi signal drift, sudden changes in lighting, and Bluetooth beacon obstruction. This involves analyzing the potential impact of environmental and equipment factors on evidence quality by considering the weights and dependencies of evidence nodes marked as redundant or anomalous within the map. Potential interference factors are environmental or equipment factors that may cause a decrease in single-modal confidence or anomalies in the evidence chain. For example, WiFi signal drift refers to the deviation of wireless signal strength from expectations due to environmental changes; sudden changes in lighting refer to short-term fluctuations in lighting conditions during image acquisition; and Bluetooth beacon obstruction refers to the blocking or attenuation of Bluetooth signals, leading to positioning or connection anomalies. By identifying potential interference factors, targeted corrections can be made to optimize instructions.

[0064] Based on potential interference factors, the attendance optimization instructions are iteratively revised until the attendance record utility test passes. This means that after identifying potential interference factors, the feature encoding dimensions, weight allocation, threshold parameters, and scheduling strategies in the optimization instructions are continuously adjusted, and the attendance record utility is recalculated after each adjustment until the utility index meets the preset test criteria. Iterative revision refers to enabling the model to adapt to environmental disturbances and modal conflicts through cyclical feedback and parameter updates; attendance record utility test refers to evaluating the accuracy, conflict rate, and correction rate of the finally generated seamless attendance records to ensure the reliability of the identity judgment results.

[0065] In summary, the deep learning-driven, mutual-verification, seamless attendance record generation method provided in this application has the following technical effects: by achieving the technical goal of seamless attendance record generation based on dynamic identity mutual verification and conflict adaptive processing of multimodal perception data, it can still accurately and in real time generate employee attendance records under modal anomalies or environmental interference, and improve the stability of attendance data, reduce the need for manual correction, and optimize attendance management efficiency and system reliability.

[0066] Example 2: Based on the same inventive concept as the deep learning-driven mutual verification seamless attendance record generation method in the foregoing examples, this application also provides a deep learning-driven mutual verification seamless attendance record generation system. Please refer to the appendix. Figure 2The system includes: a multimodal attendance perception data acquisition module 1, used to acquire multimodal attendance perception data including face capture images, WiFi probe signal strength, Bluetooth beacon connection records, and access control card swipe logs; a single-modal confidence determination module 2, used to perform multimodal identity verification analysis based on the multimodal attendance perception data, configure identity verification freedom, and determine the single-modal confidence corresponding to each modality by combining the employee historical attendance profile database and attendance rules and strategies; a trusted attendance information determination module 3, used to configure attendance optimization instructions that drive conflict resolution based on the identity verification freedom and the single-modal confidence corresponding to each modality, and use the attendance optimization instructions to adjust the underlying configuration parameters of the dual-channel identity verification model to determine trusted attendance information; and a seamless attendance record generation module 4, used to determine valid attendance events based on the trusted attendance information and the current attendance scenario type, and generate structured seamless attendance records using the employee identifier, attendance time, and attendance location in the valid attendance events.

[0067] Furthermore, the deep learning-driven mutual authentication seamless attendance record generation system is also used for: collecting attendance record call data and corresponding review and correction snapshots within historical attendance periods; extracting spatiotemporal consistency features and identity verification latency indicators corresponding to parallel processing tasks in the parallel inference path based on the attendance record call data and corresponding review and correction snapshots; and optimizing the decision nodes of the dual-channel identity mutual authentication model according to the spatiotemporal consistency features and identity verification latency indicators.

[0068] Furthermore, the deep learning-driven mutual authentication seamless attendance record generation system is also used to: obtain the spatiotemporal consistency deviation during the synchronous parsing of multimodal attendance perception data, and use the task backlog depth and priority inversion number of the identity verification queue as auxiliary indicators to measure attendance response efficiency; and adaptively adjust the parallel inference scheduling strategy of the dual-channel identity mutual authentication model through the spatiotemporal consistency deviation, task backlog depth and priority inversion number.

[0069] Furthermore, the deep learning-driven mutual verification seamless attendance record generation system is also used to: dynamically adjust the feature encoding dimension of the explicit channel and the behavior trajectory mining depth of the potential channel based on attendance optimization instructions that include explicit identity verification schemes, potential behavior trajectory mining strategies and attendance evidence complexity adaptation parameters; the dual-channel identity mutual verification model includes explicit channels and potential channels.

[0070] Furthermore, the deep learning-driven mutual verification-based seamless attendance record generation system is also used to: optimize the attendance evidence complexity stratification threshold based on the seamless interference index under normal user activity; and dynamically suppress the feature encoding dimension of high interference modalities and reduce the mining time window of potential behavioral trajectories based on the attendance evidence complexity stratification threshold.

[0071] Furthermore, the deep learning-driven mutual verification seamless attendance record generation system is also used to: record model state snapshots and utility difference data, and locate key configuration items that cause the attendance record review correction rate to increase; generate compensation adjustment instructions based on the key configuration items; and use the compensation adjustment instructions to update nodes and correct weights in the attendance evidence chain dependency graph.

[0072] Furthermore, the deep learning-driven mutual verification seamless attendance record generation system is also used to bind the temporal correspondence between the attendance evidence chain dependency graph and the multimodal attendance perception data, the constraints of the identity mutual verification degree of freedom, and the decision priority of the single-modal confidence degree corresponding to each modality.

[0073] Furthermore, the deep learning-driven mutual authentication seamless attendance record generation system is also used to: analyze the modal differences of each sensing terminal in a multi-terminal collaborative attendance scenario; dynamically allocate authentication tasks and synchronously optimize the attendance parameter configuration between each sensing terminal based on the modal differences of each sensing terminal; and enable the authentication load migration mechanism when the attendance response time difference between each sensing terminal exceeds the time difference threshold.

[0074] Furthermore, the deep learning-driven, mutually verifiable, seamless attendance record generation system is also used for: configuring an attendance credibility influence factor graph by referring to the attendance evidence chain dependency graph; upon receiving a manual correction request, predicting the attendance utility change trend based on the attendance credibility influence factor graph, issuing optimization suggestion alarms, and identifying redundant evidence items through a graph backtracking mechanism; based on the redundant evidence items, mining potential interference factors in conjunction with the attendance credibility influence factor graph, the potential interference factors including WiFi signal drift, sudden changes in illumination, and Bluetooth beacon obstruction; and iteratively correcting the attendance optimization instructions based on the potential interference factors until the attendance record utility verification passes.

[0075] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The deep learning-driven mutual verification seamless attendance record generation method and specific examples in the aforementioned embodiment one are also applicable to the deep learning-driven mutual verification seamless attendance record generation system of this embodiment. Through the foregoing detailed description of the deep learning-driven mutual verification seamless attendance record generation method, those skilled in the art can clearly understand the deep learning-driven mutual verification seamless attendance record generation system of this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0076] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0077] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A deep learning-driven, mutually verified, seamless attendance record generation method, characterized in that, The method includes: Acquire multimodal attendance perception data, including face capture images, WiFi probe signal strength, Bluetooth beacon connection records, and access control card swipe logs; Multimodal identity verification analysis is performed based on the multimodal attendance perception data, the degree of freedom for identity verification is configured, and the confidence level of each modality is determined by combining the employee historical attendance profile database and attendance rules and strategies. Based on the degrees of freedom of identity verification and the single-modal confidence level corresponding to each modality, configure attendance optimization instructions to drive conflict resolution, and use the attendance optimization instructions to adjust the underlying configuration parameters of the dual-channel identity verification model to determine reliable attendance information; Based on the reliable attendance information, valid attendance events are determined in conjunction with the current attendance scenario type, and structured, seamless attendance records are generated using the employee identifier, attendance time, and attendance location in the valid attendance events.

2. The deep learning-driven, mutually verified, seamless attendance record generation method as described in claim 1, characterized in that, Collect attendance record data and corresponding review and correction snapshots within the historical attendance period; Based on the attendance record call data and the corresponding audit correction snapshot, extract the spatiotemporal consistency features and identity verification latency indicators of the parallel processing tasks in the parallel inference path. The decision nodes of the dual-channel identity verification model are optimized based on the spatiotemporal consistency characteristics and identity verification delay index.

3. The deep learning-driven, mutually verified, seamless attendance record generation method as described in claim 2, characterized in that, The method for optimizing the decision nodes of the dual-channel identity verification model based on the spatiotemporal consistency characteristics and identity verification latency index includes: The spatiotemporal consistency deviation during the synchronous parsing of multimodal attendance perception data is obtained, and the task backlog depth and priority inversion number of the identity verification queue are used as auxiliary indicators to measure attendance response efficiency. The parallel inference scheduling strategy of the dual-channel identity verification model is adaptively adjusted by considering the spatiotemporal consistency deviation, task backlog depth, and priority inversion count.

4. The deep learning-driven, mutually verified, seamless attendance record generation method as described in claim 1, characterized in that, The method of adjusting the underlying configuration parameters of the dual-channel identity verification model using the attendance optimization command includes: Based on attendance optimization instructions that include explicit identity verification schemes, potential behavior trajectory mining strategies, and attendance evidence complexity adaptation parameters, the feature encoding dimension of explicit channels and the behavior trajectory mining depth of potential channels are dynamically adjusted. The dual-channel identity verification model includes an explicit channel and a potential channel.

5. The deep learning-driven, mutually verified, seamless attendance record generation method as described in claim 4, characterized in that, The method includes: Based on the non-obtrusive interference index under normal user activities, optimize the stratified threshold for attendance evidence complexity. Based on the aforementioned attendance evidence complexity stratification threshold, the feature encoding dimension of high-interference modalities is dynamically suppressed, and the mining time window for potential behavioral trajectories is reduced.

6. The deep learning-driven, mutually verified, seamless attendance record generation method as described in claim 4, characterized in that, Before using the attendance optimization command to adjust the underlying configuration parameters of the dual-channel identity verification model, the method further includes: Record model status snapshots and utility difference data to identify key configuration items that lead to an increase in the attendance record review and correction rate. Based on the aforementioned key configuration items, a compensation adjustment instruction is generated; The compensation adjustment instruction is used to update the nodes and correct the weights of the attendance evidence chain dependency graph.

7. The deep learning-driven, mutually verified, seamless attendance record generation method as described in claim 6, characterized in that, The attendance evidence chain dependency graph is bound to the temporal correspondence of the multimodal attendance perception data, the constraints of the freedom of identity mutual verification, and the decision priority of the single-modal confidence degree corresponding to each modality.

8. The deep learning-driven, mutually verified, seamless attendance record generation method as described in claim 7, characterized in that, The method further includes: In a multi-terminal collaborative attendance scenario, the modal differences of each sensing terminal are analyzed; Based on the modal differences of each sensing terminal, identity verification tasks are dynamically allocated and attendance parameter configurations among each sensing terminal are optimized synchronously. When the time difference between attendance response times of various sensing terminals exceeds the time difference threshold, the authentication load migration mechanism is activated.

9. The deep learning-driven, mutually verified, seamless attendance record generation method as described in claim 7, characterized in that, The method further includes: Configure an attendance credibility influence factor graph based on the attendance evidence chain dependency graph described above; Upon receiving a manual correction request, the system predicts the trend of attendance effectiveness based on the attendance credibility influencing factor map, issues an optimization suggestion alarm, and identifies redundant evidence items through the map backtracking mechanism. Based on the redundant evidence items, potential interference factors are mined by combining the attendance credibility influence factor map. The potential interference factors include WiFi signal drift, sudden changes in illumination, and Bluetooth beacon blockage. Based on the potential interference factors, the attendance optimization instructions are iteratively modified until the attendance record effectiveness test is passed.

10. A deep learning-driven, mutually verified, seamless attendance record generation system, characterized in that: The steps for implementing the deep learning-driven, mutual-verification, contactless attendance record generation method according to any one of claims 1 to 9 include: The multimodal attendance perception data acquisition module is used to acquire multimodal attendance perception data, including face capture images, WiFi probe signal strength, Bluetooth beacon connection records, and access control card swipe logs. The single-modal confidence determination module is used to perform multimodal identity verification analysis based on the multimodal attendance perception data, configure the degree of freedom of identity verification, and determine the single-modal confidence corresponding to each modality by combining the employee historical attendance profile database and attendance rules and strategies. The trusted attendance information determination module is used to configure attendance optimization instructions that drive conflict resolution based on the identity mutual authentication degree of freedom and the single-modal confidence degree corresponding to each modality, and to use the attendance optimization instructions to adjust the underlying configuration parameters of the dual-channel identity mutual authentication model to determine trusted attendance information. The seamless attendance record generation module is used to determine valid attendance events based on the trusted attendance information and the current attendance scenario type, and to generate structured seamless attendance records using the employee identifier, attendance time, and attendance location in the valid attendance events.

Images