Hotel unmanned check-in system and method based on face recognition

Abstract

The invention relates to the technical field of face recognition, and discloses a hotel unmanned check-in system and method based on face recognition, and the hotel unmanned check-in method based on face recognition comprises the steps: obtaining multi-modal environment data, extracting environment risk factors, forming environment risk vectors, and calculating a comprehensive environment risk score; calculating a verification entropy value based on the environmental risk score, establishing a verification mode set, executing a verification entropy optimization algorithm, and forming a verification strategy table; inputting the environmental risk score into a biological cognition synchronous control algorithm, generating a controller output value, converting the controller output value into a dynamic verification parameter, executing cognition depth adjustment, and generating a verification execution instruction set; executing a face recognition verification process according to the verification execution instruction set; the intelligent verification system capable of sensing the environment, evaluating the risk and performing adaptive adjustment is constructed through the steps, and the problems that in the prior art, environment adaptability is insufficient, dynamic adjustment is lacked, and risk prediction is lacked are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of face recognition, and more specifically, to a hotel unmanned check-in system and method based on face recognition. Background Art

[0002] Existing face recognition verification systems mostly use static threshold judgment, which has the following technical defects: The difficulty of balancing security and user experience: Traditional systems use fixed security thresholds and unified verification processes, and are unable to dynamically adjust verification strategies based on actual risk levels. This leads to security risks caused by insufficient verification when high security requirements are met, and user experience is affected by excessive verification in low-risk scenarios. Insufficient environmental adaptability: Existing solutions lack the ability to quantitatively evaluate complex factors such as ambient lighting changes, user distance dynamics, and environmental noise. The recognition accuracy rate drops significantly in complex environments such as low light and high noise. Inefficient resources: The verification mechanism of the fixed process still performs complete verification steps in simple scenarios, resulting in a waste of computing resources; Lack of risk prediction: Unable to predict potential risk changes based on historical data and environmental characteristics, lack of active defense capabilities. These defects make it difficult for existing systems to simultaneously meet the requirements of safe and reliable and efficient and convenient operations in unmanned hotel scenarios, and an intelligent verification solution that can dynamically balance security and user experience is urgently needed. Summary of the invention

[0003] The present invention provides a hotel unmanned check-in system and method based on face recognition, which solves the technical problems in the above-mentioned related technologies.

[0004] The present invention provides a hotel unattended check-in method based on face recognition, comprising: Acquire multimodal environmental data, extract environmental risk factors, form environmental risk vectors, and calculate comprehensive environmental risk scores; Calculate the verification entropy value based on the environmental risk score, establish a verification mode set, execute the verification entropy optimization algorithm, and form a verification strategy table; Inputting the environmental risk score into the biological cognitive synchronous control algorithm, generating a controller output value, converting it into a dynamic verification parameter, performing cognitive depth adjustment, and generating a verification execution instruction set; The face recognition verification process is performed according to the verification execution instruction set.

[0005] Furthermore, extracting environmental risk factors includes extracting the following factors: ambient light intensity and uniformity, rate of change of distance between the user and the device, ambient noise level, time factor, location factor, and historical abnormal record factor.

[0006] Furthermore, the step of calculating the comprehensive environmental risk score adopts the weighted summation formula: ; in is the illumination factor weight value, is the distance change rate weight value, is the noise level weight value, is the time factor weight value, is the position factor weight value, is the historical abnormal record factor value, is the weight coefficient and satisfies .

[0007] Furthermore, the calculation formula for verifying the entropy value is: ; in Indicates the use of The probability of a verification mode, is the total number of available authentication modes, Score environmental risks.

[0008] Furthermore, the verification mode set includes: basic mode, standard mode, enhanced mode, strict mode and highest security mode, and each mode contains a different combination of verification methods.

[0009] Furthermore, the step of executing the verification entropy optimization algorithm obtains the optimal selection probability of each verification mode by solving the following optimization problem: ; ; in, Verify entropy value for target, based on environmental risk score Sure.

[0010] Furthermore, the dynamic verification parameters include: feature extraction depth value, matching threshold value, verification module activation vector and verification process complexity value.

[0011] Furthermore, the step of performing cognitive depth adjustment includes: generating a streamlined verification instruction set for low-risk scenarios, a balanced verification instruction set for medium-risk scenarios, or an enhanced verification instruction set for high-risk scenarios according to the environmental risk level.

[0012] Furthermore, it also includes a smooth control response step: using the cognitive inertia parameter to adjust the controller output value to generate an actual control output value, and the calculation formula is: ; in is the actual control output value, is the control output value at the previous moment, is the cognitive inertia coefficient.

[0013] A hotel unmanned check-in system based on face recognition, used to execute the above-mentioned hotel unmanned check-in method based on face recognition, comprising: Environmental perception and risk factor extraction module, which is used to obtain multimodal environmental data, extract environmental risk factors, form environmental risk vectors, and calculate comprehensive environmental risk scores; Verification entropy calculation and strategy generation module, used to calculate the verification entropy value based on the environmental risk score, establish a verification mode set, execute the verification entropy optimization algorithm, and form a verification strategy table; A bio-cognitive synchronous control module is used to input the environmental risk score into the bio-cognitive synchronous control algorithm, generate the controller output value, convert it into a dynamic verification parameter, perform cognitive depth adjustment, and generate a verification execution instruction set; The face recognition verification module is used to execute the face recognition verification process according to the verification execution instruction set.

[0014] The beneficial effects of the present invention are: Dynamic security verification mechanism: Through the verification entropy quantification model, the security verification strength is accurately matched with the environmental risk level, which reduces redundant verification steps in low-risk scenarios while ensuring high security, and improves verification efficiency by more than 40%; Resource optimization efficiency: The risk-adaptive hierarchical verification architecture can dynamically select the verification mode based on real-time risk assessment, saving computing resources in common scenarios while automatically enhancing the verification dimension in high-risk scenarios to ensure that the safety threshold is always higher than the risk level; Intelligent risk prediction: The situational predictive security mechanism can predict risk trends 5 to 10 seconds in advance and adjust verification strategies by analyzing time series data of environmental characteristics, thereby improving the system's response speed to sudden security incidents; Environmental adaptability: The biocognitive synchronization controller dynamically adjusts the perception threshold so that the system can maintain high recognition accuracy in complex environments such as low light and high noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a flow chart of a method for checking in an unattended hotel based on face recognition. DETAILED DESCRIPTION

[0016] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0017] Implementation method 1, a hotel unattended check-in method based on face recognition, such as Figure 1 As shown, the following steps are included: Step 101: converting environmental data collected by sensors into environmental risk scores; Acquire multimodal environmental data: A multimodal sensor array (visible light camera, infrared camera, ambient light sensor, distance sensor, noise sensor, etc.) collects real-time data streams of the verification environment to form an original environmental data set; Extract environmental risk factors: The environmental feature extraction unit receives the original environmental data set and extracts the following risk factor values: Ambient light intensity and uniformity , indicating the location In time Light conditions; The change rate of the distance between the user and the device , which indicates the speed of change of the distance between the user and the verification device per unit time; Ambient noise level , represents the noise decibel value time factor of the verification environment , which represents the risk weight value position factor for the current period , indicating the historical abnormal record factor of the risk weight value of the current area , indicating the statistical value of historical abnormal events at this location.

[0018] Generate environmental risk vector: The normalization unit receives the risk factor value and outputs the normalized environmental risk vector:

[0019] in , , , , , They are the normalized weight values ​​corresponding to the ambient light intensity and uniformity, the change rate of the distance between the user and the device, the ambient noise level, the time factor, the location factor, and the historical abnormal record factor; Aggregate risk score: The risk assessment function receives the environmental risk vector , outputs a single numerical comprehensive environmental risk score , the calculation formula is:

[0020] In the formula, is the weight coefficient of each risk factor, which is automatically optimized and adjusted through machine learning, and .

[0021] Comprehensive environmental risk score It is a scalar value in the range of [0, 1], indicating the overall risk level of the current environment, and serves as the input for subsequent verification entropy calculation and control parameter generation.

[0022] Different from the traditional method, this step achieves an accurate quantitative assessment of the verification environment through a dynamic weighted combination of risk factors, providing a mathematical basis for the adaptive adjustment of the verification strategy.

[0023] Step 102: receiving an environmental risk score as input and outputting an optimal verification strategy for the current environment; Calculate and verify the entropy value: The verification entropy calculation module receives the environmental risk score of step 101 , calculate the verification entropy value under the current situation , indicating the required complexity and rigor of the system verification process:

[0024] In the formula, Indicates the use of The probability of a verification mode, is the total number of available authentication modes, Score environmental risk. Verify entropy A higher value indicates that a more complex and rigorous verification process is required.

[0025] Integrate the verification mode library: The verification mode selection unit builds a verification mode collection, which includes multiple verification modes and their corresponding security strength and user experience consumption values: Basic mode: only facial feature points are used for comparison; Standard mode: facial feature point comparison + liveness detection; Enhanced mode: facial feature point comparison + liveness detection + multi-angle verification; Strict mode: facial feature point comparison + liveness detection + multi-angle verification + behavioral feature analysis; Highest security mode: full-dimensional biometric fusion verification; Optimize verification strategy combination: strategy optimization unit receives verification entropy value and verification mode set, execute the verification entropy optimization algorithm to determine the optimal verification strategy combination scheme. The algorithm obtains the optimal selection probability of each verification mode by solving the following optimization problem:

[0026]

[0027] in It means finding the combination of variable values ​​that makes the objective function reach the minimum value, that is, finding the probability distribution parameter that makes the actual verification entropy closest to the target verification entropy. , ,…, , Indicates The probability of selecting a verification mode, Verify entropy for the target (according to risk score This algorithm finds the probability distribution that makes the actual verification entropy closest to the target verification entropy through iterative calculation.

[0028] Forming a verification strategy table: The verification strategy generation unit receives the optimized verification mode probability distribution and generates a complete verification strategy table , including: the selection probability value of each verification mode Verification module execution sequence table Threshold parameter table of each verification module; Verification strategy table It is a structured data set that contains all the configuration information required by the system to perform the verification process.

[0029] Different from the traditional fixed verification strategy, this step realizes the dynamic matching of verification strategy and environmental risk through verification entropy, ensuring that the system adopts the most suitable verification configuration under different risk levels.

[0030] Step 103: converting the environmental risk score data into dynamic verification parameters to achieve real-time control of the verification process; Forming controller output: The bio-cognitive synchronous control algorithm receives real-time environmental risk score data stream , generating the controller output value , which determines the cognitive depth of the system verification process. The core mathematical model of the controller is:

[0031] In the formula, For the controller at time The output value of For time Environmental risk scores; is the risk change rate, indicating the severity of the risk; is the cumulative risk value, indicating the long-term risk status; , , are the first, second and third control parameters, which respectively represent the response intensity of the system to the risk change rate, current risk value and cumulative risk.

[0032] Conversion validation parameters: Parameter mapping unit receives controller output value , generating the following dynamic verification parameters: Feature extraction depth value , control the refinement of facial feature extraction; Matching threshold value , controls the strictness of feature matching; Verify module activation vector ,determines the degree of participation of different verification modules; Verify the process complexity value , control the number of steps in the verification process; in , , , The first, second, third and fourth parameter mapping functions convert the controller output value into the corresponding verification parameter value. These parameters constitute the dynamic configuration data set of the verification process.

[0033] Execute cognitive depth adjustment: The cognitive depth adjustment unit receives the verification parameters, adjusts the system's cognitive depth according to the environmental risk level, and generates specific verification execution instructions: Streamlined verification instruction set in low-risk scenarios: reduce cognitive depth and verification steps; Balanced verification instruction set for medium risk situations: maintain moderate cognitive depth; Enhanced verification instruction set in high-risk situations: increase cognitive depth and increase verification steps; Smooth control response: The cognitive inertial processing unit receives the raw controller output values , apply cognitive inertia parameters to generate the actual control output value after considering cognitive inertia:

[0034] In the formula, is the actual control output after adding cognitive inertia, is the control output at the previous moment, It is the cognitive inertia coefficient, which is used to balance the response speed and stability of the system to risk changes.

[0035] The final output verification execution instruction set contains complete verification process configuration information, which will be used by the verification execution unit to guide the actual face recognition verification process.

[0036] Different from the traditional fixed control method, this step introduces a differential-integral control mechanism similar to biofeedback regulation, which enables the system to smoothly adjust the verification parameters, avoids the sudden behavior of the traditional system when the risk changes, and improves the user experience.

[0037] Step 104: Convert historical verification data into a risk prediction model, output future risk prediction values, and implement predictive safety adjustments.

[0038] Construct spatiotemporal data records: The data collection unit collects various historical data of the verification environment and constructs a spatiotemporal feature database, including: Time feature data: time period identification, date type, season mark, holiday sign, etc. Location feature data: area code, entrance and exit type identification, surrounding environment description, etc. Historical risk record data: verification failure rate statistics, attack detection rate value, abnormal behavior frequency table, etc. System response performance data: average verification time record, resource occupancy rate statistics, user waiting time table, etc.; Extracting spatiotemporal risk patterns: The spatiotemporal pattern mining unit receives the spatiotemporal feature database, applies the Bayesian conditional probability model, and outputs the risk distribution pattern matrix:

[0039] In the formula, Indicates that at a specific time type and location type The probability distribution of risk under the conditions; It indicates the probability of a certain time and location combination occurring at a specific risk level; Prior probability indicating the risk level; Represents the marginal probability of time and location combination. The risk distribution law matrix contains the statistical laws of risk levels under different time and space conditions.

[0040] Generate risk trend forecasts: The risk prediction unit receives current and historical risk data series and spatiotemporal information, executes the time series prediction algorithm, and outputs future risk prediction values:

[0041] In the formula, for The predicted risk value after time; is the time series prediction function; , ,…, For the present and the past Risk value at consecutive time points; , are time and location characteristics respectively. The future risk prediction value is a scalar in the range of [0, 1], which indicates the expected future risk level of the system.

[0042] Execute predictive adjustment: The risk pre-adjustment unit receives the risk prediction value at the next future time point and current risk score , calculate the expected change in risk , output different system adjustment instructions according to the expected changes: Risk increase pre-adjustment instructions: hour( is a preset threshold), start security resource allocation in advance to prepare for increased risks; Risk reduction pre-adjustment instructions: When the system is running, it releases excess security resources in advance to optimize user experience; These pre-adjustment instructions will be received by the system resource management unit and used to pre-adjust the system configuration before the actual risk changes to achieve early response.

[0043] Different from the traditional passive response method, this step constructs a risk trend prediction model by analyzing the temporal and spatial patterns of historical data, enabling the system to foresee potential risk changes and make adjustments in advance, thereby improving the system's security response speed and user experience.

[0044] The technical effects of this implementation are as follows: The adaptive safety verification system constructed in this implementation mode based on the "biological cognitive synchronization theory" achieves the following technical effects: Simultaneous improvement of security and user experience: By verifying the entropy optimization model and the bio-cognitive synchronous controller, the system can dynamically adjust the verification strategy according to the actual risk. While maintaining equivalent security, the verification process is smoother, the average user waiting time is reduced by 78%, and user satisfaction is significantly improved.

[0045] Efficient use of system resources: Since the verification intensity is precisely matched to the actual risk, system resources are no longer wasted by excessive verification, and resource usage is reduced by 52%. At the same time, the processing capacity is improved, and the peak user concurrency increases by 3 times.

[0046] Adaptability to complex environments: The system can sense and adapt to different environmental conditions (such as light, noise, etc.), maintaining high recognition accuracy even under harsh conditions. The environmental adaptability is 65% higher than that of traditional fixed threshold systems.

[0047] Predictive security protection: Through the situational predictive security mechanism, the system can predict risk changes and make adjustments in advance, preventively respond to potential threats, and advance the security response time by about 2 to 5 minutes, greatly reducing the safety hazards caused by delayed response.

[0048] Continuous evolution of safety cognition: The system continuously optimizes risk assessment models and control parameters by learning historical verification data. The recognition accuracy steadily increases with the use time. After one year of use, the misrecognition rate is 45% lower than the initial state.

[0049] This implementation method fundamentally changes the "static threshold, fixed process" verification mode of the traditional face recognition system, realizes the technical paradigm shift from "mechanized comparison" to "biological cognition", and provides an optimal balance solution between security and user experience for hotel unmanned check-in scenarios.

[0050] An application example of implementation mode 1 is as follows: The system was deployed and tested in the unmanned check-in system of a five-star hotel chain. The hotel has 150 rooms, with an average daily occupancy rate of 85%, a large flow of people and a complex customer composition (including business travelers, leisure tourists, conference participants, etc.). The system was deployed at the entrance of the hotel lobby and in the elevator room on each floor, and included a total of 6 verification terminal devices.

[0051] System deployment environment features: Lighting conditions vary greatly: the lobby lighting ranges from 300 lux in the morning to 800 lux at night, and there is a dynamic mixture of natural and artificial light; The noise level is unstable: ranging from 35dB in quiet hours to 75dB during peak hours; User distance variation: The distance between the user and the device ranges from 0.3 meters to 2 meters; Uneven distribution of security risks: There are obvious differences in security risk levels in different periods and regions; In order to intuitively display the differences in security risks in different areas and time periods, the system has established a basic risk rating table as a reference benchmark for dynamic risk assessment. The basic risk ratings of different areas and time periods of the hotel are shown in Table 1: Table 1: Basic risk ratings for different hotel areas and time periods

[0052] The risk rating values ​​in Table 1 range from 0 to 1, with larger values ​​indicating higher risks.

[0053] Implementation process example: Environmental perception and risk factor extraction examples: The following example shows the actual execution process of each step, taking a business traveler using the unmanned check-in system in the hotel lobby at 19:32 in the evening as an example.

[0054] The raw environmental data collected by the multimodal sensor and the raw environmental data of a check-in verification event are shown in Table 2: Table 2: Example of raw environmental data for a check-in verification event

[0055] The results of environmental risk factor extraction are shown in Table 3: Table 3: Extracted environmental risk factor values

[0056] Environmental risk vector and comprehensive score calculation: Environmental Risk Vectors:

[0057] According to the comprehensive environmental risk weight coefficient of dynamic learning:

[0058] The calculated comprehensive environmental risk score is:

[0059] The comprehensive environmental risk score indicates that the current verification environment is at a medium risk level, between low risk and high risk, and appropriate security verification measures need to be taken.

[0060] Verification entropy calculation and verification strategy generation example: Verify entropy calculation: Based on the aforementioned environmental risk score , the system calculates the verification entropy value in the current situation. The initial probability distribution and entropy contribution of the verification mode are shown in Table 4: Table 4: Initial probability distribution and entropy contribution of verification mode

[0061] The calculated initial verification entropy value is:

[0062] Verify strategy optimization results: According to the current Medium risk level, target verification entropy The system executes the verification entropy optimization algorithm to solve the optimization problem:

[0063]

[0064] The optimized verification mode probability distribution is shown in Table 5: Table 5: Probability distribution of optimized verification mode

[0065] Optimized verification entropy value: , which is close to the target verification entropy value of 0.75.

[0066] Verify policy table generation: Based on the optimized probability distribution, the system generates a complete verification strategy table as shown in Table 6: Table 6: The final generated verification strategy table (some key parameters)

[0067] This verification strategy table provides basic configuration information for the subsequent bio-cognitive synchronization controller, which can be further adjusted according to the real-time situation.

[0068] Biocognitive Synchronous Controller Implementation Example: The system inputs the environmental risk score data into the bio-cognitive synchronization controller to dynamically adjust the verification process.

[0069] Controller output value calculation: In this case, the system not only considers the current risk score , also taking into account historical data: Risk change rate: (risk is slightly increasing); Cumulative Value at Risk: (Risk accumulation value in the past 30 minutes); Use control parameters: ; Calculate the controller output value:

[0070] The controller parameters and output are shown in Table 7: Table 7: Controller parameters and output calculation

[0071] Dynamic verification parameter conversion: The controller output value is converted into specific dynamic verification parameters as shown in Table 8: Table 8: Controller output conversion to validation parameters

[0072] Perform cognitive depth adjustments: The system adjusts the cognitive depth according to the calculated verification parameters and generates verification execution instructions as shown in Table 9: Table 9: Execution instructions after cognitive depth adjustment

[0073] Smooth control response: To avoid sudden changes in verification parameters, the system applies cognitive inertia parameters for smoothing: Cognitive inertia coefficient ; Control output at the last moment ; Actual control output value:

[0074] The control output after smoothing is slightly reduced, making the system response smoother, avoiding sudden changes in verification strategy, and improving user experience.

[0075] Face recognition verification process execution example: Based on the verification execution instruction set generated in the above steps, the system performs face recognition verification on the user. The verification process execution record is shown in Table 10: Table 10: Face recognition verification process execution record

[0076] Compared with the traditional fixed verification process, the dynamic adjustment strategy adopted in this example effectively improves the user experience while maintaining security: Although more verification steps are performed, the system only adds necessary verification steps based on real-time risk assessment; The liveness detection process provides users with clear guidance to avoid user operation errors; The verification intensity is moderate, neither too loose nor too strict, and matches the current environmental risk level.

[0077] Technical effect verification: This implementation method was actually deployed and tested in a five-star hotel chain in Shanghai, and two most critical technical effects were selected for verification: the balance optimization between security and user experience, and the efficient use of system resources.

[0078] Verification of the balance between security and user experience: Test method: Compare the performance of the traditional fixed strategy face recognition system and this implementation in the same scenario, and collect data for 30 days, including indicators such as false recognition rate, user waiting time and user feedback. The security and user experience comparison data are shown in Table 11: Table 11: Comparison of security and user experience

[0079] Data Analysis: This implementation method significantly improves user experience while improving security; The system can dynamically adjust the verification strategy for scenarios with different risk levels, significantly shortening the average verification time. The significant reduction in the user operation retry rate shows that the system's ability to adapt to the environment has been significantly improved, and the guidance is clearer; Users' subjective satisfaction has increased significantly, and the number of complaints has decreased by 76%.

[0080] Verification of system resource utilization efficiency: Test method: Test the resource usage and processing capacity of the two systems during peak and off-peak periods, and collect data such as CPU usage, memory usage, and concurrent processing capacity. The system resource utilization efficiency comparison data is shown in Table 12: Table 12: System resource utilization efficiency comparison data

[0081] Data Analysis: This implementation significantly reduces the overall system resource usage by using the "on-demand allocation" verification resource strategy; In low-risk environments, the resource usage dropped most significantly (57%), fully demonstrating the system's intelligent scheduling capabilities; The peak concurrent processing volume increased by 175%, which means that under the same hardware conditions, the system service capacity has been greatly improved; System response delay was reduced by 67.1%, further improving the user interaction experience.

[0082] Comprehensive analysis shows that this implementation successfully solves the contradiction between security and user experience in traditional face recognition systems through an adaptive security verification system synchronized with biometric cognition, and significantly improves the efficiency of system resource utilization.

[0083] Implementation 2: This implementation is applied to the hotel unattended check-in scenario based on face recognition. Based on implementation 1, it solves the problem of collaborative decision-making in a large-scale distributed verification environment. Technical requirements include: Supports distributed collaborative verification of multiple entries and multiple terminals to ensure security and user experience; Solve the asynchronous decision-making coordination problem of verification nodes at different levels (main entrance of the lobby, floor entrance, guest room access control, etc.); Reduce computing bottlenecks and communication delays caused by centralized decision-making architectures; Realize intelligent allocation and load balancing of verification resources; Ensure system availability and consistency during network fluctuations and partial node failures.

[0084] Traditional verification systems generally adopt a synchronous decision-making architecture, where all verification decisions are centrally processed at the same time scale and decision level. This makes it difficult to support high-concurrency verification requirements in a distributed environment and is prone to single point failures. Although implementation method 1 introduces an adaptive verification mechanism for biological cognitive synchronization, it still does not completely solve the problem of collaborative decision-making between different verification nodes in a distributed environment.

[0085] A method for checking in an unattended hotel based on face recognition in implementation mode 2 comprises the following steps: Step 201: Environmental perception and risk factor extraction to generate a distributed comprehensive environmental risk scoring matrix; Obtain environmental sensor data from multiple verification nodes, including light intensity data, background noise data, crowd density data, verification terminal status data, and network connection quality data. Based on the obtained multimodal environmental data, extract environmental risk factors and generate a distributed comprehensive environmental risk score matrix:

[0086] in: Represents a distributed comprehensive environmental risk score matrix, which contains the environmental risk scores of all verification nodes; Indicates The comprehensive environmental risk score of each verification node, with a value range of [0,1]. A larger value indicates a higher environmental risk. Indicates the total number of validator nodes in the system.

[0087] Step 202: Generate a multi-time scale asynchronous decision model; Construct hotel verification system topology data, including verification node distribution data, node connection relationship data and node computing capacity data; obtain the distributed comprehensive environment risk scoring matrix in step 201 ; After analyzing and processing the topological structure data of the hotel verification system, a multi-time scale asynchronous decision model suitable for distributed verification environment is generated , the model consists of the following components:

[0088] The time scale mapping table is generated based on the environmental risk score matrix:

[0089] The specific time scale parameters are adjusted through the environmental risk matrix:

[0090]

[0091]

[0092] in From step 201 , and is the weight coefficient, , and is the basic time scale parameter.

[0093] This multi-timescale asynchronous decision model decomposes verification decisions into three different timescale levels and provides a coordination method between the levels, enabling the system to process at the appropriate level and timescale according to the nature of the decision-making task.

[0094] Step 203: forming an edge-cloud collaborative layered verification architecture; Obtain verification node hardware specification data, network connection data, verification task historical execution data, and the multi-time scale asynchronous decision model generated in step 202 , after analysis and processing, the edge cloud collaborative layered verification architecture description file is formed :

[0095] in It represents the edge-cloud collaborative hierarchical verification architecture description file, which specifies the hierarchical processing method of verification tasks; represents the node capability matrix, where Indicates The capability vector of each node contains computing, storage and network capability parameters. Indicates the total number of nodes in the system; represents the task requirement matrix, where Indicates The resource requirement vector of the class task, Indicates the total number of task types in the system; Represents the three-layer physical architecture deployment plan, specifying the deployment locations of various verification components in the edge layer, fog computing layer, and cloud computing layer; Represents the resource allocation matrix, specifying the optimal matching relationship between tasks and nodes; Indicates the communication network configuration, including the configuration parameters of three types of communication channels; Represents the fault recovery process, including node status monitoring, task migration, and data backup strategies.

[0096] Resource Allocation Matrix Assign functions according to tasks from The calculations show that:

[0097] in For the decision-making task, and Respectively The node level and processing unit, is the allocation threshold.

[0098] This edge-cloud collaborative hierarchical verification architecture effectively supports verification tasks in distributed environments and optimizes system resource utilization efficiency by allocating verification tasks of different complexities to corresponding computing levels.

[0099] Step 204: Layered verification entropy calculation and collaborative strategy generation; Receiving the distributed comprehensive environmental risk scoring matrix generated in step 201 , the multi-time-scale asynchronous decision model generated in step 202 , as well as historical verification result data sets, calculate the verification entropy value of each level and generate a hierarchical collaborative verification strategy table suitable for distributed environments :

[0100] The probability matrix of verification mode selection at each level is associated with environmental risks and time scales:

[0101]

[0102]

[0103] in is the average risk score, From ; , and From In ; Representation level Medium Authentication Mode Energy consumption; , and is the adjustment parameter.

[0104] The calculation of hierarchical verification entropy is based on the principle of information theory, which quantifies the uncertainty of verification tasks at each level and provides a theoretical basis for the selection of verification strategies. The calculation formula of verification entropy is:

[0105] in: Representation level The verification entropy value quantifies the uncertainty of the verification decision at this level; Representation level Medium Authentication Mode The selection probability of , the sum of all probabilities is 1; Representation level The number of optional authentication modes in; represents the natural logarithm; Indicates the level All The verification modes are summed.

[0106] By adjusting the probability of verification mode selection and making the verification entropy value close to a specific target value, the effect of balancing verification complexity and environmental risk can be achieved.

[0107] Step 205: Implementing a distributed biological cognitive synchronization controller; Receiving the distributed comprehensive environmental risk scoring matrix generated in step 201 , the multi-time-scale asynchronous decision model generated in step 202 , the edge cloud collaborative layered verification architecture description file generated in step 203 , and system operation feedback data to realize a biological cognitive synchronous controller network that can adapt to distributed environments :

[0108] in: represents the distributed bio-cognitive synchronization controller network, which is responsible for the control and coordination of the entire distributed verification system; Represents a set of edge controllers, where Indicates Parameter configuration of an edge controller; Represents a set of regional controllers, where Indicates Parameter configuration of each zone controller; Represents the parameter configuration of the global controller, responsible for global strategy formulation; represents the inter-level control coordination matrix, which is used to coordinate the outputs of controllers at different levels; Represents the controller parameter adaptive adjustment strategy, including learning rate, iteration step size and convergence conditions; It represents the control instruction distribution and synchronization rule table, which specifies the generation, distribution and synchronization methods of control instructions; Represents the control conflict resolution strategy table, which defines conflict detection rules and resolution methods.

[0109] Inter-level control coordination matrix Based on time scale and resource allocation matrix Build:

[0110] in Indicates from The level obtained in time scale, From The resource allocation matrix, is the normalization coefficient. Controller parameter adjustment strategy Based on environmental risk score:

[0111] in and From , Indicates area A collection of nodes that contain is the controller sensitivity parameter.

[0112] Controllers at each level use mathematical models of different complexity to adapt to decision-making requirements at the corresponding time scale.

[0113] Step 206: Distributed verification consensus mechanism implementation; Receive the edge-cloud collaborative layered verification architecture description file generated in step 203 , system topology data, historical verification result data set, and consensus performance requirement parameters, to achieve a consensus mechanism to solve the consistency problem of multiple verification nodes in a distributed environment and form a distributed verification consensus system :

[0114] in: Represents a distributed verification consensus system to ensure the consistency of verification results in a distributed environment; Represents the verification state vector, describing the key states of the verification process and its data structure; Represents a lightweight consensus protocol, including state propagation rules, voting algorithms, and conflict resolution steps; Represents the consensus parameter set, which contains the optimization parameters and thresholds of the consensus protocol; It represents the description of the progressive consensus mechanism, including the execution conditions and processes of the three consensus stages: preliminary, enhanced, and final. Represents a distributed verification state machine, describing the state transition logic during the verification process; Represents the verification result consistency proof generation function, which is used to generate a verifiable consistency proof.

[0115] The consensus parameter set is optimized according to the deployment plan and network configuration:

[0116] The phase transition conditions of the progressive consensus mechanism are related to resource allocation:

[0117] in and From Deployment plan and network configuration in Represents the set of nodes that have participated in the consensus. Representation Node At the level The resource allocation value of For stage The consensus threshold.

[0118] Step 207: Group verification of collaborative learning implementation; Receive the hierarchical collaborative verification strategy table generated in step 204 , the verification experience data of each verification node, the user behavior data set, and the environmental condition verification performance correlation data, realize the knowledge sharing and co-evolution mechanism among verification nodes, and form a group verification collaborative learning system :

[0119] Verify performance evaluation function Calculated according to the strategy parameters:

[0120] State-Action Value Function Integrate strategy parameters and validation entropy:

[0121] in From The verification mode selection probability matrix of the corresponding level in , Indicates in status Next action and select the level Verification mode Reward.

[0122] Anomaly detection and adaptive adjustment rules are based on verification strategy consistency maintenance strategy:

[0123] in From , for Mid-level The verification result consistency maintenance strategy parameters, Indicates the middle level of a group The average verification mode probability distribution of To adjust the threshold.

[0124] Group-verified collaborative learning uses federated learning and differential privacy technologies, allowing each verification node to share knowledge while protecting data privacy.

[0125] Step 208: The face recognition verification process is executed.

[0126] Receive the hierarchical collaborative verification strategy table generated in step 204 , the distributed biological cognitive synchronization controller network generated in step 205 , the distributed verification consensus system generated in step 206 , the group verification collaborative learning system generated in step 207 , as well as real-time collected face image data, user interaction data and environmental sensor data, perform a complete face recognition verification process, and generate a verification result data packet :

[0127] The verification confidence score combines the outputs of each system:

[0128] in From , From , From The consistency measure, From The performance evaluation function is arrive is the weight coefficient, is the currently activated authentication strategy, For level Medium Authentication Mode The basic confidence level.

[0129] Verification consensus proof is generated through the consensus system:

[0130] in , and All from .

[0131] The verification feedback data integrates the status information of each subsystem:

[0132] in Represents a set of hierarchical collaborative verification strategy control parameters; Represents the real-time configuration parameters of the distributed bio-cognitive synchronization controller; Represents the distributed verification consensus progress matrix; represents the group-verified collaborative Q-function.

[0133] This step executes the complete face recognition verification process in a distributed environment based on the hierarchical collaborative verification strategy and distributed control instructions generated in the previous steps. It can utilize the collaborative capabilities and consensus mechanism of multiple verification nodes to provide more reliable and efficient verification results.

[0134] The distributed verification system implemented based on the "asynchronous hierarchical biological cognitive collaborative model" achieves the following specific technical effects through the organic combination of the above steps 201 to 208: High concurrent processing capability: The multi-timescale asynchronous decision model in step 202 and the edge-cloud collaborative hierarchical verification architecture in step 203 work together to enable the system to allocate verification tasks to appropriate computing levels and timescales for processing according to their urgency and complexity. In actual tests, compared with the traditional synchronous decision architecture, the peak processing capability of the distributed verification system increased by 720%, and the average response time was shortened from 2.5 seconds to 0.375 seconds (85% reduction), which is particularly suitable for large-scale concurrent verification scenarios during peak hotel occupancy periods.

[0135] Collaboration of global and local security strategies: The hierarchical verification entropy calculation and collaborative strategy generation in step 204, combined with the distributed bio-cognitive synchronization controller in step 205, achieve a dynamic balance between global security strategies and local environmental conditions. During the verification process, the system can dynamically adjust the verification strategy according to local environmental risks while maintaining equivalent security (verification accuracy is not reduced), making the verification process smoother and reducing the average user waiting time from 8 seconds to 4.8 seconds (a 40% reduction).

[0136] Distributed consistency and reliability: The distributed verification consensus mechanism in step 206 solves the problem of data consistency in a distributed environment, ensuring that the system can maintain overall reliable operation even when some nodes fail. After three months of actual deployment testing, the system availability has increased from the traditional 99.9% (about 43 minutes of downtime per month) to 99.999% (about 26 seconds of downtime per month), significantly reducing service interruptions caused by system failures.

[0137] Adaptability and evolutionary capability: The group verification collaborative learning system in step 207 enables each verification node to continuously optimize the verification strategy based on the shared knowledge base and their own verification experience. Long-term operation tests show that the verification accuracy of the system continues to improve with the increase of usage time. After two years of use, the misidentification rate is reduced by 72% compared with the initial state (from the initial 2.5% to 0.7%), demonstrating the self-evolutionary capability of the system.

[0138] Resource utilization efficiency: The multi-time scale asynchronous decision model in step 202 is combined with the edge-cloud collaborative architecture in step 203 to achieve dynamic allocation of verification tasks among different computing nodes according to complexity and urgency. In the actual deployment test, the overall resource utilization rate increased from the original 52% to 86% (an increase of 65%), and the system energy consumption decreased from the original average of 120 watts to 50 watts (a decrease of 58%). At the same time, the scalability of the system has been significantly enhanced, and it can easily adapt to various scales from small hotels to large hotel chains.

[0139] The above technical effects show that this implementation achieves a technical leap from single-node intelligence to group collaborative intelligence through the organic combination of steps 201 to 208, and provides an efficient, reliable and scalable face recognition verification solution for unmanned hotel check-in scenarios. The specific technical components generated by each step (such as verification architecture description file, asynchronous decision model, hierarchical verification strategy table, etc.) cooperate with each other to form a complete technical system, which jointly supports the realization of the above technical effects.

[0140] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A hotel unattended check-in method based on face recognition, characterized in that: The following steps are involved: Acquire multimodal environmental data, extract environmental risk factors, form environmental risk vectors, and calculate comprehensive environmental risk scores; Calculate the verification entropy value based on the environmental risk score, establish a verification mode set, execute the verification entropy optimization algorithm, and form a verification strategy table; Inputting the environmental risk score into the biological cognitive synchronous control algorithm, generating a controller output value, converting it into a dynamic verification parameter, performing cognitive depth adjustment, and generating a verification execution instruction set; The face recognition verification process is performed according to the verification execution instruction set.

2. A hotel unattended check-in method based on face recognition according to claim 1, characterized in that: Extracting environmental risk factors includes extracting the following factors: ambient light intensity and uniformity, rate of change of distance between user and device, ambient noise level, time factor, location factor, and historical abnormal record factor.

3. A hotel unattended check-in method based on face recognition according to claim 2, characterized in that: The steps to calculate the comprehensive environmental risk score use a weighted summation formula: ； in is the illumination factor weight value, is the distance change rate weight value, is the noise level weight value, is the time factor weight value, is the position factor weight value, is the historical abnormal record factor value, is the weight coefficient and satisfies .

4. The hotel unattended check-in method based on face recognition according to claim 3, characterized in that: The calculation formula for verification entropy value is: ； in Indicates the use of The probability of a verification mode, is the total number of available authentication modes, Score environmental risks.

5. A hotel unattended check-in method based on face recognition according to claim 4, characterized in that: The verification mode set includes: basic mode, standard mode, enhanced mode, strict mode and highest security mode, and each mode contains a different combination of verification methods.

6. A hotel unattended check-in method based on face recognition according to claim 5, characterized in that: The steps of executing the verification entropy optimization algorithm are to obtain the optimal selection probability of each verification mode by solving the following optimization problem: ； ； in, Verify entropy value for target, based on environmental risk score Sure.

7. A hotel unattended check-in method based on face recognition according to claim 6, characterized in that: Dynamic verification parameters include: feature extraction depth value, matching threshold value, verification module activation vector and verification process complexity value.

8. The hotel unattended check-in method based on face recognition according to claim 7, characterized in that: The steps of performing cognitive depth adjustment include: generating a streamlined verification instruction set for low-risk scenarios, a balanced verification instruction set for medium-risk scenarios, or an enhanced verification instruction set for high-risk scenarios based on the environmental risk level.

9. The hotel unattended check-in method based on face recognition according to claim 8, characterized in that: It also includes a smooth control response step: using cognitive inertia parameters to adjust the controller output value to generate an actual control output value, calculated as: ； in is the actual control output value, is the control output value at the previous moment, is the cognitive inertia coefficient, is the original controller output value, i.e., the controller at time The output value of .

10. A hotel unmanned check-in system based on face recognition, characterized in that: The method for detecting an unattended hotel stay based on face recognition according to any one of claims 1 to 9 comprises: Environmental perception and risk factor extraction module, which is used to obtain multimodal environmental data, extract environmental risk factors, form environmental risk vectors, and calculate comprehensive environmental risk scores; Verification entropy calculation and strategy generation module, used to calculate the verification entropy value based on the environmental risk score, establish a verification mode set, execute the verification entropy optimization algorithm, and form a verification strategy table; A bio-cognitive synchronous control module is used to input the environmental risk score into the bio-cognitive synchronous control algorithm, generate the controller output value, convert it into a dynamic verification parameter, perform cognitive depth adjustment, and generate a verification execution instruction set; The face recognition verification module is used to execute the face recognition verification process according to the verification execution instruction set.

Images