Hotel check-in method, device and equipment based on face recognition and storage medium

Abstract

The invention relates to the technical field of smart hotel access control, and discloses a hotel check-in method, device and equipment based on face recognition and a storage medium, and the method comprises the steps: collecting radio frequency coupling strength and communication response time delay data during the interaction period of a radio frequency card reading module and a first near field communication unit of an intelligent card, resolving an environment coupling fluctuation value and a protocol time sequence fluctuation value, and configuring a dynamic isolation duration according to the environment coupling fluctuation value and the protocol time sequence fluctuation value so as to delay activation of the second near field communication unit; calculating an interaction stability parameter according to the identification interval duration and the fluctuation value, and mapping the identity label of the first near field communication unit into a preset face orientation angle; collecting a face deflection angle of a current user, calculating an angle matching score, and dynamically adjusting an angle allowable deviation range and a stability weight by utilizing an interaction stability parameter; and finally calculating an access control composite value to drive the door lock to open.

Description

Technical Field

[0001] This invention relates to the field of smart hotel access control technology, and more specifically, to a hotel check-in method, apparatus, equipment, and storage medium based on facial recognition. Background Technology

[0002] With the development of smart hotels, contactless radio frequency identification (RFID / NFC) and facial recognition technologies are widely used in guest room door lock systems. Existing high-security check-in processes typically require guests to possess a room key and cooperate with facial verification. To enhance anti-counterfeiting capabilities and the interactive experience, some smart room keys have adopted embedded dual communication units (dual chips), attempting to trigger different verification logic by detecting the activation order of the two communication units (e.g., recognizing unit A before recognizing unit B, or vice versa), such as requiring the user to cooperate in completing a liveness detection of the left or right side of their face.

[0003] However, in actual hotel engineering scenarios, the above solution faces challenges. First, hotel door lock panels are often made of metals such as zinc alloys. This material environment can cause severe de-resonance effects in near-field communication in the 13.56MHz band, leading to decreased and uneven antenna coupling efficiency. Second, when a room card with dual communication units enters the reader's sensing area, it triggers the anti-collision and card selection process according to the ISO / IEC-14443 standard. Under metallic interference, the reader's selection of the two communication units no longer solely depends on the user's card-waving direction, but is affected by both environmental coupling fluctuations and the randomness of communication protocol time slots.

[0004] Furthermore, guests often make brief, paused movements during check-in or door opening, such as quickly waving their cards or covering their faces with their hands. During these rapid and unstable operations, the signal strength and response latency of the card reader fluctuate significantly. If the system relies solely on a simple sequence of recognition to determine the user's intent and trigger corresponding facial orientation verification, signal drift can easily lead to misjudgments (for example, if a user wants to trigger left-face verification, the system might read the wrong chip first due to signal fluctuations and request right-face verification instead). This can result in verification failure or inability to open the door, severely impacting the guest's check-in experience and access efficiency. Summary of the Invention

[0005] This invention provides a hotel check-in method, apparatus, equipment, and storage medium based on facial recognition, which solves the technical problems mentioned in the background art.

[0006] Firstly, a hotel check-in method based on facial recognition is applied to an access control system that includes an RFID card reader module and a camera acquisition module, including: During the establishment of a communication link between the RF card reader module and the first near-field communication unit in the smart card, the RF coupling strength dataset and the communication response delay dataset are collected, the environmental coupling fluctuation value and the protocol timing fluctuation value are calculated, and the dynamic isolation duration is configured according to the environmental coupling fluctuation value and the protocol timing fluctuation value to delay the activation operation of the RF card reader module on the second near-field communication unit in the smart card. Based on the identification interval between the first near-field communication unit and the second near-field communication unit by the radio frequency card reader module, as well as the environmental coupling fluctuation value and the protocol timing fluctuation value, the interaction stability parameter is calculated, and the identity identifier of the first near-field communication unit is converted into an activation sequence code to map and generate a preset facial orientation angle. The camera acquisition module acquires the current user's facial deflection angle, calculates the angle matching score for the preset facial orientation angle, and dynamically adjusts the allowable deviation range and stability weight using the interaction stability parameter. The access control composite value is calculated by combining the stability weight, angle matching score and face similarity. When the access control composite value meets the system threshold condition, an opening command is sent to the door lock drive unit.

[0007] Secondly, a hotel check-in device based on facial recognition, applied to an access control system including an RFID card reader module and a camera acquisition module, implements the steps of any of the facial recognition-based hotel check-in methods described above, including: The dynamic isolation control module is used to collect radio frequency coupling strength dataset and communication response delay dataset during the period when the radio frequency card reader module establishes a communication link with the first near-field communication unit in the smart card, calculate the environmental coupling fluctuation value and the protocol timing fluctuation value, and configure the dynamic isolation duration according to the environmental coupling fluctuation value and the protocol timing fluctuation value to delay the activation operation of the radio frequency card reader module on the second near-field communication unit in the smart card. The parameter calculation and mapping module is used to calculate the interaction stability parameters based on the identification interval of the first near-field communication unit and the second near-field communication unit by the radio frequency card reader module and the environmental coupling fluctuation value and protocol timing fluctuation value, and to convert the identity of the first near-field communication unit into the activation order code to map and generate a preset facial orientation angle. The visual analysis and adjustment module is used to acquire the current user's facial deflection angle through the camera acquisition module, calculate the angle matching score for the preset facial orientation angle, and dynamically adjust the allowable deviation range and stability weight of the angle using the interaction stability parameter. The access control decision execution module is used to calculate the access control composite value by combining the stability weight, angle matching score and face similarity. When the access control composite value meets the system threshold condition, it sends an opening command to the door lock drive unit.

[0008] Thirdly, an electronic device includes: Memory, processor, RFID card reader module, and camera acquisition module; The memory is used to store computer programs; The processor is used to execute the computer program to implement the steps of any of the facial recognition-based hotel check-in methods described above.

[0009] Fourthly, a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the hotel check-in method based on facial recognition as described in any one of the claims. Attached Figure Description

[0010] Figure 1 This is a flowchart of the hotel check-in method based on facial recognition of the present invention; Figure 2 This is a schematic diagram of a specific implementation scenario of the present invention. Detailed Implementation

[0011] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0012] Example 1: As Figure 1 As shown, a hotel check-in method based on facial recognition is applied to an access control system that includes an RFID card reader module and a camera acquisition module, comprising: During the establishment of a communication link between the RF card reader module and the first near-field communication unit in the smart card, the RF coupling strength dataset and the communication response delay dataset are collected, the environmental coupling fluctuation value and the protocol timing fluctuation value are calculated, and the dynamic isolation duration is configured according to the environmental coupling fluctuation value and the protocol timing fluctuation value to delay the activation operation of the RF card reader module on the second near-field communication unit in the smart card. Based on the identification interval between the first near-field communication unit and the second near-field communication unit by the radio frequency card reader module, as well as the environmental coupling fluctuation value and the protocol timing fluctuation value, the interaction stability parameter is calculated, and the identity identifier of the first near-field communication unit is converted into an activation sequence code to map and generate a preset facial orientation angle. The camera acquisition module acquires the current user's facial deflection angle, calculates the angle matching score for the preset facial orientation angle, and dynamically adjusts the allowable deviation range and stability weight using the interaction stability parameter. The access control composite value is calculated by combining the stability weight, angle matching score and face similarity. When the access control composite value meets the system threshold condition, an opening command is sent to the door lock drive unit.

[0013] The access control system, comprising an RFID card reader module and a camera acquisition module, consists of an RFID card reader module, a camera acquisition module, a main control processor, a memory, and a door lock drive unit. It is specifically designed to achieve collaborative access control using dual-chip smart card recognition and facial recognition. The RFID card reader module conforms to the ISO / IEC-14443 standard and features near-field communication signal transmission and reception, received signal strength detection, and communication response delay recording. It can establish a link with a smart card embedded with dual near-field communication units, collect RFID coupling strength datasets and communication response delay datasets, and activate and identify the two near-field communication units sequentially by sending suspend commands and controlling dynamic isolation durations, providing foundational data for subsequent calculation of interaction stability parameters. The camera acquisition module is fixed near the door lock panel and supports real-time image capture including the user's face. It extracts two-dimensional facial key points and matches them with preset three-dimensional head model key points, calculating the user's facial deflection angle and providing visual data support for angle matching scoring and facial feature extraction. The main control processor is responsible for executing the core algorithms, including solving environmental coupling fluctuation values, protocol timing fluctuation values, and interaction stability parameters; dynamically adjusting the allowable deviation range of angles and stability weights; calculating angle matching scores, face similarity, and access control composite values; and finally determining whether to send an opening command to the door lock drive unit through a step function. The memory is used to store smart card chip identifiers, guest face template vectors, system preset parameters, and computer programs to ensure the continuity and reliability of data processing and command execution throughout the entire process. The overall system achieves a highly secure and adaptable access control verification function in hotel scenarios.

[0014] Preferably, during the establishment of a communication link between the RF card reader module and the first near-field communication unit in the smart card, RF coupling strength datasets and communication response delay datasets are collected, and environmental coupling fluctuation values ​​and protocol timing fluctuation values ​​are calculated, including: Set a fixed sampling time window With sampling frequency Determine the number of samples to be collected. ; In the fixed sampling time window The radio frequency coupling strength dataset is collected using the following formula. With communication response delay dataset : ; ; in, The sampling point number, The value of the received signal strength register. For the maximum range of the register, For the first The moment when the response to the next sample is interrupted. For the first The start time of sending the sampling instruction. This is the conversion factor for microseconds to milliseconds; The environmental coupling fluctuation value is calculated using the following formula. With protocol timing fluctuation value : ; ; in, This represents the function for calculating sample variance. The normalized communication response delay dataset is calculated using the following formula: , This is a preset reference delay scale.

[0015] The fixed sampling time window is the continuous sampling period set by the RFID reader module when acquiring data. The preferred value is 200 milliseconds, based on the typical interaction time of 13.56 MHz near-field communication, which can balance sampling integrity and response speed.

[0016] The sampling frequency is the number of samples taken per unit of time. A preferred value is 200 Hz, based on the consideration of balancing data density and hardware processing capabilities, and avoiding data redundancy or missing data.

[0017] The number of samples is the total number of data points collected within a fixed sampling time window, calculated from the fixed sampling time window and the sampling frequency.

[0018] The value in the received signal strength register is the value corresponding to the register in the RF reader module that stores the strength of the received signal. It can be obtained by reading the registers of RF front-end chips that support received signal strength measurement, such as the ST25R3916B.

[0019] The command transmission start time is the point in time when the RFID reader module sends a communication command to the smart card. This can be obtained by recording a timestamp at the moment the command is sent using a microsecond-level hardware timer.

[0020] The interrupt response time is the moment when the RFID reader module receives the smart card's feedback response signal and triggers the interrupt. This can be obtained by recording a timestamp at the instant the interrupt is triggered using a microsecond-level hardware timer.

[0021] The preset reference delay scale is a baseline value used to normalize communication response delay data. A preferred value of 5 milliseconds is used, based on the typical response delay range of near-field communication protocols.

[0022] The radio frequency coupling strength dataset is a set of continuous data formed by normalizing the received signal strength register values, reflecting the coupling stability of radio frequency communication.

[0023] The communication response delay dataset is a set of continuous data formed by converting the time difference between the start time of instruction transmission and the interrupt time of response reception, reflecting the timing characteristics of the communication protocol.

[0024] The normalized communication response delay dataset is standardized data obtained by dividing the communication response delay dataset by a preset reference delay scale, and is used to eliminate the influence of dimensions.

[0025] The environmental coupling fluctuation value is the square root of the variance of the radio frequency coupling strength dataset, used to quantify the degree of coupling fluctuation caused by environmental factors.

[0026] The protocol timing fluctuation value is the square root of the variance of the normalized communication response delay dataset, used to quantify the degree of timing fluctuation caused by the communication protocol.

[0027] The value range of the received signal strength register is typically 0 to 15. Dividing this value at each sampling point by 15 normalizes the signal strength to between 0 and 1, forming an RF coupling strength dataset. For example, if the register value at a sampling point is 12, the normalized result is 0.8.

[0028] A microsecond-level counter can achieve a timing accuracy of 1 microsecond. After recording the time difference between sending a command and receiving a response, dividing by 1000 converts the unit to milliseconds, which conforms to engineering time measurement practices. For example, a time difference of 800 microseconds is converted to 0.8 milliseconds, becoming a data point in the communication response delay dataset.

[0029] The square root of variance, or standard deviation, directly reflects the degree of dispersion of data. Calculating the standard deviation of a dataset of radio frequency coupling strength directly quantifies the magnitude of coupling fluctuations caused by factors such as metallic environments; a larger value indicates more severe fluctuations.

[0030] Communication response delay is affected by protocol mechanisms, and its numerical range may vary. Normalizing by dividing by a preset reference delay scale of 5 milliseconds before calculating the standard deviation ensures that the dimensions of protocol timing fluctuations are consistent, facilitating analysis in conjunction with other fluctuation values. For example, a delay data point of 10 milliseconds, after normalization, becomes 2, which is then used in variance calculation.

[0031] The fixed sampling time window is preferably set to 200 milliseconds. This value is based on the typical interaction time of 13.56 MHz near-field communication. It can fully collect the signal changes during a communication process without causing response delay due to excessive time, which is suitable for the fast interaction scenario of hotel door locks.

[0032] The sampling frequency is preferably set to 200 Hz, which means sampling once every 5 milliseconds. This frequency can capture subtle changes in signal strength and response delay, while not generating too much redundant data to increase the hardware processing pressure, which is in line with the processing capabilities of conventional RF modules.

[0033] The preset reference delay scale is preferably set to 5 milliseconds, referencing the typical response delay range of near-field communication in the ISO / IEC-14443 standard, to ensure that the normalized delay data is within a reasonable range of 0 to 2, and to avoid extreme values ​​affecting the fluctuation calculation results.

[0034] The hardware model adapted to the received signal strength register is the mainstream RF front-end chip such as ST25R3916B. The maximum range of the register of this type of chip is designed to be 15, which is a common industry standard. The value is reasonable and universal, and those skilled in the art can directly select the adapted hardware.

[0035] Preferably, a dynamic isolation duration is configured based on the environmental coupling fluctuation value and the protocol timing fluctuation value to delay the activation operation of the second near-field communication unit in the smart card by the RF card reader module, including: The radio frequency coupling strength dataset is used according to the following formula. Calculate motion disturbance value : ; ; in, For the first Rate of change of each sample The adjacent sampling interval is equal to the sampling frequency. The reciprocal, This is a preset scaling constant; Calculate the total jitter scale using the following formula. : ; in, The environmental coupling fluctuation value, This refers to the timing fluctuation value of the protocol. The minimum positive number is preset. The dynamic isolation duration is calculated using the following formula. : ; in, To preset the basic beat duration, This is the preset beat delay ratio coefficient; Send a suspend command to the first near-field communication unit in the smart card, and wait for the dynamic isolation time. After completion, the activation operation of the second near-field communication unit in the smart card is initiated.

[0036] The rate of change between adjacent sampling times is the ratio of the numerical difference between two consecutive sampling points in the RF coupling strength dataset to the adjacent sampling interval, reflecting how fast the coupling strength changes instantaneously.

[0037] The preset scaling constant is a baseline constant used to convert motion disturbance-related calculation results into dimensionless values. A value of 50 is preferred, based on the numerical range of the radio frequency coupling strength change rate, ensuring that the motion disturbance value is within a reasonable calculation range.

[0038] The motion disturbance value is an indicator that quantifies the instability of a guest's card-waving action. It is obtained by adjusting the root mean square value of the rate of change of the radio frequency coupling strength at adjacent sampling times by a preset scaling constant.

[0039] Total jitter is a unified quantitative indicator of fluctuations that integrates environmental coupling fluctuations, protocol timing fluctuations, action disturbances, and extremely small positive numbers. It reflects the combined effect of various factors that affect communication stability.

[0040] The preset beat delay ratio is a parameter that converts the total jitter scale into a time delay. A preferred value is 250 milliseconds, chosen to balance jitter compensation and communication response speed, adapting to the interactive characteristics of near-field communication.

[0041] The preset base clock duration is the baseline time for dynamic isolation. A preferred value of 350 milliseconds is used, based on the minimum interaction period for near-field communication in the ISO / IEC-14443 standard, to ensure that the base latency meets the chip's state switching requirements.

[0042] The dynamic isolation duration is the specific time length for delaying the activation of the second near-field communication unit, used to widen the time interval between two chip recognitions and avoid signal conflicts.

[0043] The adjacent sampling interval is the reciprocal of the sampling frequency. For example, when the sampling frequency is 200 Hz, the interval is 5 milliseconds. The rate of change is obtained by subtracting the previous sampling value from the current sampling point's RF coupling strength value, and then dividing by the interval. For example, if a sampling point has a value of 0.8, the previous sampling point had a value of 0.6, and the interval is 5 milliseconds, the rate of change is (0.8-0.6) / 0.005=40. The square root of the average square of all rates of change is then calculated to obtain the root mean square value. Dividing this by a preset scaling constant of 50 yields the final motion disturbance value. This logic can accurately quantify the instability of the card-waving motion.

[0044] The environmental coupling fluctuation value reflects the influence of the metal environment, the protocol timing fluctuation value reflects the randomness of the communication protocol, and the action disturbance value reflects the instability of the card waving action. The three can be directly summed to add up the effects of various fluctuations. The addition of a preset minimum positive number is to avoid the calculation abnormality of the denominator being 0. This synthesis method can comprehensively reflect the overall jitter level of the system.

[0045] The preset base beat duration of 350 milliseconds ensures the minimum time requirement for chip state switching. The total jitter scale multiplied by a scaling factor of 250 milliseconds can dynamically increase the delay time according to the actual fluctuation situation. For example, if the total jitter scale is 0.4, the dynamic isolation time is 350 + 250 × 0.4 = 450 milliseconds. Through this dynamic adjustment, the interference of fluctuations on the two recognition sequences can be effectively offset.

[0046] The suspend command uses the HLTA command in the ISO / IEC-14443-3 standard. The command frame format is a fixed binary sequence. After being sent, the first near-field communication unit will enter a sleep state and stop responding to the reader's polling commands, thereby ensuring that the reader will subsequently activate the second near-field communication unit and avoid identification conflicts caused by the two chips responding at the same time.

[0047] The default scaling constant is set to 50. This value is determined based on the typical range of the rate of change of the radio frequency coupling strength. When the root mean square value of the rate of change is between 0 and 50, the motion disturbance value will fall within a reasonable range of 0 to 1, which is convenient for superimposing and calculating with other fluctuation values ​​and conforms to the quantitative habits in engineering.

[0048] The default minimum positive number is 10 to the power of -6. This value is small enough that it will not affect the calculation result of the total jitter scale. At the same time, it can completely avoid the abnormal situation of the denominator being 0 in subsequent division operations. It is a commonly used protective value in engineering calculations.

[0049] The preset beat delay ratio coefficient is preferably set to 250 milliseconds. This value can match the adjustment range of the dynamic isolation time with the total jitter scale, so that the delay is not too short to offset the fluctuation, nor too long to affect the user interaction experience.

[0050] The preset basic beat duration is preferably set to 350 milliseconds. This duration is based on the state switching time of mainstream near-field communication chips, which ensures that the first near-field communication unit has enough time to complete the state transition after receiving the suspend command, thus avoiding signal interference when the second chip is activated later.

[0051] The suspend command uses the HLTA command specified in the ISO / IEC-14443-3 standard. The command frame consists of a start bit, a command code, and a check bit. The command code is a fixed hexadecimal value. After the card reader sends this command, the first near-field communication unit will stop sending response signals until it receives a wake-up command.

[0052] Preferably, based on the identification interval between the first near-field communication unit and the second near-field communication unit by the RFID reader module, and the environmental coupling fluctuation value and protocol timing fluctuation value, the interaction stability parameters are calculated, including: During the activation and identification of the second near-field communication unit in the smart card by the RF card reader module, the second RF coupling strength dataset is collected. With the second communication response delay dataset ; The comprehensive environmental coupling fluctuation value is calculated using the following formula. Integrated protocol timing fluctuation value and comprehensive motion disturbance value : ; ; ; in, , , These are the environmental coupling fluctuation value, the protocol timing fluctuation value, and the action disturbance value, respectively. This is the variance calculation function; The normalized second communication response delay dataset is calculated using the following formula: ; For based on Based on the calculated second motion disturbance value; The interaction stability parameter is calculated using the following formula. : ; in, The recognition interval duration, As a preset time normalization reference constant, It is a preset minimum positive number.

[0053] The second RF coupling strength dataset is a normalized continuous signal strength dataset collected when the RF reader module activates the second near-field communication unit, reflecting the coupling stability of the second communication. It can be obtained by reading the received signal strength register of an RF front-end chip such as the ST25R3916B and normalizing it.

[0054] The second communication response delay dataset is a continuous dataset formed by converting the time difference between the instruction transmission start time and the response reception interrupt time when the second near-field communication unit is activated. It reflects the timing characteristics of the second communication. It can be obtained by recording timestamps using a microsecond-level hardware timer and then converting the data.

[0055] The second action disturbance value is an indicator that quantifies the instability of the card-waving action during the second communication, and is calculated based on the second radio frequency coupling strength dataset.

[0056] The comprehensive environmental coupling fluctuation value is a quantitative indicator that superimposes the environmental coupling fluctuations of the first and second communications, reflecting the comprehensive interference of the entire environment on the coupling.

[0057] The overall protocol timing fluctuation value is a quantitative indicator that combines the timing fluctuations of the first and second communications, reflecting the overall interference of the protocol on timing throughout the entire process.

[0058] The overall motion disturbance value is a quantitative indicator that superimposes the effects of motion disturbances during the two communications, reflecting the overall instability of the entire card-waving action.

[0059] The identification interval is the time difference between the identification of the first near-field communication unit and the second near-field communication unit by the RFID reader module. It can be obtained by recording the two identification times and calculating the difference using a microsecond-level hardware timer.

[0060] The preset time normalization reference constant is a baseline constant used to standardize the identification interval to a dimensionless value. A preferred value of 500 milliseconds is chosen, based on the typical interval range of two identifications in near-field communication, to ensure the normalized value is reasonable.

[0061] The normalized interval is a standardized value obtained by dividing the identification interval length by a preset time normalization reference constant, and is used to eliminate the influence of time units.

[0062] Interaction stability parameter is a core indicator for quantifying the controllability of the two chip recognition sequences, reflecting the ease or difficulty for users to control the recognition sequence through actions.

[0063] The square of the environmental coupling fluctuation value represents the environmental coupling fluctuation energy of the first communication, and the variance of the second radio frequency coupling strength dataset represents the environmental coupling fluctuation energy of the second communication. Adding the two together and taking the square root allows for the superposition and quantification of the two fluctuation energies, yielding the comprehensive environmental coupling fluctuation value. For example, if the environmental coupling fluctuation value is 0.3, and the variance of the second radio frequency coupling strength dataset is 0.04, the sum of the squares (0.09 + 0.04) equals 0.13. Taking the square root, the comprehensive environmental coupling fluctuation value is approximately 0.36.

[0064] The second communication response delay dataset is first normalized by dividing it by a preset reference delay scale of 5 milliseconds, and then the variance is calculated to obtain the protocol timing fluctuation energy of the second communication. This energy is then added to the square of the protocol timing fluctuation value of the first communication and the square root is taken to achieve the superposition of the two protocol timing fluctuations, ensuring uniformity of dimensions. For example, if the protocol timing fluctuation value is 0.2, the variance of the normalized second communication response delay dataset is 0.03, the sum of squares is 0.04 plus 0.03 equals 0.07, and the combined protocol timing fluctuation value after taking the square root is approximately 0.26.

[0065] The square of the motion disturbance value represents the motion disturbance energy of the first communication, and the square of the second motion disturbance value represents the motion disturbance energy of the second communication. Adding them together and taking the square root provides a comprehensive quantification of the motion instability during the two communications, consistent with the engineering calculation logic of superimposing the energy of random variables. For example, if the first motion disturbance value is 0.25 and the second motion disturbance value is 0.2, the sum of their squares is 0.0625 plus 0.04 equals 0.1025, and the square root of this sum results in a comprehensive motion disturbance value of approximately 0.32.

[0066] The identification interval is divided by a preset time normalization reference constant of 500 milliseconds to obtain a dimensionless normalized interval. This normalized interval is then divided by the sum of the integrated environmental coupling fluctuation value, the integrated protocol timing fluctuation value, the integrated action disturbance value, and a preset minimum positive number to form the interaction stability parameter. This model correlates the identification interval with the integrated fluctuations to quantify the controllability of the sequence. For example, if the normalized interval is 1, the integrated fluctuation sum is 0.5, and the interaction stability parameter is 2.

[0067] The sampling time window and sampling frequency of the second radio frequency coupling strength dataset are consistent with those of the first radio frequency coupling strength dataset, which are 200 milliseconds and 200 Hz, respectively. This ensures that the time density and length of the two sampling data are consistent, which facilitates subsequent fluctuation superposition calculation and meets the consistency requirements of data acquisition.

[0068] The normalization reference scale of the second communication response delay dataset is consistent with that of the first communication response delay dataset, which is 5 milliseconds. This maintains a unified normalization standard, avoids fluctuation quantization errors caused by different reference scales, and ensures the comparability of the two time series data.

[0069] The preset time normalization reference constant is preferably set to 500 milliseconds. This value is based on the typical time interval between two card swipe recognitions in a hotel scenario, which can keep the normalization interval in a reasonable range of 0.5 to 2 in most scenarios, avoiding extreme values ​​from affecting the calculation of interaction stability parameters.

[0070] The calculation of the second motion disturbance value is based on a preset scaling constant D, which is set to 50. This standardizes the calculation of motion disturbance values ​​and ensures that the dimensions of the two motion disturbance data are consistent. The combined motion disturbance value after superposition can accurately reflect the degree of instability of the entire motion.

[0071] Preferably, the identity identifier of the first near-field communication unit is converted into an activation sequence code to map and generate a preset facial orientation angle, including: Pre-store the first chip identifier With the second chip identifier and the corresponding first sequence target angle Angle of the second sequence target ; The activation sequence code is determined according to the following formula. : ; in, This serves as the identifier for the first near-field communication unit. For positive cell values, Negative unit value; The preset facial orientation angle is generated using the following formula. : ; in, The first linear interpolation weights are... The weights are for the second linear interpolation.

[0072] The first chip identifier is a unique identification symbol for the first near-field communication unit in the smart card. The preferred value is the standard unique identifier format for the near-field communication chip, conforming to the ISO / IEC-14443 standard to ensure the universality and uniqueness of identification.

[0073] The second chip identifier is a unique identification symbol for the second near-field communication unit in the smart card. It is preferably chosen to be a unique identifier for the near-field communication standard that differs from the first chip identifier, in order to distinguish the two communication units and avoid identity confusion.

[0074] The first sequence target angle is the preset facial orientation angle corresponding to the first activation of the first near-field communication unit. The preferred value is any value among -30 degrees, 0 degrees, and +30 degrees, based on the fact that this angle range can clearly distinguish the left face, front face, and right face, which is suitable for the engineering implementation of head pose estimation.

[0075] The second sequence target angle is the preset facial orientation angle corresponding to the first activation of the second near-field communication unit. The preferred value is a value that is different from the first sequence target angle among -30 degrees, 0 degrees, and +30 degrees, based on the requirement that different facial orientations need to be collected for the two recognitions.

[0076] The identifier of the first near-field communication unit is a unique symbol of the near-field communication unit first identified by the RFID card reader module. It can be obtained by performing the anti-collision card selection process of the ISO / IEC-14443-3 standard through the RFID card reader module.

[0077] The positive cell value is the encoded value used to identify the first near-field communication unit to be activated. It is preferentially set to 1 to simplify the encoding logic and facilitate subsequent linear interpolation weight calculation.

[0078] The negative cell value is a coded value used to identify the second near-field communication unit that is activated first. It is preferentially set to negative 1 to clearly distinguish it from the positive cell value and ensure the uniqueness of the sequential encoding.

[0079] The activation order code is a code that reflects the identity of the first near-field communication unit. It takes a positive or negative unit value and is used to associate the corresponding preset facial orientation angle.

[0080] The first linear interpolation weight is a coefficient calculated based on the activation order encoding and is used to allocate the proportion of the first sequence target angle in the preset facial orientation angle.

[0081] The second linear interpolation weight is a coefficient calculated based on the activation order encoding and is used to allocate the proportion of the second sequence target angle in the preset facial orientation angle.

[0082] The preset facial orientation angle is a standard value of facial orientation that the user needs to present, determined according to the recognition order, and is used for subsequent angle matching score calculation.

[0083] After the RFID reader module reads the identity identifier of the first near-field communication unit, it compares it byte-by-byte with the pre-stored first chip identifier and second chip identifier. If they match the first chip identifier exactly, the activation sequence code is assigned a value of 1; if they match the second chip identifier exactly, the value is assigned a negative 1. For example, the first chip identifier is a 7-byte standard unique identifier. When the read identity identifier matches this value exactly, the code is 1. This logic implements a direct mapping from chip identity to sequence code.

[0084] When the activation order code is 1, the first linear interpolation weight is calculated as (1 + 1) divided by 2, which equals 1, and the second linear interpolation weight is calculated as (1 - 1) divided by 2, which equals 0. When the code is negative 1, the first weight is 0 and the second weight is 1. This design achieves selective activation of the target angles of two sequences through arithmetic operations, without the need for complex logical judgments.

[0085] When the activation order is encoded as 1, the preset facial orientation angle is equal to the first sequence target angle multiplied by 1 plus the second sequence target angle multiplied by 0, which equals the first sequence target angle. When the encoding is negative 1, it equals the second sequence target angle. For example, if the first sequence target angle is 0 degrees and the second is positive 30 degrees, the preset angle is 0 degrees when the encoding is 1 and positive 30 degrees when the encoding is negative 1. This model achieves angle mapping.

[0086] The target angles of the first and second sequences are fixed at -30 degrees, 0 degrees, and +30 degrees, corresponding to the left, front, and right sides of the face, respectively. This range is determined based on engineering practices of head pose estimation, which facilitates user cooperation while ensuring the distinguishability of different angles and avoiding recognition confusion due to excessively small angle differences.

[0087] The first and second chip identifiers adopt the 7-byte unique identifier format specified in the ISO / IEC-14443 standard and are stored in the configuration table of the door lock controller, bound to the corresponding room information and user facial information. Encrypted storage is used to ensure that the identifier information cannot be tampered with.

[0088] The identification of the first near-field communication unit follows the anti-collision procedure of the ISO / IEC-14443-3 standard. First, a wake-up command is sent, then the first near-field communication unit is selected through the anti-collision algorithm, and its unique identifier is read. The verification logic includes identifier length verification and consistency verification with the pre-stored identifier to ensure the accuracy of the reading result.

[0089] The difference between the first and second sequence target angles is constrained to an absolute value of no less than 30 degrees. For example, if the first sequence target angle is 0 degrees, the second sequence target angle can only be -30 degrees or +30 degrees. This constraint ensures that there is a significant difference in facial orientation between the two recognitions, meeting the core requirements of the verification process.

[0090] Preferably, the current user's facial deflection angle is acquired through a camera acquisition module, and an angle matching score is calculated for the preset facial orientation angle, including: The camera acquisition module captures images containing the user's face, extracts two-dimensional key points, and calculates the rotation matrix by combining these with preset three-dimensional head model key points. The facial deflection angle is extracted using the following formula. : ; in, and For the rotation matrix matrix elements, It is the arctangent function in the four quadrants; The angle matching score is calculated using the following formula. : ; in, The preset facial orientation angle, The allowable deviation range for the angle is... It is an exponential function with the natural constant as its base.

[0091] The user's facial image is image data containing the user's complete face, captured by the camera acquisition module. It can be obtained in real time through a high-definition camera module on the door lock panel.

[0092] Two-dimensional keypoints are sets of two-dimensional coordinate points extracted from a user's facial image that reflect the location of facial features. They can be detected and extracted from facial images using open-source algorithms such as the Dlib library.

[0093] The preset key points of the 3D head model are the coordinates of feature points on the pre-defined 3D human head model. The preferred values ​​are 6 standard points, including the left and right corners of the eyes, the left and right corners of the mouth, the tip of the nose, and the tip of the chin, based on the need to adapt to the fast solution requirements of the PnP algorithm, while taking into account both accuracy and efficiency.

[0094] The coordinate system of the camera acquisition module defines the spatial position of the camera acquisition module. A right-handed coordinate system is preferred because it conforms to common standards in the field of computer vision and facilitates algorithm compatibility.

[0095] The rotation matrix is ​​a 3x3 matrix that represents the spatial orientation of the user's head relative to the coordinate system of the camera acquisition module, reflecting the rotation state of the head.

[0096] The elements of a rotation matrix are the values ​​at specific positions within the rotation matrix. R21 is the element in the second row and first column of the matrix, and R11 is the element in the first row and first column of the matrix.

[0097] Facial yaw angle is the angle of rotation of the user's head in the horizontal direction, used to represent the orientation of the face, i.e., left or right face or frontal face.

[0098] The allowable deviation range of the angle is the range of error within which the facial deflection angle is judged to meet the requirements. The preferred value is 6 to 10 degrees, based on the balance between recognition accuracy and the difficulty of user cooperation.

[0099] The preset constant 2 is a fixed value used to calculate the normalization factor, namely the number 2.

[0100] The deviation squared term is the square of the difference between the facial deflection angle and the preset facial orientation angle, used to quantify the degree of angular deviation.

[0101] The normalization factor is a calculation factor used for the squared term of the standardized deviation, which is obtained by multiplying the square of the allowable angular deviation range by a preset constant.

[0102] The Gaussian exponent is the negative of the squared deviation term divided by the normalization factor, and serves as the exponential input of the Gaussian function.

[0103] Angle matching score is an indicator that quantifies the degree to which the current facial deflection angle matches the preset facial orientation angle, with a value ranging from 0 to 1.

[0104] First, 68 2D keypoints on the face are extracted using algorithms such as the Dlib library. Six points corresponding to keypoints in a pre-defined 3D head model are then selected. The coordinates of the 2D and 3D keypoints are input into a PnP algorithm to obtain the head's rotation and translation vectors. Then, a Rodrigues transform is used to convert the rotation vectors into a 3x3 rotation matrix, which fully represents the head's spatial pose. For example, when the user's face is directly facing the camera, the rotation matrix is ​​an identity matrix, with each element conforming to the definition of an identity matrix.

[0105] The four-quadrant arctangent function can handle coordinate values ​​from different quadrants, avoiding the limitations of angle calculation range. By using the R21 element of the rotation matrix as the ordinate and the R11 element as the abscissa, inputting it into this function directly yields facial deflection angles from -180 degrees to 180 degrees. For example, when R21 is 0.2 and R11 is 0.98, the calculated facial deflection angle is approximately 11.5 degrees, accurately reflecting the horizontal rotation of the head.

[0106] The Gaussian function, with its continuous and smooth characteristics, can convert angular deviations into a continuous score between 0 and 1. The smaller the angular deviation, the closer the score is to 1, indicating a higher degree of fit; the larger the deviation, the closer the score is to 0. For example, if the preset facial orientation angle is 0 degrees, the user's facial deflection angle is 3 degrees, and the allowable angular deviation range is 6 degrees, the angle matching score calculated by substituting into the formula is approximately 0.77, which intuitively reflects the degree of fit.

[0107] The preset 3D head model uses 6 standard rigid points as key points, with coordinates based on a general 3D human head model. For example, the coordinates of the nose tip are 0,0,0; the coordinates of the left and right eye corners are -30,0,20 and 30,0,20 respectively; the coordinates of the left and right mouth corners are -25,-30,15 and 25,-30,15 respectively; and the coordinates of the chin tip are 0,-60,0. The model type is a simplified homogeneous head model, free from interference factors such as hair and facial expressions, and is suitable for rigid transformation solutions using the PnP algorithm.

[0108] The camera acquisition module's coordinate system is defined as a right-handed coordinate system, with the origin located at the optical center of the camera lens. The X-axis is parallel to the lens horizontally to the right, the Y-axis is parallel to the lens vertically downwards, and the Z-axis is along the lens's optical axis forward, conforming to the OpenCV coordinate system standard.

[0109] The allowable angular deviation range is dynamically adjusted by the interactive stability parameter, with a default value of 8 degrees when there is no initial adjustment signal. Its minimum tolerance is fixed at 6 degrees, and the variable tolerance range is 10 degrees.

[0110] Two-dimensional facial key points were extracted using the 68-point detection algorithm from the Dlib library, an industry-standard solution. The 68 key points cover 17 facial contour points, 5 points each for the left and right eyebrows, 6 points each for the left and right eyes, 9 points for the nose, 12 points for the mouth, and 1 point for the chin, which can completely represent facial features and meet the accuracy requirements of pose calculation.

[0111] Preferably, the allowable deviation range of the angle and the stability weight are dynamically adjusted using the interactive stability parameter. The access control composite value is calculated by combining the stability weight, angle matching score, and face similarity. When the access control composite value meets the system threshold condition, an opening command is sent to the door lock drive unit, including: Calculate the permissible deviation range of the angle using the following formula. : ; in, To preset the minimum tolerance constant, To preset the variable tolerance range, For the preset attenuation coefficient, For the interaction stability parameter, It is a natural constant; The stability weight is calculated using the following formula. : ; in, The preset slope coefficient, To preset the center point of the parameters, It is an exponential function with the natural constant as its base; Calculate the access control composite value using the following formula. : ; in, This is the smart card validity verification value; it is set to this value when the smart card verification passes. Otherwise, the value is ; The stability weight; Score the angle matching. The formula for calculating facial feature similarity is: , The facial feature vector of the current user. For pre-stored guest registration template vectors; The activation command is generated according to the following formula. : ; in, To enable the preset system threshold, For step function, when hour This indicates that an open command is sent to the door lock drive unit; otherwise... This indicates that the action will not be performed.

[0112] The preset attenuation coefficient is a factor that adjusts the degree to which the interactive stability parameter affects the allowable range of angle deviation. A value of 1.5 is preferred, based on the principle of balancing adjustment sensitivity and tolerance stability to avoid excessive tolerance fluctuations.

[0113] The preset variable tolerance range is the upper limit of the dynamic adjustment of the allowable deviation range of the angle. The preferred value is 10 degrees, which is based on the reasonable fluctuation range of the facial deflection angle, taking into account both recognition accuracy and user experience.

[0114] The preset minimum tolerance constant is a fixed lower limit of the allowable angular deviation range. A value of 6 degrees is preferred to ensure minimum recognition accuracy and avoid misjudgment due to excessive tolerance.

[0115] The preset slope coefficient is a factor that adjusts the rate at which the interaction stability parameter affects the stability weight. A value of 3.0 is preferred, based on the principle of ensuring the weight changes smoothly within a reasonable range to adapt to different stability scenarios.

[0116] The preset parameter center point is the benchmark point for changes in stability weights. The preferred value is 1.0, based on the typical value range of the reference interaction stability parameter, so that the weight adjustment is more in line with the actual scenario.

[0117] The stability weight is a coefficient that quantifies the influence of the controllability of the two chip recognition sequences on the unlocking decision, and its value ranges from 0 to 1.

[0118] The smart card validity verification value is a binary indicator representing the legitimacy of the smart card. It can be obtained by verifying the room number, validity period, and encrypted verification code on the card through an RFID reader module.

[0119] The current user's facial feature vector is a high-dimensional numerical vector extracted from the user's facial image, used to represent facial features. It can be extracted from facial images using deep learning networks such as ResNet.

[0120] The pre-stored guest registration template vector is the guest's facial feature vector stored at check-in time, serving as a benchmark for identity verification. The preferred value is a 512-dimensional vector with the same dimensions as the current user's facial feature vector, based on compliance with mainstream industry feature extraction standards.

[0121] Facial feature similarity is an indicator of the degree of matching between the current user's facial feature vector and the pre-stored template vector, with a value ranging from -1 to 1.

[0122] Normalized face similarity is a standardized index that maps facial feature similarity to a value between 0 and 1, making it easier to integrate with other scoring factors for calculation.

[0123] The access control composite value is a core indicator for unlocking decisions that integrates smart card validity, stability weights, angle matching scores, and normalized facial similarity.

[0124] The preset system unlock threshold is the critical value for determining whether to allow unlocking. A value of 0.55 is preferred, based on the optimal balance between security and availability verified through extensive experiments.

[0125] The step function is a decision function used to determine whether the composite value of the access control meets the opening conditions. The input is a numerical value, and the output is 0 or 1.

[0126] The unlock command is a control signal that drives the door lock to perform the unlocking action. When the output is 1, the unlocking is performed; when the output is 0, the unlocking is not performed.

[0127] In the dynamic adjustment logic of the allowable deviation range of the angle, the higher the interaction stability parameter, the smaller the exponent value, the smaller the dynamic tolerance, and the closer the total tolerance is to the preset minimum tolerance constant; conversely, the lower the dynamic tolerance, the larger the total tolerance, and the closer the total tolerance is to the sum of the minimum tolerance and the variable tolerance amplitude. For example, when the interaction stability parameter is 2.0, the exponent is -1.5 × 2.0 = -3.0, the dynamic tolerance is approximately 10 × 0.05 = 0.5 degrees, and the total tolerance is 6 + 0.5 = 6.5 degrees, which is suitable for scenarios with high controllability of the adaptation sequence.

[0128] In the sigmoid mapping model with stability weights, the weight is 0.5 when the interaction stability parameter equals the preset parameter center point of 1.0; when the parameter is higher than 1.0, the weight approaches 1, strengthening the contribution of sequence controllability to unlocking; when the parameter is lower than 1.0, the weight approaches 0, weakening the influence of unstable sequences. For example, when the interaction stability parameter is 2.0, the weight is approximately 0.95, fully recognizing the effectiveness of stable sequences.

[0129] In the multi-factor multiplication logic of the access control composite value, when the smart card validity verification value is 0, the composite value is directly 0, and unlocking is refused; when all other factors are 1, the composite value reaches its maximum. For example, if the smart card verification is successful (value 1), the stability weight is 0.95, the angle matching score is 0.8, and the normalized face similarity is 0.9, the composite value is 1 × 0.95 × 0.8 × 0.9 = 0.684.

[0130] In the branchless decision logic of the step function, when the combined access control value is greater than or equal to the preset system opening threshold of 0.55, the output is 1, and an opening command is sent; otherwise, the output is 0, and no command is sent. This logic avoids complex branch judgments, ensuring fast and consistent decision-making.

[0131] The preset attenuation coefficient is preferably set to 1.5, the preset variable tolerance amplitude is preferably set to 10 degrees, and the preset minimum tolerance constant is preferably set to 6 degrees. The combination of these three values ​​allows the allowable angle deviation range to be dynamically adjusted between 6 and 10 degrees, adapting to different interactive stability scenarios.

[0132] The preset slope coefficient is preferably set to 3.0, and the preset parameter center point is preferably set to 1.0. This ensures that the stability weight smoothly transitions from 0.05 to 0.95 when the interactive stability parameter is between 0 and 2.0. The adjustment logic is clear and reproducible.

[0133] The facial feature vector has a dimension of 512. The extraction algorithm uses the ResNet50 deep learning network, which is a common solution in the industry and can efficiently extract stable facial features to meet the accuracy requirements of identity recognition.

[0134] The pre-stored guest registration template vector is stored in binary format, with the same dimension as the current user's facial feature vector, and is stored in the encrypted storage area of ​​the door lock controller; the update logic is that when a guest checks in, facial features are re-collected and the original template is overwritten.

[0135] The preset system unlock threshold is set to 0.55. This value has been determined through extensive scenario testing and can effectively prevent unauthorized unlocking while reducing the verification failure rate of legitimate users.

[0136] The smart card validity verification standard includes three items: the room number stored in the card is consistent with the room number bound to the door lock; the card's validity period is within the current time range; and the encryption verification uses the AES algorithm, with the verification code being consistent with the pre-stored code of the door lock.

[0137] Specifically, when guests check in, staff will install a smart card with embedded first and second near-field communication (NFC) units. Simultaneously, the device will collect the guest's facial information, generating and storing a pre-stored guest registration template vector. Staff will also assign two different sets of target facial angles to the guest: a first sequence target angle and a second sequence target angle. These correspond to the activation order of the two NFC units in the smart card. For example, if the first NFC unit is activated first, it corresponds to a frontal face angle; if the second NFC unit is activated first, it corresponds to a right-facing face angle. The two sets of angles are clearly different to ensure verification validity.

[0138] After arriving at the room door, the guest brings their smart card close to the RFID reader module on the door lock. The RFID reader module first establishes a communication link with a near-field communication unit (NFC) on the smart card (this unit is the first NFC unit). During this process, the RFID reader module collects data related to the RF coupling strength and communication response delay. This data reflects the interference of the hotel's metal door lock environment on communication and the timing fluctuations of the communication protocol itself. The system configures a dynamic isolation duration based on this data. The purpose is to delay the activation of the other NFC unit (the second NFC unit) to avoid identification conflicts caused by simultaneous responses from both units. For example, when the user quickly waves the card, the dynamic isolation duration provides sufficient time for the chip state to switch.

[0139] After the dynamic isolation period ends, the RFID reader module activates and identifies the second near-field communication unit, while recording the interval between the two identifications (i.e., the identification interval). The system combines the previously collected fluctuation data and the identification interval to calculate an interaction stability parameter. This parameter is used to determine the controllability of the chip identification sequence; for example, the parameter value is more stable when the user's card-waving action is smooth. Subsequently, the system compares the identity identifier of the first near-field communication unit with the pre-stored first and second chip identifiers, converting this into an activation order code. Then, it generates a preset facial orientation angle through code mapping. For example, when the first near-field communication unit is the pre-stored first chip identifier, a preset angle for a frontal face is generated.

[0140] The door lock's camera module simultaneously captures the guest's facial image, extracts 2D facial key points from the image, and combines them with preset 3D head model key points to calculate the guest's facial deflection angle, i.e., the actual facial orientation. The system calculates the degree of fit between this actual angle and the preset facial orientation angle, obtaining an angle matching score. Simultaneously, it dynamically adjusts the allowable deviation range of the angle based on the interaction stability parameter; that is, the more stable the parameter, the stricter the deviation range, and vice versa. For example, when the recognition sequence fluctuates significantly, a slightly larger deviation in the guest's facial orientation is allowed to improve the user experience.

[0141] At the same time, the system will verify the legitimacy of the smart card. The verification includes whether the room number stored on the card matches the room number bound to the door lock, and whether the card's validity period is within the current time range. If the verification passes, the smart card's validity verification value is 1; otherwise, it is 0. Simultaneously, the system extracts the current user's facial feature vector and compares it with the pre-stored guest registration template vector to obtain facial feature similarity, which is then normalized.

[0142] Finally, the system combines the smart card validity verification value, the stability weight calculated from the interaction stability parameter, the angle matching score, and the normalized face similarity to obtain the access control composite value. When this composite value reaches the preset system opening threshold, the system sends an opening command to the door lock drive unit, the door lock performs the unlocking action, and the guest can enter the room; if the composite value does not meet the threshold, no opening command is sent, and the door lock remains locked.

[0143] like Figure 2 As shown, Figure 2 The core components, interaction relationships, and key parameters of a hotel check-in access control system based on facial recognition are presented. The radio frequency / near field communication card reader is used to establish a communication link with the near field communication unit A and near field communication unit B embedded in the smart card. The camera acquisition module is responsible for capturing the user's facial image to obtain the current deflection angle. The indicator light is used to provide feedback on the system's operation or verification status. The dynamic isolation duration is a time parameter configured by the system to delay the activation of the second near field communication unit. The preset orientation angle is the target facial angle bound to the chip activation order. The door lock drive unit receives system commands to execute the unlocking action.

[0144] Example 2: A hotel check-in device based on facial recognition, applied to an access control system including an RFID card reader module and a camera acquisition module, implementing the steps of any of the facial recognition-based hotel check-in methods described above, including: The dynamic isolation control module is used to collect radio frequency coupling strength dataset and communication response delay dataset during the period when the radio frequency card reader module establishes a communication link with the first near-field communication unit in the smart card, calculate the environmental coupling fluctuation value and the protocol timing fluctuation value, and configure the dynamic isolation duration according to the environmental coupling fluctuation value and the protocol timing fluctuation value to delay the activation operation of the radio frequency card reader module on the second near-field communication unit in the smart card. The parameter calculation and mapping module is used to calculate the interaction stability parameters based on the identification interval of the first near-field communication unit and the second near-field communication unit by the radio frequency card reader module and the environmental coupling fluctuation value and protocol timing fluctuation value, and to convert the identity of the first near-field communication unit into the activation order code to map and generate a preset facial orientation angle. The visual analysis and adjustment module is used to acquire the current user's facial deflection angle through the camera acquisition module, calculate the angle matching score for the preset facial orientation angle, and dynamically adjust the allowable deviation range and stability weight of the angle using the interaction stability parameter. The access control decision execution module is used to calculate the access control composite value by combining the stability weight, angle matching score and face similarity. When the access control composite value meets the system threshold condition, it sends an opening command to the door lock drive unit.

[0145] Example 3: An electronic device, comprising: Memory, processor, RFID card reader module, and camera acquisition module; The memory is used to store computer programs; The processor is used to execute the computer program to implement the steps of any of the facial recognition-based hotel check-in methods described above.

[0146] Example 4: A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the above-described hotel check-in methods based on face recognition.

[0147] It should be noted that the user privacy-related parameters involved in this solution include, but are not limited to, pre-stored guest registration template vectors, current user facial feature vectors, user facial images, facial 2D key points, facial deflection angles, and other facial biometric information, as well as smart card-related identity information such as the first chip identifier, second chip identifier, and first near-field communication unit identifier, and guest-related information such as room number and validity period. The collection, uploading, and use of all privacy parameters are contingent upon the user's explicit consent. Guests can choose whether to authorize the relevant information for access control verification during check-in, and the authorization process is fully traceable and auditable. The parameter uploading process uses an encrypted transmission protocol, and the storage phase uses encrypted storage. Additional security measures, such as de-identification processing and access control, are implemented for sensitive privacy data such as facial biometric information. The entire implementation process is solely for hotel access control verification, and unauthorized use, illegal disclosure, or provision to third parties is strictly prohibited.

[0148] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A hotel check-in method based on facial recognition, applied to an access control system including an RFID card reader module and a camera acquisition module, characterized in that, include: During the establishment of a communication link between the RF card reader module and the first near-field communication unit in the smart card, the RF coupling strength dataset and the communication response delay dataset are collected, the environmental coupling fluctuation value and the protocol timing fluctuation value are calculated, and the dynamic isolation duration is configured according to the environmental coupling fluctuation value and the protocol timing fluctuation value to delay the activation operation of the RF card reader module on the second near-field communication unit in the smart card. Based on the identification interval between the first near-field communication unit and the second near-field communication unit by the radio frequency card reader module, as well as the environmental coupling fluctuation value and the protocol timing fluctuation value, the interaction stability parameter is calculated, and the identity identifier of the first near-field communication unit is converted into an activation sequence code to map and generate a preset facial orientation angle. The camera acquisition module acquires the current user's facial deflection angle, calculates the angle matching score for the preset facial orientation angle, and dynamically adjusts the allowable deviation range and stability weight using the interaction stability parameter. The access control composite value is calculated by combining the stability weight, angle matching score and face similarity. When the access control composite value meets the system threshold condition, an opening command is sent to the door lock drive unit.

2. The hotel check-in method based on facial recognition according to claim 1, characterized in that, During the establishment of a communication link between the RF card reader module and the first near-field communication unit in the smart card, RF coupling strength datasets and communication response delay datasets are collected, and environmental coupling fluctuation values ​​and protocol timing fluctuation values ​​are calculated, including: Set a fixed sampling time window and sampling frequency, and determine the number of samples to be sampled; Within the fixed sampling time window, the radio frequency coupling strength dataset is obtained by reading the received signal strength register of the radio frequency card reader module and performing normalization processing. A microsecond-level counter is used to record the time difference from the start of instruction transmission to the interruption of response reception, thereby obtaining a communication response delay dataset. Calculate the square root of the variance of the radio frequency coupling strength dataset as the environmental coupling fluctuation value; The communication response delay dataset is normalized by dividing it by a preset reference delay scale, and the square root of the variance of the normalized communication response delay dataset is calculated as the protocol timing fluctuation value.

3. The hotel check-in method based on facial recognition according to claim 2, characterized in that, Based on the environmental coupling fluctuation value and the protocol timing fluctuation value, a dynamic isolation duration is configured to delay the activation operation of the second near-field communication unit in the smart card by the RF card reader module, including: The rate of change between adjacent sampling times is calculated using the radio frequency coupling strength dataset, and the root mean square value of the rate of change is calculated. After adjustment by a preset scaling constant, the motion disturbance value is obtained. The total jitter scale is obtained by summing the environmental coupling fluctuation value, the protocol timing fluctuation value, the action disturbance value, and a preset minimum positive number. Multiply the total jitter scale by the preset beat delay ratio coefficient and add it to the preset basic beat duration to obtain the dynamic isolation duration; A suspend command is sent to the first near-field communication unit in the smart card, and after the dynamic isolation period has ended, the activation operation of the second near-field communication unit in the smart card is initiated.

4. The hotel check-in method based on facial recognition according to claim 3, characterized in that, Based on the identification interval between the first and second near-field communication units by the RFID reader module, and the environmental coupling fluctuation value and protocol timing fluctuation value, the interaction stability parameters are calculated, including: During the activation and identification of the second near-field communication unit in the smart card by the radio frequency card reader module, the second radio frequency coupling strength dataset and the second communication response delay dataset are collected. Calculate the variance of the second radio frequency coupling strength dataset, add it to the square of the environmental coupling fluctuation value, and take the square root to obtain the comprehensive environmental coupling fluctuation value; The second communication response delay dataset is normalized and its variance is calculated. The variance is then added to the square of the protocol timing fluctuation value and the square root is taken to obtain the comprehensive protocol timing fluctuation value. The second motion disturbance value is calculated based on the second radio frequency coupling strength dataset, and the square root of the sum of the square of the motion disturbance value and the square of the second motion disturbance value is calculated to obtain the comprehensive motion disturbance value. The normalized interval is obtained by dividing the identification interval by a preset time normalization reference constant, and the normalized interval is then divided by the sum of the integrated environmental coupling fluctuation value, the integrated protocol timing fluctuation value, the integrated action disturbance value, and a preset minimum positive number to obtain the interaction stability parameter.

5. The hotel check-in method based on facial recognition according to claim 4, characterized in that, The identity identifier of the first near-field communication unit is converted into an activation order code to map and generate a preset facial orientation angle, including: The first chip identifier and the second chip identifier are pre-stored, and the first sequence target angle and the second sequence target angle are set. The identity identifier of the first near-field communication unit is compared with the first chip identifier and the second chip identifier. If they match the first chip identifier, the activation order code is assigned a positive unit value; if they match the second chip identifier, the activation order code is assigned a negative unit value. The first linear interpolation weight and the second linear interpolation weight are calculated based on the activation order code, wherein the first linear interpolation weight is half of the sum of the activation order code and the second linear interpolation weight is half of the difference between the activation order code and the activation order code. Calculate the first product of the first sequence target angle and the first linear interpolation weight, and the second product of the second sequence target angle and the second linear interpolation weight, and add the first product and the second product to obtain the preset facial orientation angle.

6. The hotel check-in method based on facial recognition according to claim 5, characterized in that, The camera acquisition module acquires the current user's facial deflection angle and calculates an angle matching score for the preset facial orientation angle, including: The camera acquisition module acquires an image containing the user's face, extracts two-dimensional key points, and combines them with the preset three-dimensional head model key points to calculate the rotation matrix relative to the coordinate system of the camera acquisition module. The elements of the rotation matrix are used to calculate the arctangent value in the four quadrants to obtain the facial deflection angle. Calculate the square of the difference between the facial deflection angle and the preset facial orientation angle, and use it as the deviation squared term; Calculate the product of the square of the allowable deviation range of the angle and a preset constant two, and use it as a normalization factor; The negative of the quotient of the squared deviation term divided by the normalization factor is used as the Gaussian exponent. The angle matching score is obtained by calculating the function value with the natural constant as the base and the Gaussian exponent as the exponent.

7. The hotel check-in method based on facial recognition according to claim 6, characterized in that, The allowable deviation range of the angle and the stability weight are dynamically adjusted using the interactive stability parameters. The access control composite value is calculated by combining the stability weight, angle matching score, and face similarity. When the access control composite value meets the system threshold condition, an opening command is sent to the door lock drive unit, including: Calculate the negative of the product of the interactive stability parameter and the preset attenuation coefficient, and use it as the first exponential term; calculate the function value with the natural constant as the base and the first exponential term as the exponent, and multiply the function value by the preset variable tolerance range, and then add it to the preset minimum tolerance constant to obtain the allowable angular deviation range; Calculate the difference between the interaction stability parameter and the preset parameter center point, and take the negative of the product of the difference and the preset slope coefficient as the second exponential term; calculate the function value with the natural constant as the base and the second exponential term as the exponent, and add the function value to the constant to obtain the intermediate sum value, and calculate the reciprocal of the intermediate sum value to obtain the stability weight; Perform a validity check on the smart card. If the check passes, set the smart card validity check value to one; otherwise, set it to zero. Extract the current user's facial feature vector and the pre-stored guest registration template vector, and calculate the cosine similarity as the facial feature similarity; add the facial feature similarity to a constant one and divide by a constant two to obtain the normalized facial similarity; The combined value of the access control system is obtained by multiplying the smart card validity verification value, the stability weight, the angle matching score, and the normalized face similarity. The difference between the access control composite value and the preset system opening threshold is calculated, and the difference is judged using a step function. When the difference is greater than or equal to zero, an opening command is generated and sent to the door lock drive unit.

8. A hotel check-in device based on facial recognition, applied to an access control system including an RFID card reader module and a camera acquisition module, implementing the steps of the hotel check-in method based on facial recognition as described in any one of claims 1-7, characterized in that, include: The dynamic isolation control module is used to collect radio frequency coupling strength dataset and communication response delay dataset during the period when the radio frequency card reader module establishes a communication link with the first near-field communication unit in the smart card, calculate the environmental coupling fluctuation value and the protocol timing fluctuation value, and configure the dynamic isolation duration according to the environmental coupling fluctuation value and the protocol timing fluctuation value to delay the activation operation of the radio frequency card reader module on the second near-field communication unit in the smart card. The parameter calculation and mapping module is used to calculate the interaction stability parameters based on the identification interval of the first near-field communication unit and the second near-field communication unit by the radio frequency card reader module and the environmental coupling fluctuation value and protocol timing fluctuation value, and to convert the identity of the first near-field communication unit into the activation order code to map and generate a preset facial orientation angle. The visual analysis and adjustment module is used to acquire the current user's facial deflection angle through the camera acquisition module, calculate the angle matching score for the preset facial orientation angle, and dynamically adjust the allowable deviation range and stability weight of the angle using the interaction stability parameter. The access control decision execution module is used to calculate the access control composite value by combining the stability weight, angle matching score and face similarity. When the access control composite value meets the system threshold condition, it sends an opening command to the door lock drive unit.

9. An electronic device, characterized in that, include: Memory, processor, RFID card reader module, and camera acquisition module; The memory is used to store computer programs; The processor is used to execute the computer program to implement the steps of the hotel check-in method based on face recognition as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the hotel check-in method based on face recognition as described in any one of claims 1-7.

Images