Big data analysis-based second gate physical and mental health screening method

Abstract

This invention discloses a method for screening physical and mental health at two-gate access control systems based on big data analysis, belonging to the field of intelligent access control technology. The method involves acquiring, filtering, and re-collecting verification data of target personnel at each gate port within the two-gate area. Based on the unique identifier of each target person, a spatiotemporal correlation data collection record set is constructed. The data within this record set is then aligned along a timeline to extract multi-dimensional physiological features and generate independent physical and mental health scores for each target person. These scores are then aggregated across all gate ports for multi-port collaborative verification, identifying collaborative anomalies. The distribution patterns of all individuals with collaborative anomalies within the two-gate area are statistically analyzed to identify hotspot ports and intersection ports for group anomalies, generating dynamic access control commands. This method achieves multi-port data correlation verification and hierarchical management of individual and group anomalies.

Description

Technical Field

[0001] This invention relates to the field of intelligent access control technology, specifically a two-door physical and mental health screening method based on big data analysis. Background Technology

[0002] The field of physical and mental health screening technology includes fundamental components such as monitoring equipment, sensors, data transmission protocols, and processing platforms. With real-time environmental status perception, multi-source data integration and analysis, and abnormal event early warning and response as its core, the system architecture typically consists of a front-end data acquisition layer, a network transmission layer, a data storage layer, and an application service layer. In recent years, there has been a growing trend of integrating health assessment functions into access control systems.

[0003] However, existing technologies still have the following shortcomings in practical applications: each gate port operates independently, and data from the same person on different ports cannot be automatically correlated and merged, forming data silos; the system lacks real-time verification of the validity of the collected data, as people may be looking down, turning their faces to the side, or passing through quickly, making assessments based on low-quality data unreliable; the sampling frequencies of facial images, body surface temperature, vibration, and other data are different, and existing technologies have not solved the problem of time synchronization of multi-source heterogeneous data; the system only performs single-point anomaly detection and cannot distinguish between single-port misjudgments and real anomalies; and it lacks the ability to collaboratively perceive group abnormal events, making it difficult to detect abnormal symptoms in multiple people at the same time.

[0004] Therefore, in response to the above problems, there is an urgent need for a two-stage physical and mental health screening method based on big data analysis. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a two-tiered physical and mental health screening method based on big data analysis, which solves the problems of isolated and uncorrelated multi-port data in two-tiered screening scenarios, lack of verification of collection quality, inability to synchronize heterogeneous data, difficulty in distinguishing between single-point misjudgment and real abnormalities, and inability to perceive group abnormalities.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a two-gate physical and mental health screening method based on big data analysis, comprising the following steps: acquiring target personnel verification data at each gate port within the two-gate area; filtering and re-collecting the target personnel verification data; uploading the qualified data after filtering and re-collection to the central management port; the central management port, based on the unique identity of the target personnel, retrieves qualified data from all gate ports verified by the target personnel within a preset time window, constructs a spatiotemporal correlation collection record set of the target personnel, and aligns the data in the spatiotemporal correlation collection record set with the time axis; extracting multidimensional physiological features from the spatiotemporal correlation collection record set after time axis alignment, generating independent physical and mental state scores of the target personnel at each gate port, aggregating the independent physical and mental state scores of all gate ports verified by the target personnel within the preset time window for multi-port collaborative verification, and marking collaborative anomalies; statistically analyzing the distribution patterns of all personnel marked as having collaborative anomalies within the two-gate area, identifying hotspot ports and intersection ports of group anomalies, and generating dynamic access control instructions based on the multi-port collaborative verification results and the identification status of hotspot ports and intersection ports of group anomalies.

[0007] Furthermore, the screening and re-collection process includes: the target personnel verification data includes facial image data, body surface temperature data, and vibration data; multi-dimensional judgment is performed using the facial image data, including image clarity judgment, integrity judgment of key facial features, and facial posture angle judgment; when there are cases where multi-dimensional judgments are unqualified, a re-collection command is issued to the gate port, and the target personnel are prompted by voice to adjust their facial angle or the acquisition distance to re-collect facial image data; if, after re-collection attempts within a preset re-collection time window, the re-collected facial image data is still unqualified after multi-dimensional judgment, the acquisition record of that gate port is discarded, and a voice prompt is issued to the target personnel to switch gate ports to re-collect facial image data.

[0008] Furthermore, the time axis alignment includes: for the spatiotemporal correlation acquisition record set of the target personnel, taking the facial image data acquisition time as the reference time axis, for scenarios where the sampling frequency of body surface temperature data and vibration data is lower than the sampling frequency of facial images, extracting two adjacent acquisition time points before and after the reference time point, and calculating the fitted data values ​​of body surface temperature data and vibration data at the reference time point through linear interpolation; using the fitted data values ​​as data gaps on the reference time axis, so that facial image data, body surface temperature data, and vibration data have corresponding correlation data at each reference time point.

[0009] Furthermore, the extraction steps of the multidimensional physiological features include: acquiring facial image data, body surface temperature data, and vibration data at each reference time point within a preset acquisition window; analyzing facial muscle features in the facial image data at each reference time point, and extracting the contraction frequency and amplitude features of the muscles around the eyes and mouth; extracting the steady-state value and temperature fluctuation trend of the body surface temperature data at each reference time point; extracting the vibration spectrum features of the facial and neck regions in the vibration data at each reference time point, and identifying specific vibration components; and combining the contraction frequency and amplitude features of the muscles around the eyes and mouth, the steady-state value and temperature fluctuation trend of the body temperature, and the specific vibration components to form multidimensional physiological features for assessing physical and mental state.

[0010] Furthermore, the physical and mental state assessment adopts an intelligent recognition model trained on big data. The intelligent recognition model learns from physiological characteristic samples of known healthy and known abnormal populations to establish a mapping relationship between physiological appearance and internal physical and mental state. The multidimensional physiological characteristics of the physical and mental state assessment are input into the intelligent recognition model, and the independent physical and mental state score of the target person at each gate port is output. The independent physical and mental state score refers to the physical and mental health status of the target person assessed at a single gate port.

[0011] Furthermore, the multi-port collaborative verification includes: comparing the independent physical and mental state scores of all turnstile ports in the two-door area verified by the target person within a preset time window with preset scoring benchmark values; marking turnstile ports with independent physical and mental state scores lower than the scoring benchmark values ​​as single-port abnormal records; when single-port abnormal records exist, extracting the physical and mental state assessment records of other turnstile ports verified by the target person within the preset time window (excluding the current turnstile port), and obtaining the proportion of the number of single-port abnormal records to the total number of turnstile ports passed by the target person; when the proportion exceeds or equals a preset abnormal proportion, marking it as a collaborative abnormality; when the proportion is lower than the preset abnormal proportion, marking the single-port abnormal record as a single-point suspected abnormality.

[0012] Furthermore, the identification of abnormal hotspot ports and converging ports includes: obtaining the number of all personnel identified as having coordinated abnormalities at each gate port within a preset time window's two-gate area; comparing the number with a preset abnormal group threshold; marking gate ports with a number exceeding the abnormal group threshold as abnormal hotspot ports, indicating that the gate port has a clustering risk; obtaining the gate ports pointed to by all the spatial passage paths of personnel with coordinated abnormalities; and marking the gate port as a converging port when the number of points to the same gate port exceeds a preset number threshold.

[0013] Furthermore, the dynamic access control commands include: generating an interception command and triggering a voice warning for coordinated abnormal personnel; generating a monitoring command for single suspected abnormal personnel, pushing the personnel's unique identifier to the central management port, which then distributes the personnel's unique identifier to all turnstile ports for focused monitoring, wherein focused monitoring refers to enhancing the data collection strategy to a preset high-alert strategy when the turnstile port receives a verification request for the unique identifier; and generating an area control command for group abnormal hotspot ports and converging ports, closing the verification of the turnstile port, and using voice guidance to direct personnel to switch turnstile ports for verification.

[0014] Furthermore, the time axis alignment also includes data cleaning: before performing the time axis alignment operation, invalid impact signals that exceed the range of human physiological response in the vibration data and extreme outliers caused by external heat source interference in the body surface temperature data are screened out; the associated data at each reference time point after time axis alignment is smoothed to eliminate data jitter when missing data is introduced.

[0015] Furthermore, it also includes generating a source tracing report for abnormal group events: based on the spatiotemporal distribution characteristics of abnormal group hotspot ports and convergence ports, the starting area and key diffusion period of the abnormal event are determined, and then a handling recommendation report containing the starting area and key diffusion period of the abnormal event is generated.

[0016] The present invention has the following beneficial effects: This two-gate physical and mental health screening method based on big data analysis retrieves qualified data from all gate ports verified by the target person within a preset time window through a central management port based on the target person's unique identity. It then constructs a spatiotemporal correlation collection record set, which solves the problem that each gate port operates independently and the data of the same person on different ports cannot be automatically correlated and merged. This allows the data of the same person passing through multiple ports to form a complete spatiotemporal physical and mental health profile.

[0017] By performing multi-dimensional judgments on facial image data in the target personnel verification data, and issuing re-collection instructions and voice prompts to the target personnel to adjust their posture and re-collect when the judgment is unqualified, the problem of lacking a real-time verification mechanism for data validity is solved, and erroneous control decisions are avoided based on low-quality data.

[0018] By using the sampling time point of facial image data as the reference time axis, linear interpolation is used to align the sampling data in body surface temperature data and vibration data that are not aligned with the reference time axis, thus solving the data synchronization and alignment problem caused by different sampling frequencies of different sensors.

[0019] By co-validating the physical and mental status assessment results of a single port with those of other turnstile ports, a collaborative anomaly is marked when the proportion of abnormal records in a single port to the total number of turnstile ports exceeds a preset anomaly threshold. This solves the problem of misjudgment in single ports and improves the accuracy of anomaly detection.

[0020] By statistically analyzing the number of coordinated abnormal individuals at each gate port, hotspot ports for group anomalies can be identified. Furthermore, by overlaying and analyzing the spatial passage paths of all coordinated abnormal individuals, intersecting ports can be identified. This enables the timely detection of group events in which multiple people exhibit abnormal symptoms simultaneously, providing a basis for decision-making in emergency response.

[0021] By generating interception commands for coordinated abnormal personnel, monitoring commands for suspected abnormal personnel at single points, and area control commands for abnormal hotspot ports and intersection ports in groups, dynamic access control with hierarchical and categorized management is achieved.

[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0023] Figure 1 This is a flowchart of the two-stage physical and mental health screening method based on big data analysis according to the present invention.

[0024] Figure 2 This is a flowchart of step S1 in the two-stage physical and mental health screening method based on big data analysis of the present invention.

[0025] Figure 3 This is a flowchart of step S2 in the two-stage physical and mental health screening method based on big data analysis of the present invention.

[0026] Figure 4 This is a flowchart of step S3 in the two-stage physical and mental health screening method based on big data analysis of the present invention.

[0027] Figure 5 This is a flowchart of step S4 in the two-stage physical and mental health screening method based on big data analysis of the present invention. Detailed Implementation

[0028] This application embodiment uses a two-tiered physical and mental health screening method based on big data analysis to realize a complete screening scheme that associates multi-port data with identity identifiers and constructs a spatiotemporal collection record set. Through quality verification, time synchronization, multi-port collaborative verification, and group anomaly perception, it generates hierarchical and classified dynamic access control instructions.

[0029] The problem addressed in this application's embodiments can be summarized as follows: Each gate port within the second-level access area is treated as a data acquisition terminal. Qualified data from all ports is aggregated through a central management port. The unique identification of the target personnel is used as the association identifier to achieve automatic association and fusion of cross-port data. Specifically, a real-time data quality verification and re-sampling mechanism ensures that all data entering the evaluation process is valid. A multi-node data spatiotemporal association mechanism solves the time synchronization problem of heterogeneous data. A multi-port collaborative verification and group anomaly perception mechanism solves the problems of single-point misjudgment and group anomaly event identification. Based on the evaluation results, hierarchical and categorized dynamic access control instructions are generated.

[0030] Please see Figure 1 This invention provides a technical solution: a two-tiered physical and mental health screening method based on big data analysis, comprising the following steps: Step S1: Obtain the target personnel verification data of each gate port in the second gate area, filter and re-collect the target personnel verification data, and upload the qualified data after filtering and re-collection to the central management port.

[0031] Step S2: Based on the unique identity of the target personnel, the central management port retrieves the qualified data of all gate ports verified by the target personnel within a preset time window, constructs a spatiotemporal correlation collection record set of the target personnel, and aligns the data in the spatiotemporal correlation collection record set with the time axis.

[0032] Step S3: Extract multidimensional physiological features from the spatiotemporal correlation acquisition record set after time axis alignment, generate independent physical and mental state scores of the target personnel at each gate port, aggregate the independent physical and mental state scores of the target personnel at all gate ports verified within the preset time window for multi-port collaborative verification, and mark collaborative anomalies.

[0033] Step S4: Statistically analyze the distribution pattern of all personnel marked as having collaborative anomalies within the second-door area, identify the hotspot ports and intersection ports of the group anomalies, and generate dynamic access control instructions based on the multi-port collaborative verification results and the identification status of the hotspot ports and intersection ports of the group anomalies.

[0034] The above steps will be analyzed and expanded in detail below in the context of specific application scenarios.

[0035] Suppose a large industrial park has two gates at its entrance. Within these gates, four turnstiles are arranged from left to right, numbered Turnstile A, Turnstile B, Turnstile C, and Turnstile D. Each turnstile is equipped with a facial recognition camera, an infrared thermometer, and a vibration sensor. Park employees must pass through these two gates to enter and exit the park daily. The following detailed description of the specific implementation of this invention will focus on the process of an employee passing through the two gates.

[0036] Please see Figure 2 Step S1: Obtain the target personnel verification data collected by each gate port in the second gate area, and ensure that the data entering the subsequent processing process are all qualified data through screening and re-collection mechanisms.

[0037] S11, Deployment and Data Acquisition Configuration of Turnstile Ports: Deploy turnstile ports at each passage location in the two-gate area. Each turnstile port is equipped with a data acquisition terminal, which includes at least a facial image acquisition device, a body surface temperature acquisition device, and a vibration acquisition device.

[0038] The facial image acquisition device uses a high-definition camera to continuously acquire facial image data of the target person. Its frame rate is set to no less than 25 frames per second to ensure the capture of dynamic features such as facial muscle contractions and changes in the curvature of the corners of the mouth. The body surface temperature acquisition device is set to sample once every 5 seconds; the vibration acquisition device is set to sample 5 times per second. It should be noted that the facial image acquisition device, body surface temperature acquisition device, and vibration acquisition device can all be adjusted by those skilled in the art.

[0039] Each gate port is equipped with a clock synchronization device to periodically calibrate with the reference clock of the central management port, so that the time deviation between the local clock of each gate port and the reference clock of the central management port does not exceed a preset synchronization error threshold. The synchronization error threshold can be set to no more than 10 milliseconds to ensure that the timestamps of the data collected by each port are comparable when the data is correlated in the future. The specific settings can be preset by professionals in the field.

[0040] S12, Target Personnel Verification Data Collection Process: When a target personnel arrives at any gate in the second-level access area, the facial recognition device at that gate first identifies the target personnel. The facial recognition device captures the target personnel's facial image and compares it with a pre-stored personnel identity database to determine the target personnel's unique identifier. The unique identifier is a code that uniquely distinguishes different target personnel within the system; for example, it can be an employee number, employee ID card, or employee mobile phone number.

[0041] After successful identity verification, the data acquisition terminal at the gate port initiates data collection to obtain a set of verification data for the target person at that port. The verification data includes at least the following: Facial image data consists of facial images continuously captured by a facial image acquisition device as the target person passes through the gate port. Each frame of facial image is accompanied by a timestamp of the acquisition of that frame.

[0042] Body surface temperature data is a sequence of body surface temperature values ​​collected by a body surface temperature acquisition device as the target person passes through the gate port. Each body surface temperature value is accompanied by a timestamp of the acquisition.

[0043] Vibration data consists of a sequence of vibration signals collected by a vibration acquisition device as a target person passes through the gate port. Each vibration signal value is accompanied by a timestamp of the acquisition.

[0044] S13, Verification Data Screening and Re-collection: Each gate port sends the collected verification data to the central management port. After receiving the target personnel verification data sent by each gate port, the central management port extracts facial image data from each set of verification data and uses the facial image data to perform multi-dimensional judgment to determine whether the set of verification data is qualified.

[0045] The multi-dimensional assessment includes image sharpness assessment, facial key feature integrity assessment, and facial pose and angle assessment, among which: The image sharpness judgment analysis is as follows: The central management port calculates the sharpness index for each frame of facial image. Specifically, it calculates the change range of gray values ​​between adjacent pixels in the image. The larger the change range, the sharper the image. The central management port takes the average of the sharpness index of all frames to obtain the image sharpness value of the set of verification data. The image sharpness value is compared with the preset sharpness threshold. If it is lower than the sharpness threshold, it is judged that the image sharpness is unqualified. The sharpness threshold is a pre-set critical value used to distinguish between image sharpness and blurriness. For example, the sharpness threshold can be set to a certain value. When the average change range of gray values ​​between adjacent pixels in the image is lower than this value, it indicates that the image is blurry overall and the details are not clear.

[0046] The sharpness threshold is set using an adaptive adjustment method based on ambient light intensity. The central management port collects the current ambient light intensity value in real time through an ambient light sensor installed at the gate port. When the ambient light intensity is lower than the preset low light threshold, for example, below 50 lux, it indicates that the current ambient light is dark, the overall grayscale value of the image is low, and the sharpness index naturally decreases. At this time, the central management port lowers the sharpness threshold by a preset adjustment step, for example, by 10%. When the ambient light intensity is higher than the preset high light threshold, for example, above 1000 lux, it indicates that the current ambient light is too strong and may cause overexposure. At this time, the central management port raises the sharpness threshold by a preset adjustment step, for example, by 10%. When the ambient light intensity is within the normal range, the sharpness threshold remains unchanged. Through the above adaptive adjustment, the sharpness threshold is matched with the current ambient light conditions, avoiding misjudgment caused by changes in ambient light. The adjustment step and the light threshold can be preset by those skilled in the art according to the actual application scenario.

[0047] The specific analysis for determining the completeness of key facial features is as follows: The central management port performs facial region detection on each frame of facial image to determine the proportion of the facial region in the image and whether each key facial feature is completely presented within the image area. Key facial features include the eye region, mouth region, and nose region. If the proportion of the facial region in a frame is lower than a preset proportion threshold, or if any key feature in the eye, mouth, or nose region is not fully presented, then that frame is marked as an incomplete frame. The proportion threshold is a pre-set critical proportion value used to determine whether the facial region completely occupies the image area; for example, it can be set to a proportion of at least 60% of the total image area. The central management port counts the proportion of incomplete frames in all frames. If this proportion exceeds the preset incomplete frame proportion threshold, then the completeness of the key facial features is deemed unqualified. The incomplete frame proportion threshold is a pre-set upper limit for the proportion of incomplete frames allowed in the total number of frames; for example, it can be set to no more than 10%.

[0048] The incomplete frame ratio threshold is set using an adaptive adjustment method based on time periods. The central management port obtains the current system time and determines whether it is currently in a preset peak or off-peak period. Peak periods refer to times with high pedestrian traffic density, such as 8:00-9:00 AM and 5:00-6:00 PM on weekdays. During these times, people pass through the turnstiles quickly, increasing the difficulty of facial image acquisition and naturally increasing the probability of incomplete frames. The central management port increases the incomplete frame ratio threshold by a preset adjustment step, such as 5%. During off-peak periods, pedestrian traffic density is lower, and acquisition conditions are better, so the incomplete frame ratio threshold returns to the default value. Through this adaptive adjustment based on time periods, the threshold is matched with the actual acquisition conditions of different time periods, avoiding excessive resampling caused by overly strict thresholds during peak periods.

[0049] The facial pose angle judgment analysis is as follows: The central management port estimates the facial pose of each frame of facial image, determining the deflection angle of the target person's face relative to the image acquisition device, including left-right deflection angles and up-down deflection angles. If the left-right deflection angle of the face in a certain frame exceeds a preset left-right angle threshold, or the up-down deflection angle exceeds a preset up-down angle threshold, then that frame is marked as a large-angle pose frame. The left-right angle threshold can be set, for example, to a left-right deflection of no more than 30°, and the up-down angle threshold can be set, for example, to an up-down deflection of no more than 20°; both can be adjusted and preset by those skilled in the art. The central management port counts the proportion of large-angle pose frames in all frames. If this proportion exceeds a preset large-angle pose proportion threshold, then the facial pose angle is judged to be unqualified. The large-angle pose proportion threshold is a preset upper limit for the proportion of large-angle pose frames in the total number of frames, for example, it can be set to no more than 15%.

[0050] The large-angle posture ratio threshold adopts an adaptive adjustment method based on the speed of personnel passing through. The central management port obtains the speed of the target person passing through the gate port through the infrared beam sensor or video motion detection algorithm. When the speed exceeds the preset high-speed passage threshold, for example, greater than 1.5 meters per second, it indicates that the target person is passing through the gate quickly, and there is a high probability that their facial posture is difficult to keep facing the camera. The central management port increases the large-angle posture ratio threshold by a preset adjustment step, for example, by 5%. When the speed is lower than or equal to the high-speed passage threshold, it returns to the default threshold. Through the above adaptive speed adjustment, excessive resampling caused by rapid personnel passage is avoided.

[0051] When image clarity, facial key feature integrity, and facial pose angle are all judged as qualified, the central management port marks the set of verification data as qualified data. If any one of the above three dimensions is judged as unqualified, the central management port marks the set of verification data as unqualified data and performs the following re-sampling process: The central management port sends a re-collection command to the gate port corresponding to the set of verification data. The re-collection command contains the unique identifier of the target person and requires the gate port to re-collect the target person's verification data before the target person leaves the gate area.

[0052] Upon receiving a re-acquisition command, the gate port issues a voice prompt to the target person, requesting them to adjust their facial angle or distance from the acquisition device to re-acquire facial image data. After the target person adjusts their posture, the gate port re-acquires facial image data, simultaneously re-acquiring body surface temperature and vibration data, and sends the re-acquisitioned verification data to the central management port.

[0053] After receiving the re-collected verification data, the central management port performs the above multi-dimensional judgment again. If the re-collected verification data is qualified after multi-dimensional judgment, the original unqualified data is replaced with the re-collected verification data, the re-collected verification data is marked as qualified data, and sent to the subsequent processing flow.

[0054] If, after multiple re-collection attempts within the preset re-collection time window, the re-collected verification data still fails multi-dimensional evaluation, the central management port discards the set of verification data, records the discard event, and prompts the target personnel to switch to another gate port to re-collect verification data via a voice prompt device. The re-collection time window is a preset allowed re-collection operation time length, for example, it can be set to 30 seconds. The discard event must include at least the target personnel's unique identifier, the gate port's port identifier, and the discard time.

[0055] Please see Figure 3 Step S2: Based on the unique identity of the target personnel, the central management port retrieves the qualified data of all gate ports verified by the target personnel within the preset time window, constructs the spatiotemporal correlation collection record set of the target personnel, and aligns the data in the collection record set with the time axis.

[0056] S21, Multi-port data retrieval based on unique identifier: The central management port uses the unique identifier of the target personnel as the retrieval primary key, and within a preset time window, retrieves all qualified data carrying that unique identifier from the qualified data uploaded by all gate ports. The preset time window is a pre-set time period length used to limit the scope of data retrieval, for example, it can be set to the 30 minutes prior to the current time, and the specific preset can be adjusted by professionals in this field.

[0057] Within a 30-minute time window, a target individual may pass through multiple turnstile ports sequentially, with each port uploading qualified data carrying the individual's unique identifier. The central management port aggregates all retrieved qualified data carrying the same unique identifier, forming a multi-port qualified data set for that individual. This multi-port qualified data set contains multiple sets of verification data for the individual at different turnstile ports and at different collection times.

[0058] S22, Construction of the spatiotemporal correlation acquisition record set: The central management port parses the acquisition timestamp of each set of grid data from the multi-port qualified data set and the port identifier of the gate port corresponding to the set of data.

[0059] The central management port sorts the data from each combination of cells in the multi-port qualified data set according to the order of their collection timestamps, generating a time-series data collection record sequence for the target person. This time-series data collection record sequence reflects the order in which the target person passes through each gate port within a preset time window.

[0060] The central management port queries the physical layout information of each turnstile port in the secondary gate area based on the turnstile port identifiers carried in the data of each combination grid. It then arranges the turnstile ports that the target person actually passed through according to their physical layout order, generating a spatial passage path sequence for that person. The physical layout order refers to the actual spatial arrangement of the turnstile ports in the secondary gate area from left to right or from entrance to exit. For example, if turnstile ports A, B, C, and D are arranged from left to right in the secondary gate area, the physical layout order is A, B, C, D.

[0061] The central management port associates the time-series data acquisition sequence with the spatial access path sequence to generate a spatiotemporal associated data acquisition record set for the target person. Specifically, for each set of grid data in the time-series data acquisition sequence, it is bound to the corresponding gate port location information in the spatial access path sequence, so that each set of grid data is simultaneously associated with its acquisition time and the physical location of its corresponding gate port. In the spatiotemporal associated data acquisition record set, each set of verification data contains four dimensions of information: the target person's unique identifier, the acquisition timestamp, the gate port identifier and its physical location, and the content of the verification data.

[0062] S23, Time axis alignment of data within the spatiotemporal correlation acquisition record set: The central management port parses each set of verification data in the spatiotemporal correlation acquisition record set, extracts multiple frames of facial images and their corresponding sampling times from facial image data, extracts multiple body surface temperature values ​​and their corresponding sampling times from body surface temperature data, and extracts multiple vibration signal values ​​and their corresponding sampling times from vibration data.

[0063] The sampling frequencies of the facial image acquisition device, the body surface temperature acquisition device, and the vibration acquisition device are different. Therefore, the sampling point distribution density of these three types of sampling data on the time axis is different, and the sampling times are not one-to-one. Therefore, it is necessary to align these data from different acquisition sources on the time axis.

[0064] The specific method for timeline alignment is as follows: the central management port uses the sampling time point of the facial image data as the reference timeline. The reason for choosing facial image data as the reference is that facial image data is the main source for subsequent extraction of facial muscle activity features, and its sampling frequency is moderate, providing sufficient temporal resolution.

[0065] For body surface temperature data, the central management port checks if there is a body surface temperature value whose sampling time is exactly equal to a certain time point on the reference time axis. If not, it takes the preceding and following body surface temperature values ​​whose sampling time is closest to the reference time point. Using these two adjacent body surface temperature values ​​and their corresponding sampling times, a fitted body surface temperature value at the reference time point is calculated using linear interpolation. Specifically, the linear interpolation method uses the preceding and following body surface temperature values ​​as endpoints, and the proportion of the time difference between the reference time point and the preceding sampling time point to the total time difference between the two sampling times is used as the weight to calculate the fitted body surface temperature value at the reference time point. This fitted value is then used as the aligned body surface temperature data at the reference time point.

[0066] For vibration data, the central management port processes it in the same way as for body surface temperature data: it takes the vibration signal value before and after the sampling time closest to the reference time point in the vibration data, calculates the vibration signal fitting value at the reference time point through linear interpolation, and uses it as the aligned vibration data at the reference time point.

[0067] After time axis alignment, facial image data, body surface temperature data, and vibration data all have corresponding data values ​​at each time point on the reference time axis.

[0068] It should be noted that before performing timeline alignment, the central management port also filters out invalid impact signals from the vibration data that exceed the range of human physiological responses. Invalid impact signals refer to signals whose vibration amplitude exceeds the physical upper limit of normal vibration amplitude in the human face and neck area. For example, sudden high-intensity vibration signals generated by the target person's body colliding with the gate or external objects hitting the data acquisition device. These signals do not originate from the target person's own physiological activities, and retaining them will interfere with the accurate extraction of subsequent physiological characteristics. The central management port also filters out extreme outliers in the body surface temperature data caused by interference from external heat sources. Extreme outliers refer to body surface temperature values ​​that deviate significantly from the normal human body temperature range, such as localized abnormal temperature readings caused by nearby hot air equipment. These values ​​do not represent the target person's true body surface temperature state, and retaining them will affect the accurate calculation of steady-state body temperature values ​​and body temperature fluctuation trends.

[0069] The physical upper limit of normal vibration amplitude in the human face and neck region is determined based on physiological statistics. Specifically, under normal walking posture, the normal vibration amplitude generated by muscle activity and blood circulation in the human face and neck region typically does not exceed 50 micrometers. When the vibration signal amplitude detected by the vibration acquisition device exceeds 100 micrometers, it is determined to be an invalid impact signal exceeding the range of human physiological response. The 100-micrometer threshold is set as the physical upper limit of normal vibration amplitude, which is twice 50 micrometers, to retain a certain redundancy range and avoid misjudging extreme vibrations that are still within the physiological range as invalid signals. This threshold can be appropriately adjusted by those skilled in the art based on the actual acquisition environment and equipment characteristics.

[0070] After timeline alignment, the central management port performs smoothing processing on the associated data at each reference time point to eliminate data jitter introduced by interpolation. Specifically, the smoothing process involves comprehensively considering the data values ​​at each reference time point and their adjacent reference time points to avoid unreasonable abrupt changes in data values ​​at a particular time point due to interpolation calculations. This results in smoother data changes across consecutive time points that conform to the continuous variation patterns of physiological signals.

[0071] It should be noted that this embodiment is configured as a conventional hardware solution where the sampling frequency of body surface temperature data and vibration data is lower than that of facial image sampling frequency, and linear interpolation is used to fill the timing gap; if the vibration sampling frequency or body surface temperature sampling frequency is higher than that of facial image sampling frequency in actual engineering design, then downsampling operation is performed on the high-frequency acquired data to match the facial image reference time axis.

[0072] The specific steps for downsampling and matching the reference time axis for high-frequency data are as follows: the acquisition timestamp of each frame of facial image is used as the global reference time point; the original sampling sequences of high-frequency vibration and body surface temperature are traversed, and all original sampling data falling within the interval of two adjacent reference timestamps are selected; the mean of all high-frequency original data within the interval is calculated as the physiological characteristic value corresponding to the reference time point; after all reference time points have been downsampled, data smoothing is performed uniformly to eliminate time series data fluctuations and jitter, and to complete the alignment of the time axis of multi-source heterogeneous data.

[0073] Please see Figure 4 Step S3: Extract multidimensional physiological features from the spatiotemporal correlation acquisition record set after time axis alignment, generate independent physical and mental state scores of the target personnel at each gate port, aggregate the independent physical and mental state scores of the target personnel at all gate ports verified within the preset time window for multi-port collaborative verification, and mark collaborative anomalies.

[0074] S31, Extraction of multidimensional physiological features: The central management port acquires facial image data, body surface temperature data, and vibration data at various reference time points within a preset acquisition window. The preset acquisition window is a pre-defined time period for feature extraction, such as 5 seconds before and after the moment the target person passes through the gate, or a 10-second window centered on the moment the target person passes through the gate.

[0075] The central management port analyzes facial muscle features from facial image data at various reference time points, extracting the contraction frequency and amplitude features of the periorbital muscles and the perioral muscles. Specifically, the central management port performs facial key point localization on each frame of the facial image, identifying key feature points in the eye region and the mouth region. By tracking the positional changes of key feature points in the eye region across multiple consecutive frames, the number of changes and the magnitude of these changes in distance between key feature points per unit time are calculated; the number of changes represents the contraction frequency, and the magnitude of these changes represents the contraction amplitude. Similarly, by tracking the positional changes of key feature points in the mouth region across multiple consecutive frames, the number of changes and the magnitude of these changes in distance between key feature points in the mouth region are calculated per unit time.

[0076] The central management port analyzes the steady-state and fluctuation trends of body surface temperature data at various baseline time points. The steady-state value is the average or median body surface temperature of the target individual during their passage through the gate. The fluctuation trend refers to the direction and magnitude of body surface temperature change over time within a preset data collection window. Specifically, it determines whether the body temperature is rising, falling, or remaining stable by comparing the body surface temperature value at the beginning and end of the window.

[0077] The central management port analyzes the vibration spectral characteristics of the facial and neck regions from vibration data at various reference time points. Specifically, the central management port performs spectral analysis on the vibration signal sequence, converting the vibration signal from the time domain to the frequency domain to obtain the energy distribution of the vibration signal at different frequencies. The frequency components with concentrated energy are identified from the spectral distribution as the dominant vibration frequencies, and the vibration amplitude corresponding to the dominant vibration frequencies is extracted. The central management port also identifies anomalous vibration components that deviate from the normal vibration pattern from the spectral distribution. Anomalous vibration components refer to vibration components whose frequency or amplitude significantly exceeds the historical vibration baseline range of the target individual.

[0078] The historical vibration baseline is established as follows: For newly hired personnel or those passing through the second gate for the first time, the system collects vibration data during their first 10 passages through the gate (during the first week of work), under normal passage conditions. Each data collection session lasts at least 2 seconds. This collected vibration data is stored in chronological order as the initial vibration dataset for that personnel. The central management port performs spectral analysis on each vibration data point in the initial dataset, extracting the dominant vibration frequency and corresponding amplitude for each collection. It calculates the average and standard deviation of the dominant vibration frequency and amplitude across all collected data. The range of the average dominant vibration frequency plus or minus twice the standard deviation is used as the historical vibration frequency baseline range for that personnel, and the range of the average vibration amplitude plus or minus twice the standard deviation is used as the historical vibration amplitude baseline range. As the number of subsequent passages increases, the central management port updates the historical vibration baseline using a sliding window method. This means that the baseline's average and standard deviation are always recalculated using the vibration data from the most recent 30 passages, allowing the baseline to update slowly and naturally with changes in the personnel's physiological state.

[0079] The quantitative standard for "significantly exceeding" is as follows: if the main vibration frequency deviates from the historical vibration frequency baseline range by more than 50% of the baseline range width, or if the vibration amplitude deviates from the historical vibration amplitude baseline range by more than 50% of the baseline range width, it is considered a significant exceedance. When the above conditions are met, the central management port will mark the current vibration component as an anomalous vibration component.

[0080] The contraction frequency and amplitude characteristics of the muscles around the eyes and mouth, the steady-state value and fluctuation trend of body temperature, the principal vibration frequency and its corresponding vibration amplitude, and the specific vibration components extracted above are combined to form a multidimensional physiological feature for assessing physical and mental state.

[0081] S32, Generation of Independent Physical and Mental State Score: The central management port inputs multidimensional physiological characteristics into the physical and mental state assessment unit, which then outputs the independent physical and mental state score of the target person at that gate port and at that collection time. The independent physical and mental state score refers to the score of the target person's physical and mental health status obtained from a single gate port assessment, used to characterize the target person's physical and mental state at that time and place.

[0082] The physical and mental state assessment unit employs an intelligent recognition model trained on big data. The training process and operation of the intelligent recognition model are as follows: During the training phase, a large number of multidimensional physiological characteristic samples were first collected from individuals with known healthy and known abnormal states. Individuals with known healthy states refer to those whose physical and mental health has been confirmed through professional medical assessment, while individuals with known abnormal states refer to those whose physical or psychological abnormalities have been confirmed through professional medical assessment. Each training sample contains the individual's multidimensional physiological characteristic vector and the corresponding physical and mental state label, i.e., healthy or abnormal.

[0083] It should be noted that the collection process for multidimensional physiological characteristic samples strictly adheres to relevant laws and regulations regarding personal information protection and ethical review. Before collection begins, the collection team fully explains the purpose of the data collection to the sample participants: it will be used to train a physical and mental state assessment model; the types of data collected include facial images, body surface temperature, vibration signals, etc.; the method of data use, such as using the data solely for model training and not for other purposes; and the data storage and protection measures. After fully understanding the above information, the sample participants sign an informed consent form in writing, confirming their voluntary participation in data collection and agreeing that their data will be used for model training. The entire collection plan must be reviewed and approved before implementation.

[0084] The specific sample size is as follows: no fewer than 5,000 samples of multidimensional physiological characteristics from individuals in known healthy states, and no fewer than 3,000 samples of multidimensional physiological characteristics from individuals in known abnormal states. Each sample refers to a complete set of validation data and its corresponding physical and mental state label collected when a target individual passes through a turnstile once. This sample size ensures sufficient statistical power for model training, enabling the model to learn a robust mapping relationship between physiological characteristics and physical and mental states.

[0085] The intelligent recognition model is trained using the aforementioned training samples. The training objective is to enable the model to learn the mapping relationship between multidimensional physiological feature vectors and mental / physical state labels. Specifically, the training method involves inputting the multidimensional physiological feature vectors from the training samples into the model, which then outputs a prediction of the mental / physical state of that sample. This prediction is compared with the sample's true label, the prediction error is calculated, and the model's internal parameters are adjusted based on the error to ensure more accurate predictions in subsequent iterations. This process is iterated until the model's prediction accuracy reaches a preset convergence criterion.

[0086] During operation, the central management interface inputs the multidimensional physiological feature vectors of the target individuals into the trained intelligent recognition model. The model then outputs a physical and mental state score for the target individual based on the learned mapping relationship. The score is a numerical value; a higher score indicates a healthier physical and mental state, while a lower score indicates a greater deviation from a healthy state. For example, the score range can be set from 0 to 100, where 90 to 100 represents a healthy state, 80 to 90 represents a slight deviation, 70 to 80 represents a moderate deviation, and below 70 represents a severe deviation. The specific range and settings are adjusted and preset by professionals in the field.

[0087] The central management port performs the aforementioned independent physical and mental state score generation process for each gate port verified by the target personnel within a preset time window, thereby obtaining the independent physical and mental state scores of the target personnel at each gate port.

[0088] In this embodiment, the intelligent recognition model employs a deep neural network architecture, consisting of an input layer, three hidden layers, and an output layer. The input layer receives multidimensional physiological feature vectors, and each hidden layer sequentially performs nonlinear transformations and feature abstraction on the input data. The output layer outputs the predicted value of the physical and mental state score. Network training utilizes an error backpropagation algorithm, specifically: the multidimensional physiological feature vectors from the training samples are input into the input layer, processed through the forward computation of each hidden layer to obtain the output layer's prediction result, and compared with the sample's true physical and mental state label to obtain the prediction error. This prediction error is then propagated backward from the output layer to each hidden layer, adjusting the connection weights between neurons based on their contribution to the error, so that the network outputs a prediction result closer to the true label in the next round of forward computation. This process is repeated, and each completion of forward computation and backpropagation of all training samples is called a training round.

[0089] The convergence determination of the intelligent recognition model training adopts a dual standard: In five consecutive training epochs, the fluctuation range of the model's prediction accuracy on the validation set does not exceed a preset fluctuation threshold, for example, not exceeding 0.5%; the model's prediction accuracy on the validation set reaches a preset accuracy target value, for example, not less than 95%. When both conditions are met, the model training is considered converged and training stops. The validation set is a pre-defined subset of samples from all training samples, which does not participate in the weight adjustment process and is only used to evaluate the model's generalization ability; it accounts for 20% of all training samples.

[0090] S33, Multi-port Collaborative Verification and Collaborative Anomaly Marking: The central management port compares the independent physical and mental state scores of all turnstile ports within the two-gate area verified by the target personnel within a preset time window with preset scoring benchmark values. The scoring benchmark value is a pre-set critical score value used to distinguish between normal and abnormal states. It can be preset by professionals in the field and adjusted according to actual work needs; for example, it can be set to 80 points. Turnstile ports with independent physical and mental state scores below the scoring benchmark value are marked as single-port anomaly records.

[0091] The scoring benchmark adopts a scenario-based adaptation method based on the job safety level. The central management interface pre-stores the job safety level information for each target personnel. Job safety levels are categorized according to the risk level of the work environment. For example, high-risk jobs, such as underground mining operations or chemical plant operations, are classified as Level 1, while ordinary jobs are classified as Level 2. For Level 1 personnel, because their physical and mental condition directly affects production safety, the central management interface sets a higher scoring benchmark value, such as 85 points, to more strictly judge anomalies. For Level 2 personnel, the scoring benchmark value is set to a standard value, such as 80 points. Through this scenario adaptation, the strictness of anomaly judgment is matched to the safety requirements of different jobs.

[0092] When at least one single-port anomaly record is detected, the central management port initiates multi-port collaborative verification. Specific collaborative verification methods include: Extract the physical and mental state assessment records of all other turnstile ports verified by the target personnel within the preset time window, and count the total number of turnstile ports the target personnel passed through within the preset time window. Count the number of turnstile ports marked as single-port abnormal by the target personnel within the preset time window, and calculate the proportion of single-port abnormal records to the total number of turnstile ports passed by the target personnel. Compare this proportion with a preset abnormality proportion threshold, which is a pre-set threshold for determining whether a coordinated abnormality has occurred, for example, it can be set to 50%. When this proportion exceeds or equals the preset abnormality proportion threshold, it indicates that the target personnel was not only judged as abnormal at a single turnstile port, but also at multiple turnstile ports, and the possibility of a single-port misjudgment is low. At this time, the central management port marks the target personnel as having a coordinated abnormality. When the ratio is lower than the preset abnormal ratio threshold, it means that the target person is only judged as abnormal at a few turnstile ports. This may be due to a data acquisition quality problem or environmental interference at a certain turnstile port, resulting in a single port misjudgment. At this time, the central management port marks the single port abnormal record as a single point suspected abnormality and stores the target person's unique identity and corresponding independent physical and mental status score in the review queue for manual review by the management personnel.

[0093] The anomaly rate threshold is adaptively adjusted based on the total number of turnstile ports traversed by the target person. When the total number of turnstile ports traversed by the target person within a preset time window is large, such as 5 or more, the impact of a single port misjudgment on the overall rate is small, and the anomaly rate threshold can be set to a lower value, such as 40%. When the total number of turnstile ports traversed by the target person is small, such as less than 3, the impact of a single port misjudgment on the overall rate is large, and the anomaly rate threshold should be set to a higher value, such as 60%, to avoid misjudgments due to data quality issues at a single port. This adjustment method can be expressed as follows: the anomaly rate threshold decreases as the total number of turnstile ports traversed increases, and the specific mapping relationship can be preset by those skilled in the art based on the actual application scenario.

[0094] Please see Figure 5 Step S4: Statistically analyze the distribution pattern of all personnel marked as having collaborative anomalies within the second-door area, identify the hotspot ports and intersection ports of the group anomalies, and generate dynamic access control instructions based on the multi-port collaborative verification results and the identification status of the hotspot ports and intersection ports of the group anomalies.

[0095] S41, Identification of abnormal hotspot ports: The central management port obtains the unique identifiers of all personnel marked as having abnormal coordination within the two-gate area within a preset time window, counts the number of personnel marked as having abnormal coordination at each gate port within the preset time window, and generates a count of personnel with abnormal coordination for each gate port. The preset time window can be set to the most recent 1 hour, the most recent 2 hours, or the most recent shift; the specific duration can be configured according to the actual application scenario.

[0096] The count of coordinated abnormal personnel at each turnstile port is compared with a preset group abnormality threshold. The group abnormality threshold is a pre-set critical value for the number of coordinated abnormal personnel used to determine whether a certain turnstile port constitutes a group abnormality hotspot; for example, it can be set to 6 people.

[0097] The group anomaly threshold adopts an adaptive adjustment method based on the average daily number of people passing through the secondary gate area. The central management port counts the average daily number of people passing through the secondary gate area within a preset period, such as the past 7 working days. When the average daily number of people passing through is large, such as exceeding 5,000, the group anomaly threshold is set to a higher value, such as 10, to avoid false flagging due to an excessively large base number. When the average daily number of people passing through is small, such as less than 1,000, the group anomaly threshold is set to a lower value, such as 3, to ensure timely detection of group anomalies in scenarios with a small base number of people. The specific value of the group anomaly threshold can be determined according to the preset mapping relationship between the average daily number of people passing through.

[0098] When the number of individuals exhibiting abnormal behavior at a particular turnstile port exceeds the group abnormality threshold, it indicates that a large number of individuals with abnormal behavior have appeared at that turnstile port within a short period of time, potentially indicating a clustered health risk event, such as multiple people simultaneously experiencing physical discomfort or a group psychological stress reaction. In this case, the turnstile port is marked as a group abnormality hotspot port. A group abnormality hotspot port indicates a cluster risk at that turnstile port, and its identification is based on the statistical count of individuals exhibiting abnormal behavior at a single turnstile port.

[0099] S42, Identification of abnormal intersection ports of the group: The central management port obtains the spatiotemporal correlation collection record set of all personnel marked as collaborative abnormality, and extracts the spatial travel path sequence of each collaborative abnormality personnel within a preset time range from each spatiotemporal correlation collection record set.

[0100] An overlay analysis is performed on the spatial access path sequences of all personnel with abnormal coordination. Specifically, the overlay analysis involves summarizing and statistically analyzing the gate ports traversed by all personnel with abnormal coordination according to their spatial access path sequences, and identifying the gate ports that are commonly pointed to by multiple spatial access path sequences, i.e., those that are commonly traversed.

[0101] The central management port counts the number of coordinated abnormal individuals pointing to the same gate port. When the number of pointing to the same gate port exceeds a preset threshold, the gate port is marked as a convergence port. The threshold is a pre-set critical value for the number of paths used to determine whether a particular gate port constitutes a convergence port for a group of abnormal events; for example, it can be set to be a path sequence that passes through the same point. A convergence port reflects the overlapping node on the spatial movement trajectory of multiple coordinated abnormal individuals. Its identification is based on the superposition analysis of the spatial travel paths of coordinated abnormal individuals, which is completely different from the identification basis of hotspot ports for group abnormal events. This overlapping node may be a common source of abnormal events or a key location that they pass through together.

[0102] The pointer quantity threshold uses an adaptive adjustment method based on the total number of individuals marked as having abnormal collaboration. The central management port obtains the total number of individuals marked as having abnormal collaboration within a preset time window. When the total number of individuals with abnormal collaboration is large, for example, exceeding 20, the pointer quantity threshold is set to a higher value, such as 5, to filter out intersection nodes with statistical significance. When the total number of individuals with abnormal collaboration is small, for example, less than 10, the pointer quantity threshold is set to a lower value, such as 2, to ensure that potential intersection nodes can be identified even in small-scale group abnormal events. The specific value of the pointer quantity threshold can be determined according to the preset mapping relationship between the total number of individuals with abnormal collaboration.

[0103] S43, Generation of Dynamic Access Control Commands: Based on the multi-port collaborative verification results and the identification status of abnormal hotspot ports and converging ports, the central management port generates different types of dynamic access control commands: For individuals exhibiting abnormal behavior during collaboration, the central management port generates an interception command. This command includes the individual's unique identifier and the port identifier of the gate they are currently at, or the port identifier of the gate they are about to reach. The central management port then sends the interception command to the corresponding gate. Upon receiving the command, the gate that detects the arrival of the individual remains closed, preventing their passage, and simultaneously triggers a voice warning, issuing a prompt such as: "Your status is abnormal; please cooperate with staff for verification."

[0104] For individuals suspected of abnormal activity at a single point, the central management port generates a monitoring instruction. This instruction contains the individual's unique identifier. The central management port pushes this identifier to its internal management interface, which then distributes it to all turnstiles within the secondary gate area, requiring all turnstiles to prioritize monitoring of this individual. This prioritized monitoring means that when a turnstile receives a verification request for this unique identifier, it automatically upgrades its data collection strategy to a preset high-alert strategy. This high-alert strategy is preset by professionals in the field and includes at least one or more of the following measures: increasing the frame rate of facial image data acquisition, for example, from 25 frames per second to 50 frames per second; extending the data collection time when the individual passes through the turnstile; and performing more detailed feature extraction and analysis on the individual's body surface temperature and vibration data. It should be noted that if the same individual is suspected of abnormal activity three times consecutively, the alert is automatically upgraded to a collaborative anomaly flag, triggering an interception and review.

[0105] The specific optimization methods for the high-alert data collection strategy include the following aspects: At the hardware acquisition level: the frame rate of facial image acquisition has been increased from 25 frames per second to 50 frames per second to capture more subtle changes in facial muscle dynamics; the resolution of facial images has been improved from megapixel levels to higher resolutions, such as from 1280×720 to 1920×1080, to obtain more precise facial detail information; the sampling frequency of body surface temperature acquisition has been increased from once per second to twice per second to capture more subtle body temperature fluctuations; and the sensitivity of vibration acquisition has been improved by increasing the gain parameter of the vibration acquisition device to make it more sensitive to weak vibration signals.

[0106] At the feature extraction level: the acquisition window is extended from the usual ten seconds to twenty seconds to obtain a longer physiological signal sequence for analysis; the feature dimensions are expanded, in addition to the conventionally extracted features of the frequency and amplitude of muscle contraction around the eyes and mouth, the steady-state value and fluctuation trend of body temperature, and the vibration spectrum features, auxiliary features such as the frequency of facial micro-expression changes, the rate of body temperature change, and the time-domain statistical features of vibration signals are extracted; the feature analysis is refined, the analysis of vibration spectrum features is expanded from identifying the main vibration frequency to identifying the secondary vibration frequency and its harmonic components, and the criteria for judging specific vibration components are tightened from significantly exceeding the historical baseline to exceeding the historical baseline by a certain percentage, for example, exceeding by 30% is marked.

[0107] For abnormal hotspot ports and converging ports, the central management port generates a zone control command. This command includes a list of port identifiers for the gate ports requiring control. The central management port then distributes the command to the corresponding gate ports. Upon receiving the command, the gate port disables its verification function, suspending all verification requests. Simultaneously, the central management port uses a voice guidance system to issue voice prompts to individuals near the affected gate port, guiding them to other gate ports for verification. By disabling the verification functions of abnormal hotspot ports and converging ports, the further spread and propagation of abnormal events within the secondary gate area is prevented.

[0108] It should be noted that the central management port also generates a group anomaly event tracing report based on the spatiotemporal distribution characteristics of the hotspot ports and converging ports of the group anomaly. The specific generation method for this report is as follows: the central management port analyzes the chronological order in which the hotspot ports and converging ports of the group anomaly were marked, and combines this with the spatial movement sequence of the coordinated anomaly personnel to determine the initial occurrence area and critical diffusion period of the anomaly event. The initial occurrence area is the gate port location where the coordinated anomaly personnel first gathered, and the critical diffusion period is the time interval during which the number of coordinated anomaly personnel rapidly increased. The above analysis results are then summarized to generate a handling recommendation report containing the initial occurrence area and critical diffusion period of the anomaly event, which is then pushed to the management terminal for management personnel's reference.

[0109] In summary, this application has at least the following effects: By retrieving qualified data from all gate ports verified within a preset time window using the unique identifier of the target individual, and constructing a spatiotemporally correlated data acquisition record set, the problem of independent operation of each port and inability to correlate data is solved, enabling the data of the same person passing through multiple ports to form a complete spatiotemporal profile. By performing a triple judgment on facial image data—image clarity, integrity of key facial features, and facial posture angle—and triggering resampling when a judgment is unqualified, erroneous control decisions based on low-quality data are avoided. Using the sampling time of facial image data as the baseline time axis, interpolation is used to align the time of body surface temperature data and vibration data, solving the data synchronization problem caused by different sampling frequencies of multi-source heterogeneous data. The results of single-port physical and mental state assessments are collaboratively verified with the assessment results of other ports; when the proportion of abnormal records exceeds a preset threshold, it is marked as a collaborative anomaly, reducing the false judgment rate of single ports. By statistically analyzing the number of collaboratively abnormal personnel at each port, hotspot ports for group anomalies are identified, and the spatial passage paths of collaboratively abnormal personnel are overlaid to identify intersecting ports, enabling the timely detection of group anomaly events. Interception commands, monitoring commands, and area control commands are generated for coordinated abnormal personnel, suspected abnormal personnel at a single point, and abnormal ports in a group, respectively, realizing hierarchical and classified dynamic access control.

[0110] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0111] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0112] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0113] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0114] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0115] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A two-stage physical and mental health screening method based on big data analysis, characterized in that: Includes the following steps: Obtain the target personnel verification data of each gate port in the second gate area, filter and re-collect the target personnel verification data, and upload the qualified data after filtering and re-collection to the central management port. The central management port retrieves qualified data from all gate ports verified by the target personnel within a preset time window based on the target personnel's unique identity identifier, constructs a spatiotemporal correlation collection record set of the target personnel, and aligns the data in the spatiotemporal correlation collection record set with the time axis. Extract multidimensional physiological features from the spatiotemporally correlated collection record set after time axis alignment, generate independent physical and mental state scores of the target personnel at each gate port, aggregate the independent physical and mental state scores of the target personnel at all gate ports verified within a preset time window for multi-port collaborative verification, and mark collaborative anomalies; The distribution patterns of all personnel marked as having collaborative anomalies within the second-level gate area are statistically analyzed to identify hotspot ports and intersection ports of group anomalies. Dynamic access control commands are generated based on the multi-port collaborative verification results and the identification status of hotspot ports and intersection ports of group anomalies.

2. The two-stage physical and mental health screening method based on big data analysis according to claim 1, characterized in that, The screening and resampling include: The target personnel verification data includes facial image data, body surface temperature data, and vibration data; Multi-dimensional judgments are made using facial image data, including image clarity judgment, integrity judgment of key facial features, and facial posture and angle judgment. When there are multiple unqualified judgments, a re-sampling command is sent to the gate port, and the target person is prompted by voice to adjust the facial angle or the collection distance to re-sample the facial image data. If, after a re-collection attempt within the preset re-collection time window, the re-collected facial image data is still unqualified after multi-dimensional evaluation, the collection record of that gate port will be discarded, and a voice prompt will be issued to the target personnel to switch gate ports for facial image data re-collection.

3. The two-stage physical and mental health screening method based on big data analysis according to claim 2, characterized in that, The timeline alignment includes: For the spatiotemporal correlation collection record set of the target personnel, with the facial image data collection time as the reference time axis, for scenarios where the sampling frequency of body surface temperature data and vibration data is lower than the sampling frequency of facial images, two adjacent collection time points before and after the reference time point are extracted, and the fitted data values ​​of body surface temperature data and vibration data at the reference time point are calculated by linear interpolation. By using the fitted data values ​​as data gaps on the baseline time axis, facial image data, body surface temperature data, and vibration data are ensured to have corresponding associated data at each baseline time point.

4. The two-stage physical and mental health screening method based on big data analysis according to claim 3, characterized in that, The extraction steps of the multidimensional physiological features include: Acquire facial image data, body surface temperature data, and vibration data at various reference time points within the preset acquisition window; Analyze facial muscle features in facial image data at various baseline time points, and extract the contraction frequency and amplitude features of muscles around the eyes and mouth. Extract the steady-state body temperature value and body temperature fluctuation trend from the body surface temperature data at each baseline time point; Vibration spectrum features of the face and neck regions were extracted from vibration data at various reference time points to identify specific vibration components; Based on the contraction frequency and amplitude characteristics of the muscles around the eyes and mouth, the steady-state value and fluctuation trend of body temperature, as well as specific vibration components, a multidimensional physiological characteristic for assessing physical and mental state is constructed.

5. The two-stage physical and mental health screening method based on big data analysis according to claim 4, characterized in that, The physical and mental state assessment uses an intelligent recognition model trained on big data; The intelligent recognition model learns from physiological characteristic samples of known healthy and known abnormal individuals to establish a mapping relationship between physiological appearance and internal physical and mental state. The multidimensional physiological characteristics of physical and mental state assessment are input into the intelligent recognition model, and the output is an independent physical and mental state score of the target person at each gate port. The independent physical and mental state score refers to the physical and mental health status of the target person assessed at a single gate port.

6. The two-stage physical and mental health screening method based on big data analysis according to claim 1, characterized in that, The multi-port collaborative verification includes: The independent physical and mental state scores of all turnstile ports in the two-door area verified by the target personnel within the preset time window are compared with the preset scoring benchmark value. Turnstile ports with independent physical and mental state scores lower than the scoring benchmark value are marked as single-port abnormal records. When there is a single port abnormal record, extract the physical and mental status assessment records of the target person in other turnstile ports other than the current turnstile port that have been verified within the preset time window, and obtain the proportion of the number of single port abnormal records to the total number of turnstile ports passed by the target person. When the ratio exceeds or equals the preset abnormal ratio, it is marked as a collaborative abnormality; when the ratio is lower than the preset abnormal ratio, the single-port abnormality record is marked as a single-point suspected abnormality.

7. The two-stage physical and mental health screening method based on big data analysis according to claim 1, characterized in that, The identified abnormal hotspot ports and confluence ports include: The number of all personnel identified as having abnormal coordination is obtained at each gate port within the two-door area of ​​the preset time window. The number is compared with a preset group abnormality threshold. Gate ports with a number exceeding the group abnormality threshold are marked as group abnormality hotspot ports. The group abnormality hotspot port indicates that the gate port has a clustering risk. Obtain the gate ports pointed to by all collaborative abnormal personnel's spatial access paths. When the number of gate ports pointing to the same one exceeds a preset threshold, mark that gate port as an intersection port.

8. The two-stage physical and mental health screening method based on big data analysis according to claim 6, characterized in that, The dynamic access control commands include: For individuals exhibiting abnormal behavior during collaboration, generate an interception command and trigger a voice warning; For a single suspected abnormal person, a monitoring instruction is generated, and the person's unique identifier is pushed to the central management port. The central management port then distributes the person's unique identifier to all gate ports for key monitoring. The key monitoring refers to enhancing the data collection strategy to a preset high-alert strategy when the gate port receives a verification request for the unique identifier. For abnormal hotspot ports and converging ports, generate area control commands to disable the verification of the gate port, and use voice guidance to direct personnel to switch gate ports for verification.

9. The two-stage physical and mental health screening method based on big data analysis according to claim 3, characterized in that, The timeline alignment also includes data cleaning: Before performing time axis alignment, invalid impact signals that exceed the range of human physiological response in vibration data and extreme outliers caused by external heat source interference in body surface temperature data are screened out. After the timeline is aligned, the associated data at each reference time point is smoothed to eliminate data jitter when missing data is introduced.

10. The two-stage physical and mental health screening method based on big data analysis according to claim 1, characterized in that, It also includes generating a source tracing report for group anomaly events: Based on the spatiotemporal distribution characteristics of abnormal hotspot ports and confluence ports, the originating area and critical diffusion period of abnormal events are determined, and a handling recommendation report containing the originating area and critical diffusion period of abnormal events is generated.

Images