Subway station cross construction operation personnel safety intelligent management and control system

Abstract

This invention discloses an intelligent safety management system for personnel involved in cross-construction work at subway stations, relating to the field of construction worker management technology. It comprises a hardware layer, a software layer, and an interaction layer, forming a closed-loop management system for the entire process. The hardware layer employs Bluetooth-IMU fusion positioning and multi-level distributed base station deployment, combined with RSSI ranging, tunnel attenuation correction, and the least squares method to achieve accurate positioning; personnel tags support adaptive wake-up and offline alarms. The software layer, through multi-source data processing, spatiotemporal alignment, and spatial adaptive grid mapping, implements six-fold consistency verification—permissions, location, qualifications, plans, process conflicts, and density overload—through a core processing module; the early warning module integrates trajectory prediction to achieve tiered advanced early warning; permission management adopts a zero-trust dynamic permission mechanism, deeply bound to the construction request plan, enabling dynamic allocation and automatic revocation. This invention can achieve intelligent management of personnel throughout their entire lifecycle, significantly improving the efficiency of safety management during cross-construction work in subway stations.

Description

Technical Field

[0001] This invention relates to the field of construction worker management technology, and more specifically to an intelligent safety management and control system for workers engaged in cross-construction work at subway stations. Background Technology

[0002] Currently, in the field of subway construction worker management, mainstream technical solutions mostly adopt a multi-system, decentralized, and independently operating architecture, separating personnel positioning, area access control, and data management into independent system modules. Among these, real-time control of on-site workers' work areas and high-risk work locations is essentially nonexistent, relying solely on indirect safety inspections by on-site safety officers. Personnel data management is mostly stored through offline documents or independent database systems, covering information such as entry registration, education and training, safety briefings, and wage payments, supporting only basic query and statistical functions.

[0003] Such solutions lack a unified core control center, with data from various modules fragmented and lacking a real-time communication mechanism. Interaction methods primarily rely on traditional personnel basic information and gate access management, failing to incorporate real-time verification of personnel's legitimacy or implement temporary request plans and other safety control measures. Personnel data is completely disconnected from the real-time status of on-site workers (such as real-time location, access records, and violations). Management personnel must repeatedly switch between multiple systems or documents to obtain complete personnel management information. Furthermore, existing on-site safety control systems mostly use GPS positioning, which suffers severe signal attenuation or even failure in complex construction environments such as subway tunnels, underground shafts, and enclosed corridors, making continuous and stable positioning impossible. Moreover, most positioning devices have high power consumption and require frequent charging, failing to meet the demands of long-distance, uninterrupted, and highly reliable construction control in subway systems.

[0004] Therefore, how to break down the traditional barriers of modular dispersion and realize intelligent management of the entire process of cross-construction personnel in subway stations, from data filing, permission allocation, on-site control to violation warning, is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides an intelligent safety management and control system for personnel engaged in cross-construction work in subway stations, which solves the problems existing in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A smart safety management and control system for personnel engaged in cross-construction work at subway stations includes a hardware layer, a software layer, and an interaction layer. Each layer connects in real time with data to construct a closed-loop management system for the entire process, as detailed below: The hardware layer, as a multi-dimensional sensing terminal, is used to collect operator identity information, location data, behavioral images, machine operating status and environmental parameters, and transmits the raw data to the software layer in real time through a standardized interface. The software layer adopts an edge-cloud collaborative architecture, serving as the core processing hub of the system. It receives raw data transmitted from the hardware layer and performs standardized processing, analysis, storage, and correlation integration to achieve job status determination, violation identification, risk prediction, and early warning instruction generation. The processed valid data and instructions are then pushed to the interaction layer and hardware layer, respectively. The interaction layer serves as the management interaction carrier, receiving valid data pushed by the software layer and providing a visual display of personnel information, work status, early warning information, and equipment status, as well as a data query portal. This allows managers to intuitively grasp the on-site control situation and perform interactive operations. The software layer includes: a data acquisition module, a permission management module, a location visualization module, a data management module, an early warning module, and a core processing module; The data acquisition module is configured to perform collaborative fusion processing of multi-source heterogeneous data, which is used to standardize, normalize and align timestamps and perform spatial adaptive grid mapping on various types of raw data to form a standardized dataset with a unified spatiotemporal benchmark. The core processing module is configured with a six-fold consistency check algorithm for execution permissions, location, qualifications, plans, process conflicts, and personnel density overload. This algorithm is used to determine in real time the matching status between the current location of personnel and the allowed work area, the validity of personnel qualifications, the construction plan time period, the spatiotemporal compatibility of processes, and the personnel density in the area. It automatically determines violations such as boundary crossing, unplanned operation, unqualified operation, process conflicts, and personnel gathering. The early warning module is configured to execute a composite early warning analysis method based on dynamic thresholds, duration, confidence filtering, and LSTM trajectory prediction. It is used to conduct graded early warnings and advanced early warnings based on personnel management ratio, on-site supervision duration, temporary personnel accompaniment distance, AI violation confidence, and personnel trajectory trends, and push them to the corresponding management personnel according to the risk level.

[0007] Optionally, the hardware layer includes: a gate module, a Bluetooth-IMU fusion positioning base station module, a personnel positioning tag module, an AI capture module, a mechanical sensing radar module, and an environmental monitoring module. Each module is deployed in a distributed manner to cover key construction nodes and work areas. The gate module is used for real-time comparison of access permissions and recording of access data; The personnel positioning tag module has a built-in identity recognition chip, role identification chip, gyroscope and accelerometer, supports two-way communication with the base station and receives warning commands to trigger vibration prompts; at the same time, it adopts an adaptive sampling and wake-up mechanism for work status, low-power sleep sampling during non-work periods, and automatically switches to high-frequency positioning mode when entering the work area; when the tag falls off, is powered off or the signal is blocked, an offline alarm is automatically triggered. The AI ​​capture module is used to collect on-site images and detect violations and personnel fatigue. The Bluetooth-IMU fusion positioning base station module adopts a distributed deployment approach, consisting of a main base station, a work site base station, and a hazard source base station, to achieve full-area coverage without blind spots. It achieves tight-coupled and accurate Bluetooth-IMU positioning through RSSI ranging, triangulation, tunnel multipath attenuation correction, and the least squares method. The distance calculation formula for Bluetooth-IMU fusion positioning is as follows:

[0008] In the formula: RSSI This indicates the received signal strength; a negative value is used. A This indicates the signal strength when the transmitter and receiver are 1 meter apart. n This represents the environmental attenuation factor, which is dynamically corrected based on tunnel curvature, wall material, and distance from the arch. Position compensation and drift correction are performed using IMU inertial navigation data, and finally, the precise coordinates of the workers are obtained using the least squares method. The mechanical sensing radar module is used to detect the position, operating posture and movement trajectory of construction machinery in real time, so as to realize the collaborative perception of personnel-machine collision avoidance; The environmental monitoring module is used to collect real-time data on toxic gas concentrations, temperature, humidity, and dust parameters in the work area, and trigger environmental risk warnings in conjunction with personnel location.

[0009] Optionally, the data acquisition module supports multi-protocol access and performs dedicated data cleaning for subway construction scenarios on the raw data, including deduplication, noise reduction, signal drift correction, and invalid data removal, and uses AES-256 encrypted transmission; the edge terminal realizes positioning calculation, real-time early warning and network outage self-healing, while the cloud is responsible for big data analysis, trajectory backtracking, report generation and model iteration.

[0010] Optionally, timestamp normalization alignment specifically involves aligning all collected data to a uniform 5-second time slice:

[0011] In the formula: For the aligned standard timestamp, This is the raw timestamp provided by the hardware during data acquisition, with 5 seconds as the normalization step size. This is the rounding function; The spatial adaptive grid mapping specifically involves dynamically dividing the grid cells according to the risk level of the work area: the hazardous area uses a 0.5m×0.5m high-precision grid, the ordinary area uses a 2m×2m standard grid, and the passageway area uses a 1m×4m long strip grid. The mapping formula is as follows:

[0012] In the formula: These are the mapped standard grid coordinates. The actual location of the person determined by Bluetooth-IMU fusion positioning. L , W For dynamic grid width and length, This is the floor function; Based on the above operations, a unique correspondence is achieved for the same person, the same time, the same location, and the same behavior.

[0013] Optionally, the core processing module's six-fold consistency verification algorithm includes: verifying whether the personnel's current location belongs to the permitted work area allocated by the access management module; verifying whether the personnel's qualifications are valid; verifying whether the current time is within the permitted work period of the construction plan; verifying whether there are on-site personnel at high-risk work sites; verifying whether there are spatiotemporal conflicts between the current work process and overlapping work processes; and verifying whether the personnel density in the current area exceeds the safety threshold. Any mismatch is marked as abnormal, and continuous... N If the cycles do not match, it is considered a violation.

[0014] Optionally, the composite early warning analysis method executed by the early warning module includes: If the ratio of management personnel to operational personnel (R) is less than 10% and continues to exceed the set duration, an alert for insufficient management capacity will be triggered. If a personnel supervising a high-risk operation are absent for more than 10 minutes, a major risk warning will be triggered. If a temporary visitor / management staff member is more than 30 minutes away from management personnel without being accompanied, a visitor safety alert is triggered. When the confidence level of AI-captured violation is ≥95%, it is determined as a valid violation and an alert is triggered; Based on the LSTM model, short-term prediction of personnel trajectories is performed, and the tendency to cross the boundary is identified 10 seconds in advance and an early warning is triggered. If the distance between personnel and machinery is less than the safety threshold and the duration exceeds the set value, a collision avoidance warning is triggered. An environmental safety warning was triggered when environmental parameters in the work area exceeded the standard and personnel were in that area.

[0015] Optionally, the permission management module adopts a zero-trust dynamic permission mechanism, which is deeply bound to the construction request plan. It automatically derives and dynamically allocates work permissions based on construction procedures, work areas, risk levels, personnel roles and qualifications. It performs a full-dimensional verification every 30 seconds. When the construction plan changes, the permissions are automatically updated synchronously. When the plan ends, the qualification expires, the location is out of bounds, or the procedure conflicts, the permissions are automatically revoked.

[0016] Optionally, the positioning visualization module is based on the 3D model of the construction area and adopts a triple mapping rule of role-color-coordinate to dynamically map the real-time position, trajectory, density, machine position, and environmental parameters of personnel to the 3D scene. It also supports trajectory playback, view zooming, rotation, and highlighting of risk areas.

[0017] Optionally, the data management module uses the unique ID of each person as an index to construct a safety knowledge graph that links static files with dynamic work status. This graph links basic personnel information, qualification documents, safety briefings, education records, attendance and wages, location trajectory, pre-shift safety speeches, access records, violation records, and fatigue work records in real time. It also connects to the safety big data model interaction module, supporting compliance queries, risk analysis, and automatic generation of rectification suggestions.

[0018] Optionally, the interaction layer is centered around a large visual screen, using a two-level display interface of overview and details; The overview interface displays the number of personnel, role distribution, number of violation warnings, area personnel density, machinery operating status, and environmental parameters in real time, and highlights violators and risk areas with flashing markers. The details page displays complete personnel files, real-time location, historical trajectory, work plan, violation handling records, fatigue status, and environmental data. It supports filtering and querying by region, role, and time.

[0019] As can be seen from the above technical solution, compared with the prior art, the present invention discloses an intelligent safety management and control system for personnel engaged in cross-construction work in subway stations, which has the following beneficial effects: (1) This invention innovatively integrates core functions such as safety knowledge graph for each worker, entry legality verification, Bluetooth-IMU fusion positioning calculation, AI violation and fatigue detection, collaborative perception of construction machinery, environmental risk monitoring, early warning and post-event safety education management. Through multi-source heterogeneous data collaborative fusion, unified spatiotemporal alignment, and six-fold consistency verification of permissions-location-qualification-plan-process conflict-personnel density overload, a unified system hub is constructed, which completely solves the problem of information silos in traditional solutions where each module operates independently and data is not shared. All management operations can be completed within the same system without cross-platform switching. Managers can coordinate and check personnel legal entry information, location monitoring, violation handling, human-machine collision prevention and post-event safety education and training through a visual large screen. The process simplification rate is over 70%, which greatly improves management convenience and overall efficiency, and perfectly adapts to the management needs of cross-construction operations in subway stations involving multiple units, multiple types of work, multiple links, high density and high dynamics. (2) This invention relies on Bluetooth-IMU tight coupling positioning, dynamic correction of tunnel multipath attenuation, least squares coordinate optimization, spatial adaptive grid division, three-dimensional graphic mapping and role-color-coordinate triple mapping rules to intuitively present the real-time location, trajectory, density, construction machinery status, environmental parameters, and risk areas of personnel on a large screen, replacing the single display form of traditional text reports and two-dimensional maps; managers do not need to analyze complex data, and can quickly grasp the dynamics of the entire site by zooming and rotating the large screen, and identify situations such as personnel crowding in high-risk areas, absence of key positions, time and space conflicts in work processes, and human-machine proximity risks in real time, reducing the decision response time from "hours" to "seconds", effectively reducing the safety risks caused by information lag; at the same time, the system adopts a composite early warning mechanism of dynamic threshold, duration, confidence filtering, and LSTM trajectory advance prediction, combined with dual push of large screen and message and hierarchical handling process, to ensure that managers receive information and quickly close the loop, upgrading from "post-event handling" to "pre-event prevention"; (3) The hardware layer of this invention adopts a distributed Bluetooth-IMU fusion positioning architecture of main base station-work face base station-hazard source base station, combined with mechanical sensing radar and environmental monitoring module, which can fully adapt to closed and complex scenarios such as tunnels, underground, and corridors, and the positioning is stable and without blind spots; the software layer adopts an edge-cloud collaborative architecture, with the edge end realizing local calculation, real-time early warning and network outage self-healing, and the cloud end realizing big data analysis, model iteration and knowledge graph management. Through multi-source data standardization, spatiotemporal alignment, adaptive grid mapping and other special processing methods, the system can achieve deep data fusion and precise control; the system supports flexible access of multiple types of terminals such as gates, fusion positioning base stations, AI capture, radar, and environmental sensors. Functional modules can be added according to the scale and stage requirements of the construction project. Regardless of different construction scenarios such as ground work areas and underground tunnels, or different scale requirements such as small sections and large comprehensive projects, the system can be quickly adapted through hardware deployment adjustment and software parameter configuration without reconstructing the architecture. It has a wide range of application scenarios and promotion value. In particular, this invention uses Bluetooth-IMU fusion positioning to replace traditional GPS, overcoming the problem of GPS signal failure in enclosed environments such as subway tunnels and underground mines. It achieves blind-spot-free positioning through multi-level coverage of main base station, work face base station, and hazard source base station. At the same time, it adopts an ultra-low power consumption and adaptive sampling wake-up design, requiring only the main base station to be externally powered, while the other devices can be charged for two years without charging. Tag detachment, power failure, and shielding can trigger alarms in seconds, greatly reducing the difficulty of on-site operation and maintenance, and perfectly adapting to the construction management and control requirements of long sections, uninterrupted operation, and high reliability in subways. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0021] Figure 1 This is an architecture diagram of the intelligent safety management and control system for personnel involved in cross-construction operations at subway stations provided by the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] To address the issues raised in the background section, this embodiment requires the establishment of an integrated intelligent personnel safety management system. This system includes "person-to-person file creation + entry legality verification + Bluetooth-IMU fusion positioning and calculation for work areas + AI violation and fatigue detection + collaborative perception of construction machinery + early warning + post-event safety education data management." Through deep real-time computing at the hardware, software, and interaction layers, this system breaks down the barriers of traditional modular architecture and enables intelligent management of personnel involved in cross-construction work at subway stations, from data filing and permission allocation to on-site control and violation warnings. The specific technical solution is explained in detail from four dimensions: system architecture, core modules, key technologies, and data flow.

[0024] I. System Overall Architecture Design

[0025] This invention discloses an intelligent safety management and control system for personnel engaged in cross-construction work at subway stations, such as... Figure 1 As shown, the system comprises a hardware layer, a software layer, and an interaction layer. Each layer works in real-time with data to create a closed-loop control system for the entire process. The system adopts a "three-layer architecture + edge-cloud collaboration + full-link closed-loop" model, with data interoperability at its core, achieving seamless integration of hardware acquisition, software processing, and interactive display. Its specific architectural layer design is as follows: The hardware layer, as a multi-dimensional sensing terminal, is used to collect operator identity information, location data, behavioral images, machine operating status and environmental parameters, and transmits the raw data to the software layer in real time through a standardized interface to complete basic functions such as identity recognition, location acquisition, violation capture, machine sensing and environmental monitoring. The software layer adopts an edge-cloud collaborative architecture, serving as the core processing hub of the system. It receives raw data transmitted from the hardware layer and performs standardized processing, analysis, storage, and correlation integration to achieve job status determination, violation identification, risk prediction, and early warning instruction generation. The processed valid data and instructions are then pushed to the interaction layer and hardware layer, respectively. The interaction layer serves as the management interaction carrier, receiving valid data pushed by the software layer. It provides a visual display of personnel information, work status, early warning information, equipment status, and environmental parameters, as well as a data query portal. This allows managers to intuitively grasp the on-site control situation and perform interactive operations, achieving a "what you see is what you get" management experience.

[0026] Each level is connected through a fixed relationship of "hardware layer - data acquisition module - core processing module - interaction layer / early warning module" to build a data closed loop, ensuring that the entire process of acquisition, processing, display and early warning is seamless and lag-free.

[0027] II. Detailed Description of Core Functions

[0028] 1. Core device functions at the hardware layer

[0029] In this embodiment, the hardware layer includes: a gate module (deployed at the entrance / wellhead), a Bluetooth-IMU fusion positioning base station module (distributed coverage of the work area), a personnel positioning tag module (personally worn by each person), an AI capture module (deployed in key areas), a mechanical sensing radar module, and an environmental monitoring module. Each module is deployed in a distributed manner to cover key construction nodes and work areas.

[0030] The gate module is used for real-time comparison of access permissions and recording of access data. This module is deployed at key nodes such as construction entrances, wellhead passages, and section boundaries to complete real-time comparison of access permissions: when permissions match, passage is automatically granted and data such as passage time and passage location are recorded; when permissions do not match, passage is denied and a local prompt is triggered.

[0031] The Bluetooth-IMU fusion positioning base station module receives location signals and achieves precise coordinate calculation through tight coupling of Bluetooth and IMU using RSSI ranging, triangulation, tunnel multipath attenuation correction, and least squares method. This module employs a distributed deployment strategy, including a main base station, a work surface base station, and a hazard source base station, covering ground work areas, underground tunnels, and various construction sections to ensure no positioning blind spots and provide data support for real-time positioning and trajectory tracking. In specific implementation, given the Bluetooth beacon locations at three positions and the location of the distance measurement device, the RSSI signal strength of the Bluetooth positioning device is converted into distance using the following formula:

[0032] In the formula: RSSI This indicates the received signal strength; a negative value is used. A This indicates the signal strength when the transmitter and receiver are 1 meter apart. n This represents the environmental attenuation factor, which is dynamically corrected based on tunnel curvature, wall material, and distance from the arch. Position compensation and drift correction are performed using IMU inertial navigation data, and finally, the precise coordinates of the workers are obtained using the least squares method.

[0033] This system employs Bluetooth-IMU fusion positioning technology instead of traditional GPS positioning. This is because GPS signals are prone to attenuation and interruption in enclosed and complex environments such as subway tunnels and underground mines, making continuous and reliable positioning impossible. Bluetooth-IMU fusion positioning, on the other hand, provides a more stable signal with no attenuation or blind spots in enclosed environments, resulting in stronger positioning continuity and effectively solving the drift problem caused by multipath reflections in tunnels. Simultaneously, the system boasts extremely low overall power consumption; only the main base station requires an external power supply. The work site base station, hazardous source base station, and personnel positioning tags all feature ultra-low power design, enabling two years of continuous operation without charging. This perfectly adapts to the long-distance, uninterrupted, and complex construction management scenarios of two subway stations and one section. The main base station supports uploading positioning data to the software layer via 4G IoT cards, Wi-Fi, or wired connections, ensuring stable and reliable data transmission.

[0034] The personnel positioning tag module is used to bind personnel identity and role information and send location signals. This module assigns a unique tag to each worker, manager, safety officer, and temporary construction worker. It incorporates an identity recognition chip, role identification chip, gyroscope, and accelerometer, has an IP68 protection rating (waterproof and drop-proof), weighs ≤50g, and can be integrated into a safety helmet or worn on the wrist. The tag binds basic information such as personnel ID, name, position, and qualification number, supports two-way communication with a base station, and can receive warning commands and trigger vibration alerts. It automatically triggers an offline alarm when the tag falls off, is powered off, or the signal is blocked. Furthermore, it employs an adaptive sampling and wake-up mechanism based on work status, using low-power sleep sampling during non-work periods and automatically switching to high-frequency positioning mode when entering the work area.

[0035] The personnel positioning tag adopts an adaptive sampling and wake-up mechanism based on the work status. Specifically, the tag obtains its own location information in real time. When it is determined to be in a non-work area or non-work period, it automatically enters a low-power sleep mode and works with a low-frequency sampling method of 30 seconds / time. When the tag enters the work area designated by the construction plan, it immediately wakes up automatically and switches to a high-frequency positioning mode of 5 seconds / time to ensure real-time positioning. After leaving the work area, it resumes sleep mode, achieving ultra-long battery life while ensuring positioning accuracy.

[0036] The AI-powered image capture module is used to collect on-site images and detect violations and personnel fatigue. Deployed in high-risk work areas, passageway corners, and key construction locations, this module supports real-time acquisition of on-site personnel images. Through image recognition algorithms, it automatically detects behaviors such as not wearing reflective clothing or safety helmets, achieving an accuracy rate of ≥95%. The module simultaneously uploads violation images along with time and location information to the software layer. In specific implementation, AI high-definition image capture cameras are deployed in high-risk work areas, passageway corners, and locations with high violation rates. These cameras support real-time video stream acquisition, with an image resolution ≥1080P and a recognition frame rate ≥25fps. The cameras must be adaptable to low-light environments and upload violation images along with location and time information in real-time via the network.

[0037] The mechanical sensing radar module is used to detect the position, operating posture and movement trajectory of construction machinery in real time, and realize the collaborative perception of personnel-machine collision avoidance. When the distance between personnel and machinery is less than the safety threshold and the duration exceeds the set value, the collision avoidance warning is immediately triggered.

[0038] The environmental monitoring module is used to collect real-time data on the concentration of toxic gases, temperature, humidity, and dust parameters in the work area. It is linked to personnel location to trigger environmental risk warnings. When environmental parameters exceed the standard and personnel are in the area, warning instructions are automatically pushed.

[0039] Temporary construction worker module: By combining temporary construction plans, supervising and accompanying personnel, and an early warning mechanism, the safety risks of temporary construction workers are minimized. Temporary workers wear visitor safety helmets (with built-in positioning tags), and the system uses positioning technology to monitor in real time whether safety management personnel are accompanying the visitors. If the system detects that the safety management personnel have left the temporary workers for 10 minutes, it will immediately trigger an early warning mechanism and push the project department's safety management department to urge safe operation.

[0040] 2. Core Module Functions of the Software Layer

[0041] In this embodiment, the software layer includes: a data acquisition module, a permission management module, a location visualization module, a data management module, an early warning module, and a core processing module. In specific implementation, the software system adopts a B / S architecture, supports access from both web and mobile devices, and after system deployment, a firewall and access permissions are configured to allow only authorized personnel to log in and operate.

[0042] The data acquisition module supports multi-protocol data access (compatible with formats such as HTTP, Bluetooth, video streams, radar, and environmental sensors) and is used to standardize and encrypt the raw data for transmission (AES-256 encryption). This module acts as a bridge between the hardware and software layers, configured to perform collaborative fusion processing of multi-source heterogeneous data. It receives real-time gate access data, fused positioning base station location data, AI capture device image data, radar mechanical data, environmental monitoring data, and construction site schedule data. It then performs unified standardization, timestamp normalization and alignment, and spatial adaptive grid mapping on all types of raw data to form a standardized dataset with a unified spatiotemporal reference.

[0043] The data acquisition module supports multi-protocol access and performs specialized data cleaning for subway construction scenarios on the raw data, including deduplication, noise reduction, signal drift correction, and invalid data removal, and uses AES-256 encrypted transmission. The specialized data cleaning for subway construction scenarios specifically includes: 1) Deduplication: Merge multiple location data and access data uploaded repeatedly by the same person within the same time slice, retaining only one valid data; 2) Noise Reduction: Removes abnormal coordinate jumps and invalid values ​​caused by Bluetooth signal interference and instantaneous device jitter; 3) Signal drift correction: Combine IMU inertial navigation data with historical trajectories to fit and calibrate continuously offset positioning points to eliminate positioning drift caused by tunnel multipath reflection; 4) Invalid data removal: Discard invalid data collected that is outside the construction area, outside the working time period, or when the equipment is offline; The cleaned and valid data is encrypted using the AES-256 algorithm for transmission and storage to ensure data security and integrity.

[0044] The edge device enables location calculation, real-time early warning, and self-healing from network outages, while the cloud device is responsible for big data analysis, trajectory backtracking, report generation, and model iteration, ensuring that core on-site control functions remain functional even when the network is interrupted.

[0045] Timestamp normalization alignment specifically involves aligning all collected data to a uniform 5-second time slice:

[0046] In the formula: For the aligned standard timestamp, This is the raw timestamp provided by the hardware during data acquisition, with 5 seconds as the normalization step size. This is the rounding function; The spatial adaptive grid mapping specifically involves dynamically dividing the grid cells according to the risk level of the work area: the hazardous area uses a 0.5m×0.5m high-precision grid, the ordinary area uses a 2m×2m standard grid, and the passageway area uses a 1m×4m long strip grid. The mapping formula is as follows:

[0047] In the formula: These are the mapped standard grid coordinates. The actual location of the person determined by Bluetooth. L , W For dynamic grid width and length, This is the floor function; Based on the above operations, a unique correspondence is achieved for the same person, the same time, the same location, and the same behavior.

[0048] The access control module employs a zero-trust dynamic access control mechanism, deeply integrated with the construction task plan. It automatically derives and dynamically allocates work permissions based on construction procedures, work areas, risk levels, personnel roles, and qualifications. A full-dimensional verification is performed every 30 seconds. Permissions are automatically updated when the construction plan changes, and automatically revoked when the plan ends, qualifications expire, locations exceed boundaries, or procedures conflict. This module allows administrators to input construction task information (work area, work time period, and participant list). The system automatically assigns corresponding work permissions based on construction procedures, work areas, risk levels, personnel roles, and qualification levels (e.g., laborers are limited to specific sections, safety officers can inspect across areas). In practical implementation, automatic permission synchronization trigger conditions can be set, such as synchronization within 3 seconds of a construction plan modification.

[0049] The positioning visualization module is based on a 3D model of the construction area (a 1:1 recreation of real-world scenes such as the ground, underground, and tunnels). It employs a role-color-coordinate triple mapping rule to dynamically map the real-time location, trajectory, density, machinery location, and environmental parameters of personnel to the 3D scene. It supports trajectory playback, view zooming, rotation, and highlighting of risk areas. This module maps personnel location data collected by positioning base stations to the 3D model in real time, using a "role-color" mapping rule to distinguish personnel types: laborers are marked in blue, managers in purple, safety officers in red, and visitors in yellow. It supports personnel trajectory playback, allowing managers to intuitively view the real-time location of individual personnel, the distribution density of multiple personnel, and their movement trajectories through zooming and rotating the view on a large screen. Internal trajectory playback functionality is also supported.

[0050] The data management module is used to build a personnel lifecycle data archive. Using a unique personnel ID as an index, this module constructs a one-person-one-file safety knowledge graph that links static records with dynamic work status. It links basic personnel information, qualification documents, safety briefings, training records, attendance and pay records, location tracking, pre-shift safety speeches, access records, violation records, and fatigue work records in real time. It also connects to the safety big data model interaction module, supporting compliance queries, risk analysis, and automatic generation of rectification suggestions, achieving full lifecycle traceability.

[0051] The early warning module is configured to execute a composite early warning analysis method based on dynamic thresholds, duration, confidence filtering, and LSTM trajectory prediction. It is used to provide tiered and proactive early warnings based on personnel management ratios, on-site monitoring duration, temporary staff accompaniment distance, AI violation confidence level, personnel trajectory trends, human-machine distance, and environmental parameters, and pushes the warnings to the corresponding management personnel according to risk level. When the AI ​​capture module detects a violation or location data shows that a person has crossed the boundary, an early warning is immediately triggered: first, a scrolling display is shown on a large screen; second, an early warning notification (including violation details and handling suggestions) is sent to DingTalk management personnel.

[0052] The composite early warning analysis method executed by the early warning module includes: 1) If the ratio of management personnel to operational personnel (R) is less than 10% and continues for more than the set time (e.g., 30 seconds), an alert for insufficient management personnel will be triggered. 2) If the absence of on-site personnel for a high-risk operation exceeds 10 minutes, a high-risk warning will be triggered; 3) If temporary personnel / visitors are more than 10 meters away from management personnel and remain unaccompanied for more than 30 minutes, a visitor safety alert will be triggered; 4) When the confidence level of AI-captured violation is ≥95%, it is determined as a valid violation and an alert is triggered; 5) Based on the LSTM model, short-term prediction of personnel trajectories is performed, and boundary crossing trends are identified 10 seconds in advance and an early warning is triggered; 6) If the distance between personnel and machinery is less than the safety threshold and the duration exceeds the set value, a collision avoidance warning will be triggered; 7) If environmental parameters in the work area exceed the standard and personnel are in the area, an environmental safety warning will be triggered.

[0053] The system will only trigger the corresponding warning when the above thresholds and duration conditions are met, thus avoiding frequent false alarms.

[0054] The core processing module is configured with a six-fold consistency verification algorithm covering permissions, location, qualifications, plans, process conflicts, and personnel density overload. This algorithm is used to determine in real-time the matching status between personnel's current location and permitted work areas, personnel qualification validity, construction plan time periods, spatiotemporal compatibility of processes, and regional personnel density. It automatically identifies violations such as boundary violations, unplanned work, unqualified work, process conflicts, and personnel gathering. As the "brain" of the software layer, this module integrates permission management and location analysis functions. It comprehensively processes access data, location data, image data, radar data, environmental data, and construction request plan data uploaded by the data acquisition module; verifies the consistency between personnel access permissions and location areas to determine if boundary violations exist; correlates personnel information with work status to ensure only qualified personnel can participate in work; and synchronizes the processing results to the interaction layer and early warning module to drive visualization and early warning triggering.

[0055] The core processing module's six-fold consistency verification algorithm includes: verifying whether the personnel's current location is within the permitted work area allocated by the access management module; verifying whether the personnel's qualifications are valid; verifying whether the current time is within the permitted work period of the construction plan; verifying whether there are on-site personnel at high-risk work sites; verifying whether there are any spatiotemporal conflicts between the current work process and overlapping work processes; and verifying whether the personnel density in the current area exceeds the safety threshold. Any mismatch is marked as abnormal, and continuous errors are detected. N If a cycle mismatch occurs, it is considered a violation. In specific implementation, continuous... N The system is set to determine a violation only if a mismatch occurs for three consecutive positioning cycles (i.e., 15 consecutive seconds). This multi-cycle confirmation mechanism avoids misjudgments caused by instantaneous signal fluctuations or brief false starts, thereby improving the accuracy and stability of violation identification.

[0056] 3. Interaction Layer Functions

[0057] In this embodiment, the interaction layer is centered around a large visual screen, and adopts a two-level display interface of overview and details.

[0058] Overview interface: Real-time display of key indicators such as number of personnel, role distribution, number of violation warnings, area personnel density, machinery operating status, and environmental parameters, and flashing markings for violating personnel and risk areas; Details page: Clicking the personnel tag in the overview page will take you to the details page, which displays the personnel's complete profile (basic information, qualification profile, salary record, fatigue status), real-time location, historical trajectory, work plan, violation handling record, and environmental data. It supports filtering and querying by region, role, and time.

[0059] In practice, the interactive layer screen is divided into a core indicator area (total number of personnel, role distribution, number of violations), a 3D positioning view area, a passage record area, and an early warning area, with a set data refresh frequency (real-time data refreshed once every 5 seconds).

[0060] III. Key Technological Support

[0061] Multi-source data fusion technology: unifies the data transmission of gates, integrates positioning base stations, AI capture devices, radar, environmental sensors, and intelligent early warning mechanisms, solves the problem of data format differences between different hardware, and realizes real-time interoperability and correlation analysis of identity data, location data, image data, machinery data, and environmental data, providing data support for the safety and violation determination of operators.

[0062] Role-based 3D visualization mapping technology: Establishes multiple mapping relationships between "personnel role - label color - 3D model coordinates - machine / environment status", accurately projects positioning data onto the 3D scene through coordinate calculation algorithms, and supports interactive operations such as scene scaling, translation, and rotation to ensure the intuitiveness and accuracy of visualization.

[0063] Personnel data and work status linkage technology: By constructing a unique association key through personnel ID, the static file data (qualification, salary) of the data management module is deeply bound with the dynamic work data (location, access, violation, fatigue, environment), realizing the linkage effect of "querying personnel information to display work status, and viewing work status to associate file data".

[0064] Bluetooth-IMU low-power multi-level base station cooperative positioning technology: Through the distributed layout of the main base station, the work surface base station, and the hazard source base station, combined with IMU inertial navigation supplementation and tunnel multipath attenuation correction, it achieves blind-spot-free signal coverage in complex environments such as tunnels and underground mines. Compared with GPS positioning, it is more stable and has no attenuation. With the low-power hardware design, it achieves the technical effects of stable positioning, ultra-long battery life, convenient deployment, and offline second-level alarm, fully meeting the construction control requirements of two subway stations and one section.

[0065] Edge-cloud collaborative computing technology: The edge is responsible for local positioning and calculation, real-time early warning, and self-healing after network outage, while the cloud is responsible for big data analysis, model iteration, and knowledge graph management, solving industry pain points such as unstable network and data transmission delay in subway stations.

[0066] LSTM trajectory prediction and early warning technology: Based on the historical trajectory of personnel, short-term trend prediction is performed to achieve early identification and proactive warning of dangerous behaviors, upgrading from "post-event handling" to "pre-event prevention".

[0067] IV. Data Integration and Processing Methods

[0068] After completing hardware-level data acquisition, this system transforms raw information into usable controllable information through core integration methods such as multi-source data standardization, spatiotemporal alignment, unique identity binding, and regional permission matching. First, it unifies the formats and converts the protocols of gate access data, Bluetooth-IMU fusion positioning data, AI-analyzed image data, radar machinery data, environmental monitoring data, and construction site planning data, converting data from different sources and frequencies into a system-recognizable standard structure, completing data cleaning, deduplication, and verification. Based on this, it unifies and aligns timestamps, enabling synchronous association of personnel location, behavior status, access records, machinery status, environmental parameters, and planning information at the same time. Then, using the personnel ID as a unique index, it deeply binds static data such as basic personnel information, qualification files, education and training, attendance, and wages with dynamic data such as real-time location, work area, violation records, trajectory paths, and fatigue status, forming a complete data entity with a unique file for each person. Simultaneously, the system will integrate positioning coordinates with the 3D model of the construction site for spatial mapping, completing the conversion from coordinates to areas and from areas to risk levels. In conjunction with the construction site plan, it will perform real-time matching and updating of personnel work permissions, permitted work periods, and permitted activity ranges, forming a standardized dataset that can be used for location display, permission judgment, violation identification, risk prediction, and early warning analysis.

[0069] V. System Early Warning Analysis Methods

[0070] This system employs a composite analysis method combining rule-based judgment, threshold calculation, scene matching, and LSTM trajectory prediction for early warning analysis. A multi-level early warning model is established based on the risk characteristics of overlapping construction at subway stations. The system reads and integrates location data, access control data, area boundary data, AI recognition results, radar data, and environmental data in real time. First, it determines whether personnel are within the permitted work area according to the work plan. If they exceed the preset boundary, it triggers an out-of-bounds work warning. Based on the LSTM model, it predicts personnel out-of-bounds trends in advance, achieving proactive early warning. Second, it performs statistical analysis of the on-site personnel structure. When the ratio of management personnel to workers falls below a set threshold, it triggers an on-site management shortage warning. For high-risk work areas, the system monitors in real time whether there are on-site management personnel present. If no on-site management personnel are present for a set period, a high-risk early warning is activated. The system continuously monitors temporary workers and visitors to determine if they are accompanied by management personnel throughout their visit. If no accompaniment is provided for an extended period, a visitor safety warning is automatically triggered. Simultaneously, it monitors for spatiotemporal conflicts in work processes and overloaded personnel density in areas, promptly triggering gathering and conflict warnings. The AI ​​image capture module identifies behaviors such as not wearing safety helmets, not wearing reflective clothing, or illegally entering dangerous areas, and immediately uses image feature matching and behavioral modeling to generate unsafe behavior warnings. Radar monitors the distance between personnel and machinery in real time, triggering collision avoidance warnings when thresholds are reached. The environmental monitoring module collects parameters exceeding standards, and if personnel are in the corresponding area, an environmental safety warning is triggered. All warnings are categorized according to risk level and pushed to the corresponding management personnel via preset paths, forming a complete analytical closed loop of warning generation, push, handling, feedback, and archiving.

[0071] VI. Complete Data Flow and Execution Process

[0072] Step 1: System Initialization

[0073] Deploy hardware layer equipment (gates, Bluetooth-IMU fusion positioning base stations, AI capture devices, mechanical sensing radar, and environmental monitoring modules) to ensure coverage of all work areas; complete the configuration of software layer modules, connect to data acquisition interfaces, and set permission rules, visualization mapping rules, early warning thresholds, and LSTM prediction model parameters.

[0074] Step 2: Basic Data Entry and Permission Assignment

[0075] Managers log in to the system via the web interface, access the data management module, and enter basic personnel information (name, age, contact information, role type), qualification information (safety briefing records, scanned copies of Level 3 education certificates, special operation certificate numbers), salary payment records, etc.

[0076] Enter the access control module and input the construction plan information, including the construction project name, work area (marked with boundary coordinates), work time period, and list of participants (associated by personnel ID). The system will automatically generate a unique access list based on personnel roles and the construction plan (e.g., visitors are limited to the ground viewing area, while safety officers can access all areas).

[0077] After the permission list is generated, the system automatically triggers a synchronization command, pushes the permission information to the corresponding turnstile through the data acquisition module, stores the mapping relationship between personnel ID and allowed access area, writes the allowed work area information into the positioning tag, and the system sends a "synchronization successful" prompt after the synchronization is completed.

[0078] Step 3: On-site data acquisition and real-time data transfer

[0079] Construction workers wear location tags to go to the work area. When passing through the turnstile, if the permissions match, the turnstile allows passage and records the passer ID, time, and channel location data, which is then uploaded to the data acquisition module in real time; if the permissions do not match, passage is denied, and a prompt is triggered in the local system.

[0080] After personnel enter the work area, the positioning tag sends a signal to the surrounding fusion positioning base station every 5 seconds. Combined with IMU inertial navigation data, the positioning base station calculates the precise coordinates of the personnel through RSSI ranging + tunnel multipath attenuation correction + triangulation positioning algorithm, uploads it to the data acquisition module, and pushes it to the positioning visualization module after edge processing, and maps it to the 3D model in real time.

[0081] The AI ​​capture module continuously collects on-site images and analyzes in real time the wearing status of personnel's protective equipment, work positions, and fatigue status. If a violation is detected, it immediately uploads the violation image, time, and location data to the data acquisition module, triggering the activation of the early warning module.

[0082] The mechanical sensing radar module collects the location and operating posture of construction machinery in real time, while the environmental monitoring module collects regional environmental parameters simultaneously and uploads them to the data acquisition module for fusion processing.

[0083] Step 4: Visualization and Early Warning Response

[0084] The interactive layer's large screen receives information pushed by the data acquisition module in real time. The core indicator area dynamically updates the total number of personnel, the number of each role, the cumulative number of passes, the number of violation warnings, the operating status of machinery, and environmental parameters. The 3D positioning view area displays the real-time location of personnel, the location of machinery, and risk areas. Different roles are marked with preset colors, and the labels of violators and risk areas flash continuously. The passage record area displays personnel passage details in chronological order. The warning prompt area pops up violation information (including personnel labels, violation type, risk level, and violation image).

[0085] Managers can perform interactive operations through the large screen: clicking on the personnel tag in the 3D model will take them to the details page to view the personnel's complete information (basic information, qualification file, salary record, fatigue status) and work status (current location, travel trajectory, violation history, environmental data); entering the personnel ID or name will allow them to quickly retrieve information about the target personnel; and they can view the distribution of personnel, machinery, and environmental status in different areas.

[0086] Once the warning module is triggered, in addition to the pop-up notification on the large screen, the system automatically sends an SMS notification to the corresponding manager. After receiving the warning, the manager can view the details of the violation through a mobile or web terminal, arrange on-site personnel to verify and handle the matter, and enter the handling result into the system after the handling is completed (such as "ordered to rectify, safety education has been conducted"), thus forming a closed loop for warning handling.

[0087] Step 5: Data Traceability and System Maintenance

[0088] Data traceability: Managers can query historical data through the system, including personnel's work trajectory, access records, violation records, fatigue records, environmental data, as well as personnel information change records and construction plan adjustment records. It supports filtering by time, region, personnel ID and other conditions.

[0089] In summary, based on the above description, the technical solution of this embodiment can achieve the following technical effects: (1) The integrated system architecture of “one person, one file safety knowledge graph construction + entry legal verification + regional work surface Bluetooth-IMU fusion positioning technology + AI violation and fatigue detection + construction machinery collaborative perception + environmental risk monitoring + post-event safety education data management” breaks through the traditional module dispersion barrier through deep collaboration of hardware layer, software layer and interaction layer, realizes multi-functional centralized integration and data interoperability, and solves the “information island” problem of existing technologies.

[0090] (2) The fixed connection and data flow mechanism of “hardware layer → data acquisition module → core processing module → interaction layer / early warning module”, combined with edge-cloud collaboration and network outage self-healing, constructs a closed loop of the whole link from data acquisition, processing to display, early warning and risk prediction, to ensure efficient linkage of each link.

[0091] (3) Role-based 3D positioning visualization display method: By embedding role identifiers in the positioning tags, a multi-mapping rule of "role type - preset color - 3D model coordinates - machine position - environmental parameters" is established to realize intuitive differentiation and location presentation of different roles (labor / management / safety / visitor), construction machinery, and environmental status.

[0092] (4) Six-fold consistency verification of permissions, location, qualifications, plans, process conflicts, and personnel density overload enables accurate judgment of violations such as overstepping boundaries, lack of qualifications, lack of plans, process conflicts, and personnel gathering, greatly improving the accuracy of safety management and control of cross-construction.

[0093] (5) The "overview-details" two-level interactive architecture of the visualization screen supports real-time display of personnel information, machine status and environmental parameters, viewing of three-dimensional positioning view and related query of details page, realizing the "what you see is what you get" management interaction experience and improving the efficiency of on-site status control.

[0094] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0095] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A smart safety management and control system for personnel engaged in cross-construction work at subway stations, characterized in that, It includes a hardware layer, a software layer, and an interaction layer. Each layer works together in real time to build a closed loop for full-process control, as detailed below: The hardware layer, as a multi-dimensional sensing terminal, is used to collect operator identity information, location data, behavioral images, machine operating status and environmental parameters, and transmits the raw data to the software layer in real time through a standardized interface. The software layer adopts an edge-cloud collaborative architecture, serving as the core processing hub of the system. It receives raw data transmitted from the hardware layer and performs standardized processing, analysis, storage, and correlation integration to achieve job status determination, violation identification, risk prediction, and early warning instruction generation. The processed valid data and instructions are then pushed to the interaction layer and hardware layer, respectively. The interaction layer serves as the management interaction carrier, receiving valid data pushed by the software layer and providing a visual display of personnel information, work status, early warning information, and equipment status, as well as a data query portal. This allows managers to intuitively grasp the on-site control situation and perform interactive operations. The software layer includes: a data acquisition module, a permission management module, a location visualization module, a data management module, an early warning module, and a core processing module; The data acquisition module is configured to perform collaborative fusion processing of multi-source heterogeneous data, which is used to standardize, normalize and align timestamps and perform spatial adaptive grid mapping on various types of raw data to form a standardized dataset with a unified spatiotemporal benchmark. The core processing module is configured with a six-fold consistency check algorithm for execution permissions, location, qualifications, plans, process conflicts, and personnel density overload. This algorithm is used to determine in real time the matching status between the current location of personnel and the allowed work area, the validity of personnel qualifications, the construction plan time period, the spatiotemporal compatibility of processes, and the personnel density in the area. It automatically determines violations such as boundary crossing, unplanned operation, unqualified operation, process conflicts, and personnel gathering. The early warning module is configured to execute a composite early warning analysis method based on dynamic thresholds, duration, confidence filtering, and LSTM trajectory prediction. It is used to conduct graded early warnings and advanced early warnings based on personnel management ratio, on-site supervision duration, temporary personnel accompaniment distance, AI violation confidence, and personnel trajectory trends, and push them to the corresponding management personnel according to the risk level.

2. The metro station cross operation personnel safety intelligent management and control system according to claim 1, characterized in that, The hardware layer includes: a gate module, a Bluetooth-IMU fusion positioning base station module, a personnel positioning tag module, an AI capture module, a mechanical sensing radar module, and an environmental monitoring module. Each module is deployed in a distributed manner to cover key construction nodes and work areas. The gate module is used for real-time comparison of access permissions and recording of access data; The personnel positioning tag module has a built-in identity recognition chip, role identification chip, gyroscope and accelerometer, supports two-way communication with the base station and receives warning commands to trigger vibration prompts; at the same time, it adopts an adaptive sampling and wake-up mechanism for work status, low-power sleep sampling during non-work periods, and automatically switches to high-frequency positioning mode when entering the work area; when the tag falls off, is powered off or the signal is blocked, an offline alarm is automatically triggered. The AI ​​capture module is used to collect on-site images and detect violations and personnel fatigue. The Bluetooth-IMU fusion positioning base station module adopts a distributed deployment approach, consisting of a main base station, a work site base station, and a hazard source base station, to achieve full-area coverage without blind spots. It achieves tight-coupled and accurate Bluetooth-IMU positioning through RSSI ranging, triangulation, tunnel multipath attenuation correction, and the least squares method. The distance calculation formula for Bluetooth-IMU fusion positioning is as follows: In the formula: RSSI This indicates the received signal strength; a negative value is used. A This indicates the signal strength when the transmitter and receiver are 1 meter apart. n This represents the environmental attenuation factor, which is dynamically corrected based on tunnel curvature, wall material, and distance from the arch. Position compensation and drift correction are performed using IMU inertial navigation data, and finally, the precise coordinates of the workers are obtained using the least squares method. The mechanical sensing radar module is used to detect the position, operating posture and movement trajectory of construction machinery in real time, so as to realize the collaborative perception of personnel-machine collision avoidance; The environmental monitoring module is used to collect real-time data on toxic gas concentrations, temperature, humidity, and dust parameters in the work area, and trigger environmental risk warnings in conjunction with personnel location.

3. The metro station cross operation personnel safety intelligent management and control system according to claim 1, characterized in that, The data acquisition module supports multi-protocol access and performs dedicated data cleaning for subway construction scenarios on the raw data, including deduplication, noise reduction, signal drift correction, and invalid data removal, and uses AES-256 encrypted transmission; the edge terminal realizes positioning calculation, real-time early warning and network outage self-healing, while the cloud is responsible for big data analysis, trajectory backtracking, report generation and model iteration.

4. The intelligent safety management and control system for personnel engaged in cross-construction work at subway stations according to claim 1, characterized in that, Timestamp normalization alignment specifically involves aligning all collected data to a uniform 5-second time slice: In the formula: is the aligned standard timestamp, is the original timestamp with hardware collection, and 5s is the normalization step, is a rounding function. The spatial adaptive grid mapping specifically involves dynamically dividing the grid cells according to the risk level of the work area: the hazardous area uses a 0.5m×0.5m high-precision grid, the ordinary area uses a 2m×2m standard grid, and the passageway area uses a 1m×4m long strip grid. The mapping formula is as follows: In the formula: is the mapped standard grid coordinate, is the real position of the personnel located by the Bluetooth-IMU fusion, L , W is the dynamic grid width and length, is a floor function; Based on the above operations, a unique correspondence is achieved for the same person, the same time, the same location, and the same behavior.

5. The intelligent safety management and control system for personnel engaged in cross-construction work at subway stations according to claim 1, characterized in that, The core processing module's six-fold consistency check algorithm includes: verifying whether the personnel's current location is within the permitted work area allocated by the access management module; verifying whether the personnel's qualifications are valid; verifying whether the current time is within the permitted work period of the construction plan; verifying whether there are on-site personnel at high-risk work sites; verifying whether there are spatiotemporal conflicts between the current work process and overlapping work processes; and verifying whether the personnel density in the current area exceeds the safety threshold. Any mismatch is marked as an anomaly, and this is continued... N If the cycles do not match, it is considered a violation.

6. The intelligent safety management and control system for personnel engaged in cross-construction work at subway stations according to claim 1, characterized in that, The composite early warning analysis methods executed by the early warning module include: If the ratio of management personnel to operational personnel (R) is less than 10% and continues to exceed the set duration, an alert for insufficient management capacity will be triggered. If a personnel supervising a high-risk operation are absent for more than 10 minutes, a major risk warning will be triggered. If a temporary visitor / management staff member is more than 30 minutes away from management personnel without being accompanied, a visitor safety alert is triggered. When the confidence level of AI-captured violation is ≥95%, it is determined as a valid violation and an alert is triggered; Based on the LSTM model, short-term prediction of personnel trajectories is performed, and the tendency to cross the boundary is identified 10 seconds in advance and an early warning is triggered. If the distance between personnel and machinery is less than the safety threshold and the duration exceeds the set value, a collision avoidance warning is triggered. An environmental safety warning was triggered when environmental parameters in the work area exceeded the standard and personnel were in that area.

7. The intelligent safety management and control system for personnel engaged in cross-construction work at subway stations according to claim 1, characterized in that, The access control module adopts a zero-trust dynamic access control mechanism, which is deeply integrated with the construction schedule. It automatically derives and dynamically assigns work permissions based on construction procedures, work areas, risk levels, personnel roles and qualifications. It performs a full-dimensional verification every 30 seconds. When the construction schedule changes, the permissions are automatically updated. When the schedule ends, the qualification expires, the location is out of bounds, or there is a conflict in the procedures, the permissions are automatically revoked.

8. The intelligent safety management and control system for personnel engaged in cross-construction work at subway stations according to claim 1, characterized in that, The positioning visualization module is based on a 3D model of the construction area and uses a triple mapping rule of role-color-coordinate to dynamically map the real-time location, trajectory, density, machine position, and environmental parameters of personnel to the 3D scene. It also supports trajectory playback, view zooming, rotation, and highlighting of risk areas.

9. The intelligent safety management and control system for personnel engaged in cross-construction work at subway stations according to claim 1, characterized in that, The data management module uses a unique personnel ID as an index to construct a safety knowledge graph that links static files with dynamic work status, creating a one-person-one-file system. This system links personnel basic information, qualification documents, safety briefings, education records, attendance and wages, location tracking, pre-shift safety talks, access records, violation records, and fatigue work records in real time. It also connects to the safety big data model interaction module, supporting compliance queries, risk analysis, and automatic generation of rectification suggestions.

10. The intelligent safety management and control system for personnel engaged in cross-construction work at subway stations according to claim 1, characterized in that, The interaction layer is centered around a large visual screen, employing a two-level display interface of overview and details; The overview interface displays the number of personnel, role distribution, number of violation warnings, area personnel density, machinery operating status, and environmental parameters in real time, and highlights violators and risk areas with flashing markers. The details page displays complete personnel files, real-time location, historical trajectory, work plan, violation handling records, fatigue status, and environmental data. It supports filtering and querying by region, role, and time.

Images