An edge-computing-based high-altitude worker safety intelligent monitoring method and system

Abstract

The application discloses a kind of high-altitude operation personnel safety intelligent monitoring method and system based on edge computing, it is related to high-altitude operation safety monitoring technical field, comprising: acquisition original perception data;Original perception data is time-synchronized and space-aligned, and obtains alignment perception data;Based on the alignment perception data fusion obtains fusion situation data;The fusion situation data is judged, and obtains risk determination result;According to the risk determination result, generate disposal instruction on edge side;The disposal instruction, risk determination result and original perception data are associated storage, and obtain traceability record.The application continuously associates safety belt state, attitude change, position trajectory, dangerous boundary and on-site image, completes progressive risk determination and hierarchical alarm on edge side, whereby can identify high-altitude operation high-risk state such as unhooking, instability, overrunning and edge approaching earlier, and avoid misjudgment caused by instantaneous jitter, short-time shielding or single-point exception.

Description

Technical Field

[0001] This invention relates to the field of high-altitude operation safety monitoring technology, specifically to a method and system for intelligent monitoring of high-altitude operation personnel safety based on edge computing. Background Technology

[0002] In recent years, existing high-altitude operation safety monitoring technologies have gradually evolved from simple manual inspections and video surveillance to a comprehensive regulatory model integrating wearable terminals, video monitoring, location tracking, and remote alarms. Some solutions can collect information on the wearing status, location, or on-site images of workers and upload the data to a platform for centralized management via wireless communication, thereby improving the visualization and manageability of high-altitude operations to a certain extent. With the development of edge computing technology, some monitoring devices can also complete preliminary identification and alarm processing on-site, reducing the transmission pressure on the cloud and improving response efficiency.

[0003] However, existing technologies typically focus on judging single data sources or discrete events, lacking collaborative analysis of the continuous relationship between wearer status, posture changes, location trajectory, dangerous boundaries, and on-site images. This results in the identification of high-altitude operation risks such as detachment, instability, boundary crossing, and approaching the edge still relying heavily on single-point anomalies, which are easily affected by instantaneous shaking, short-term obstruction, or network fluctuations, thus affecting the accuracy of alarms and the continuity of handling.

[0004] In response to this, this application proposes an intelligent monitoring method and system for the safety of high-altitude workers based on edge computing, in order to solve the above-mentioned problems. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for intelligent monitoring of the safety of high-altitude workers based on edge computing. This addresses the problem that existing technologies focus on judging single data sources or discrete events, lacking collaborative analysis of the continuous relationship between wearing status, posture changes, location trajectory, dangerous boundaries, and on-site images. As a result, the identification of high-altitude work risks such as desnagging, instability, boundary crossing, and approaching the edge still relies heavily on single-point anomalies and is easily affected by instantaneous shaking, short-term obstruction, or network fluctuations, thus affecting the accuracy of alarms and the continuity of handling.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Firstly, this application provides a method for intelligent monitoring of the safety of high-altitude workers based on edge computing, including:

[0008] Collect data on the clothing status, posture, location trajectory, and on-site images of high-altitude workers to obtain raw sensing data;

[0009] The original sensing data is time-synchronized and spatially aligned to obtain aligned sensing data;

[0010] Based on the alignment perception data, the operation height, attachment point status, and dangerous boundary information are fused to obtain fused situational data;

[0011] The fused situational data is analyzed to determine whether it is detached, unstable, crosses the boundary, or approaches a dangerous boundary, and the risk assessment results are obtained.

[0012] At the edge, based on the risk assessment results, a tiered alarm and linkage response command are generated to obtain the response command.

[0013] The disposal instructions, the risk assessment results, and the original perception data are associated and stored to obtain a traceability record.

[0014] Furthermore, the raw sensing data is collected synchronously by wearable terminals, head-mounted terminals, vision terminals, and ranging terminals, and a unified high-altitude coordinate benchmark is established based on the current hanging point position of the personnel, the edge position of the working surface, and the position of the support surface under their feet, so as to eliminate clock drift and spatial deviation of multi-source data.

[0015] Furthermore, the alignment perception data is formed by using the relative poses between the person's head, chest and back, hands and hanging points as constraints, and correcting for continuous occlusion, instantaneous frame loss or posture jumps, so that the alignment perception data maintains a continuous high-altitude operation action chain.

[0016] Furthermore, the fused situational awareness data is obtained by calculating the fall margin index M, which satisfies:

[0017] M=α·(Hh) / H+β·d / D+γ·T-δ·θ

[0018] Where H is the working reference height, h is the current height of the personnel, d is the minimum distance from the personnel to the danger boundary, D is the characteristic scale of the working area, T is the effective attachment duration, θ is the attitude offset angle, and α, β, γ, and δ are weighting coefficients; when M is lower than the preset threshold, the fused situational data enters the instability risk branch.

[0019] Furthermore, the risk determination result is jointly triggered by abnormal attitude, changes in hanging point state, and approach of dangerous boundary in several consecutive frames. Any single frame anomaly only forms a candidate risk, and the risk determination result is only output when multiple adjacent frames satisfy the same direction of cumulative anomaly.

[0020] Furthermore, the tiered alarm and linkage response instructions include at least wearable terminal vibration alerts, head-mounted terminal audio and visual prompts, on-site voice prompts, and work stop instructions;

[0021] When the risk assessment result continues to exceed the time threshold, the handling instruction is upgraded from the reminder level to the forced evacuation level.

[0022] Furthermore, when the communication link is interrupted, the edge side performs event-based caching of the original sensing data and risk determination results, and retains only the key frames, key pose segments and key location points related to the risk; after the communication is restored, the event-based cache is transmitted back in chronological order to avoid the full data being uploaded repeatedly.

[0023] Furthermore, the hazardous boundary information is dynamically generated based on the operation type, height level, support structure form, and environmental slope. Different operation types correspond to different boundary threshold combinations, so that the risk assessment results in the fused situational data match the actual high-altitude scenario.

[0024] Furthermore, the traceability records are associated and numbered according to the event sequence of attachment point switching—attitude change—boundary approach—alarm triggering—response execution, and the corresponding on-site image fragments, location trajectories and risk assessment results are retained to form a replayable high-altitude operation safety chain.

[0025] Secondly, this application provides an intelligent safety monitoring system for high-altitude workers based on edge computing, comprising:

[0026] The sensing data acquisition module is used to collect the wearing status, posture status, position trajectory and on-site images of high-altitude workers to obtain raw sensing data;

[0027] The data preprocessing module is used to perform time synchronization and spatial alignment on the raw sensing data to obtain aligned sensing data;

[0028] The situation fusion module is used to fuse the operation height, attachment point status and danger boundary information based on the alignment perception data to obtain fused situation data;

[0029] The risk assessment module is used to assess the fused situational data for detachment, instability, boundary crossing, and proximity to dangerous boundaries, and to obtain risk assessment results.

[0030] The edge processing module is used to generate graded alarms and linkage processing instructions at the edge based on the risk assessment results, and obtain processing instructions;

[0031] The traceability storage module is used to associate and store the disposal instructions, the risk assessment results, and the original sensing data to obtain traceability records.

[0032] Compared with existing technologies, this invention provides a method and system for intelligent monitoring of high-altitude work safety based on edge computing. This invention continuously correlates safety belt status, posture changes, position trajectory, hazardous boundaries, and on-site images, completing progressive risk assessment and graded alarms at the edge. This enables earlier identification of high-risk conditions in high-altitude operations such as detachment, instability, boundary crossing, and approaching edges, and avoids misjudgments caused by momentary vibrations, short-term obstructions, or single-point anomalies, forming a closed loop for risk detection, alerts, and handling. This invention moves key judgments forward to the edge computing side, maintaining rapid local response even in situations with unstable communication, complex on-site environments, or concurrent access from multiple terminals. It also preserves the key data chain before and after alarms through event-based caching and associated storage. This ensures uninterrupted on-site alarms and provides complete evidence for subsequent playback analysis, responsibility verification, safety training, and management optimization, thereby elevating high-altitude work supervision from "post-event recording" to a technical model that emphasizes both "real-time intervention and full-process traceability." Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0034] Figure 1 A flowchart of a method for intelligent monitoring of the safety of high-altitude workers based on edge computing, provided in an embodiment of the present invention;

[0035] Figure 2 This is a block diagram of a high-altitude worker safety intelligent monitoring system based on edge computing, provided as an embodiment of the present invention. Detailed Implementation

[0036] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0037] As attached Figure 1 As shown:

[0038] Example 1:

[0039] A method for intelligent monitoring of the safety of high-altitude workers based on edge computing includes:

[0040] S1. Collect the clothing status, posture status, position trajectory and on-site images of high-altitude workers to obtain raw perception data;

[0041] In step S1, the raw sensing data is collected synchronously by wearable terminal, head-mounted terminal, vision terminal and ranging terminal, and a unified high-altitude coordinate reference is established based on the current hanging point position of the personnel, the edge position of the working surface and the position of the support surface under their feet, so as to eliminate the clock drift and spatial deviation of multi-source data.

[0042] Specifically, raw perception data refers to first-hand on-site data that has not yet been fused or judged. This typically includes seatbelt buckle status, attachment point connection status, helmet wearing status, human posture parameters, real-time position coordinates, distance values ​​to hazard boundaries, and on-site video / image frames. In implementation, wearable terminals continuously output status quantities, visual terminals output image sequences, ranging terminals output spatial distances, and positioning terminals output trajectory points, all collected synchronously at a unified sampling period. After collection, noise reduction, outlier removal, encoding compression, and field standardization are performed to form a data stream suitable for subsequent processing.

[0043] S2. Perform time synchronization and spatial alignment on the original sensing data to obtain aligned sensing data;

[0044] In step S2, the alignment perception data is formed by using the relative poses between the person's head, chest and back, hands and hanging points as constraints, and correcting the data that is continuously occluded, momentarily lost frames or changes in posture, so that the alignment perception data maintains a continuous high-altitude operation action chain.

[0045] Specifically, step S2 transforms data from different devices, installation locations, and sampling frequencies into comparable data at the same time and under the same spatial semantics. Time synchronization can be achieved through a unified clock source, periodic time correction, and timestamp recalibration; spatial alignment can be achieved through a preset high-altitude operation coordinate system, installation point calibration, and target position mapping. For example, the footage captured by the safety helmet camera, the distance measured by radar, and the vital signs uploaded by the wristband should correspond to the same person and the same work point, rather than existing independently. After alignment, continuous and complete aligned sensing data can be obtained.

[0046] S3. Based on the alignment perception data, the operation height, attachment point status, and dangerous boundary information are fused to obtain fused situational data;

[0047] In step S3, the fused situational data is obtained by calculating the fall margin index M, which satisfies the following:

[0048] M=α·(Hh) / H+β·d / D+γ·T-δ·θ

[0049] Where H is the working reference height, h is the current height of the personnel, d is the minimum distance from the personnel to the danger boundary, D is the characteristic scale of the working area, T is the effective attachment duration, θ is the attitude offset angle, and α, β, γ, and δ are weighting coefficients; when M is lower than the preset threshold, the fused situational data enters the instability risk branch;

[0050] Specifically, during the fusion process, a risk relationship chain is established for high-altitude operations. The working height characterizes the personnel's current spatial level above the ground, platform, or supporting structure; the attachment point status determines whether the safety rope is effectively stressed, whether a switch has occurred, and whether there is a loose or detached attachment; hazard boundary information includes edges, holes, outer edges, restricted areas, and other high-risk spatial ranges. During processing, changes in height, attachment points, and boundary distances can be incorporated into the same risk scenario, enabling the system to determine whether the personnel are moving normally, approaching the boundary, or have entered a high-risk fall state, thus obtaining fused situational data.

[0051] S4. Determine whether the fused situational data is detached, unstable, crosses the boundary, or approaches a dangerous boundary to obtain a risk assessment result;

[0052] In step S4, the risk determination result is jointly triggered by abnormal attitude, changes in hanging point state and approach of dangerous boundary in several consecutive frames. Any single frame anomaly only forms a candidate risk, and the risk determination result is only output when multiple adjacent frames satisfy the same direction of cumulative anomaly.

[0053] The hazardous boundary information is dynamically generated based on the operation type, height level, support structure form and environmental slope. Different operation types correspond to different boundary threshold combinations so that the risk assessment results in the fused situational data match the actual high-altitude scenario.

[0054] Specifically, "detachment" refers to the failure of safety protection relationships, such as hook detachment, loose connection points, or incomplete attachment point switching; "instability" refers to a significant shift in the person's center of gravity, increased body swaying, and insufficient support surface, indicating an impending fall; "crossing the boundary" refers to a person entering a pre-defined danger zone; and "approaching the danger boundary" refers to a person's distance from edges, holes, outer scaffolding, etc., falling below a threshold. In implementation, candidate identification can be performed on a single frame of data first, followed by confirmation based on continuous changes across several preceding and following frames, preventing false alarms caused by momentary vibrations, obstructions, wind, etc. The resulting risk assessment includes not only the risk type but also the risk level and duration.

[0055] S5. Generate graded alarms and linkage handling instructions on the edge side based on the risk assessment results, and obtain handling instructions;

[0056] In step S5, the tiered alarm and linkage response instructions include at least the wearable terminal vibration reminder, the head-mounted terminal audio and visual prompt, the on-site voice prompt, and the operation pause instruction;

[0057] When the risk assessment result continues to exceed the time threshold, the handling instruction is upgraded from the reminder level to the mandatory evacuation level;

[0058] When the communication link is interrupted, the edge device caches the original sensing data and risk assessment results in an event-based manner, and only retains key frames, key pose segments and key location points related to the risk; after the communication is restored, the event-based cache is sent back in chronological order to avoid uploading all data repeatedly.

[0059] Specifically, edge computing emphasizes local computing closer to the work site, aiming to move high-risk assessments forward and reduce cloud communication latency. Tiered alerts take different actions based on the severity of the risk: low-level alerts primarily provide warnings, medium-level alerts include voice, visual, or vibration prompts, and high-level alerts can trigger mandatory evacuation, work suspension, or notification of management personnel.

[0060] The coordinated response command transforms the judgment result into an executable command, such as instructing the safety helmet to issue a voice announcement, the wristband to vibrate, the on-site equipment to issue an audible and visual alarm, and the platform to push alarm information. This enables a synchronous closed loop of risk detection, on-site alerts, and management intervention.

[0061] S6. Associate and store the disposal instruction, the risk assessment result, and the original sensing data to obtain a traceability record;

[0062] In step S6, the traceability records are associated and numbered according to the event sequence of attachment point switching—attitude change—boundary approach—alarm triggering—handling execution, and the corresponding on-site image segments, location trajectories and risk assessment results are retained to form a replayable high-altitude operation safety chain;

[0063] Specifically, step S6 focuses on association. The system links key images, location points, attitude changes, attachment point status, judgment conclusions and handling results before and after the alarm according to the same event number, forming a replayable event chain.

[0064] To balance storage efficiency and traceability integrity, only key frames, key trajectory points, and key state variables from a few seconds before and after the risk is triggered can be retained, rather than retaining all the original data. This supports post-event analysis, accountability verification, and training reviews, while also ensuring data manageability over long-term operation.

[0065] As can be seen from the above, this invention does not only judge a single wearing state or a single location data, but continuously correlates the safety belt status, posture changes, position trajectory, dangerous boundaries, and on-site images to complete progressive risk assessment and graded alarms at the edge. This enables earlier identification of high-risk conditions in high-altitude operations such as detachment, instability, boundary crossing, and approaching the edge, and avoids misjudgments caused by momentary shaking, short-term obstruction, or single-point anomalies, thus forming a closed loop for risk detection, alerts, and handling.

[0066] This invention moves key decision-making to the edge computing side, maintaining rapid local response even in situations with unstable communication, complex on-site environments, or concurrent access from multiple terminals. It also preserves the critical data chain before and after an alarm through event-based caching and associated storage. This ensures uninterrupted on-site alarms while providing complete evidence for subsequent playback analysis, accountability verification, safety training, and management optimization. Thus, the supervision of high-altitude operations is elevated from "post-event recording" to a technical model that emphasizes both "real-time intervention and full-process traceability."

[0067] Example 2:

[0068] This embodiment is specifically applied to the inspection, cleaning, and maintenance of the exterior facade of high-rise buildings.

[0069] Workers wear smart safety belt monitoring terminals, smart wristbands, and smart safety helmets. Visual acquisition devices, spatial ranging devices, and edge computing nodes are deployed in the work area. The edge computing nodes establish a data interaction relationship with the cloud platform, thereby forming a complete processing link from on-site perception to cloud storage.

[0070] Before the operation begins, the management personnel preset the task number in the platform and enter the identity information of the operators, the working height range, the location of the dangerous boundary and the permitted activity area. At the same time, each terminal is bound to the corresponding operator so that the subsequent data can be associated and stored based on the same task number.

[0071] In step S1, the intelligent safety belt monitoring terminal is used to collect information on the safety belt buckle status, attachment point connection status, and attachment point switching; the intelligent wristband is used to collect personnel vital sign information; the intelligent safety helmet is used to collect images of the work site and personnel head posture information; the visual acquisition device is used to collect overall images of the work area; and the spatial ranging device is used to collect distance information between personnel and the work boundary.

[0072] Various types of data are collected synchronously according to a preset sampling period. The sampling frequency for seat belt status can be set to 10Hz, the sampling frequency for attitude data can be set to 5Hz, the sampling frequency for near-boundary distance data can be set to 10Hz, the video frame rate can be set to 25fps, the sampling frequency for vital signs can be set to once every 30 seconds, and the sampling frequency for positioning data can be set to once every 5 to 10 seconds.

[0073] The collected data is first processed for noise reduction, format standardization, and data compression to form raw sensory data.

[0074] In step S2, the raw sensing data undergoes time synchronization and spatial alignment processing. Specifically, the timestamps of data collected by different terminals are corrected using a unified time reference to ensure that all types of data are in the same time reference system; and a high-altitude operation spatial coordinate system is established to map the visual target position, ranging data, and positioning information to unify them into the same spatial reference frame.

[0075] After this processing, the data output by each terminal can be mapped to the same worker, the same work site, and the same hazard boundary, thus forming aligned perception data.

[0076] The purpose of this step is to eliminate data misalignment caused by inconsistent device sampling times, different installation locations, or differences in transmission delays.

[0077] In step S3, the alignment sensing data is fused with the operation height, attachment point status, and danger boundary information.

[0078] The working height is used to reflect the spatial position of personnel in the vertical direction, the hanging point status is used to reflect whether the safety protection connection is effective, and the danger boundary information is used to describe the location range of the edge area, hole area, restricted area or other high-risk area.

[0079] The system constructs a spatial relationship between "personnel, attachment point, and boundary" by uniformly modeling the operation height, attachment point status, and boundary distance. This allows the integrated situational data to reflect not only the current location of personnel but also the risk relationship status of personnel.

[0080] Therefore, subsequent risk assessment no longer relies on a single image or a single distance value, but rather on a continuous situation assessment based on multi-source information.

[0081] In step S4, the fused situational data is assessed for detachment, instability, boundary crossing, and proximity to dangerous boundaries.

[0082] Specifically, the system analyzes data from multiple frames within a continuous time window. When it detects abnormal connection of the attachment point, a continuous increase in attitude deviation, or a continuous decrease in the distance between personnel and the danger boundary, it determines the corresponding risk state and outputs the risk determination result and the corresponding risk level.

[0083] To reduce false alarms caused by wind disturbances, short-term obstruction, or normal body movements, the system adopts a continuity constraint when making judgments. That is, single-frame or short-term anomalies are only considered as candidate risks, and the risk judgment result is only confirmed and output when the same-direction anomaly accumulation condition is met in multiple adjacent time windows.

[0084] This method can improve the stability of risk identification while ensuring response speed.

[0085] In step S5, the edge computing node generates graded alarms and linkage handling instructions based on the risk assessment results.

[0086] When the risk level is low, the wearable terminal's vibration alert or voice reminder is triggered first; when the risk level increases, an on-site audio-visual alarm is triggered, and an alarm message is pushed to the cloud platform; when the risk persists or reaches a high-risk threshold, a mandatory handling command is generated to prompt personnel to evacuate, suspend operations, or request on-site support.

[0087] Since the determination and alarm process is completed at the edge, it can significantly reduce the round-trip latency to the cloud, improve the timeliness of alarms, and is suitable for high-altitude operation scenarios with unstable communication conditions or complex on-site environments.

[0088] In step S6, the disposal instruction, risk assessment result, and original sensing data are associated and stored.

[0089] Specifically, using the task number or event number as an index, key image frames, location trajectories, attitude changes, attachment point status, and handling actions before and after the alarm occur are associated to form a complete traceability record.

[0090] The traceability records can be used for subsequent playback analysis, safety debriefing and accountability verification, and can also provide a data foundation for work training, risk assessment and management optimization.

[0091] In the event of communication interruption, edge computing nodes can first cache key frames, key location points and key pose segments related to the event locally, and then upload them to the cloud platform after communication is restored, thereby ensuring the integrity of the event chain.

[0092] In one specific embodiment, taking the maintenance of the facade of a high-rise building as an example, there are a total of 8 workers, the working height is 18m to 96m, and the working time is 6.5 hours.

[0093] Four AI monitoring balls, two high-altitude operation radars, and two edge computing nodes were deployed on site. Each worker was equipped with a smart safety belt monitoring terminal, a smart wristband, and a smart safety helmet.

[0094] In each terminal, the sampling frequency for seat belt status is 10Hz, the sampling frequency for attitude data is 5Hz, the sampling frequency for near-boundary distance data is 10Hz, the video frame rate is 25fps, the sampling frequency for vital signs is once every 30 seconds, and the sampling frequency for positioning data is once every 5 to 10 seconds.

[0095] After being denoised, compressed, and formatted, the raw data enters the time synchronization and spatial alignment process to form aligned perception data, and further completes the construction of fused situational data and risk assessment.

[0096] In this embodiment, the system's average data reporting latency is 0.58s, the average edge alarm trigger time is 0.31s, and the total number of alarms is 24, of which 23 are valid alarms, 1 is a false alarm, and 0 are missed alarms. In the event of a short-term network interruption, the system can cache the event data locally for 12 minutes, and the retransmission success rate after network recovery is 100%.

[0097] In this embodiment, the edge warning threshold is set to 1.0m, and the strong alarm threshold is set to 0.6m; when the seat belt attachment point is continuously without force for more than 2 seconds, it is determined to be an abnormal attachment point; when the pitch offset angle of the person's torso continuously exceeds 30° and is accompanied by a continuous reduction in the boundary distance, it is determined to be an instability risk.

[0098] During the test, the system recorded 3 hook point switching, 7 edge approach incidents, and 2 short-term attitude imbalances. Among them, 2 were identified as early warning events and 1 was upgraded to a strong alarm event. All of these were handled in real time at the edge and did not evolve into a substantial risk of falling.

[0099] Because the system uses a continuous time window judgment method, it can effectively suppress short-term misjudgments caused by instantaneous jitter, wind disturbance or screen obstruction, and maintain a high alarm effectiveness.

[0100] In another embodiment, taking the operation scenario of maintenance and hoisting on the top of a steel structure factory as an example, the work area spans 68m, there are 6 workers, the work time is 4.2 hours, and 3 AI control balls, 1 high-altitude operation radar and 1 edge computing node are deployed on site.

[0101] Because the scene is characterized by strong metallic reflection, numerous local obstructions, and dynamic changes in the hoisting path, the system uses the roof edge, inspection holes, and hoisting path as a combined danger boundary and includes the swing radius control area of ​​the hoisted object in the dynamic no-entry zone.

[0102] In this embodiment, the number of hoisting operations was 5, the preset danger boundary classification zone was 0.8m to 2.5m, the radius of the dynamic restricted area was 2.5m, the average edge determination time was 0.34s, the average handling trigger time was 0.49s, the total number of alarms was 11, of which 11 were valid alarms and 0 were false alarms; in the event of a short network interruption, the system cached 19 events and the retransmission success rate was 100%.

[0103] In this embodiment, the system sets the roof edge warning threshold to 0.8m and the strong alarm threshold to 0.5m; the no-entry distance within the hoisting operation path is set to 2.5m.

[0104] During the test, the system accurately identified two instances of personnel accidentally entering the hoisting danger zone, four instances of personnel approaching the roof edge, and five instances of posture adjustments that triggered potential risks. Among these, four potential risks were suppressed because they did not meet the continuity condition, and no false alarms were generated. The two instances of personnel actually entering the hoisting restricted area were directly upgraded to strong alarms, triggering on-site voice reminders and platform notifications.

[0105] Therefore, this invention can balance the accuracy of risk identification and the timeliness of response in complex high-altitude operation environments, and can ensure the integrity of the event chain through edge-side caching and subsequent retransmission mechanisms.

[0106] As can be seen from the above, by continuously collecting, synchronizing, fusion judging, and edge linkage the personnel's wearing status, posture status, position trajectory, danger boundary, and on-site images in different high-altitude operation scenarios, it is possible to achieve real-time identification, immediate handling, and full-process traceability of high-altitude operation risks, and maintain good feasibility and adaptability under conditions of unstable communication or strong scene disturbances.

[0107] like Figure 2 As shown, in one embodiment, the present invention also provides a high-altitude worker safety intelligent monitoring system based on edge computing, comprising:

[0108] The sensing data acquisition module is used to collect the wearing status, posture status, position trajectory and on-site images of high-altitude workers to obtain raw sensing data;

[0109] The data preprocessing module is used to perform time synchronization and spatial alignment on the raw sensing data to obtain aligned sensing data;

[0110] The situation fusion module is used to fuse the operation height, attachment point status and danger boundary information based on the alignment perception data to obtain fused situation data;

[0111] The risk assessment module is used to assess the fused situational data for detachment, instability, boundary crossing, and proximity to dangerous boundaries, and to obtain risk assessment results.

[0112] The edge processing module is used to generate graded alarms and linkage processing instructions at the edge based on the risk assessment results, and obtain processing instructions;

[0113] The traceability storage module is used to associate and store the disposal instructions, the risk assessment results, and the original sensing data to obtain traceability records.

[0114] The beneficial effects of this method are the same as those of the embodiment of the intelligent monitoring method for the safety of high-altitude workers based on edge computing, and will not be repeated here.

[0115] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for intelligent monitoring of the safety of high-altitude workers based on edge computing, characterized in that, Includes the following steps: Collect data on the clothing status, posture, location trajectory, and on-site images of high-altitude workers to obtain raw sensing data; The original sensing data is time-synchronized and spatially aligned to obtain aligned sensing data; Based on the alignment perception data, the operation height, attachment point status, and dangerous boundary information are fused to obtain fused situational data; The fused situational data is analyzed to determine whether it is detached, unstable, crosses the boundary, or approaches a dangerous boundary, and the risk assessment results are obtained. At the edge, based on the risk assessment results, a tiered alarm and linkage response command are generated to obtain the response command. The disposal instructions, the risk assessment results, and the original perception data are associated and stored to obtain a traceability record.

2. The intelligent monitoring method for high-altitude worker safety based on edge computing according to claim 1, characterized in that, The raw sensing data is collected synchronously by wearable terminals, head-mounted terminals, vision terminals, and ranging terminals. A unified high-altitude coordinate benchmark is established based on the current hanging point position of the personnel, the edge position of the working surface, and the position of the support surface under their feet to eliminate clock drift and spatial deviation of multi-source data.

3. The intelligent monitoring method for high-altitude worker safety based on edge computing according to claim 1, characterized in that, The alignment perception data is formed by using the relative poses of the person's head, chest and back, hands and hanging points as constraints, and correcting for continuous occlusion, instantaneous frame loss or posture jumps, so that the alignment perception data maintains a continuous high-altitude operation action chain.

4. The intelligent monitoring method for high-altitude worker safety based on edge computing according to claim 1, characterized in that, The fused situational data is obtained by calculating the fall margin index M, which satisfies the following: M=α·(Hh) / H+β·d / D+γ·T-δ·θ Where H is the working reference height, h is the current height of the personnel, d is the minimum distance from the personnel to the danger boundary, D is the characteristic scale of the working area, T is the effective attachment duration, θ is the attitude offset angle, and α, β, γ, and δ are weighting coefficients; when M is lower than the preset threshold, the fused situational data enters the instability risk branch.

5. The intelligent monitoring method for high-altitude worker safety based on edge computing according to claim 1, characterized in that, The risk determination result is jointly triggered by abnormal attitude, changes in hanging point status, and approach of dangerous boundaries in several consecutive frames. Any single frame anomaly only forms a candidate risk, and the risk determination result is only output when multiple adjacent frames satisfy the same direction of cumulative anomaly.

6. The intelligent monitoring method for high-altitude worker safety based on edge computing according to claim 1, characterized in that, The tiered alarm and linkage response instructions include at least: wearable terminal vibration alerts, head-mounted terminal audio and visual alerts, on-site voice prompts, and work stoppage instructions; When the risk assessment result continues to exceed the time threshold, the handling instruction is upgraded from the reminder level to the forced evacuation level.

7. The intelligent monitoring method for high-altitude worker safety based on edge computing according to claim 1, characterized in that, When the communication link is interrupted, the edge side performs event-based caching of the original sensing data and risk determination results, and retains only the key frames, key pose segments and key location points related to the risk. After communication is restored, the event-based cache will be sent back in chronological order to avoid duplicate uploading of all data.

8. The intelligent monitoring method for high-altitude worker safety based on edge computing according to claim 1, characterized in that, The hazardous boundary information is dynamically generated based on the operation type, height level, support structure form, and environmental slope. Different operation types correspond to different boundary threshold combinations to ensure that the risk assessment results in the fused situational data match the actual high-altitude scenario.

9. The intelligent monitoring method for the safety of high-altitude workers based on edge computing according to claim 1, characterized in that, The traceability records are associated and numbered according to the event sequence of attachment point switching—attitude change—boundary approach—alarm triggering—response execution, and the corresponding on-site image fragments, location trajectories and risk assessment results are retained to form a replayable high-altitude operation safety chain.

10. A high-altitude worker safety intelligent monitoring system based on edge computing, characterized in that, include: The sensing data acquisition module is used to collect the wearing status, posture status, position trajectory and on-site images of high-altitude workers to obtain raw sensing data; The data preprocessing module is used to perform time synchronization and spatial alignment on the raw sensing data to obtain aligned sensing data; The situation fusion module is used to fuse the operation height, attachment point status and danger boundary information based on the alignment perception data to obtain fused situation data; The risk assessment module is used to assess the fused situational data for detachment, instability, boundary crossing, and proximity to dangerous boundaries, and to obtain risk assessment results. The edge processing module is used to generate graded alarms and linkage processing instructions at the edge based on the risk assessment results, and obtain processing instructions; The traceability storage module is used to associate and store the disposal instructions, the risk assessment results, and the original sensing data to obtain traceability records.

Images