A construction site intelligent security monitoring method and system, a storage medium and a program product

Abstract

A construction site intelligent security monitoring method and system, a storage medium and a program product. The method comprises the following steps: constructing a digital twin of a temporary elevated work platform according to an engineering information model and real-time video images, and pre-acquiring a stress threshold of each structural interface; tracking the three-dimensional human posture in the video to calculate the active dynamic force exerted by the worker, and simultaneously calculating the wind thrust acting on the worker in combination with the real-time wind conditions; superimposing the two force vectors into a composite force, and applying the composite force to the digital twin, and solving the real-time multi-axis load of each interface through simulation conduction; and when the load component of any interface exceeds the preset threshold, an alarm information is sent. The technical scheme provided by the application is used in the construction work to improve the accuracy of the dangerous behavior early warning of the temporary elevated work platform caused by the dynamic behavior of the worker in the monitoring video.

Description

Technical Field

[0001] This application relates to the field of electronic digital data processing, and in particular to a method, system, storage medium, and program product for intelligent security monitoring of construction sites. Background Technology

[0002] Currently, temporary elevated work platforms are crucial facilities in building construction, indispensable in stages such as main structure construction, decoration and finishing, and exterior wall work. With the acceleration of urbanization and the trend of building projects becoming taller and more complex, ensuring the stability and safety of these temporary support structures throughout the entire construction cycle is of paramount importance for protecting the lives of on-site personnel and ensuring the smooth progress of the project.

[0003] In related technologies, construction site safety management largely relies on centralized video surveillance systems. These systems utilize cameras deployed throughout the site, applying artificial intelligence video analytics to real-time video feeds to automate the monitoring of various safety regulations. For example, they can automatically identify whether workers are correctly wearing safety helmets and reflective vests, detect falls, prolonged periods of inactivity, and whether personnel have entered pre-defined danger zones.

[0004] However, on temporary elevated work platforms at construction sites, when workers perform seemingly routine operations such as pushing, pulling, and prying, these actions themselves are not considered violations, and the relevant technology struggles to identify any irregularities from the monitoring videos. However, the dynamic forces instantaneously applied to the platform by workers in specific exertion postures, or the combined force resulting from the independent actions of multiple workers, and even the influence of ambient wind forces due to the nature of working at heights, can potentially create concentrated loads in a specific area of ​​the platform that far exceed design standards. Because the monitoring system's hazard warning accuracy is insufficient, this structural overload risk caused by seemingly normal work behavior is difficult to identify, potentially leading to localized overload risks in the platform structure and resulting in safety accidents. Summary of the Invention

[0005] This application provides a construction site intelligent security monitoring method, system, storage medium, and program product, which is used to improve the accuracy of early warning of dangerous behaviors caused by the dynamic behavior of workers in the monitoring video on temporary high-altitude work platforms during construction site operations.

[0006] The first aspect of this application provides a method for intelligent security monitoring at construction sites, the method comprising:

[0007] Interface information is obtained based on the engineering information model of the temporary elevated work platform; a real-time digital twin of the operation is constructed based on the real-time video images of the temporary elevated work platform and the engineering information model; preset key force points of the human body in the three-dimensional human posture in the real-time video images are tracked, and the real-time motion parameters of the key force points of the human body are calculated; combined with the preset human biomechanical calculation model and real-time motion parameters, the active dynamic force vector applied by the worker to the temporary elevated work platform is calculated; real-time wind speed and real-time wind direction at the temporary elevated work platform are obtained; the windward force area of ​​the worker is calculated based on the real-time wind direction and the three-dimensional human posture; based on the preset aerodynamic model, and combined with the real-time wind speed and windward force area, the wind thrust vector acting on the worker is calculated; the composite force vector is applied to the corresponding contact point in the real-time digital twin of the operation, and the transmission process of the composite force vector is simulated in the real-time digital twin of the operation to obtain the multi-axis load components corresponding to multiple interfaces in the temporary elevated work platform; when the multi-axis load component of any interface is greater than the corresponding interface force threshold, an alarm message is issued.

[0008] In the above embodiments, by adopting the aforementioned technical solution, the shortcomings of existing technologies in identifying seemingly routine but potentially dangerous behaviors of workers, such as pushing and pulling, that could lead to structural overload are addressed. First, by tracking three-dimensional human posture and combining it with a biomechanical model, the dynamic behaviors of workers, invisible in the monitoring video, are quantified into calculable active dynamic force vectors. Next, the wind thrust vector during high-altitude operations is further coupled to form a composite force vector. Finally, the composite force is applied to a digital twin and its transmission is simulated to calculate the real-time load on the platform interface caused by the actual dynamic behavior of personnel (rather than static estimation). This upgrades the basis for early warning judgment from behavioral recognition to mechanical analysis, thereby improving the accuracy of early warnings for dangerous behaviors on temporary high-altitude work platforms caused by the dynamic behavior of workers in monitoring videos during construction site operations.

[0009] In conjunction with some embodiments of the first aspect, in some embodiments, after applying the composite force vector to the corresponding contact point in the real-time operation digital twin, simulating the transmission process of the composite force vector in the real-time operation digital twin, and obtaining the multi-axis load components corresponding to multiple interfaces in the temporary high-altitude operation platform, the method further includes:

[0010] When the multiaxial load component of any interface is not greater than the corresponding interface stress threshold, the non-overloaded multiaxial load component of the non-overloaded interface is taken as one stress cycle of the non-overloaded interface; according to the preset material stress-life curve and the preset fatigue cumulative damage calculation formula, the micro-fatigue damage degree caused by the stress cycle to the non-overloaded interface is calculated; the micro-fatigue damage degree is accumulated into the fatigue damage accumulation counter of the non-overloaded interface; when the value in the fatigue damage accumulation counter of any interface exceeds the corresponding preset fatigue life threshold, a fatigue warning is issued.

[0011] In the above embodiments, each multiaxial load component that does not cause instantaneous overload is considered as a stress cycle on the interface, and based on a preset fatigue cumulative damage model, the minute damage caused by these cycles is quantified and continuously accumulated. This process can reveal the hidden structural performance degradation caused by seemingly safe dynamic behavior, thereby expanding the early warning dimension from strength verification to life prediction. Ultimately, it can be used in construction site operations to improve the accuracy of early warnings of dangerous behaviors caused by the dynamic behavior of workers in monitoring videos on temporary high-altitude work platforms.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, after applying the composite force vector to the corresponding contact point in the real-time operation digital twin, simulating the transmission process of the composite force vector in the real-time operation digital twin, and obtaining the multi-axis load components corresponding to multiple interfaces in the temporary high-altitude operation platform, the method further includes:

[0013] By combining a real-time operational digital twin, the theoretical displacement response of the interface center point in three-dimensional space under multi-axis load components is calculated. The changes of edge pixels of each interface over time in the real-time video image are analyzed to obtain the actual displacement response of the center point of each interface. The ratio of the theoretical displacement response to the actual displacement response is calculated as the stiffness degradation coefficient. When the stiffness degradation coefficient of any interface center point is greater than the preset stiffness degradation threshold, the corresponding interface is determined to be a stiffness degradation interface. The preset stiffness degradation interface fatigue life threshold corresponding to the stiffness degradation interface in the preset structural safety standard database is obtained. The fatigue life threshold of the stiffness degradation interface is corrected with the stiffness degradation coefficient.

[0014] In the above embodiments, by comparing the theoretical displacement in the digital twin with the actual displacement in the video surveillance, a stiffness degradation coefficient is calculated, thereby quantifying the actual structural damage that is difficult to observe directly. Furthermore, this coefficient is used to dynamically correct the fatigue life threshold, so that the assessment of fatigue damage is no longer a fixed theoretical calculation, but is linked to the actual health condition of the structure in real time. Ultimately, this is used in construction site operations to improve the accuracy of early warnings of dangerous behaviors on temporary high-altitude work platforms caused by the dynamic behavior of workers in surveillance videos.

[0015] In conjunction with some embodiments of the first aspect, in some embodiments, when the multi-axis load component of any interface exceeds the corresponding interface force threshold, an alarm message is issued, specifically including:

[0016] When the multi-axis load component of any interface exceeds the stress threshold of the corresponding overload interface, the bearing capacity of the corresponding overload interface is set to zero in the real-time operation digital twin to simulate the failure of the overload interface's bearing capacity. Based on structural mechanics analysis, the simulated multi-axis load components corresponding to multiple interfaces in the temporary high-altitude operation platform are determined. When the simulated multi-axis load component of any interface exceeds the preset stress threshold of the corresponding interface in the preset structural safety standard database, a failure path diagram containing the initial failure point and the predicted secondary failure point is generated, and the first alarm message is issued based on all nodes in the failure path diagram.

[0017] In the above embodiments, when an initial overload is detected, the load-bearing capacity of the interface is set to zero in the digital twin to simulate a real failure, and a mechanical reanalysis is immediately performed. This process can proactively deduce the chain reaction of load transfer to other interfaces caused by a single-point failure, and visually present the collapse process of the entire structure in the form of a failure path diagram. Therefore, the early warning information is upgraded from single-point overload to prediction of the overall continuous failure of the structure, and ultimately used in construction site operations to improve the accuracy of early warning of dangerous behaviors caused by the dynamic behavior of workers in monitoring videos on temporary high-altitude work platforms.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, after calculating the active dynamic force vector applied by the worker to the temporary elevated work platform by combining a preset human biomechanical calculation model and real-time motion parameters, the method further includes:

[0019] Identify the model of the work tool held by the worker in the real-time video image; retrieve the inherent mechanical parameters of the corresponding work tool model from the tool physical property database; calculate the tool force vector based on the tool's inherent mechanical parameters; and synthesize the tool force vector with the dynamic force vector to obtain a composite work force vector, which replaces the active dynamic force vector.

[0020] In the above embodiments, by identifying the type of tool held by the worker and retrieving its inherent mechanical parameters, the additional force exerted by the tool can be quantified into an independent tool force vector. This force vector is then vector-combined with the force exerted by the worker, thus correcting the previous consideration of only the worker's active dynamic force into a composite operational force vector. Because this composite operational force vector more accurately reflects the actual load applied to the platform by the worker using the tool, the basis for subsequent mechanical analysis and early warning judgments is more precise. Ultimately, this improves the accuracy of early warnings regarding dangerous behaviors on temporary elevated work platforms caused by the dynamic behavior of workers in monitoring videos during construction site operations.

[0021] In conjunction with some embodiments of the first aspect, in some embodiments, the tool force vector is calculated based on the tool's inherent mechanical parameters, specifically including:

[0022] When the working tool is a preset lever-type tool and there is a tool contact point between the working tool and the temporary high-altitude working platform, the power point, resistance point, and fulcrum of the working tool are identified in the real-time video image; the length of the power arm between the power point and the fulcrum, and the length of the resistance arm between the resistance point and the fulcrum are measured; based on the lever principle formula, and based on the dynamic force vector, the lengths of the power arm and the resistance arm, the supporting force borne by the fulcrum is calculated; and the supporting force is used as the tool force vector.

[0023] In the above embodiments, for specific scenarios where workers use lever-type tools such as crowbars, by identifying the power point, fulcrum, and resistance point in the video and combining this with the lever principle for mechanical modeling, the supporting force acting on the platform's fulcrum can be calculated from the relatively small dynamic force applied by the worker. This process deepens the calculation of tool forces from macroscopic parameter retrieval to microscopic mechanical analysis of specific work behaviors, making the final synthesized load closer to physical reality. Ultimately, this improves the accuracy of early warnings for dangerous behaviors on temporary elevated work platforms caused by the dynamic behavior of workers in monitoring videos during construction site operations.

[0024] In conjunction with some embodiments of the first aspect, in some embodiments, after constructing a real-time operation digital twin based on real-time video images and an engineering information model of the temporary elevated work platform, the method further includes:

[0025] When there are multiple workers, the spatial distance between each worker is calculated; when the spatial distance is less than the preset safe distance threshold for personnel, an alarm is issued to prompt the workers to disperse and stand in their designated positions.

[0026] In the above embodiments, by calculating the spatial distance between workers, the risk assessment is deepened from simple total load monitoring to real-time analysis of load spatial distribution. Since personnel gathering can cause a sharp increase in local stress, this solution proactively intervenes through alarms before the structure bears dangerous loads, prompting personnel to disperse and optimize load distribution. This shifts the early warning from a passive mechanical response to proactive behavioral intervention, ultimately improving the accuracy of early warnings for dangerous behaviors on temporary elevated work platforms caused by the dynamic behavior of workers in monitoring videos during construction site operations.

[0027] Secondly, embodiments of this application provide a construction site intelligent security monitoring system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the construction site intelligent security monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0028] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a construction site intelligent security monitoring system, cause the construction site intelligent security monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0029] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a construction site intelligent security monitoring system, cause the construction site intelligent security monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0030] Understandably, the intelligent security monitoring system for construction sites provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the intelligent security monitoring method for construction sites provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0031] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0032] 1. This application addresses the shortcomings of existing technologies in identifying seemingly routine but potentially dangerous behaviors of workers, such as pushing and pulling, that could lead to structural overload. First, by tracking three-dimensional human posture and combining it with a biomechanical model, the dynamic behaviors of workers, invisible in surveillance videos, are quantified into calculable active dynamic force vectors. Next, the wind thrust vector during high-altitude operations is further coupled to form a composite force vector. Finally, the composite force is applied to a digital twin and its transmission is simulated to calculate the real-time load on the platform interface caused by the actual dynamic behavior of personnel (rather than static estimation). This upgrades the basis for early warning judgment from behavioral recognition to mechanical analysis, thereby improving the accuracy of early warnings for dangerous behaviors caused by the dynamic behavior of workers in surveillance videos on temporary high-altitude work platforms during construction site operations.

[0033] 2. This application treats each multiaxial load component that does not cause instantaneous overload as a stress cycle on the interface, and quantifies and continuously accumulates the minute damage caused by these cycles based on a preset fatigue cumulative damage model. This process can reveal the hidden structural performance degradation caused by seemingly safe dynamic behavior, thereby expanding the early warning dimension from strength verification to life prediction. Ultimately, it can be used in construction site operations to improve the accuracy of early warning of dangerous behaviors caused by the dynamic behavior of workers on temporary high-altitude work platforms as seen in monitoring videos.

[0034] 3. This application calculates the stiffness degradation coefficient by comparing the theoretical displacement in the digital twin with the actual displacement in video surveillance, thereby quantifying the actual structural damage that is difficult to observe directly. Furthermore, this coefficient is used to dynamically correct the fatigue life threshold, so that the assessment of fatigue damage is no longer a fixed theoretical calculation, but is linked to the actual health condition of the structure in real time. Ultimately, this is used in construction site operations to improve the accuracy of early warnings of dangerous behaviors on temporary high-altitude work platforms caused by the dynamic behavior of workers in surveillance videos. Attached Figure Description

[0035] Figure 1 This is a flowchart illustrating a construction site intelligent security monitoring method in an embodiment of this application;

[0036] Figure 2 This is another flowchart illustrating the intelligent security monitoring method for construction sites in this application embodiment;

[0037] Figure 3 This is an exemplary hardware structure diagram of a construction site intelligent security monitoring system in the embodiments of this application. Detailed Implementation

[0038] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to and includes any or all possible combinations of one or more of the listed items.

[0039] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0040] In this embodiment of the application, a construction site intelligent security monitoring system may include a data acquisition terminal, a data processing and analysis server, and an alarm and display terminal.

[0041] The data processing and analysis server, as the core computing unit of the system, is used to receive and process field data from the data acquisition terminal, run digital twin, biomechanical and aerodynamic models, perform structural mechanics analysis and risk assessment, and finally send instructions to the alarm and display terminal.

[0042] Data acquisition devices, such as cameras and anemometers deployed near temporary elevated work platforms, are used to capture real-time three-dimensional posture videos of workers and on-site environmental parameters, and transmit these multi-source data to the server.

[0043] The alarm and display terminal is used to receive early warning information from the server and present it to on-site management personnel.

[0044] Through the above system architecture, this solution transforms invisible structural stress risks into visualized data and early warnings, providing decision support for managers and thus preventing platform structural failure accidents caused by dynamic loads.

[0045] In related technologies, artificial intelligence video analytics is widely used in construction site safety management, primarily focusing on monitoring easily visually identifiable violations. For example, it can automatically identify whether workers are wearing safety helmets, entering dangerous areas, or detecting obvious abnormalities such as falls. However, when workers perform seemingly routine actions like pushing, pulling, or prying on temporary elevated platforms (such as scaffolding), these actions themselves do not constitute pre-defined violation patterns, making it difficult to identify potential hazards. In reality, the instantaneous dynamic forces generated by these actions, the combined force of multiple workers, and the combined effect of wind forces in the high-altitude environment can create concentrated loads in localized areas of the platform structure far exceeding its design load-bearing capacity. Therefore, the deficiency of this technology lies in its warning logic being based on behavioral pattern matching, which fails to detect invisible structural overload risks caused by seemingly normal work behaviors, thus posing a safety hazard.

[0046] In this embodiment, to address the aforementioned shortcomings, this solution tracks three-dimensional human posture and combines it with a biomechanical model to quantify the exertion behavior of workers—which cannot be directly observed in monitoring videos—into a calculable active dynamic force vector. Building upon this, environmental factors are further introduced. By analyzing real-time wind speed and the worker's windward posture, a wind thrust vector is calculated, and these two are combined to form a composite force vector that more closely approximates real-world working conditions. Finally, this composite force is applied to a real-time digital twin of the operation. By simulating the transmission process within the platform structure, the real-time load on each key interface is calculated. This shifts the basis for early warning judgment from superficial behavioral recognition to underlying structural stress analysis, thereby enabling the identification of structural overload risks caused by personnel dynamic behavior and improving the accuracy of early warnings for dangerous behaviors on temporary elevated work platforms.

[0047] Figure 1 This is a flowchart illustrating the construction site intelligent security monitoring method used in the embodiments of this application, including the following steps:

[0048] S101. Obtain interface information based on the engineering information model of the temporary high-altitude operation platform.

[0049] As can be understood, a temporary high-altitude work platform refers to a structure temporarily erected at a construction or industrial site to support personnel and materials for high-altitude operations, such as scaffolding or operating platforms. An engineering information model (EIM) is a pre-provided, structured engineering dataset that defines the ideal geometric dimensions, layout, and connection methods of the temporary high-altitude work platform, and records inherent physical properties that cannot be directly perceived visually, such as the material specifications and mechanical properties (e.g., yield strength) of each component, as well as the grade standards of connectors (e.g., bolts). An interface represents the specific location where two or more structural members are connected to each other by means of bolts, welding, or clips. The interface stress threshold refers to the maximum multidimensional force and moment limit that the interface can withstand without permanent deformation or fracture, calculated according to material mechanics and structural design specifications.

[0050] Specifically, the process begins by loading and parsing the digitized engineering information model file. Through layer analysis, symbol recognition, and other technologies, all critical load-bearing interfaces in the platform structure are automatically identified, or with manual assistance. For each identified interface, the built-in structural safety standard database is queried based on the connector type (e.g., bolt grade, weld height) and the properties of the parent material as indicated on the drawings. This database pre-stores allowable stresses or ultimate loads for different materials and connection methods in various stress directions (e.g., tension, shear, bending, and torsion). Through query matching, a stress threshold is assigned to each degree of freedom (three translation axes and three rotation axes) for each interface, and this information is bound to the interface's coordinates in the 3D model.

[0051] S102. Construct a real-time operation digital twin based on the real-time video images and engineering information model of the temporary high-altitude operation platform.

[0052] It is understandable that real-time video images refer to dynamic video streams continuously captured and transmitted by one or more cameras deployed around the work platform; a real-time work digital twin refers to a dynamic three-dimensional simulation model that is highly synchronized with the real work platform and environment in the physical world in virtual space.

[0053] Specifically, firstly, the engineering information model information parsed from S101 is transformed into a basic 3D geometric model. Then, multi-view geometric or single-view depth estimation algorithms are used to analyze the video stream to achieve camera self-localization and scene 3D reconstruction. Based on the real-time spatial position and orientation of each camera, pixels in the video frame are mapped to 3D spatial coordinates. Building upon this, target detection and segmentation techniques are further used to identify the components of the temporary elevated work platform in the video. These identified components are then placed in real-time into their corresponding positions within the 3D geometric model, and combined with the materials of each component in the engineering information model, a digital twin is constructed and continuously updated.

[0054] To overcome the inherent limitations of fixed cameras, such as limited monitoring range and blind spots, and to adapt to the dynamic and complex working environment of construction sites, mobile monitoring capabilities are deeply integrated. Mobile monitoring devices can include drone-borne cameras, helmet-integrated cameras, or handheld intelligent inspection terminals for management personnel. For mobile monitoring, a robust real-time positioning and mapping technology is employed to ensure accurate video data and digital twin fusion even in dynamically changing and signal-unstable construction site environments.

[0055] Firstly, on mobile devices, Visual-Inertial Odometry (VIO) technology is prioritized. This technology tightly couples visual information captured by the camera with high-frequency motion data from the device's built-in inertial measurement unit (IMU). Faced with common challenges on construction sites such as ground bumps, severe image jitter caused by personnel movement, and rapid changes in perspective, VIO can provide high-frequency, low-latency, and smooth six-DOF pose (position and attitude) estimation, which is fundamental for achieving stable tracking and analysis.

[0056] Secondly, VIO provides the pose relative to the starting point. To integrate it into a unified digital twin model of the construction site, Simultaneous Localization and Mapping (SLAM) technology is employed. SLAM continuously analyzes static environmental features (such as building outlines and fixed equipment) seen by the moving camera, constructs a 3D map of the environment in real time, and locates itself within this map. To address the challenges of numerous dynamic objects (people, vehicles) and high feature similarity in construction site scenarios, this solution's SLAM algorithm incorporates semantic information, enabling it to identify and eliminate interference from dynamic objects, thus enhancing the robustness of positioning. Furthermore, the accumulated error of SLAM can be periodically calibrated by identifying pre-set fixed markers in the scene (such as QR codes and specific structural landmarks) or intermittently utilizing GPS / BeiDou satellite positioning signals, ensuring long-term global accuracy.

[0057] Ultimately, once the real-time global pose of the mobile camera is accurately determined, the video stream it captures can be precisely registered in the digital twin scene. From this mobile, flexible perspective, the same precise analysis as with a fixed camera can be performed, such as reconstructing the 3D pose of the workers and identifying their interactions with the environment. This mobile monitoring capability enables this solution to achieve comprehensive, blind-spot-free coverage of the work site, allowing for close observation of key work points (such as manhole openings) or continuous monitoring of moving workers from a follow-up perspective, thus improving the flexibility, coverage, and effectiveness of data collection.

[0058] To address the challenges of extremely confined, dark, and signal-free working environments, such as underground pipelines, this solution further supports the integration of specialized mobile monitoring equipment, such as pipeline inspection mobile monitoring devices or serpentine mobile monitoring devices equipped with cameras. These devices feature built-in active light sources (e.g., high-intensity LED lights as supplementary lighting for the cameras) to ensure clear video images are captured even in completely dark pipelines. More importantly, the mobile monitoring equipment can integrate various environmental sensors, such as sensors for harmful gases like hydrogen sulfide, carbon monoxide, and methane, as well as oxygen concentration sensors.

[0059] Inside the pipeline, VIO and SLAM technologies are primarily used, utilizing pipe walls, interfaces, and sediment as visual features for location. To ensure stable data transmission, mobile monitoring equipment is typically connected to a base station at the wellhead via a tethered connection. High-definition video and sensor data are transmitted in real time to a ground server via fiber optic or cable, solving the problem of underground wireless signal shielding.

[0060] In this solution, the engineering information model serves as the static foundation and data kernel for constructing the digital twin. Based on this model, a basic mechanical analysis model is built. Then, real-time video images are combined to calibrate and fuse the actual geometric shape of the platform, ensuring that the twin and the physical entity are consistent in appearance.

[0061] In some embodiments, the construction and synchronization of digital twins can be achieved in various ways: Optionally, a marker-based registration and fusion method can be used: First, uniquely identifiable visual markers (such as QR codes or special patterns) are pre-attached or set at several key, fixed locations on the real-world operating platform; second, the video analysis system continuously detects and identifies these markers, and uses their known physical world coordinates to accurately calculate the real-time pose of the camera through a perspective transformation algorithm; finally, the static model generated by the engineering information model is aligned according to the calculated camera pose, and the dynamic objects identified in the video are superimposed onto the model. It is understood that other methods can also be used to construct digital twins, such as combining other sensors like LiDAR for multimodal data fusion, which is not limited here.

[0062] In some embodiments, after constructing a real-time digital twin of the operation containing multiple workers, the three-dimensional spatial distance between each worker can be monitored and calculated in real time to proactively manage the distribution of personnel positions and prevent potential structural overload and operational safety risks caused by localized personnel gathering.

[0063] First, in real-time video images, a continuously running human posture recognition algorithm acquires the three-dimensional spatial coordinates of each worker on the platform. For ease of calculation, the position of each worker is typically abstracted as the geometric center point of a human skeletal model. Then, all worker pairings on the platform are traversed, and the straight-line spatial distance between each pair is calculated using the Euclidean distance formula in three-dimensional space. This calculation is real-time and continuous, dynamically reflecting changes in worker positions. Next, each calculated distance value is compared with a pre-set personnel safety distance threshold. This threshold is comprehensively set based on building safety codes, ergonomics, and load distribution principles in the platform structural design, defining the minimum permissible distance to ensure operational safety and avoid excessive load concentration. Once the spatial distance between any pair of workers is detected to be less than the threshold, an alarm is triggered prompting workers to disperse.

[0064] The above-mentioned technical steps are a proactive, behavioral intervention-based early warning system. Instead of waiting until the structure has been subjected to excessive pressure before responding, they point out the unsafe behavior itself (i.e., standing too close) rather than merely monitoring the consequences (i.e., structural overload). This allows for intervention at the nascent stage of danger, preventing the risk of platform plate breakage or support point failure due to excessive local load.

[0065] By monitoring and managing the spacing between human workers in real time, the risk of localized concentrated loads caused by excessive gathering of people can be identified and warned. Intervention can be carried out at the source of danger, rather than waiting for the dangerous consequences to occur, thereby improving the accuracy and foresight of warnings for such dynamic dangerous behaviors.

[0066] S103. Track the preset key force points of the human body in the three-dimensional human body posture in the real-time video image, and calculate the real-time motion parameters of the key force points of the human body.

[0067] Understandably, three-dimensional human posture refers to the set of three-dimensional coordinate points representing the human skeletal structure, calculated from video images by algorithms. These typically include key joints such as the head, neck, shoulders, elbows, wrists, hips, knees, and ankles. Key force points in the human body refer to the body parts selected from the above postures that primarily transmit force when performing tasks such as pushing, pulling, lifting, and prying. These are typically the wrists, shoulders, ankles, and the core area of ​​the waist. Real-time motion parameters refer to the physical quantities used to describe the motion state of these key points, mainly including their instantaneous velocity, instantaneous acceleration, and displacement distance in three-dimensional space.

[0068] Specifically, the process involves identifying workers in real-time video images and, based on predefined rules (e.g., defining wrists and ankles as primary points of contact and force application), selecting key points of force application for focused analysis. For each tracked key point, its three-dimensional spatial coordinates are recorded on consecutive video frames (or time points). The instantaneous velocity vector at that point is calculated using numerical differentiation by dividing the difference in coordinate displacement between two adjacent time points by the time interval. Based on the obtained velocity data, the velocity is further numerically differentiated by dividing the difference in velocity vectors between two adjacent time points by the time interval to calculate the instantaneous acceleration vector at that point.

[0069] S104. Based on the preset human biomechanical calculation model and real-time motion parameters, calculate the active dynamic force vector applied by the operator to the temporary high-altitude work platform.

[0070] Understandably, the human biomechanical computational model is a mathematical model based on Newton's second law and human anatomical features. It describes the relationship between the mass, inertia, joint forces, and muscle forces of various parts of the human body.

[0071] Specifically, applying the principles of inverse dynamics, the real-time acceleration of the worker's key force application points (such as hands or feet) acquired by the S103, along with a pre-set human biomechanical model (which includes parameters such as the mass, center of mass, and moment of inertia of each standard human limb segment), is used as input. According to Newton's second law, the total force (i.e., the resultant force of the joints) required to produce the observed acceleration of that limb segment (such as the forearm and hand) is first calculated. This total force includes the force generated by the muscles and the force exerted by the external environment. By calculating joint by joint from bottom to top (e.g., from hand to elbow, then to shoulder), the model can solve for a series of internal joint forces and torques. When calculating the key force application point where the human body contacts the platform, the reaction force exerted by the external environment on the human body is isolated. According to Newton's third law, the active dynamic force exerted by the worker on the platform is equal in magnitude to this reaction force calculated by the model, but in the opposite direction. This calculated force is a three-dimensional vector describing the pushing, pulling, or supporting force exerted by the worker on the platform at that moment.

[0072] In some embodiments, the calculation of active dynamic forces can be achieved in several ways: Optionally, a rigid body dynamics model can be used for calculation: First, the human body is simplified as a multi-body system composed of multiple rigid rods (representing bones) connected by ideal hinges (representing joints); second, each rod is assigned standard anthropometric parameters (mass, inertia, etc.) obtained from a lookup table; finally, based on kinematic data, the Newton-Euler equations or Lagrange equations are used to recursively calculate from the end effector (hand or foot) towards the body core, ultimately solving for the contact force at the point of contact between the end effector and the environment. Optionally, a machine learning-based regression model can be used for calculation: First, in a laboratory environment, the test subject wears force sensors and motion capture equipment to perform a series of standard work actions, simultaneously collecting their actual force application data and motion posture data; second, this massive amount of data is used to train a deep neural network to establish an end-to-end mapping relationship from human motion parameters to the actual magnitude and direction of applied force; finally, in practical applications, real-time motion parameters are directly input into the trained model, and the model directly outputs the predicted active dynamic force vector without the need for complex dynamic modeling. It is understandable that other methods can be used to calculate dynamic forces, such as more complex musculoskeletal models, which are not limited here.

[0073] In some embodiments, after calculating the active dynamic force vector applied by the operator to the platform, the tools held by the operator can be further identified and mechanically analyzed to take into account the additional force generated by the tools, thereby obtaining a composite working force that integrates the combined action of the human body and the tools.

[0074] First, a deep learning-trained object detection model is invoked, specifically designed to identify various handheld tools commonly found on construction sites. This model continuously analyzes high-resolution video image slices extracted from real-time video footage, focusing on the areas around the workers' hands. Through convolutional neural networks, the model extracts and compares image features, enabling it not only to pinpoint the tool's location but also to classify it and output specific tool models, such as "large adjustable wrench" or "handheld electric impact drill."

[0075] After identifying the tool model, the model number is used as the query key to access a pre-built database of tool physical properties. This database is a structured table that stores the standardized physical and mechanical properties of each tool. For all tools, the database includes its own weight as a basic parameter. Furthermore, for power tools, the database also stores the characteristic forces generated by the tool itself under typical operating conditions. For example, for a handheld impact drill, parameters might include the average amplitude and dominant frequency of the high-frequency vibration force it generates during operation; for a handheld cutter, it might include the reaction torque generated by the high-speed rotation of the grinding wheel. These parameters are pre-entered using data provided by the manufacturer or laboratory calibration, and they represent additional force sources generated by the tool itself, independent of the user.

[0076] When it is identified that the operator is using a lever-type tool such as a crowbar or wrench, and image analysis determines that the tool has a stable contact point (i.e., fulcrum) with the work platform or a fixed object on it, image segmentation technology is used to separate the tool outline from the background. Combined with depth information or multi-view geometry, three core mechanical points are located on the tool's 3D model: the center position of the operator's hand applying force, i.e., the power point; the position where the tool contacts the object being worked on (such as a stone slab that needs to be pried), i.e., the resistance point; and the support position where the tool itself contacts the platform structure (such as a crossbar), i.e., the fulcrum.

[0077] In real-time video images, the three-dimensional straight-line distance from the power point to the fulcrum is identified and calculated to obtain the length of the power arm; similarly, the three-dimensional straight-line distance from the resistance point to the fulcrum is calculated to obtain the length of the resistance arm.

[0078] In this scenario, the active dynamic force vector is defined as the driving force acting on the fulcrum. Based on the static lever balance principle (M_fulcrum = 0), a torque balance equation is established: Force × Force Arm = Resistance × Resistance Arm. Simultaneously, according to the principle of force balance, the supporting force at the fulcrum is equal to the sum of the force and resistance. The supporting force at the platform fulcrum, significantly amplified by the lever, is calculated. The magnitude and direction of this supporting force represent the actual pressure applied to the platform by the operator using the tool.

[0079] The forces calculated in the preceding steps are uniformly encapsulated as tool force vectors. If the tool being analyzed is a lever, the tool force vector is the calculated fulcrum support force; if the tool being analyzed is a non-lever tool, the tool force vector is the vector sum of its own weight and vibration force. Subsequently, the tool force vector is superimposed with the active dynamic force vector in three-dimensional space, and the composite working force vector replaces the original active dynamic force vector.

[0080] By identifying the tools used in the operation and calculating the tool force based on their physical characteristics (especially the lever principle), the active dynamic force that was previously only considered from the human body can be corrected to a composite operating force that includes the amplification effect of the tool. This makes the force analysis closer to physical reality and avoids underestimating the actual load-bearing pressure of the platform interface due to ignoring the mechanical gain of the tool. It can more accurately simulate the transmission and distribution of loads in the platform structure, thereby improving the accuracy of early warning of overload risks caused by dangerous operating behaviors.

[0081] In some embodiments, after calculating the active dynamic force vector applied by the operator to the platform, the operator's safety protection behavior and related safety equipment can be further identified and mechanically analyzed to take into account the additional force generated by the safety equipment under specific dangerous conditions, thereby obtaining a composite working force that integrates multiple potential loads.

[0082] First, the system identifies whether workers in real-time video images are wearing personal protective equipment (PPE) such as safety ropes and harnesses. If PPE is worn, the system locates the connection points between the PPE and the platform. Using deep learning object detection and image segmentation algorithms, the system not only identifies the human body but also the safety equipment attached to key nodes of the human body's 3D model (such as the waist and back). For example, it identifies the safety rope connected to the worker's safety harness and tracks its other end to the fixed anchor point on the temporary elevated work platform.

[0083] Secondly, based on the real-time motion parameters of the workers, it is determined whether a dangerous situation such as a fall has occurred. The three-dimensional motion trajectory, speed, and acceleration of the workers are continuously tracked. When it is detected that the worker's center of gravity begins to move outward from the edge of the platform, and the instantaneous acceleration in the vertical direction significantly exceeds the normal range of motion (e.g., approaching or exceeding the acceleration due to gravity), it is determined that the worker has entered a fall state.

[0084] Upon determining that a fall has occurred, the impact force vector generated by the safety rope on the platform anchor point is calculated. A pre-defined fall impact mechanics model is invoked. This model, which incorporates the worker's weight (obtainable from a database or using standard values), the fall coefficient (dependent on the anchor point height and worker position), and the material properties and cushioning performance of the safety rope (retrieved from the database based on the identified model), calculates the peak impact force generated on the platform anchor point when the safety rope is taut and in operation during the fall. This impact force is a three-dimensional force vector with precise calculation of both magnitude and direction.

[0085] Finally, this impact force vector is treated as a special equipment force vector and vector-synthesized with the active dynamic force vector to obtain a composite operating force vector, which replaces the active dynamic force vector.

[0086] In this way, the scope of monitoring is expanded from routine operational activities to the prediction of the mechanical consequences of sudden hazardous events such as falls. By identifying critical protective equipment such as safety ropes and combining this with human kinematics analysis, the extreme but lethal load of a fall impact can be quantified. This upgrades the mechanical model, which previously only considered the forces applied during operations, to a comprehensive model capable of assessing the reaction forces of the safety system on the platform structure in emergency situations. This analysis avoids a serious underestimation of the load-bearing capacity of the platform's anchor points due to neglecting the fall impact force, and can simulate the transmission of extreme loads in the platform structure, thereby improving the accuracy of early warnings for such sudden, high-risk hazardous events.

[0087] S105. Obtain the real-time wind speed and direction at the temporary elevated work platform.

[0088] Specifically, to obtain meteorological information about the location of the work platform, data communication is established with physical sensors deployed on-site. Typically, wind speed and direction sensors are installed at appropriate locations on the temporary elevated work platform (e.g., on the top or windward-facing pillars). These sensors (such as three-cup anemometers and wind vanes, or more advanced ultrasonic anemometers) continuously measure airflow at their location. The sensors transmit the measurement data periodically (e.g., once per second) to the monitoring system's server via wired or wireless means (e.g., Wi-Fi, LoRa, 4G / 5G) in a specific data packet format. Upon receiving the data, the server parses it, extracts the wind speed and direction values, and timestamps this data to ensure alignment with video image data from the same moment.

[0089] In some embodiments, meteorological data is obtained indirectly by calling a public meteorological service interface: First, the precise geographical coordinates (latitude and longitude) of the operating platform are obtained; second, a request containing these coordinates is sent via the Internet to the application programming interface (API) of a public or commercial meteorological data service provider; finally, the meteorological service provider returns real-time data from the meteorological observation station closest to these coordinates, or an estimated wind speed and direction calculated by interpolation using a regional meteorological model. It is understood that other methods can also be used to obtain meteorological data, such as analyzing the swaying patterns of flags or hanging objects in a video to estimate wind force, etc., and are not limited here.

[0090] S106. Calculate the windward force area of ​​the operator based on the real-time wind direction and three-dimensional human posture.

[0091] Specifically, a real-time 3D human posture model of the worker is retrieved and orthogonally projected onto a 2D plane perpendicular to the wind direction, along the wind direction vector. This process is similar to casting a shadow on the wall behind the person when a beam of parallel light shines on them in the direction the wind is blowing. After projection, the human model becomes a 2D silhouette. Then, image processing algorithms are applied to calculate the area of ​​this 2D silhouette.

[0092] In some embodiments, the windward area can be calculated in several ways: Optionally, a rendering and reading method based on a polygonal mesh model can be used: First, a surface mesh composed of a large number of triangular faces is applied to the three-dimensional human skeleton posture model to form a solid model; second, an orthogonal projection camera is set in the virtual three-dimensional scene, facing completely opposite to the wind direction; finally, a special rendering is performed on this scene, rendering only the silhouette of the human model (e.g., rendering it as pure white with a pure black background), and the number of white pixels in the rendered image is read. The windward area is calculated by the ratio of the number of pixels to the actual size.

[0093] S107. Based on the preset aerodynamic model and combined with real-time wind speed and windward force area, calculate the wind thrust vector acting on the operator.

[0094] It is understandable that an aerodynamic model refers to a physical formula that describes the interaction forces between a fluid (air) and an object; here, it specifically refers to the classic formula for calculating wind resistance.

[0095] Specifically, the magnitude of wind thrust is calculated using the classic hydrodynamic drag formula. This formula is: Wind drag = 0.5 × Air density × Wind speed squared × Drag coefficient × Windward contact area. In this formula, air density is a physical constant that can be adjusted according to altitude and temperature; wind speed and windward contact area are the calculation results from S105 and S106, respectively. The drag coefficient is a dimensionless parameter related to the shape of the object. For complex human postures, an approximate drag coefficient value is selected from a pre-defined lookup table based on the degree of extension of the posture. Substituting all these values ​​into the formula yields the magnitude of the wind thrust. The direction of the wind thrust is consistent with the real-time wind direction vector obtained from S105. Finally, the calculated magnitude and direction of the force are combined to form a three-dimensional wind thrust vector.

[0096] In some embodiments, for the human body with its varied shape, the drag coefficient changes in real time with the worker's posture. To achieve accurate calculation, the drag coefficient can be determined using the following methods:

[0097] A human posture-drag coefficient database is pre-established. By summarizing a large amount of wind tunnel experimental data or fluid dynamics simulation results, corresponding drag coefficient reference values ​​are calibrated for typical working postures (such as standing completely, bending at 90 degrees, squatting completely, and arms outstretched horizontally). In real-time analysis, the currently identified 3D human posture is matched with the standard postures in the database, and the closest drag coefficient value is selected for calculation.

[0098] A machine learning model (such as a neural network) is trained, taking as input key parameters of the 3D human posture (such as the angles of major joints, the degree of body extension, and the height of the center of gravity), and outputting real-time predicted drag coefficient values. This method enables more refined and continuous drag coefficient estimation for any non-standard posture, with greater adaptability and accuracy.

[0099] By substituting all these parameters acquired in real time or calculated dynamically into the formula, the magnitude of the wind thrust can be calculated. The direction of the wind thrust is consistent with the direction of the real-time wind direction vector acquired by S105. Finally, the calculated magnitude and direction of the force are combined to form a three-dimensional wind thrust vector.

[0100] S108. Apply the composite force vector to the corresponding contact point in the real-time operation digital twin, simulate the transmission process of the composite force vector in the real-time operation digital twin, and obtain the multi-axis load components corresponding to multiple interfaces in the temporary high-altitude operation platform.

[0101] It is understandable that the contact point refers to the actual point of contact between the worker's hand or foot and the surface of the temporary elevated work platform. If a worker has multiple contact points with the temporary elevated work platform, the center of gravity of the multiple points is calculated as the unique contact point.

[0102] Specifically, the active dynamic force vector and the wind thrust vector are vector-added to obtain a composite force vector. This composite force vector is then applied to the digital twin to locate the corresponding contact point between the worker and the platform surface. A built-in finite element analysis (FEA) solver is then activated. This solver treats the digital twin model of the work platform as a structure composed of a large number of finite elements, with each interface as a critical node. When the composite force is applied to the contact point, the solver calculates how the force is transmitted to each interface through the members, based on the material's elastic modulus, Poisson's ratio, and other mechanical properties, as well as the structure's geometric constraints. After the calculation is complete, the solver outputs six load components (Fx, Fy, Fz, Mx, My, Mz) experienced by each interface node in its own local coordinate system. These components describe the tensile / compressive, shear, torque, and bending moment that the interface is experiencing.

[0103] In some embodiments, the simulation of mechanical transmission can be achieved in several ways: Optionally, a real-time finite element analysis (FEA) method can be used: First, the digital twin model of the work platform is pre-divided into finite element meshes, and material properties and boundary conditions (such as ground support points) are defined; second, at each time step, the calculated composite force vector is applied as a load to the corresponding mesh nodes; finally, an efficient linear statics or dynamics solver is run to solve the stress-strain distribution of the entire structure in real time and extract the load components of all target interface nodes. Optionally, a fast solution method based on model order reduction can be used: First, in the offline stage, the complete finite element model is calculated multiple times under different working conditions, and the main deformation modes of the structure are extracted through techniques such as principal component analysis (PCA) or intrinsic orthogonal decomposition (POD), establishing a reduced-order model (ROM) with minimal computational cost; finally, the composite force vector is applied to this reduced-order model, and the load components of each interface are estimated in a very short time, achieving real-time mechanical analysis. It is understandable that other methods can be used to achieve mechanical simulation, such as using simplified matrix structural analysis methods, which are not limited here.

[0104] S109. When the multi-axis load component of any interface is greater than the corresponding interface force threshold, an alarm message is issued.

[0105] It is understandable that an alarm message is a signal used to indicate a dangerous situation. It can be in the form of visual, auditory, or data signals, and is used to indicate which interface is a dangerous interface.

[0106] Specifically, for each interface on the platform, the absolute value of the load component is compared one by one with the corresponding force threshold stored in S101. For example, the actual X-direction tensile force borne by the interface is compared with the X-direction tensile force threshold, and the Y-direction torque is compared with the Y-direction torque threshold. During this comparison process, if any load component of any interface exceeds the corresponding safety threshold, an overload event is determined to have occurred, triggering an alarm mechanism to generate and issue an alarm message. This alarm message typically includes detailed information such as the overloaded interface number, the specific direction of the overload, the percentage of overload, and the time of occurrence, so that management personnel can quickly locate the problem and take intervention measures.

[0107] In the above embodiments, by tracking three-dimensional human posture and combining it with a biomechanical model, the force exertion behavior of workers, which cannot be directly observed in the monitoring video, is quantified into a calculable active dynamic force vector. Based on this, environmental factors are further introduced; by analyzing real-time wind speed and the personnel's windward posture, a wind thrust vector is calculated, and the two are combined into a composite force vector that more closely approximates the actual working conditions. Finally, the composite force is applied to a real-time operational digital twin, and by simulating the transmission process within the platform structure, the real-time load on each key interface is calculated. This shifts the basis for early warning judgment from superficial behavioral recognition to underlying structural stress analysis, thereby enabling the identification of structural overload risks caused by personnel dynamic behavior and improving the accuracy of early warnings for dangerous behaviors on temporary high-altitude work platforms.

[0108] In other embodiments of this application, the monitoring method provided by the present invention is not limited to the structural mechanics analysis of high-altitude work platforms, but can also utilize core video analysis, human posture recognition and digital twin technologies to conduct in-depth monitoring of the compliance of personnel behavior in other work scenarios such as road and underground.

[0109] A typical application scenario is intelligent safety monitoring for operations in confined spaces such as manholes in urban areas or factory areas.

[0110] A local dynamic digital twin containing semantic information is constructed based on mobile monitoring. Managers or operators approach the wellhead using handheld terminals or helmet cameras. Utilizing the aforementioned VIO and SLAM technologies, a local 3D scene model of the area surrounding the wellhead is built in real time. Through target detection algorithms, the wellhead is automatically identified and marked as a semantic region with hazardous, confined spatial attributes in the digital twin. Simultaneously, nearby available safety rope anchor points (such as dedicated anchor piles or sturdy railings) are identified and located as compliance anchor points.

[0111] A multi-dimensional compliance analysis was conducted on the entire process of well operations.

[0112] 1. Personal Protective Equipment (PPE) Wearing Detection: When workers prepare to go down the mine, a full-body scan is performed. 3D human posture recognition locates the worker's feet area, and a specially trained fine-grained image classification model analyzes the image of this area to determine whether the worker is wearing standard safety shoes. This model is resistant to interference from mud, partial obstruction, and other factors.

[0113] 2. Safety Rope Wearing and Attachment Detection: This function simultaneously detects three targets: the worker, the safety rope, and the compliance anchor point. By analyzing the relative positions of these three elements in three-dimensional space, it determines: a) whether the safety rope is correctly connected to the worker's safety harness (i.e., one end of the rope is spatially and tightly connected to the waist / back area of ​​the human skeletal model); and b) whether the other end of the safety rope is securely attached to the pre-identified "compliance anchor point." An alarm will be immediately triggered if any of these conditions are not met.

[0114] 3. Compliance Assessment of Mine Descending Procedures (Behavioral Posture Sequence): As personnel begin descending the mine, their 3D human posture is continuously tracked, and a time sequence of key postures is recorded. A built-in standard safe mine descent behavior model defines the correct sequence of descent actions (e.g., facing the ladder, alternating hand grips, alternating foot treads). The real-time posture sequence is compared with the standard model, and a high-level alarm is triggered if dangerous behavior is detected, such as facing away from the ladder, jumping directly, or loss of balance (abnormal shift of the center of gravity).

[0115] 4. Continuous Monitoring of Operations Inside Downhole Pipelines: Once personnel enter the pipeline, the monitoring task is handed over from the mobile device at the wellhead to a pre-placed mobile monitoring device. First, the mobile monitoring device follows the personnel or enters the designated work area beforehand and begins continuous monitoring. The mobile monitoring device uses its built-in active light source to illuminate the work area and transmits real-time video and sensor data back via cable. Second, a dynamic digital twin of the pipeline's interior is constructed and updated in real time. As the mobile monitoring device moves within the pipeline, its onboard SLAM system constructs a 3D point cloud map of the pipeline's interior in real time. This map is then fused with the upper-level site digital twin model to create a complete scene model that includes the pipeline's internal geometry, the mobile monitoring device's real-time position and attitude, and the personnel's real-time position and attitude. Then, multimodal fusion risk analysis and early warning are performed. This is the core advantage of this solution in downhole monitoring—it not only analyzes video but also deeply integrates environmental sensor data with personnel behavior data.

[0116] Data collected by mobile monitoring equipment, such as the concentration of harmful gases and oxygen content, is mapped in real time to a digital twin model and visualized in the form of heat maps or digital labels. Once any indicator exceeds the safety threshold, a clear environmental hazard alarm will be immediately issued to the ground monitoring center and the workers (via the speaker on the mobile monitoring equipment or the communication device inside the safety helmet).

[0117] The system continuously analyzes the three-dimensional working postures of workers in confined spaces. By comparing these postures with a pre-set safe working posture database, it identifies prolonged poor postures that may lead to musculoskeletal injuries or the risk of getting stuck due to confined space, and issues warnings accordingly.

[0118] Combining human posture analysis with environmental data is crucial. For example, when a worker is detected to be stationary for an extended period while gas sensors on a motion monitoring device show excessively low oxygen concentrations or excessive levels of harmful gases, it can be determined that the worker is likely in an emergency due to poisoning or hypoxia, triggering the highest-level emergency rescue alarm. This approach is far more reliable and rapid than simply detecting stillness.

[0119] Through the above steps, safety monitoring is expanded from macroscopic mechanical analysis to microscopic, procedural compliance review of specific work behaviors. In particular, by strengthening the application of mobile monitoring technology, it can be flexibly deployed in any corner of the construction site to supervise temporary and dynamic operations (such as road excavation and well maintenance) that are difficult to cover with traditional monitoring. This not only improves the early warning capability for structural risks of high-altitude work platforms but also constructs a comprehensive intelligent security system that can fully protect the safety of personnel in various work scenarios on construction sites.

[0120] In other embodiments of this application, when an initial overload failure occurs at a single interface of the platform, unpredictable chain reactions may be triggered by load redistribution. Using the intelligent construction site security monitoring method provided in this application, this failure can be simulated and subsequent chain reactions can be deduced, generating a failure path diagram and escalating the alarm from a single point of danger to a prediction of the overall collapse trend.

[0121] like Figure 2 The diagram shown is another flowchart illustrating the intelligent security monitoring method for construction sites provided in this application, which includes the following steps:

[0122] S201. Obtain interface information based on the engineering information model of the temporary high-altitude operation platform.

[0123] S202. Construct a real-time operation digital twin based on the real-time video images and engineering information model of the temporary high-altitude operation platform.

[0124] S203. Track the preset key force points of the human body in the three-dimensional human posture in the real-time video image, and calculate the real-time motion parameters of the key force points of the human body.

[0125] S204. Combining the preset human biomechanical calculation model and real-time motion parameters, calculate the active dynamic force vector applied by the operator to the temporary high-altitude work platform.

[0126] S205. Obtain the real-time wind speed and direction at the temporary elevated work platform.

[0127] S206. Calculate the windward force area of ​​the operator based on the real-time wind direction and three-dimensional human posture.

[0128] S207. Based on the preset aerodynamic model and combined with real-time wind speed and windward force area, calculate the wind thrust vector acting on the operator.

[0129] S208. Apply the composite force vector to the corresponding contact point in the real-time operation digital twin, simulate the transmission process of the composite force vector in the real-time operation digital twin, and obtain the multi-axis load components corresponding to multiple interfaces in the temporary high-altitude operation platform.

[0130] Steps S201-S208 and Figure 1 Steps S101-S108 in the illustrated embodiment are similar and can be found in the descriptions of steps S101-S108, which will not be repeated here.

[0131] In some embodiments of this application, after calculating the multiaxial load components of each interface and determining that none of them exceed the instantaneous force threshold, these load fluctuations within the safe range can be continuously tracked and analyzed to assess the long-term fatigue damage to the structure, thereby achieving early warning of safety risks with hysteresis caused by material aging and cumulative damage.

[0132] The multiaxial load components of each interface are continuously recorded as a function of time. Using rainflow counting or a similar cycle counting algorithm, independent, closed stress cycle events are identified and extracted from the load history curves. A stress cycle is defined as the complete process of the load rising from a trough to a peak and back to a trough; key characteristics are the stress amplitude (half the difference between the peak and trough) and the mean stress. Calculated load fluctuation data over a certain time period is used as input to identify multiple stress cycles and their corresponding stress amplitudes.

[0133] For each identified stress cycle, a pre-defined material stress-life curve (i.e., SN curve) and fatigue cumulative damage calculation formula are invoked. The SN curve is a core concept in materials science, describing the relationship between the number of cycles a specific material can withstand under different stress amplitudes until fracture. This curve is pre-determined experimentally and stored in a database. Using the stress amplitude of the current stress cycle as input, the SN curve of the corresponding interface material is queried to obtain the total number of cycles at that stress level. According to the classic Miner's linear cumulative damage rule, the minute fatigue damage caused by a single cycle is defined as: minute fatigue damage = 1 / total number of cycles.

[0134] For each structural interface, an independent fatigue damage accumulation counter is maintained in the database. The initial value of this counter is zero. Whenever a minute amount of fatigue damage caused by stress cycle is calculated, this value is added to the counter of the corresponding interface.

[0135] Simultaneously, a preset fatigue life threshold is set for each interface. According to Miner's law, theoretically, the material will fail due to fatigue when the cumulative damage value reaches 1. In practical applications, to allow sufficient safety margin, this threshold is usually set to a value less than 1. The fatigue damage accumulation counter values ​​of all interfaces are monitored in real time. Once the cumulative damage value of any interface exceeds the corresponding fatigue life threshold, a fatigue warning is triggered. This warning differs from an instantaneous overload alarm; it indicates to management that although the interface has never been severely overloaded, the material is nearing the end of its lifespan, posing a risk of sudden brittle fracture, requiring immediate inspection, reinforcement, or replacement.

[0136] By performing fatigue cumulative damage analysis on dynamic loads without overload, it is possible to quantify the invisible microscopic damage to structures caused by routine operational activities, expanding the dimension of safety monitoring from a single strength check to lifespan warning. Because it can predict and warn of material property degradation caused by long-term, repeated, subcritical loads, it can issue early warnings before sudden fatigue-induced fracture occurs in the structure, improving the accuracy of warnings for hidden and delayed hazardous behaviors.

[0137] In some embodiments, after calculating the load components of each interface, the stiffness health status of the interface can be evaluated in real time by comparing the theoretical displacement in the digital twin with the actual displacement in the video surveillance, and its fatigue damage model can be dynamically corrected.

[0138] In the digital twin model, the multiaxial load components calculated by S208 are used as input, and the solver calculates the theoretical three-dimensional spatial displacement response at the center point of each interface under this load. This result represents the magnitude of the deformation that the structure should produce in an ideal model without any damage.

[0139] Simultaneously, high-precision computer vision algorithms are used to analyze real-time video images. Through techniques such as template matching, feature point tracking, or optical flow, the pixel position of each physical interface (e.g., bolt connections, weld edges) in the video frame is located. Combining the camera calibration parameters and depth information established by S202, the minute movements at the pixel level are converted into the actual displacement response of the interface center point in three-dimensional physical space.

[0140] By comparing the two sets of data, the stiffness degradation coefficient is obtained by calculating the ratio of the actual displacement response to the theoretical displacement response. In a brand-new, perfectly healthy structure, this ratio should be approximately equal to 1. However, over time, due to material fatigue, bolt loosening, or the formation of microcracks, the actual stiffness of the interface decreases, resulting in a larger displacement than the theoretical value under the same load. Therefore, when this ratio is significantly greater than 1, the degree of stiffness degradation of the interface is directly quantified. The stiffness degradation coefficient is compared with a preset stiffness degradation threshold; once it exceeds the threshold, the interface is determined to be a stiffness-degraded interface.

[0141] For interfaces deemed to have degraded stiffness, their health condition is worse than the standard condition, and their fatigue resistance should be correspondingly reduced. In this case, the preset fatigue life threshold corresponding to the stiffness-degraded interface is retrieved from the structural safety standard database. Dynamic correction is then performed using a stiffness degradation coefficient; a correction formula is: Corrected fatigue life threshold = Preset threshold / Stiffness degradation coefficient. The more severe the stiffness degradation of an interface (i.e., the larger the coefficient), the lower the upper limit of cumulative fatigue damage it can withstand. This lower, corrected threshold will be used in subsequent fatigue warning assessments.

[0142] By introducing a stiffness degradation coefficient, the abstract fatigue damage calculation is directly linked to the observable physical displacement response, allowing the fatigue life threshold to be dynamically adjusted according to the actual health condition of the structure. This overcomes the shortcomings of purely theoretical models being disconnected from physical reality, making the early warning of highly concealed structural fatigue failures caused by long-term dynamic operations more accurate, thereby improving the overall accuracy of hazard warnings.

[0143] S209. When the multi-axis load component of any interface is greater than the stress threshold of the overload interface, the bearing capacity of the corresponding overload interface is set to zero in the real-time operation digital twin to simulate the failure of the overload interface's bearing capacity.

[0144] It is understandable that an overload interface refers to a structural connection point that has been determined in the preceding steps to have actually borne a load exceeding the safety limit; zeroing the load-bearing capacity is a technical processing method in a mechanical simulation model, used to represent the complete loss of a structural unit's ability to transmit or bear any force or moment; and load-bearing capacity failure refers to the real situation in the physical world where the interface is completely destroyed, such as breaking, yielding, or falling off.

[0145] Specifically, when an interface is determined to be overloaded, the corresponding element or node in the digital twin finite element model is locked. Then, the properties of this element are modified so that it no longer has any stiffness, meaning it can no longer bear any load. This is computationally equivalent to removing the interface from the structure, thus simulating the physical process of redistributing the forces it originally bore to surrounding structural parts after the interface physically breaks completely.

[0146] S210. Based on structural mechanics analysis, determine the simulated multi-axis load components corresponding to multiple interfaces in the temporary high-altitude work platform.

[0147] Specifically, this step involves calculating the consequences of failure. The scenario is a damaged virtual structure, assessing the new impact of the original external loads. The same structural mechanics solver (such as a finite element analysis solver) as in the previous steps is invoked. The external loads applied to the platform (i.e., the combined force vector resulting from workers and wind) remain unchanged, but the force transmission paths have been altered. Based on this new structural topology and boundary conditions, the solver resolves the static equilibrium equations for the entire structure. This process calculates how the original loads are distributed and superimposed on adjacent or connected interfaces after the initial failure interface can no longer share the load. After the calculation is complete, a new load report is output, containing simulated multiaxial load components for all remaining interfaces.

[0148] S211. When the simulated multiaxial load component of any interface is greater than the preset interface stress threshold corresponding to the interface in the preset structural safety standard database, a failure path diagram containing the initial failure point and the predicted secondary failure point is generated, and a first alarm message is issued based on all nodes in the failure path diagram.

[0149] Understandably, a failure path diagram is a data or graphical representation used to show the chain of events in a structure from the initial failure point to all predicted secondary failure points; the first warning message is an alert that signals a large-scale, progressive structural failure that is about to occur or is already occurring.

[0150] Specifically, the simulated multiaxial load components of all surviving interfaces calculated in S210 are compared one by one with the original, inherent stress thresholds of each interface obtained in S201. If the new simulated load exceeds the safety threshold of any one or more previously safe interfaces, a "secondary failure" is determined to occur. At this point, a failure path diagram is generated. This failure path diagram shows the initial failure point as the starting point, as well as all predicted secondary failure points, revealing the possible paths and scope of structural collapse. Based on the failure path diagram, a first alarm message is generated and issued, conveying to management personnel an extreme danger signal that the overall structural stability has been compromised and that continuous collapse may occur in a short period of time, requiring the highest level of emergency response.

[0151] In the above embodiments, when an initial overload is detected, the load-bearing capacity of the interface is set to zero in the digital twin to simulate a real failure. A mechanical reanalysis is then performed to proactively deduce the chain reaction of load transfer to other interfaces caused by a single-point failure, and the collapse process of the entire structure is visually presented in the form of a failure path diagram. Therefore, the early warning information is upgraded from single-point overload to prediction of overall structural continuity failure. Ultimately, this is used in construction site operations to improve the accuracy of early warnings of dangerous behaviors on temporary elevated work platforms caused by the dynamic behavior of workers in monitoring videos.

[0152] The following describes an exemplary intelligent security monitoring system 300 for construction sites provided in an embodiment of this application. Figure 3 This is an exemplary hardware structure diagram of the intelligent security monitoring system 300 for construction sites provided in this application embodiment.

[0153] In some embodiments, the intelligent security monitoring system 300 at the construction site is a computer device or includes a computer device. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores data. The network interface of the computer device is used to communicate with other external terminals or servers via a network connection. In some embodiments, the network interface can be a wired network interface; in some embodiments, the network interface can also be a wireless network interface. When the computer program is executed by the processor, it implements the methods in the embodiments of this application.

[0154] Those skilled in the art will understand that Figure 3The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0155] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0156] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0157] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0158] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A construction site intelligent security monitoring method, characterized in that, The method comprises the following steps: acquiring interface information according to the engineering information model of the temporary elevated work platform; the interface information comprises an interface stress threshold corresponding to each interface; constructing a real-time work digital twin according to the real-time video image and the engineering information model of the temporary elevated work platform; tracking a preset human key force point in a three-dimensional human body posture in the real-time video image and calculating real-time motion parameters of the human key force point; calculating an active dynamic force vector exerted by the worker on the temporary elevated work platform by combining a preset human biomechanics calculation model and the real-time motion parameters; acquiring real-time wind speed and real-time wind direction at the temporary elevated work platform; calculating a windward stress area of the worker according to the real-time wind direction and the three-dimensional human body posture; calculating a wind thrust vector acting on the worker based on a preset aerodynamics model and in combination with the real-time wind speed and the windward stress area; applying a composite force vector to a corresponding contact point in the real-time work digital twin, simulating the conduction process of the composite force vector in the real-time work digital twin, and obtaining a plurality of multi-axis load components corresponding to a plurality of interfaces in the temporary elevated work platform respectively; the composite force vector is obtained by vector superposition of the active dynamic force vector and the wind thrust vector; when the multi-axis load component of any interface is greater than the corresponding interface stress threshold, an alarm information is issued.

2. The method of claim 1, wherein, After the step of applying the composite force vector to the corresponding contact point in the real-time work digital twin and simulating the conduction process of the composite force vector in the real-time work digital twin to obtain a plurality of multi-axis load components corresponding to a plurality of interfaces in the temporary elevated work platform respectively, the method further comprises the following steps: when the multi-axis load component of any interface is not greater than the corresponding interface stress threshold, the un-overloaded multi-axis load component corresponding to the un-overloaded interface is taken as a stress cycle of the un-overloaded interface; calculating a micro-fatigue damage degree caused by the stress cycle to the un-overloaded interface according to a preset material stress-life curve and a preset fatigue cumulative damage calculation formula; adding the micro-fatigue damage degree to a fatigue damage cumulative counter of the un-overloaded interface; when the value in the fatigue damage cumulative counter of any interface exceeds a corresponding preset fatigue life threshold, a fatigue warning is issued.

3. The method of claim 1, wherein, After the step of applying the composite force vector to the corresponding contact point in the real-time work digital twin and simulating the conduction process of the composite force vector in the real-time work digital twin to obtain a plurality of multi-axis load components corresponding to a plurality of interfaces in the temporary elevated work platform respectively, the method further comprises the following steps: calculating a theoretical displacement response of an interface center point in a three-dimensional space under the action of the multi-axis load component in combination with the real-time work digital twin; analyzing the change of each interface edge pixel in the real-time video image over time to obtain an actual displacement response of each interface center point; calculating a ratio of the theoretical displacement response to the actual displacement response as a stiffness degradation coefficient; When the stiffness degradation coefficient of any interface center point is greater than a preset stiffness degradation threshold, it is determined that the corresponding interface is a stiffness degradation interface, and a preset stiffness degradation interface fatigue life threshold corresponding to the stiffness degradation interface in a preset structure safety standard database is obtained; The fatigue life threshold of the stiffness degradation interface is corrected by the stiffness degradation coefficient.

4. The method of claim 1, wherein, When the multi-axis load component of any interface is greater than the corresponding interface stress threshold, an alarm information is issued, and specifically includes: When the multi-axis load component of any interface is greater than the corresponding interface stress threshold of an overload interface, the bearing capacity of the corresponding overload interface in the real-time operation digital twin is set to zero to simulate the failure of the bearing capacity of the overload interface; Based on structural mechanics analysis, the simulated multi-axis load component corresponding to the plurality of interfaces in the temporary elevated work platform is determined; When the simulated multi-axis load component of any interface is greater than the corresponding preset interface stress threshold of the corresponding interface in the preset structure safety standard database, a failure path graph including an initial failure point and a predicted secondary failure point is generated, and a first alarm information is issued based on all nodes in the failure path graph.

5. The method of claim 1, wherein, After the real-time motion parameters are combined with the preset human biomechanics calculation model, the active dynamic force vector exerted by the worker on the temporary elevated work platform is calculated, and the method further includes: Identifying the model of the work tool held by the worker in the real-time video image; Retrieving the inherent mechanical parameters corresponding to the model of the work tool in the tool physical property database; the tool inherent mechanical parameters include the tool self-weight and the vibration force generated by the tool itself in the working state; Based on the tool inherent mechanical parameters, a tool force vector is calculated; The tool force vector and the dynamic force vector are vector synthesized to obtain a complex work force vector, which replaces the active dynamic force vector.

6. The method of claim 5, wherein, The tool force vector is calculated based on the tool inherent mechanical parameters, and specifically includes: When the model of the work tool is a preset lever tool and the work tool has a tool contact point with the temporary elevated work platform, a power point, a resistance point and a fulcrum point of the work tool are identified in the real-time video image; The power arm length between the power point and the fulcrum point, and the resistance arm length between the resistance point and the fulcrum point are measured; According to the lever principle formula, the support force borne by the fulcrum point is calculated based on the dynamic force vector, the length of the power arm and the length of the resistance arm; The support force is taken as the tool force vector.

7. The method of claim 1, wherein, After the real-time work digital twin is constructed according to the real-time video image and the engineering information model of the temporary elevated work platform, the method further includes: When there are a plurality of workers, the spatial distance between each worker is calculated; When the spatial distance is less than a preset personnel safety distance threshold, an alarm is issued to prompt the workers to disperse.

8. A construction site intelligent security monitoring system, characterized in that, The construction site intelligent security monitoring system comprises one or more processors and a memory; the memory is coupled with the one or more processors, the memory is used for storing computer program codes, the computer program codes comprise computer instructions, and the one or more processors invoke the computer instructions to enable the construction site intelligent security monitoring system to execute the method in any one of claims 1-7.

9. A computer program product comprising instructions, characterized in that, When the computer program product runs on the construction site intelligent security monitoring system, the construction site intelligent security monitoring system is enabled to execute the method in any one of claims 1-7.

10. A computer-readable storage medium comprising instructions, characterized in that, When the instructions run on the construction site intelligent security monitoring system, the construction site intelligent security monitoring system is enabled to execute the method in any one of claims 1-7.

Images