A multi-sensor fusion construction safety monitoring method and system

Abstract

This application discloses a multi-sensor fusion method and system for construction safety monitoring. The method includes: acquiring data from multiple sensors installed at the construction site and fusing the data to obtain a fusion data package; acquiring a building information model (BIM) and a 3D geological model of the construction site, and constructing a 3D digital scene based on the BIM and 3D geological models; performing coordinate transformation on each type of data to map it to the 3D digital scene, obtaining a target 3D image; determining the current warning information for each type of data based on its warning threshold, and issuing construction warnings based on the current warning information; determining visualization information based on the fusion data package, the target 3D image, and the current warning information for each type of data, and conducting construction safety monitoring based on the visualization information. This application can solve the problems of isolated data processing and lack of in-depth correlation analysis.

Description

Technical Field

[0001] This invention relates to the field of construction safety monitoring, specifically to a construction safety monitoring method and system that integrates multiple sensors. Background Technology

[0002] With the increasing demands for construction safety management, existing technologies in the field of construction safety monitoring have gradually shifted from traditional manual on-site monitoring to automation and remote monitoring.

[0003] Currently, the most advanced construction monitoring technologies in the industry mainly involve installing various monitoring sensors at key construction sites to integrate multi-dimensional monitoring information such as stress, settlement, displacement, and water level. Then, through relevant transmission modules, these monitoring data are aggregated to a cloud monitoring platform, which then manages the data in a unified manner.

[0004] Although existing technologies have achieved multi-parameter monitoring and cloud server management, which is a significant improvement over traditional manual monitoring, there are still obvious limitations: they are essentially just simple data collection, lacking the ability to deeply integrate and correlate multi-source heterogeneous monitoring data, each sensor data is processed independently, and no knowledge model reflecting the mechanical relationship and spatial location relationship of the construction structure has been established, which makes it impossible to comprehensively assess the construction safety status and difficult to trace the root cause of risks. Summary of the Invention

[0005] In view of the above-mentioned defects or deficiencies in the existing technology, it is desirable to provide a construction safety monitoring method and system based on multi-sensor fusion, which can map the fused data of multiple sensors onto a three-dimensional scene constructed by building information model and three-dimensional geological model, thereby solving the problems of isolated processing of multi-source heterogeneous data, lack of in-depth correlation analysis and global safety assessment in the existing technology.

[0006] Firstly, this application provides a construction safety monitoring method based on multi-sensor fusion, the method comprising: Data is collected from various sensors installed at the construction site, and the collected data is fused to obtain a fused data package. Each type of data includes the coordinates of the collection point. Obtain the building information model and 3D geological model of the construction site, and construct a 3D digital scene based on the building information model and 3D geological model; Coordinate transformation is performed on each type of acquired data, mapping each type of acquired data to a three-dimensional digital scene to obtain a target three-dimensional map; The current warning information for each type of collected data is determined based on the warning threshold, and construction warnings are issued based on each type of current warning information. Visualization information is determined based on the fused data package, the target 3D map, and the current early warning information for each type of collected data, and construction safety monitoring is carried out based on the visualization information.

[0007] In conjunction with the first aspect, in one possible implementation, there are multiple warning thresholds for each type of collected data. Then, based on the warning threshold of each type of collected data, the current warning information for the corresponding collected data is determined, and construction warnings are issued based on each type of current warning information, specifically as follows: For each type of collected data, based on each warning threshold of the collected data, determine the current warning information of the collected data under the corresponding warning threshold; Construction warnings are issued based on each current warning information for each type of collected data.

[0008] In conjunction with the first aspect, in one possible implementation, for each type of collected data, based on each warning threshold of the collected data, the current warning information of the collected data under the corresponding warning threshold is determined, specifically as follows: When any collected data exceeds its own first warning threshold, a first-level current warning message corresponding to that collected data is generated; Each collected data point is associated with at least one other collected data point to obtain multiple association groups; When every data point collected in any associated group exceeds its own second warning threshold, a second-level current warning message corresponding to that associated group is generated. For each piece of collected data, the corresponding mechanical index is determined. When the mechanical index of the collected data is greater than the third warning threshold of the collected data, a third-level current warning information is generated for the collected data.

[0009] In conjunction with the first aspect, in one possible implementation, the metadata for each current warning information includes at least one of the following: coordinates of the warning location that generated the current warning information, collected data at the warning location, sensor number at the warning location, and warning type. After issuing a construction warning based on the current warning information, it also includes: Obtain monitoring logs, which are used to store multiple historical cases. Each historical case includes historical warning information, historical handling measures, and historical warning reasons. Based on the metadata in each current warning message, the monitoring log is matched to obtain historical cases that match the current warning message; Auxiliary diagnostic reports are generated based on successfully matched historical cases; Based on the auxiliary diagnostic report, determine the response strategy for the current early warning information, and update the monitoring log based on the response strategy and the current early warning information.

[0010] In conjunction with the first aspect, one possible implementation involves fusing data including knowledge graphs, which then involves fusing multiple types of collected data, specifically as follows: Preprocess each type of collected data; Spatiotemporal alignment and feature extraction are performed sequentially on each type of preprocessed data to obtain the feature data corresponding to each type of data. Knowledge graphs are constructed using a variety of feature data.

[0011] In conjunction with the first aspect, in one possible implementation, the acquired data includes video data. In this case, preprocessing is performed on each type of acquired data, specifically as follows: The video data is identified based on a pre-set computer vision model to obtain the recognition result; The recognition results are formatted to obtain standard data messages, which are then used as preprocessed video data.

[0012] In conjunction with the first aspect, one possible implementation involves monitoring construction safety based on visualized information, specifically as follows: Receive user requests to access target regions in the target 3D map; Based on knowledge graphs, the associated regions of the target region are identified, and a comprehensive monitoring report is generated based on the feature data of the target region and the associated regions. Construction safety monitoring is conducted based on comprehensive monitoring reports.

[0013] In conjunction with the first aspect, one possible implementation involves constructing a 3D digital scene based on a building information model and a 3D geological model, specifically as follows: The transparency of the building information model and the 3D geological model were adjusted separately; The building information model and the 3D geological model, with their transparency adjusted respectively, are overlaid and rendered to obtain a 3D digital scene.

[0014] In conjunction with the first aspect, one possible implementation involves preprocessing each type of collected data, specifically as follows: Median filtering was applied to each type of collected data.

[0015] Secondly, this application also provides a multi-sensor fusion construction safety monitoring system, the system comprising: The data acquisition module is used to acquire data from various sensors set up at the construction site, and to fuse the data to obtain a fused data package. Each data acquisition includes the coordinates of the acquisition point. The 3D module is used to acquire the building information model and 3D geological model of the construction site, and to construct a 3D digital scene based on the building information model and 3D geological model; The transformation module is used to perform coordinate transformation on each type of acquired data, mapping each type of acquired data to a three-dimensional digital scene to obtain a target three-dimensional image; The early warning module is used to determine the current early warning information for each type of collected data based on the early warning threshold, and to issue construction early warnings based on each type of current early warning information. The visualization module is used to determine visualization information based on the fused data package, the target 3D map, and the current early warning information of each type of collected data, and to conduct construction safety monitoring based on the visualization information.

[0016] Thirdly, this application also provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method of any one of the first aspects above.

[0017] Fourthly, this application also provides a computer program product containing instructions that, when executed, perform any of the methods described in the first aspect above.

[0018] Fifthly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the method of any one of the first aspects above. Attached Figure Description

[0019] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating a method in one embodiment; Figure 2 This is a structural block diagram of the system of this application in one embodiment. Detailed Implementation

[0020] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present application will now be described in detail with reference to the accompanying drawings and embodiments. Furthermore, the term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The terms "first" and "second," etc., in the specification and claims of the embodiments of this application are used to distinguish different objects, not to describe a specific order of objects.

[0022] The following is a description of the terminology used in this application: Structural components: The specific parts that form a whole system or entity and serve as support, load-bearing, or framework. In the field of engineering, it encompasses various load-bearing components and their connection nodes, such as beams, columns, slabs, and walls. These elements work together to maintain the stability and safety of the overall structure.

[0023] BIM model: A three-dimensional digital structural model constructed using building information modeling technology, which includes the spatial dimensions, component attributes, installation locations, and corresponding geographic coordinates of the engineering structure. Three-dimensional geological model: A three-dimensional digital geological model constructed based on geological survey data of the engineering site, which can reflect the site's stratigraphic distribution, rock and soil mechanical parameters, hydrogeological conditions and spatial geographic relationships; 3D digital scene: An integrated 3D visualization scene formed by transparently overlaying and rendering BIM model and 3D geological model can serve as the basic carrier for spatial mapping and intuitive display of monitoring data.

[0024] Knowledge graph: A semantic network that stores knowledge in the form of a graph structure. In this application, it represents entities (such as sensors, equipment, personnel, environmental factors, construction activities, etc.) and their relationships (such as "located in", "monitoring", "influence", "containment", etc.) as triples (entity-relationship-entity). This makes data no longer isolated numerical values ​​or signals, but information units with clear semantics and relevance, providing a solid foundation for subsequent intelligent analysis and decision-making.

[0025] In the field of safety monitoring during cofferdam construction, existing technologies have evolved from traditional manual monitoring to automation and remote monitoring. Currently, the most advanced technology involves installing various sensors at key construction sites, integrating multi-dimensional monitoring information such as stress and settlement, and converging it to a cloud monitoring platform for unified management, threshold judgment, and early warning, effectively compensating for the shortcomings of traditional manual monitoring.

[0026] Although existing technologies enable multi-parameter monitoring and cloud server management, they still have the following limitations, making it difficult to meet the needs of refined and intelligent monitoring: 1. The system is only a data collection and simple threshold alarm platform. It lacks the ability to deeply integrate and correlate multi-source heterogeneous monitoring data, and has not established a knowledge model that reflects structural mechanics and spatial location correlation. Therefore, it cannot achieve comprehensive status assessment and risk root cause tracing.

[0027] 2. The early warning logic relies on a preset static single-parameter threshold, which cannot achieve dynamic early warning with multi-parameter coupling, nor does it have the intelligent early warning capability of introducing simplified mechanical models for deduction and prediction.

[0028] 3. Data display mainly consists of two-dimensional charts and data lists, lacking visualization capabilities that are deeply integrated with the three-dimensional model of the cofferdam. It is impossible to intuitively locate sensors and correlate analysis data in three-dimensional space, which affects data analysis efficiency and user experience.

[0029] 4. The system has limited functionality, does not integrate video surveillance and visual analysis technologies, lacks management of unstructured information such as engineering documents and inspection records, and has not built an engineering knowledge base linked to monitoring data, thus limiting its decision support value.

[0030] 5. Data processing relies entirely on cloud servers, failing to leverage the advantages of edge computing in data preprocessing and real-time performance assurance. This results in poor real-time performance of monitoring data transmission and processing, reducing monitoring reliability.

[0031] In summary, existing technologies have significant shortcomings in data intelligence fusion, early warning mechanism coupling, three-dimensional interactive visualization, multimodal information integration, and edge-cloud collaborative computing. There is an urgent need for a monitoring solution that can overcome these deficiencies.

[0032] Based on this, the embodiments of this application provide a construction safety monitoring method and system based on multi-sensor fusion, which can map the fused data of multiple sensors onto a three-dimensional scene constructed by building information model and three-dimensional geological model, thus solving the problems of isolated processing of multi-source heterogeneous data, lack of in-depth correlation analysis and global safety assessment in the prior art.

[0033] The method provided in this application embodiment can be applied to, for example... Figure 2 The system shown.

[0034] One possible implementation is, such as Figure 1 As shown, this application provides a construction safety monitoring method using multi-sensor fusion, the method comprising: S1. Acquire the data collected by various sensors set up at the construction site, and perform fusion processing on the various data to obtain a fused data package. Each data includes the coordinates of the collection point. In one embodiment, if the construction project of this application is a cofferdam, then S1 includes: S11. Various sensors are deployed in the environment of the cofferdam and its surrounding area within a predetermined range to collect data. Specifically, the various sensors include sensors for monitoring the structural status of the cofferdam (such as strain gauges and displacement gauges), sensors for monitoring hydrological conditions (such as water level gauges and flow meters), and sensors for monitoring environmental parameters (such as anemometers and particulate matter monitors).

[0035] For ease of description, the above types of sensors will be referred to as structural sensors, hydrological sensors, and environmental sensors in the following description.

[0036] This application also includes an edge computing gateway, which connects to each sensor via wired or wireless communication protocols to aggregate various types of raw sensor data.

[0037] S12. Perform fusion processing on multiple collected data to obtain a fused data packet; Furthermore, this application also sets up a cloud server that communicates with the edge computing gateway. The cloud server is equipped with a data fusion analysis engine. The data fusion analysis engine is used to execute step S12, that is, to receive various pre-processed collected data, perform spatiotemporal alignment through a unified timestamp and spatial coordinate benchmark, and calculate statistical features and trend features according to a preset feature extraction algorithm to generate a fusion data package containing original data and derived features.

[0038] In one possible implementation, the fused data includes a knowledge graph, then S12 specifically refers to: S121. Preprocess each type of collected data; In one possible implementation, S121 is specifically as follows: Median filtering was applied to each type of collected data.

[0039] In one possible implementation, the edge computing gateway establishes a stable communication connection with multiple sensors via wired or wireless communication protocols. Its core function is to perform the data preprocessing operation in step S121. Specifically, the preprocessing process in step S121 includes: first, the edge computing gateway aggregates and integrates various types of collected data, unifies the format, and completes local caching of the data to avoid data loss; then, preliminary filtering and validity verification are performed on the unified formatted collected data to ensure the reliability of the subsequently uploaded data.

[0040] The preliminary filtering process is as follows: The edge computing gateway has a built-in edge computing module with certain computing capabilities. Before uploading the collected data to the cloud server, the edge computing module runs a lightweight processing algorithm locally to suppress noise by using moving average filtering or median filtering for various types of collected data. This completes the preliminary filtering operation and effectively filters out invalid noise data generated by sensor errors and environmental interference during the collection process.

[0041] After the initial filtering is completed, the edge computing module will also apply statistical methods based on standard deviation or interquartile range to initially identify and mark outliers in the collected data, and set corresponding limits for each parameter. When a parameter exceeds the preset limit, the obviously abnormal data identified can be diagnosed and marked in real time to clarify the initial anomaly type of the abnormal data. The collected data, after undergoing the aforementioned series of local preprocessing operations, will be accompanied by preliminary quality identification and diagnostic labels. It will then be compressed, packaged, and uploaded to the cloud server. This collaborative approach between local preprocessing and the cloud server not only effectively reduces the data processing load on the cloud server and lowers the data transmission bandwidth pressure, but also provides higher-quality input data for the cloud server's subsequent in-depth fusion and correlation analysis, ensuring the accuracy and efficiency of the cloud server's analysis results.

[0042] In one possible implementation, the acquired data includes video data, then S121 is specifically as follows: S1211. Based on a preset computer vision model, the video data is identified to obtain the identification result; S1212. Format the recognition results to obtain standard data messages. Use the standard data messages as preprocessed video data and incorporate them into the subsequent data fusion analysis process.

[0043] In response to the acquisition and preprocessing of video data, this system deploys video monitoring units and video intelligent analysis modules that establish communication connections with key viewpoints of the cofferdam. The video monitoring units use high-definition network cameras, which can continuously acquire real-time video streams from the cofferdam construction site to ensure the integrity and clarity of the video data. The video intelligent analysis module can be integrated into the cloud server or deployed as a standalone service on the edge computing gateway. It pre-integrates a trained computer vision model that can analyze the incoming real-time video stream frame by frame. This model can accurately identify pre-set risky behaviors at the construction site, such as personnel entering restricted areas or large machinery getting too close to the cofferdam. Furthermore, through digital image processing technology, it can automatically read the real-time readings of various instruments, such as mechanical water level gauges and pressure gauges, in the video footage, achieving automated data acquisition. The identification results or automatically read instrument values ​​obtained by the video intelligent analysis module are further formatted into standard data packets. These standard data packets, acting as a virtual sensor data source, are injected into the subsequent data analysis and processing flow. They collaborate with physical data collected by other sensors to participate in subsequent multi-source data fusion and early warning decision-making processes, thereby achieving quantitative utilization of visual information from the construction site, overcoming the limitations of traditional pure physical sensor monitoring, and improving the comprehensiveness of system monitoring.

[0044] S122. Perform spatiotemporal alignment and feature extraction on each type of preprocessed data in sequence to obtain the feature data corresponding to each type of data. S123. Construct a knowledge graph using multiple feature data.

[0045] During the implementation of steps S122 to S123, the data fusion analysis engine is specifically used to construct a knowledge graph. The knowledge graph uses each sensor entity, the type of monitored physical quantity, and the installation structure as core nodes, and the spatial affiliation, physical coupling, and data correlation between nodes as connecting edges, forming a complete networked relational structure. This knowledge graph serves as the core index for the entire system's data organization, driving the unified storage of various types of monitoring data, specifically including strain or displacement time-series data, water level or flow velocity time-series data, and wind speed or temperature time-series data, achieving standardized integration and storage of monitoring data from different types and sources. Meanwhile, the knowledge graph also supports cross-sensor type and cross-physical quantity association queries and joint analyses centered on structural parts or risk events. Among them, structural parts are derived from the BIM model and construction design documents of the construction site, corresponding to the specific structural areas of the cofferdam construction (such as cofferdam walls, supporting structures, foundation parts, etc.), which are the core areas of the construction structure that are predefined and entered into the system. Risk events are derived from preset construction safety risk rules, summaries of historical monitoring anomaly data, and abnormal scenarios identified during real-time monitoring (such as personnel entering restricted areas, machinery illegally approaching the cofferdam, monitoring data exceeding thresholds, etc.), covering various safety hazard scenarios that may occur during cofferdam construction.

[0046] This application enables the rapid association of all sensor data corresponding to a specific structural component, or various monitored physical quantity data related to a risk event. This effectively supports the assessment of the correlation between structural response and environmental loads at the data level, providing data-driven support for subsequent safety assessments and early warning decisions. By fusing multiple types of collected data, a knowledge graph reflecting the relationships between sensors, structural components, and monitoring projects is constructed. This successfully transforms monitoring data from isolated data points to a networked association, breaking down the barriers of isolated data, significantly improving the depth and efficiency of data analysis, and further enhancing the data processing and decision support capabilities of the entire monitoring system.

[0047] In one possible implementation, a 3D visualization service engine is also deployed inside the cloud server, which is used to execute step S2: S2. Obtain the building information model and 3D geological model of the construction site, and construct a 3D digital scene based on the building information model and 3D geological model; Specifically, the building information model and the 3D geological model can be extracted from their respective databases and then loaded through a graphics rendering engine. During the loading process, the two models are directly superimposed in the same coordinate system to form a 3D digital scene that includes the above-ground building structure and the underground geological structure.

[0048] In one possible implementation, a three-dimensional digital scene is constructed based on a building information model and a three-dimensional geological model, specifically as follows: The transparency of the building information model and the 3D geological model were adjusted separately; The building information model and the 3D geological model, with their transparency adjusted respectively, are overlaid and rendered to obtain a 3D digital scene.

[0049] The 3D visualization service engine in this application is built on a geographic information system platform. By loading the building information model of the cofferdam and the surrounding topographic data in the 3D geological model, a 3D digital scene is formed. The geographic coordinates of the sensors in the fusion data package are mapped to the corresponding locations in the scene, thereby realizing the spatial attachment and dynamic rendering of the monitoring data.

[0050] S3. Perform coordinate transformation on each type of collected data, mapping each type of collected data to a three-dimensional digital scene to obtain a target three-dimensional map; A target 3D map is a graphical representation formed by overlaying or mapping sensor data, after coordinate transformation, onto a 3D digital scene. This map can show the specific location and status of real-time monitoring data at the construction site.

[0051] The 3D visualization service engine provided by this application uses layered rendering technology to enable users to adjust the transparency of the 3D model layer of the cofferdam and the geological information model layer, so as to realize the transparent overlay display of the structure and the geological body, and intuitively reflect the spatial relationship between the cofferdam and the strata. Furthermore, this application relies on the sensor relationships stored in the knowledge graph, combined with 3D transparent scenes, sensor spatial attachment, and interactive association highlighting technology, to provide users with an intuitive and immersive visual monitoring experience, greatly facilitating anomaly location and status assessment.

[0052] This application constructs a knowledge graph to achieve spatiotemporal alignment, feature extraction, and associated storage and querying of multi-dimensional monitoring data, including structural, hydrological, and environmental data.

[0053] In one possible implementation, an intelligent early warning decision model is also deployed inside the cloud server. This intelligent early warning decision model has a built-in association rule library. By continuously analyzing the combination state and change pattern of multiple feature data in the fusion data package, when it meets the predefined multi-parameter association logic conditions in the library that represent different risk levels, it triggers an early warning signal of the corresponding level.

[0054] For details on the early warning method of the intelligent early warning decision model, please refer to step S4: Determine the current early warning information of the corresponding collected data based on the early warning threshold of each type of collected data, and conduct construction early warning based on each type of current early warning information; Step S4 specifically involves: based on the preset early warning threshold corresponding to each type of collected data, performing threshold comparison and status judgment on each pre-processed and fused collected data to determine the current early warning information corresponding to each type of collected data, clarifying whether each collected data is abnormal, the degree of abnormality, and the corresponding risk tendency, and then, based on the current early warning information of all collected data and combined with the early warning signal triggered by the intelligent early warning decision model, comprehensively carrying out construction early warning operations to ensure the comprehensiveness and accuracy of early warning information, and timely reminding staff to take corresponding prevention and control measures to avoid construction safety risks.

[0055] In one possible implementation, multiple warning thresholds are set for each type of collected data. Specifically, "multiple warning thresholds for each type of collected data" means that for each type of data collected at the construction site, such as displacement, settlement, vibration, temperature, and stress, instead of setting a single critical value, a series of progressively related values ​​are set. These thresholds can be determined based on engineering design specifications, historical monitoring data, risk assessment models, or expert experience. For example, they can be set as multiple levels such as "attention level," "warning level," and "danger level," each level corresponding to a specific numerical range or critical point. This multi-threshold setting aims to achieve refined hierarchical management of risks.

[0056] S4 is specifically: S41. For each type of collected data, determine the current warning information of the collected data under the corresponding warning threshold based on each warning threshold of the collected data; S42. Conduct construction early warning based on each current early warning information of each type of collected data.

[0057] In conjunction with the first aspect, in one possible implementation, S41 is specifically as follows: S411. When any collected data exceeds its own first warning threshold, a first-level current warning message corresponding to the collected data is generated. S412. Associate each collected data with at least one other collected data to obtain multiple association groups; S413. When every collected data of any associated group exceeds its own second warning threshold, generate the second-level current warning information corresponding to that associated group. To more comprehensively assess risks, this application correlates each acquired data point with at least one other acquired data point, thereby obtaining multiple correlation groups. In practice, these correlation groups can be established based on the physical location of the sensors at the construction site, the functional correlation of the monitored objects, or the mutual influence relationships within the engineering structure. For example, multiple different types of sensors (such as strain gauges, inclinometers, and crack gauges) located on the same structural component can be grouped into one correlation group to comprehensively reflect the overall condition of the component. This establishment of correlations allows the system to expand from monitoring a single data point to comprehensive analysis across multiple dimensions and factors.

[0058] The Level 2 early warning information relies on preset multi-parameter combination logic. For example, it is only triggered when the water level parameter exceeds the threshold and the strain change rate at the associated location is synchronously abnormal, thereby reducing false alarms. S414. For each piece of collected data, determine the mechanical index corresponding to the collected data. When the mechanical index of the collected data is greater than the third warning threshold of the collected data, generate the third-level current warning information of the collected data.

[0059] The three-level current warning information of this application introduces structural mechanics mechanism. By calling a simplified and parameterized overall or local mechanical calculation model of the cofferdam, the water level difference, earth pressure and other parameters in the real-time fusion data package are converted into corresponding mechanical indicators. These indicators are then compared with the third warning threshold of the corresponding mechanical indicators to infer and trigger the warning. The three-level current warning information aims to identify risks that may not yet be clearly reflected in the direct monitoring data but are about to appear through mechanical transmission.

[0060] Furthermore, Level 1 warnings may only be displayed through the system interface, reminding management personnel to pay attention; Level 2 warnings may trigger SMS or email notifications to relevant personnel and suggest on-site verification; while Level 3 warnings may activate more stringent measures such as audible and visual alarms, automatic shutdown commands, or emergency evacuation plans. This tiered warning mechanism ensures the accuracy and timeliness of the warning response, avoiding overreaction or underreaction.

[0061] Each warning threshold is determined by a person skilled in the art based on the actual engineering situation and relevant information.

[0062] This application establishes a multi-level intelligent early warning mechanism that combines single-parameter thresholds, multi-parameter coupling rules, and simplified mechanical model derivation, which significantly improves the timeliness and accuracy of early warning and achieves a leap from passive alarm to proactive prediction.

[0063] S5. Determine visualization information based on the fused data package, the target 3D map, and the current early warning information of each type of collected data, and conduct construction safety monitoring based on the visualization information.

[0064] This application sets up a remote terminal to implement step S5. The remote terminal accesses the cloud server through a web page or application programming interface, receives and parses the fused data packets, target 3D images and early warning signals sent by the cloud server, and performs graphical display and interaction in the user interface.

[0065] Visualized information refers to information that integrates data packets, target 3D maps, and current early warning information, and then presents it graphically and interactively in the form of graphics, images, animations, and reports. This information aims to help users understand the safety status of the construction site and make informed decisions.

[0066] To achieve the aforementioned graphical display and interactive functions, the visualization interface provided by the remote terminal adopts a panel layout. Each panel corresponds to specific data content sent by the cloud server, realizing accurate association and categorized presentation of data and interface. The visualization interface specifically includes: a project overview panel that centrally displays key performance indicators and the global distribution of early warnings. The key performance indicators are extracted from the core data collected in the fusion data package, and the global distribution of early warnings corresponds to the early warning signals issued by the cloud server. Combined with the spatial information of the target 3D map, it clearly presents the early warning distribution status of each region; a hydrological environment panel that dynamically displays water level, flow velocity, wind speed, and environmental quality data. All the data displayed are directly derived from the corresponding hydrological and environmental monitoring data in the fusion data package; a structural monitoring panel that specifically displays structural response data such as strain, displacement, and settlement. The data is all taken from the fusion data package. Among them, a dedicated data view designed for settlement monitoring data uses a dual Y-axis coordinate system chart. The primary Y-axis is used to draw a bar chart of daily settlement increments, and the secondary Y-axis is used to draw a curve of cumulative settlement changes from the monitoring start date to the present, so as to compare the daily changes and the overall trend in the same view; and a video and early warning panel that integrates video footage and early warning lists. The video footage comes from the video data in the fusion data package, and the early warning list corresponds to all early warning signals issued by the cloud server, realizing the linkage display of video footage and early warning information. In addition, the system provides a coordinate system switching function, which can switch the background coordinate system of various data curves from the conventional mathematical coordinate system to the construction coordinate system that is completely consistent with the cofferdam design drawings. The parameters of the construction coordinate system are taken from the target 3D map, so that the position of the data points on the chart can directly correspond to the actual position on the design drawings. This makes it easier to make an intuitive comparison between the monitoring data in the fusion data package and the design values, further improving the practicality and intuitiveness of the data display.

[0067] In other embodiments, the process of determining the visualization information is as follows: Data in the fused data package can also be used to generate a series of two-dimensional charts, such as trend charts or distribution maps. The target 3D map can be used to provide the spatial background of the construction site, and when an alert occurs, the location of the alert is indicated in the 3D scene by a simple color change (e.g., coloring the over-limit area red). The current alert information can be displayed in the form of a pop-up window or scrolling text. Users can monitor construction safety by comprehensively viewing these scattered information. As another implementation, key indicators in the fused data package can be extracted and presented in the form of simple line charts or bar charts. The target 3D map can be used to display sensor locations; when there is an alert, the corresponding sensor location is indicated by flashing red on the 3D map. Users can monitor construction safety by observing these independent charts and flashing locations.

[0068] In one possible implementation, a data penetration analysis tool is integrated into the visual interface of the remote terminal; then S5 is specifically: S51. The data penetration analysis tool receives user requests to access the target area in the target 3D map. S52. The data penetration analysis tool determines the associated regions of the target region based on the knowledge graph and generates a comprehensive monitoring report based on the characteristic data of the target region and the associated regions. Determining related regions of a target area based on a knowledge graph refers to utilizing the structured knowledge representation and reasoning capabilities provided by the knowledge graph to discover implicit relationships between the target area and other related areas on the construction site. This helps to extend isolated regional information into a broader context, thereby enabling a more comprehensive safety assessment. A knowledge graph can contain nodes (such as building components, equipment, personnel, sensors, hazard sources, construction procedures, geographical regions, etc.) and edges (representing relationships between them, such as "located in," "contains," "affects," "adjacent," "depends on," etc.). When an access request for a target area is received, the system uses the relationships in the knowledge graph for querying and reasoning. For example, if the target area is a "foundation pit," the knowledge graph can identify "adjacent" entities such as "slope monitoring points," "groundwater level sensors," "surrounding soil" that "affects" the stability of the foundation pit, and even "tower cranes" located above the foundation pit. Through semantic matching algorithms, the description of the target area is matched with entities in the knowledge graph. The graph is then traversed to find other entities with predefined relationships to that entity; the regions corresponding to these entities are the related regions. Furthermore, by combining spatial information from 3D digital scenes, knowledge graphs can also perform spatial relationship reasoning, such as identifying all sensors, devices, or structures within a certain distance of a target area. The quality of knowledge graph construction (including the richness and accuracy of entities and relationships) directly affects the accuracy and comprehensiveness of the determination of associated regions.

[0069] The comprehensive monitoring report generated based on the characteristic data of the target area and related areas refers to the summary and analysis of the overall security status of the target area and its related areas. It integrates the reasoning results of multi-source data and knowledge graphs to provide decision support for users. The report generation module extracts all characteristic data related to the target area and related areas from the fused data package, the target 3D map, current warning information, and the knowledge graph. This characteristic data may include: real-time sensor readings, historical trends, warning levels, equipment status, personnel locations, environmental parameters, structural deformation data, video analysis results, etc. The system further analyzes and evaluates the integrated characteristic data, such as comparing real-time data with warning thresholds, analyzing data change trends, assessing potential risk levels, and identifying abnormal patterns. Risk assessment can be performed using a preset rule engine or machine learning model. The report can include multiple parts, such as: area overview (location, main components), real-time data summary, historical data trends, details of current warning information, potential risk analysis, impact analysis of related areas, and recommended response measures. The report content is not limited to text and can also include visualization elements such as charts, curves, heat maps, and 3D model screenshots to intuitively display the security status. The generation of reports requires ensuring the accuracy, timeliness, and completeness of the data.

[0070] In one possible implementation, S52 specifically involves the following steps: When a user selects a closed target area of ​​any shape on the visual interface using a mouse or touch, the tool first identifies all sensors completely located within the target area based on a knowledge graph. Then, the knowledge graph is traversed again to find external sensors that, although located outside the target area, have a direct and strong connection with key components or sensors within the target area in the knowledge graph (e.g., belonging to the same support system or being on the same force path). Finally, the data penetration analysis tool includes the sensors within the area and these associated external sensors in the analysis set and automatically generates a comprehensive area monitoring analysis report. This report summarizes and displays real-time status snapshots of all sensors within the set, recent historical data change trend curves, and aggregated information on whether the area is currently in an early warning state, enabling a rapid assessment of the overall safety status of the local area.

[0071] S53. Conduct construction safety monitoring based on comprehensive monitoring reports.

[0072] Construction safety monitoring based on comprehensive monitoring reports refers to the system's ability to support more accurate and proactive construction safety monitoring and management by providing comprehensive and in-depth monitoring reports. These reports provide safety management personnel with decision-making support, such as whether to activate emergency plans, adjust construction schemes, or dispatch personnel for on-site verification. The risks and warning information clearly identified in the reports can trigger corresponding early warning mechanisms, such as sending notifications to relevant personnel or activating automated safety measures (e.g., stopping equipment operation). Safety management personnel can continuously track the safety status of the target area and its related areas based on the report content and evaluate the effectiveness of the measures taken. Comprehensive monitoring reports can also serve as historical records for accident investigations, accountability, and safety audits.

[0073] This application constructs a fusion analysis engine that integrates multi-dimensional sensing data such as structure, hydrology, environment, and video. By combining knowledge graphs and 3D visualization models, it achieves intelligent processing of the entire process from data collection, correlation analysis, intelligent early warning to 3D interactive display, thereby improving the real-time performance, accuracy, intuitiveness, and decision support capabilities of construction safety status assessment.

[0074] In one possible implementation, the metadata for each current warning information includes at least one of the following: coordinates of the warning location that generated the current warning information, the collected data at the warning location, the sensor number at the warning location, and the warning type. Then, after S5, it also includes: S6. Obtain monitoring logs. Monitoring logs are used to store multiple historical cases. Each historical case includes historical warning information, historical handling measures, and historical warning reasons. Specifically, each historical case details past warning events, including historical warning information, historical response measures, and historical warning reasons. The metadata for historical warning information is similar to that of current warning information, providing background information on past events; historical response measures record the specific actions taken to address the warning; and historical warning reasons analyze the root causes that led to the warning. By accumulating this historical experience, the monitoring logs build a valuable engineering knowledge base.

[0075] S7. Match the metadata in each current warning information in the monitoring log to obtain historical cases that match the current warning information; S8. Determine auxiliary diagnostic reports based on successfully matched historical cases; S9. Determine the response strategy for the current early warning information based on the auxiliary diagnostic report, and update the monitoring log based on the response strategy and the current early warning information.

[0076] The monitoring logs are stored in a dedicated storage area of ​​the engineering knowledge base module, which is logically integrated with the cloud server and also has an internal unstructured document parser for extracting metadata. When the intelligent early warning decision model triggers an early warning signal, it automatically triggers the retrieval interface of the engineering knowledge base module. Using the metadata attached to the current early warning signal (such as the triggering sensor number, early warning location, early warning type, and combination of related physical quantities) as query conditions, the system scans the entire engineering knowledge base module for historical cases, early warning handling reports, and other engineering documents associated with the current early warning signal, collectively categorizing them as related documents. Matching is performed by comparing the similarity between the related documents and the metadata of the current early warning signal. Upon successful matching, the system automatically generates a structured auxiliary diagnostic report. This report not only lists all types and values ​​of currently associated abnormal data collection, but also provides a textual description of the potential causes of this early warning based on the cause analysis in the matched historical cases, along with a summary of the core steps of the historical handling measures. This auxiliary diagnostic report is then pushed to remote terminals to provide decision support for management personnel.

[0077] In one possible implementation, the monitoring log also includes at least one of the following: electronic construction drawings, monitoring plans, daily inspection log text and photos, and historical early warning cases. Unstructured document parsing is used to automatically scan all document content in the monitoring log, identify and extract metadata such as engineering part identifiers, date and time information, and monitoring project keywords contained in all documents, and use these as tags to automatically establish bidirectional hyperlinks between the monitoring log and the fusion data package with spatiotemporal tags in the data fusion analysis engine in the background. This forms a structured association network between the fusion data package and the engineering documents on the timeline and spatial line, which facilitates event backtracking and experience retrieval.

[0078] This application integrates modules such as video analysis, automatic association with engineering knowledge base, edge preprocessing, and regional penetration analysis to form a complete closed loop of perception, analysis, early warning, decision-making, and backtracking, thus constructing a truly intelligent, integrated, and interactive cofferdam safety monitoring system.

[0079] The following is an example of using the method of the present invention for safety monitoring of cofferdam construction: S100: System Initialization and Sensor Network Deployment Multiple sensors are deployed at key parts of the cofferdam structure and in its surrounding environment to form a monitoring network. The structural sensors operate at frequencies... (0.1Hz~1Hz) sampling; hydrological sensors use frequency (0.0167Hz~0.1Hz) sampling; environmental sensors use frequency Sampling frequency: (0.0033Hz~0.0167Hz). Each sensor has a unique ID, type, and 3D construction coordinates. The component IDs of the cofferdam BIM model and the sensor they are installing are pre-entered into the system database. The information entered by each sensor corresponds one-to-one with the component information in the cofferdam BIM model, ensuring the accurate association between the sensor and the structural part.

[0080] S200: Edge-side data acquisition and preprocessing The edge computing gateway repeatedly executes the following steps S210~S240: S210: Receive data collected by each sensor according to the protocol. And based on the deviation between the sensor clock and the gateway standard time Perform timestamp compensation to give all data a unified absolute time label. .

[0081] S220: The collected data is packaged into a JSON format data packet with uniform fields, including the sensor's unique ID, collection timestamp, collected data value, and data type, to ensure that the cloud server can quickly parse and identify it.

[0082] S230: Performs local preprocessing on the data acquired by each sensor. It uses a length of... Median filtering is performed using a sliding time window to obtain the filtered value. Calculate the mean of the data collected within the calculation window. and standard deviation If the current data point is collected Meet the conditions ( (As a preset coefficient, usually set to 3), the quality of the collected data will be identified. Marked as "suspicious".

[0083] S240: Execute based on lightweight rule base Real-time diagnostics. A lightweight rule base pre-sets abnormal change thresholds for various sensors, based on sensor model, monitoring scenario, and construction specifications. For example, if the difference between the filtered data from the displacement gauge and the previous value exceeds the change threshold... Then a status label is generated. =“Mutation”. Finally, the preprocessed data packets are sent to the cloud server.

[0084] S300: Spatiotemporal Alignment and Feature Fusion of Multi-Source Data on Cloud Servers The cloud server receives the collected data stream from the edge computing gateway. The data fusion analysis engine then executes steps S310 to S340: S310: Precise spatiotemporal alignment. Secondary time calibration is performed based on the NTP server time. This is based on pre-stored sensor construction coordinates. Map all data to a unified construction coordinate system. The construction coordinate system is consistent with the coordinate system of the cofferdam design drawings and BIM model to ensure the spatial correlation of the data.

[0085] S320: Feature Extraction. For each sensor's acquired data stream, extract features at fixed time intervals. Using a window of 1 hour as an example, calculate the statistical characteristic: mean. ,variance Maximum value Minimum value Simultaneously, linear fitting is performed within the sliding window to obtain the trend slope. The raw acquired data values, together with these features, form the feature vector of the sensor. .

[0086] S330: Generate fused data packets. Aggregate the feature vectors of all sensors within the same time slice (e.g., 1 minute) to form a multi-dimensional fused data packet. N represents the total number of sensors. The fusion data package contains the feature information, timestamp, spatial coordinates and data quality identifier of each sensor, realizing the centralized integration of multi-source data.

[0087] S340: Update the monitoring knowledge graph .Will As an attribute update, it is sent to the corresponding sensor node. The update frequency is consistent with the generation frequency of the fusion data package, ensuring that the knowledge graph is synchronized with the real-time collected data, and providing data support for subsequent related queries and early warning judgments.

[0088] S400: 3D Scene Construction and Monitoring Data Integration The 3D visualization service engine executes steps S410~S430 as follows: S410: Obtain the geological profile map from the geological survey report, and use professional modeling software (such as AutoCAD and Revit) to construct a three-dimensional geological model based on the stratigraphic distribution and soil parameters in the geological profile map. At the same time, load the cofferdam BIM model and the three-dimensional geological model, and uniformly convert them to the target coordinate system (consistent with the construction coordinate system) through the coordinate transformation algorithm to ensure the accuracy of the model's spatial position.

[0089] S420: Set the transparency attribute of each model layer (User adjustable) ), to realize three-dimensional geological model ( ) and cofferdam BIM model ( Transparent overlay rendering.

[0090] S430: Transfer sensor coordinates Converted into icons in a 3D scene and matched with the atlas Sensor nodes in and fused data packets Dynamic links are established for real-time data, allowing users to view the real-time monitoring data of the corresponding sensor simply by clicking on the icon.

[0091] S500: Multi-level intelligent early warning and judgment The intelligent early warning decision model periodically or triggeredly processes the fused data packets. Perform steps S510~S540: S510: Execute Level 1 warning rule. Determine whether a single monitored parameter exceeds its independent statistical warning threshold. If so, generate a yellow (L1) warning signal.

[0092] S520: Execute the level 2 warning rule. Determine if the multi-parameter combination logic is satisfied. For example, the condition " "Whether both are true. If so, an orange (L2) warning signal is generated."

[0093] S530: Execute the Level 3 early warning rule. Invoke the simplified cofferdam mechanical calculation model. Use real-time data as input. Calculate the estimated value of the key response quantity Calculate the safety ratio ,in This is the design tolerance value. If... ( If the safety factor threshold is set to 0.8, a red (L3) warning signal will be generated.

[0094] S540: Integrates all triggered warnings and generates a set of warning signals. .

[0095] S600: Intelligent Video Stream Analysis and Information Fusion S610: Acquires real-time video streams from on-site high-definition network cameras via the RTSP protocol.

[0096] S620: Runs a target detection model such as YOLOv5 to identify targets in video frames. If a person is detected entering a preset "unprotected edge" electronic fence area without wearing a safety helmet, an "unsafe behavior" event message is generated, which includes the event time, location, and event type.

[0097] S630: For instruments such as water level gauges, it uses image processing algorithms such as edge detection and threshold segmentation to identify the pointer or water level line, based on pixel distance. and calibration parameters (mm / pixel) Calculation reading It generates a reading message, which includes the instrument ID, reading time, and reading value.

[0098] S640: Event messages and readout messages are treated as virtual sensor data streams, synchronized in time, and then integrated into a fusion data packet. It participates in subsequent early warning judgments.

[0099] S700: Automatic Association and Case Matching of Engineering Knowledge Base The engineering knowledge base module runs asynchronously: S710: Parses uploaded unstructured documents. Generates feature vectors for each document. , including time list (Time range covered in the document), Location list (The document covers the structural components of the cofferdam), a list of keywords. (e.g., "abnormal settlement" or "edge protection").

[0100] S720: Calculation Document With fused data packets similarity of spatiotemporal tags Calculated using Jaccard coefficients: In the formula, [T start ,T end ] represents the continuous time interval specified in the query; L list It is a list of cofferdam structural components (i.e., a set of location points) mentioned in the document, while L sensors This will retrieve the list of sensor locations of interest. The numerator is calculated as the sum of the number of points in the document's time frame that fall within the query time interval and the number of points in the document's location frame that belong to the sensor location list. The denominator is the sum of the total number of document time frames, the total number of discrete time frames within the query interval, the total number of document location frames, and the total number of sensor location frames. This ratio quantifies the overall spatiotemporal similarity between the document and the query conditions. This coefficient ranges from 0 to 1, with a higher value indicating a higher similarity.

[0101] S730: If similarity Exceeding the threshold (e.g., 0.3) then establishes a bidirectional hyperlink between the monitoring log document and the data packet, so that clicking on the fused data packet will allow you to view the associated monitoring log document, and clicking on the document will allow you to view the associated collected data.

[0102] S740: Warning When it is generated, with Metadata as query vector Search for similar historical cases in the engineering knowledge base. Calculate the weighted similarity Sim and take the top [cases]. One (e.g.) The most similar case.

[0103] S800: Automatic generation of early warning and auxiliary diagnostic reports The report generation service starts after an alert is triggered: S810: Based on the warning signal From fused data packets Data from key and related sensors is extracted for a period of time before and after the warning (e.g., 1 hour before the warning and 30 minutes after the warning) to be used for manual judgment on whether to take countermeasures against the warning signal.

[0104] S820: Retrieve results from the engineering knowledge base to obtain a summary of the root causes and remedial measures for similar cases.

[0105] S830: Fill in the predefined auxiliary diagnostic report template. The template includes fixed modules such as basic early warning information, monitoring data trends, similar case references, handling suggestions, and division of responsibilities. The system automatically fills the extracted data and summary into the corresponding modules.

[0106] S840: Package the auxiliary diagnostic report into PDF or HTML format, push it to the preset responsible person through the message service, and save the auxiliary diagnostic report as a new document back to the engineering knowledge base and associate it with the current warning event.

[0107] S900: Remote Terminal Interactive Visualization and Advanced Analytics S910: Remote terminals (such as computers and tablets) receive and render 3D scenes and fuse data in real time. and warning signals Warning information is indicated by red, orange, and yellow colors, and the warning indicators are overlaid on the corresponding monitoring point locations in the 3D scene, making it easy for users to quickly locate the warning area.

[0108] S920: Provides a structural monitoring panel. It plots dual Y-axis charts for settlement data. It supports switching the chart coordinate system to the construction coordinate system consistent with the design drawings through transformation formulas. Map the data points to their locations on the drawing, where Design the coordinates for the acquisition points, where Y′ is the corrected value or the final output value; Y s γ is the baseline or standard value, typically representing the desired reference point. γ is the correction coefficient or gain factor, used to adjust the degree of influence of the difference on the final result. y value This refers to the currently collected value or the original value actually collected.ref Reference or theoretical value, used to compare with the currently collected value to calculate the deviation.

[0109] S930: Users can select polygonal regions in 2D and 3D views. The client will Convert to polygon in construction coordinate system And send it to the data penetration analysis tool. The data penetration analysis tool performs a two-stage search: 1) Spatial search, to find the location in Internal sensor collection ;2) In the atlas In China, with Starting with the sensors within the system, a breadth-first search with a depth of 2 is performed to find the set of sensors with mechanical relationships. Final analysis set Data penetration analysis tools are based on Generate polygonal regions The comprehensive monitoring report is returned to the terminal for display. Based on the comprehensive monitoring report, the construction personnel carry out construction safety inspection and hidden danger verification work in the cofferdam construction area. Combining the monitoring data statistics, anomaly analysis and safety assessment results contained in the report, they comprehensively investigate the construction safety hazards in the corresponding area. At the same time, they optimize construction parameters and adjust construction processes according to the report prompts, and carry out construction quality verification and safety control simultaneously.

[0110] S1000: System Cycle and State Update The system continuously executes steps S200 to S900 in a loop to achieve real-time data acquisition, processing, analysis, early warning, and visualization.

[0111] Secondly, such as Figure 2 As shown, this application also provides a multi-sensor fusion construction safety monitoring system 10, the system comprising: The acquisition module 11 is used to acquire data from various sensors set up at the construction site, and to fuse the data to obtain a fused data package. Each type of data includes the coordinates of the acquisition point. 3D module 12 is used to acquire the building information model and 3D geological model of the construction site, and to construct a 3D digital scene based on the building information model and 3D geological model; The conversion module 13 is used to perform coordinate transformation on each type of acquired data, mapping each type of acquired data to a three-dimensional digital scene to obtain a target three-dimensional image; The early warning module 14 is used to determine the current early warning information of the corresponding collected data based on the early warning threshold of each type of collected data, and to issue a construction early warning based on each type of current early warning information; The visualization module 15 is used to determine visualization information based on the fused data package, the target 3D map, and the current early warning information of each type of collected data, and to conduct construction safety monitoring based on the visualization information.

[0112] It should be noted that although the operations of the method of the present invention are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all of the operations shown must be performed to achieve the desired result. On the contrary, the steps depicted in the flowchart may be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0113] On the other hand, this application also provides a computer-readable storage medium, which may be included in a computer device or exist independently without being assembled into the computer device. The aforementioned computer-readable storage medium stores one or more programs that, when used by one or more processors, execute the methods described in this application. For example, it may execute... Figure 1 The steps of the method shown.

[0114] This application provides a computer program product including instructions that, when executed, cause the method described in this application to be performed. For example, it can execute... Figure 1 The steps of the method shown.

[0115] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0116] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A construction safety monitoring method using multi-sensor fusion, characterized in that, The method includes: Data is collected from various sensors installed at the construction site, and the collected data is fused to obtain a fused data package. Each type of data includes the coordinates of the collection point. Obtain the building information model and three-dimensional geological model of the construction site, and construct a three-dimensional digital scene based on the building information model and the three-dimensional geological model; Coordinate transformation is performed on each type of acquired data, mapping each type of acquired data to the three-dimensional digital scene to obtain the target three-dimensional map; The current warning information for each type of collected data is determined based on the warning threshold, and construction warnings are issued based on each type of current warning information. Visualization information is determined based on the fused data package, the target 3D map, and the current early warning information for each type of collected data, and construction safety monitoring is performed based on the visualization information.

2. The method according to claim 1, characterized in that, There are multiple warning thresholds for each type of collected data. Then, based on the warning threshold of each type of collected data, the current warning information for the corresponding collected data is determined, and construction warnings are issued based on each type of current warning information, specifically as follows: For each type of collected data, based on each warning threshold of the collected data, determine the current warning information of the collected data under the corresponding warning threshold; Construction warnings are issued based on each current warning information for each type of collected data.

3. The method according to claim 2, characterized in that, For each type of collected data, based on each warning threshold of the collected data, the current warning information of the collected data under the corresponding warning threshold is determined, specifically as follows: When any collected data exceeds its own first warning threshold, a first-level current warning message corresponding to that collected data is generated; Each collected data point is associated with at least one other collected data point to obtain multiple association groups; When every data point collected in any associated group exceeds its own second warning threshold, a second-level current warning message corresponding to that associated group is generated. For each piece of collected data, the corresponding mechanical index is determined. When the mechanical index of the collected data is greater than the third warning threshold of the collected data, a third-level current warning information for the collected data is generated.

4. The method according to any one of claims 1-3, characterized in that, The metadata for each current warning message includes at least one of the following: coordinates of the warning location that generated the current warning message, the collected data at the warning location, the sensor number at the warning location, and the warning type. After issuing a construction warning based on the current warning message, the metadata also includes: Obtain monitoring logs, which are used to store multiple historical cases. Each historical case includes historical early warning information, historical handling measures, and historical early warning reasons. Based on the metadata in each current warning information, the monitoring log is matched to obtain historical cases that match the current warning information; Auxiliary diagnostic reports are generated based on successfully matched historical cases; Based on the auxiliary diagnostic report, a response strategy for the current early warning information is determined, and the monitoring log is updated based on the response strategy and the current early warning information.

5. The method according to claim 1, characterized in that, The fused data includes knowledge graphs, so the fusion processing of multiple collected data is performed as follows: Preprocess each type of collected data; Spatiotemporal alignment and feature extraction are performed sequentially on each type of preprocessed data to obtain the feature data corresponding to each type of data. Knowledge graphs are constructed using a variety of feature data.

6. The method according to claim 5, characterized in that, The collected data includes video data, and each type of collected data undergoes preprocessing, specifically: The video data is identified based on a preset computer vision model to obtain the identification result; The recognition result is formatted to obtain a standard data message, which is then used as preprocessed video data.

7. The method according to claim 5, characterized in that, Construction safety monitoring is conducted based on the aforementioned visualized information, specifically as follows: Receive user access requests for target areas in the target 3D map; Based on the knowledge graph, the associated regions of the target region are determined, and a comprehensive monitoring report is generated based on the feature data of the target region and the associated regions. Construction safety monitoring will be conducted based on the comprehensive monitoring report.

8. The method according to claim 1, characterized in that, A three-dimensional digital scene is constructed based on the building information model and the three-dimensional geological model, specifically as follows: The transparency of the building information model and the three-dimensional geological model are adjusted respectively; The building information model and the three-dimensional geological model, after their transparency has been adjusted, are overlaid and rendered to obtain a three-dimensional digital scene.

9. The method according to claim 5, characterized in that, Each type of collected data undergoes preprocessing, specifically as follows: Median filtering was applied to each type of collected data.

10. A construction safety monitoring system based on multi-sensor fusion, characterized in that, The system includes: The data acquisition module is used to acquire data from various sensors set up at the construction site, and to fuse the data to obtain a fused data package. Each data acquisition includes the coordinates of the acquisition point. A 3D module is used to acquire the building information model and 3D geological model of the construction site, and to construct a 3D digital scene based on the building information model and the 3D geological model. The conversion module is used to perform coordinate transformation on each type of acquired data, mapping each type of acquired data to the three-dimensional digital scene to obtain the target three-dimensional image; The early warning module is used to determine the current early warning information for each type of collected data based on the early warning threshold, and to issue construction early warnings based on each type of current early warning information. The visualization module is used to determine visualization information based on the fused data package, the target 3D map, and the current early warning information of each type of collected data, and to perform construction safety monitoring based on the visualization information.

Images