Distribution network construction air-space integrated safety monitoring method based on unmanned aerial vehicle

By constructing an integrated air-space collaborative monitoring network, combining satellite remote sensing, UAV near-ground monitoring, and the BeiDou high-precision positioning system, the immaturity of power distribution network construction safety monitoring has been resolved, enabling full-process and all-round identification and early warning of safety hazards, and improving construction safety.

CN121390873APending Publication Date: 2026-01-23STATE GRID HENAN ELECTRIC POWER CO HUIXIAN CITY POWER SUPPLY CO
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202511490979.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-10-17
Publication Date
2026-01-23

AI Technical Summary

Technical Problem

Existing power distribution network construction safety monitoring technologies are not yet fully mature, lack comprehensive coverage, and are difficult to identify and warn of safety hazards throughout the entire process and in all aspects, thus failing to meet the requirements of "inherent safety".

Method used

A collaborative monitoring network integrating air and space is constructed, combining satellite remote sensing, UAV near-ground monitoring, and BeiDou high-precision positioning system to collect multi-source data. Through data processing and fusion analysis models, real-time monitoring and early warning of construction safety are achieved.

Benefits of technology

It achieves full-process and all-round coverage of power distribution network construction, significantly improves the accuracy of safety hazard identification and the timeliness of early warning, and provides strong technical support for construction safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121390873A_ABST
    Figure CN121390873A_ABST
Patent Text Reader

Abstract

The invention discloses an air-space integrated safety monitoring method for distribution network construction based on an unmanned aerial vehicle, and the method comprises the following steps: A, constructing an air-space integrated cooperative monitoring network, and collecting the multi-source data of a distribution network construction region through the air-space integrated cooperative monitoring network; c, constructing a construction safety analysis model; the construction safety analysis model comprises a data acquisition and preprocessing module, a multi-source data fusion analysis module, a risk diagnosis and evaluation module, an intelligent prediction and early warning module, an early warning decision and disposal module and a visualization and feedback module; and D, based on the trained construction safety analysis model, collecting multi-source data by using an air-space integrated cooperative monitoring network, and carrying out online monitoring, early warning and disposal on the construction safety of the distribution network in combination with a meteorological database. Unmanned aerial vehicle near-earth monitoring and satellite remote sensing data can be fused, the distribution network construction process is effectively monitored, and potential safety hazards are reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power distribution network construction safety monitoring, and in particular to a method for integrated air-space safety monitoring of power distribution network construction based on unmanned aerial vehicles (UAVs). Background Technology

[0002] With the continuous development of my country's economy, power transmission has become one of the important guarantees for my country's economic development. The construction and upgrading of high-voltage power supply networks are increasing day by day, and distribution network construction has also become one of the important projects to ensure the normal operation of the national economy.

[0003] Due to the massive scale of my country's power distribution network construction and the complex construction environment (such as construction sites in different locations like cities, villages, and mountains), the safety risks are high (electric shock, falls from heights, falling objects, mechanical injuries, etc.). During the construction process, construction workers need to operate in a professional and strict manner in accordance with the work guidelines to reduce safety hazards and improve construction quality.

[0004] In existing technologies, video recording is commonly used. For example, multiple cameras set up at the construction site are used to monitor the construction process in real time. On the one hand, the construction process is recorded, which is convenient for future traceability and analysis. On the other hand, video recording can be used to discover and stop various violations during the construction process, such as not wearing a safety helmet or not wearing a safety rope when working at height. In recent years, with the continuous improvement of safety requirements and technological development, power distribution network construction safety monitoring is shifting from traditional "person-to-person" and "experience-based" management to information-based, visual, and intelligent management. Regulatory authorities and power companies have invested a lot of resources to improve on-site safety management, but it is still in a stage where the application of technologies is diversified but not yet fully mature, the coverage is not comprehensive, and the effect needs to be improved. There is still a gap between this and "inherent safety." Summary of the Invention

[0005] The purpose of this invention is to provide an integrated air-space safety monitoring method for power distribution network construction based on unmanned aerial vehicles (UAVs), which can integrate UAV near-ground monitoring and satellite remote sensing data to effectively monitor the power distribution network construction process and reduce safety hazards.

[0006] The present invention adopts the following technical solution: A method for integrated air-space safety monitoring of power distribution network construction based on unmanned aerial vehicles (UAVs) includes the following steps: A: Construct an integrated space-air collaborative monitoring network and use it to collect multi-source data from the power distribution network construction area. The multi-source data includes macroscopic change data of the construction area, refined scene data, three-dimensional spatial data, and infrastructure displacement data. B: Obtain historical and forecast meteorological information for the construction area and construct a meteorological database; C: Construct a construction safety analysis model; the construction safety analysis model includes a data acquisition and preprocessing module, a multi-source data fusion analysis module, a risk diagnosis and assessment module, an intelligent prediction and early warning module, an early warning decision and response module, and a visualization and feedback module; The system comprises the following modules: a data acquisition and preprocessing module, used to standardize and spatiotemporally align the input multi-source data, providing a data foundation for subsequent analysis; a multi-source data fusion and analysis module, used to deeply integrate macroscopic change data, refined scene data, 3D spatial data, infrastructure displacement data, and meteorological data of the construction area, thereby obtaining multi-source fused data and generating independent risk identification results based on data from different sources; a risk diagnosis and assessment module, used to comprehensively process the risk identification results generated from data from different sources, generating a comprehensively optimized construction safety risk assessment result; an intelligent prediction and early warning module, used to predict future construction safety risks based on the comprehensively optimized risk assessment result, combined with historical data and forecast meteorological information; an early warning decision and response module, used to generate tiered early warning information and optimized response plans based on the current risk assessment result and future risk change trends; and a visualization and feedback module, used to provide a human-computer interaction interface to display analysis results and early warning information. D: Based on the trained construction safety analysis model, multi-source data is collected using an integrated air-space collaborative monitoring network, combined with a meteorological database, to conduct online monitoring, early warning, and handling of power distribution network construction safety.

[0007] The integrated air-space collaborative monitoring network includes a satellite remote sensing monitoring system, an unmanned aerial vehicle (UAV) near-ground monitoring system, and a BeiDou high-precision positioning monitoring system. The satellite remote sensing monitoring system is used to collect macroscopic change data of the construction area; the UAV near-ground monitoring system is used to collect detailed scene data and three-dimensional spatial data of the construction area; and the BeiDou high-precision positioning monitoring system is used to collect three-dimensional displacement data of critical infrastructure in the absolute geographic coordinate system.

[0008] The data acquisition and preprocessing module includes an image processing submodule, a point cloud data processing submodule, a numerical data processing submodule, and a positioning data processing submodule; The image processing submodule is used to process satellite remote sensing image data, UAV visible light image data and infrared image data to obtain standardized image data; The point cloud data processing submodule is used to process the 3D point cloud data collected by the UAV LiDAR to obtain a structured point cloud feature dataset. The numerical data processing submodule is used to process numerical monitoring data collected by meteorological sensors or obtained from meteorological databases to obtain standardized numerical time-series datasets. The positioning data processing submodule is used to process the three-dimensional displacement data collected by the Beidou high-precision positioning monitoring system and obtain a standardized displacement time series dataset.

[0009] The multi-source data fusion and analysis module includes a macro-risk identification submodule, a refined risk detection submodule, and an infrastructure monitoring submodule; The macro-risk identification submodule is used to extract features and recognize patterns from preprocessed satellite remote sensing data, and generate macro-risk identification results and risk areas based on satellite data. The refined risk detection submodule is used to extract refined risk features, i.e. refined risk targets, of construction scenarios based on pre-processed UAV imagery data and point cloud data, using computer vision and deep learning technologies. The infrastructure monitoring submodule is used to obtain infrastructure displacement data based on BeiDou high-precision positioning data, through time series analysis and deformation recognition algorithms.

[0010] The macro-risk identification submodule includes a target detection unit, a change detection unit, and a regional risk assessment unit; The target detection unit is used to build a construction machinery recognition model based on the improved YOLOv5 deep learning framework. By introducing the CBAM attention mechanism and optimizing the anchor frame size, it can realize the automatic recognition and accurate positioning of construction machinery and output the machinery type, position coordinates and confidence level. The change detection unit adopts a semantic segmentation model based on the U-Net architecture. It extracts multi-scale features through an encoder-decoder structure, compares pixel-level differences in satellite images of different time phases, automatically identifies surface changes, and outputs vector boundaries of changed areas and change types. The regional risk assessment unit integrates target detection and change detection results to construct a construction activity intensity index. Based on the fuzzy comprehensive evaluation method, it classifies risk levels by incorporating multi-dimensional features including weighted fusion of mechanical equipment density, area of ​​change, and construction duration, and generates a macro-risk distribution heat map.

[0011] The refined risk detection submodule includes a personnel safety supervision unit, an equipment safety monitoring unit, and an environmental risk perception unit; The personnel safety supervision unit uses a multi-target tracking algorithm combined with a classification network to continuously track construction workers, and judges the wearing status of protective equipment based on an attention mechanism. It outputs personnel trajectory, behavioral norms assessment and violation alarms in real time to obtain refined risk targets. The equipment safety monitoring unit, based on LiDAR point cloud spatial analysis and combined with point cloud nearest neighbor search, calculates the spatial distance between the moving parts of mechanical equipment and power lines in real time. When the distance value is lower than the preset threshold, it triggers a collision warning. At the same time, it identifies equipment overheating abnormalities through infrared temperature field analysis. The environmental risk perception unit constructs a smoke detection model based on the YOLO framework, enhances the detection capability of small targets through multi-scale feature fusion, extracts the smoke diffusion area by combining image segmentation algorithm, and identifies environmental anomalies in the construction area by using background modeling and frame difference method, outputting the risk type, spatial location and urgency level.

[0012] The infrastructure monitoring submodule includes a displacement analysis unit, a deformation identification unit, and a stability assessment unit; The displacement analysis unit uses a time series decomposition method, decomposes the BeiDou three-dimensional coordinate sequence into trend, periodic and residual terms using the STL algorithm, identifies displacement abrupt change points based on sliding window variance analysis, and extracts vibration characteristics of different frequency bands by combining wavelet transform, thereby quantifying the overall displacement and vibration intensity of the infrastructure. The deformation identification unit uses principal component analysis to perform feature dimensionality reduction on multi-point displacement sequences, identifies similar deformation patterns through K-means clustering, constructs a displacement transformation model based on rigid body kinematics principles, distinguishes the deformation types of infrastructure, and outputs deformation pattern classification and confidence level. The stability assessment unit combines displacement rate, deformation type and structural design parameters to establish a comprehensive assessment model based on fuzzy reasoning. It quantifies the safety status of each indicator through membership function, calculates the stability index using weighted average method, and triggers graded early warning when the index exceeds the threshold, and generates a structural safety diagnosis report.

[0013] The risk diagnosis and assessment module includes a data association and registration unit, a conflict detection and resolution unit, and a comprehensive risk assessment unit; The data association and registration unit is used to establish the association mapping relationship of multi-source risk information by adopting the spatiotemporal consistency matching algorithm, through unified coordinate system transformation and temporal alignment, so as to realize the spatial registration of macro risk distribution, refined risk targets and infrastructure displacement data; The conflict detection and resolution unit is used to build a multi-source information fusion model based on DS evidence theory. It quantifies the credibility of each data source through a basic probability allocation function. When a conflict occurs in the monitoring results, a confidence weighting strategy is used to optimize the decision and improve the reliability of risk assessment. The comprehensive risk assessment unit is used to construct a multi-level fuzzy comprehensive evaluation system. By weighted integration of multi-dimensional risk indicators and rule-based reasoning from the expert knowledge base, it generates graded early warning information and risk assessment reports, and obtains a comprehensive and optimized construction safety risk assessment result.

[0014] The UAV near-ground monitoring system uses a multi-rotor UAV platform and is equipped with a high-definition visible light camera, an infrared thermal imager, and a lidar. The high-definition visible light camera captures visible light images of the construction scene; the infrared thermal imager generates thermal infrared images; and the lidar acquires the three-dimensional spatial coordinates of the target surface.

[0015] The BeiDou high-precision positioning and monitoring system uses a high-precision BeiDou / GPS dual-mode receiver that supports carrier phase differential technology.

[0016] This invention constructs an integrated space-air collaborative monitoring network, which collects macroscopic change data, refined scene data, three-dimensional spatial data, and infrastructure displacement data of the power distribution network construction area in real time. It also acquires historical and forecast meteorological information for the construction area. Supplemented by a specially constructed construction safety analysis model consisting of data acquisition and preprocessing modules, multi-source data fusion analysis modules, risk diagnosis and assessment modules, intelligent prediction and early warning modules, early warning decision-making and response modules, and visualization and feedback modules, this invention achieves a complete technical system from multi-source space-air data acquisition, intelligent risk identification, fusion situation assessment, prediction and early warning to visualized decision-making. This enables full-process and all-round coverage of power distribution network construction safety monitoring, significantly improving the accuracy of safety hazard identification and the timeliness of early warning, and providing strong technical support for the construction safety of power distribution network projects. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0018] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0019] The present invention will now be described in detail with reference to the accompanying drawings and embodiments: like Figure 1 As shown, the air-space integrated safety monitoring method for power distribution network construction based on unmanned aerial vehicles (UAVs) of the present invention includes the following steps: A: Construct an integrated space-air collaborative monitoring network to collect multi-source data from the power distribution network construction area in real time. The multi-source data includes macroscopic change data of the construction area, refined scene data, three-dimensional spatial data, and infrastructure displacement data. In this invention, the integrated air-space collaborative monitoring network includes a satellite remote sensing monitoring system, an unmanned aerial vehicle (UAV) near-ground monitoring system, and a BeiDou high-precision positioning monitoring system; Among them, the satellite remote sensing monitoring system can use my country's existing low-orbit remote sensing satellites, including high-resolution optical satellites and synthetic aperture radar satellites; optical satellites capture solar radiation energy reflected from the Earth's surface through multispectral sensors to generate images with rich spectral information; synthetic aperture radar satellites actively transmit microwaves and receive the echo signals after their interaction with the Earth's surface, and can accurately obtain information on surface deformation using interferometry technology. Satellite remote sensing monitoring systems are used to collect macroscopic change data of the construction area; macroscopic change data refers to a series of information obtained through satellite platforms that can reflect the changes in surface cover, environmental background and surface geometry of the construction area over time. Satellite remote sensing monitoring systems are designed based on the principles of satellite remote sensing imaging; optical satellites acquire continuous surface images through push-broom or swing-broom imaging, with spatial resolution reaching sub-meter level; synthetic aperture radar satellites improve azimuth resolution through synthetic aperture technology and extract phase information through interferometric processing to achieve millimeter-level deformation monitoring.

[0020] In this embodiment, the control satellite remote sensing monitoring system periodically collects data from the construction area; optical satellite images need to undergo radiometric calibration, atmospheric correction, and orthorectification to generate an orthorectified base map with accurate geographic coordinates; synthetic aperture radar data needs to undergo interferometric processing, phase unwrapping, and geocoding to extract surface deformation information; the above macroscopic change data can provide the ability to monitor the large-scale surface cover (such as earthwork excavation, filling, construction of temporary roads, vegetation removal, etc.) and periodic changes in the construction area, effectively reflecting the impact of construction activities on the surrounding environment and potential geological risks, such as slope instability and ground subsidence.

[0021] The UAV near-ground monitoring system can use existing multi-rotor UAV platforms and be equipped with various sensors such as high-definition visible light cameras, infrared thermal imagers, and lidar. The high-definition visible light camera captures visible light images of the construction scene through optical lenses and image sensors; the infrared thermal imager generates thermal infrared images by detecting the infrared radiation emitted by the object itself; and the lidar accurately obtains the three-dimensional spatial coordinates of the target surface by emitting laser pulses and measuring their return time.

[0022] The UAV near-ground monitoring system is used to collect detailed scene data and three-dimensional spatial data of the construction area. The detailed scene data refers to the high-resolution, multi-spectral visual information of the construction scene in the form of images obtained by the UAV low-altitude platform, which aims to depict the details of the site and the status of the target. The three-dimensional spatial coordinates refer to the information obtained by the UAV low-altitude platform through active detection, which can accurately characterize the three-dimensional geometric structure and spatial position of the ground objects and terrain in the construction area.

[0023] In this embodiment, the high-definition visible light camera is based on the photoelectric conversion principle, converting light signals into electrical signals to generate high-resolution digital images; the infrared thermal imager is based on the infrared radiation law, detecting the infrared radiation intensity on the surface of an object and combining it with emissivity correction to obtain temperature distribution data; the lidar is based on the time-of-flight principle, accurately measuring the time difference between laser emission and reception and combining it with the speed of light constant to calculate the distance between the sensor and the target point; the above are conventional technologies in this field and will not be described in detail here.

[0024] In this embodiment, the UAV near-ground monitoring system is designed based on the principles of near-ground remote sensing and active detection, actively acquiring refined scene data and three-dimensional spatial data. The refined scene data can reflect the visual details of the construction scene, such as the distribution and appearance of personnel, machinery, and materials, as well as the state characteristics of the target, such as whether personnel are wearing safety helmets and the working posture of machinery. At the same time, it can also reflect the thermal characteristics of the equipment, such as whether there is abnormal overheating in electrical connectors and mechanical hydraulic components.

[0025] Three-dimensional spatial data can reflect the three-dimensional geometric features of land features and terrain, including planar position and elevation information, as well as the spatial relationship features between targets, such as distance, elevation difference, volume, and surface area; it can also reflect the microscopic deformation features of the earth's surface and structures.

[0026] In this embodiment, the UAV near-ground monitoring system adopts a hierarchical data acquisition method according to a preset monitoring strategy; for high-risk construction areas, multiple video inspections are carried out daily; for the entire construction area, a detailed scan is carried out weekly; the collected detailed scene data includes visible light images, thermal infrared images, and three-dimensional spatial data includes dense three-dimensional point clouds; the above data can provide detailed information of the construction area at the centimeter or even millimeter level, accurately capturing key safety elements such as the behavior of construction personnel, equipment status, and changes in safe distance; The BeiDou high-precision positioning monitoring system can use a high-precision BeiDou / GPS dual-mode receiver that supports carrier phase differential technology. The BeiDou high-precision positioning monitoring system typically includes one or more base stations set at stable locations with known coordinates, and multiple rover stations set on key infrastructure (such as poles and substations) in the power distribution network construction area. The BeiDou high-precision positioning monitoring system is used to collect three-dimensional displacement data of key infrastructure in an absolute geographic coordinate system. The BeiDou high-precision positioning and monitoring system is designed based on the principle of satellite navigation and positioning. The base station receives satellite signals, calculates the correction data between the known coordinates and the satellite observation values, and sends the correction data to the rover station. The rover station corrects its own satellite observation values ​​based on the correction data, thereby achieving three-dimensional coordinate calculation with centimeter-level or even millimeter-level accuracy.

[0027] In this embodiment, by setting the sampling frequency, the Beidou high-precision positioning monitoring system is controlled to continuously collect three-dimensional displacement data of critical infrastructure. The three-dimensional displacement data includes timestamps, longitude data, latitude data, and elevation data. The above three-dimensional displacement data can effectively reflect the slow and cumulative three-dimensional displacement changes of critical infrastructure over a continuous period of time, such as deformations such as settlement, slippage, and torsion. In this invention, the data collected by the integrated air-space collaborative monitoring network can be cross-verified and complemented by observation data from different platforms, with different detection principles, different spatiotemporal scales, and different accuracies, forming a redundant and fault-tolerant perception network. This overcomes the limitations of a single monitoring method and enables comprehensive perception of the safety status of power distribution network construction.

[0028] In this invention, the satellite remote sensing monitoring system can perceive the macroscopic change trends and geological background risks of the construction area; the UAV near-ground monitoring system can perceive the refined dynamics and microscopic risk factors of the construction scene; and the Beidou high-precision positioning monitoring system can perceive the precise location changes of critical infrastructure in absolute geographical space.

[0029] In this embodiment, the satellite remote sensing monitoring system acquires periodic, large-scale macroscopic data, which can effectively capture regional cumulative changes and geological background risks, but the response cycle is relatively long; the UAV near-ground monitoring system acquires high-frequency, high-precision local data, which has a fast response and can monitor construction dynamics in real time, but the coverage is limited; the Beidou high-precision positioning monitoring system acquires absolute position change data of critical infrastructure, which can directly reflect the three-dimensional coordinate changes of infrastructure in space and is the ultimate evidence of displacement monitoring; by integrating the above three types of data and analyzing them through subsequent construction safety analysis models, macroscopic trend changes, microscopic real-time dynamics and absolute displacement of infrastructure can be spatiotemporally correlated and model matched.

[0030] For example, if an abnormal settlement trend is detected in a construction area via satellite, drones can be deployed for detailed inspections. LiDAR point clouds can be used to accurately measure the deformation of key facilities, while BeiDou high-precision positioning data can be combined to verify the actual displacement of the infrastructure. By using the three-dimensional coordinate displacement characteristics collected by BeiDou high-precision positioning equipment and the absolute position change sequences of longitude, latitude, and elevation, the overall displacement of the tower can be accurately obtained. Furthermore, by analyzing the time series of three-dimensional coordinates, the settlement, slippage, and torsional deformation of the tower can be effectively identified. The aforementioned data fusion can also be applied to scenarios such as safety monitoring of large machinery operations, supervision of personnel safety behavior, and infrastructure deformation monitoring, for example: In the scenario of safety monitoring of large machinery operations, satellite imagery can identify the entry and approximate distribution of large machinery such as cranes, while high-definition drone imagery combined with lidar point clouds can accurately calculate the real-time distance between the working radius of the machinery and the power lines. Beidou positioning data can monitor the impact of construction vibrations on the location of nearby towers. When the fusion analysis shows that the safe distance is close to the critical value, the system can automatically issue a collision risk warning. In the scenario of infrastructure deformation monitoring, BeiDou high-precision positioning data can monitor the three-dimensional displacement of key infrastructure such as poles and substations in real time, UAV lidar point cloud can provide a detailed three-dimensional model of the infrastructure, and satellite interferometric radar can provide a large-scale surface deformation background; when abnormal displacement of infrastructure is detected, the system can comprehensively judge the cause and degree of damage of the deformation by integrating multi-source data. B: Obtain historical and forecast meteorological information for the construction area and construct a meteorological database; In this invention, meteorological information includes key environmental parameters that affect the safety of power distribution network construction, such as wind speed, wind direction, temperature, humidity, rainfall, and visibility; historical meteorological information is used to analyze the climate patterns and risk background of the construction area; and forecast meteorological information is used to support subsequent risk trend prediction and early warning decisions.

[0031] C: Construct a construction safety analysis model; the construction safety analysis model includes a data acquisition and preprocessing module, a multi-source data fusion analysis module, a risk diagnosis and assessment module, an intelligent prediction and early warning module, an early warning decision and handling module, and a visualization and feedback module.

[0032] The system comprises the following modules: a data acquisition and preprocessing module, used to standardize and spatiotemporally align the input multi-source data, providing a data foundation for subsequent analysis; a multi-source data fusion and analysis module, used to deeply integrate macroscopic change data, refined scene data, 3D spatial data, infrastructure displacement data, and meteorological data of the construction area, thereby obtaining multi-source fused data and generating independent risk identification results based on data from different sources; a risk diagnosis and assessment module, used to comprehensively process the risk identification results generated based on data from different sources, generating a comprehensively optimized construction safety risk assessment result; an intelligent prediction and early warning module, used to predict future construction safety risks based on the comprehensively optimized risk assessment result, combined with historical data and forecast meteorological information; an early warning decision and response module, used to generate tiered early warning information and optimized response plans based on the current risk assessment result and future risk change trends; and a visualization and feedback module, used to provide a human-computer interaction interface to display analysis results and early warning information.

[0033] In this invention, the data acquisition and preprocessing module includes an image processing submodule, a point cloud data processing submodule, a numerical data processing submodule, and a positioning data processing submodule; The image processing submodule is used to process satellite remote sensing image data, UAV visible light image data and infrared image data. The image and point cloud data processing submodule obtains standardized image data through processing procedures such as radiometric calibration, atmospheric correction and geometric fine correction.

[0034] Specifically, the image processing submodule performs the following processing on the multi-source image data: Based on sensor parameters and radiative transfer models, radiometric calibration and atmospheric correction are performed on the original images to eliminate environmental interference and restore the true reflectivity of ground features. By combining geometric correction and orthorectification with a digital elevation model, image distortion is eliminated, and an orthophoto with accurate geographic coordinates is generated. Image enhancement algorithms such as wavelet transform and histogram equalization are employed to improve image quality and visual effects, providing a data foundation for subsequent feature recognition.

[0035] The point cloud data processing submodule is used to process 3D point cloud data acquired by UAV LiDAR; and outputs a structured point cloud feature dataset through point cloud denoising, filtering, registration, classification and feature extraction.

[0036] Specifically, the point cloud data processing submodule performs the following processing on the LiDAR point cloud: By using statistical filtering and radius filtering algorithms, outlier noise and abnormal data are effectively removed, thus improving the quality of point cloud data. By employing progressively encrypted triangular mesh filtering and machine learning classification methods, accurate separation of ground points and non-ground points is achieved, and key ground features are identified. Based on clustering analysis and geometric feature extraction algorithms, key three-dimensional feature parameters of power facilities, such as tower location and conductor sag, are obtained from point cloud data; The numerical data processing submodule is used to process numerical monitoring data collected by meteorological sensors or obtained from meteorological databases. Through data cleaning and repair, time series alignment and standardization, it outputs standardized numerical time series datasets.

[0037] Specifically, the numerical data processing submodule performs the following processing on the numerical monitoring data: By using the 3σ criterion and box plot analysis, outliers and gross errors in environmental parameters such as wind speed and temperature are automatically identified and removed, and cubic spline interpolation algorithm is used to repair missing data within a reasonable range. A unified time reference is established, and a linear interpolation resampling method is used to unify multi-source sensor data with different sampling frequencies to a common time axis of 1Hz, thereby achieving precise time alignment. Based on the Min-Max normalization algorithm, numerical data with different dimensions are mapped to the [0,1] interval, eliminating the difference in dimensions between parameters and generating a data quality assessment report, providing a standardized numerical time series dataset for subsequent fusion analysis.

[0038] The positioning data processing submodule is used to process the three-dimensional displacement data collected by the Beidou high-precision positioning monitoring system, including data decoding, quality inspection, gross error removal and coordinate transformation, and outputs a standardized displacement time series dataset.

[0039] Specifically, the positioning data processing submodule performs the following processing on the BeiDou high-precision positioning data: Establish a timestamp synchronization mechanism with a unified time reference to align positioning data with data from other sensors in time sequence; By utilizing the absolute positional change sequences of longitude, latitude, and elevation, the overall displacement of critical infrastructure can be obtained through time series analysis. Analyze the time series characteristics of three-dimensional coordinates and combine them with pattern recognition algorithms to identify settlement, slippage and torsional deformation of infrastructure; By combining meteorological data provided by the numerical data processing submodule, the impact of environmental factors on positioning accuracy is analyzed, and an error compensation algorithm is adopted to improve the accuracy of displacement monitoring.

[0040] In this invention, the multi-source data fusion analysis module includes a macro-risk identification submodule, a refined risk detection submodule, and an infrastructure monitoring submodule, wherein: The macro-risk identification submodule is used to extract features and identify patterns from preprocessed satellite remote sensing data, and generate macro-risk identification results based on satellite data, namely macro-risk distribution heatmaps and risk areas; The refined risk detection submodule is used to extract refined risk features, i.e. refined risk targets, from construction scenarios based on pre-processed UAV imagery data and point cloud data through computer vision and deep learning technologies. The infrastructure monitoring submodule is used to acquire infrastructure displacement data based on BeiDou high-precision positioning data, through time series analysis and deformation recognition algorithms, and to monitor the structural safety status of critical infrastructure.

[0041] In order to enable the macro-risk identification submodule, the refined risk detection submodule, and the infrastructure monitoring submodule to more efficiently perform their identification, detection, and monitoring functions, the present invention also incorporates the following special design for these submodules: In this invention, the macro-risk identification submodule includes a target detection unit, a change detection unit, and a regional risk assessment unit; The target detection unit is used to build a construction machinery recognition model based on the improved YOLOv5 deep learning framework. By introducing the CBAM attention mechanism and optimizing the anchor frame size, it can automatically identify and accurately locate large construction machinery such as cranes and excavators in satellite imagery, and output the machinery type, location coordinates and confidence level.

[0042] In practice, the target detection unit first builds a detection model based on the improved YOLOv5 framework, enhancing feature extraction capabilities by introducing the CBAM attention mechanism. Then, the anchor box sizes are adjusted to [12,16], [24,32], and [48,64] to adapt to the aspect ratio characteristics of construction machinery such as cranes and excavators. Subsequently, Mosaic data augmentation technology is used to improve the model's generalization ability through random scaling, cropping, and arrangement. Next, the non-maximum suppression algorithm is applied, setting the intersection-union ratio threshold to 0.5 to eliminate duplicate detection boxes. Finally, the detection results, including machinery type, bounding box coordinates, and confidence score, are output, with the localization error controlled within 3 pixels.

[0043] The change detection unit adopts a semantic segmentation model based on the U-Net architecture. It extracts multi-scale features through an encoder-decoder structure, compares pixel-level differences in satellite images of different time phases, and automatically identifies surface changes such as newly added construction areas, earthwork excavation, and vegetation destruction. It outputs the vector boundary of the changed area and the type of change. In practice, the change detection unit first constructs a semantic segmentation network based on U-Net, and the encoder uses a ResNet34 backbone network to extract multi-scale features. Then, the decoder gradually restores the spatial resolution through transposed convolution and combines shallow detail features with skip connections. Subsequently, the Dice loss function is used to optimize model training and enhance the learning ability of imbalanced change samples. Next, pixel-level change regions are extracted through dual-temporal image difference processing, and morphological closing operations are used to fill holes. Finally, the binary mask of the change region, the change type classification result, and the change confidence map are output.

[0044] The regional risk assessment unit integrates target detection and change detection results to construct a construction activity intensity index. By weighted fusion of multi-dimensional features such as machinery and equipment density, area of ​​change, and construction duration, it classifies high, medium, and low risk levels based on the fuzzy comprehensive evaluation method and generates a macro-risk distribution heat map. In actual operation, the regional risk assessment unit first divides the construction area based on a 500m×500m grid to establish a standardized grid assessment system; then, through the mechanical distribution data obtained by the target detection unit, it calculates the mechanical equipment density index P1 within each grid. The mechanical equipment density index P1 is the ratio of the number of engineering mechanical equipment to the grid area; subsequently, in combination with the change area data output by the change detection unit, it calculates the change intensity index P2. The change intensity index P2 is the ratio of the change area (increased or reduced area) to the initial area; then, based on the time-series monitoring data, it calculates the construction duration index P3. The construction duration index is the percentage value of the ratio of the construction time to the project start time, which can reflect the duration of construction activities; then, the analytic hierarchy process is used to determine the index weights, and the mechanical equipment density index P1, the change intensity index P2, and the construction duration index P3 are mapped to the [0,1] interval through Min-Max normalization to eliminate the dimension difference; then, it obtains the construction activity intensity index P. For example, P = 0.5×P1 + 0.3×P2 + 0.2×P3; then, according to the historical statistical data, it sets the risk level threshold: P≤0.3 is low risk, 0.3<P≤0.7 is medium risk, and P>0.7 is high risk; then, it applies the inverse distance weighted interpolation algorithm to generate a continuous risk distribution surface and outputs a risk heat map through RGB color coding; finally, it outputs a macro risk distribution heat map containing the grid risk levels.

[0045] Finally, the macro risk identification sub-module extracts features and performs pattern recognition on the satellite remote sensing data through the above steps to generate a macro risk identification result based on satellite data, that is, a macro risk distribution heat map; In the present invention, the refined risk detection sub-module includes a personnel safety supervision unit, an equipment safety monitoring unit, and an environmental risk perception unit; The personnel safety supervision unit uses the DeepSORT multi-object tracking algorithm combined with the MobileNetV3 classification network to achieve continuous tracking of construction personnel through Kalman filter prediction and Hungarian matching. Based on the attention mechanism, it judges the wearing status of protective equipment such as safety helmets and reflective vests, and outputs the personnel trajectory, behavior normality assessment, and violation warnings in real time to obtain refined risk targets; In actual operation, the personnel safety supervision unit first uses the DeepSORT multi-object tracking framework to predict the target movement trajectory through Kalman filter; then, it uses the cosine distance to measure the feature similarity and completes the association between the detection box and the trajectory in combination with the Hungarian algorithm; then, it constructs a classification network based on MobileNetV3 and focuses on key areas such as safety helmets and reflective vests through the attention mechanism; then, it sets the tracking loss frame number threshold to 30 frames, and if it exceeds the threshold, it determines that the target has disappeared and terminates the tracking; finally, it outputs the personnel ID, position coordinates, safety equipment status, and violation type warning information in real time.

[0046] The equipment safety monitoring unit, based on LiDAR point cloud spatial analysis and combined with point cloud KD-Tree nearest neighbor search, calculates the spatial distance between moving parts such as booms and buckets and power lines in real time. When the distance value is lower than a preset threshold, a collision warning is triggered. At the same time, it identifies equipment overheating abnormalities through infrared temperature field analysis. In practice, the equipment safety monitoring unit first uses high-precision 3D point cloud data collected by UAV lidar to achieve rapid nearest neighbor search of the point cloud by constructing a KD-Tree spatial index. It then uses an Euclidean distance clustering algorithm to segment the point cloud clusters of the target equipment and power lines, and calculates the minimum spatial distance between moving parts and power lines in real time. Next, it combines thermal radiation data collected by UAV infrared thermal imagers, converting grayscale values ​​into temperature distribution maps using a temperature calibration model. A region growing algorithm is then used to identify overheated areas of key equipment components, triggering an overheat warning when the temperature exceeds a preset threshold. When the spatial distance is below a safety threshold or the equipment temperature exceeds the limit, it automatically generates tiered warning information including location coordinates, distance values, and temperature readings, and pushes this information to the visualization monitoring platform.

[0047] The environmental risk perception unit constructs a smoke detection model based on the YOLO framework, enhances the detection capability of small targets through multi-scale feature fusion, extracts the smoke diffusion area by combining image segmentation algorithm, and uses background modeling and frame difference method to identify environmental anomalies such as illegal intrusion into the construction area and material stacking beyond the boundary, and outputs the risk type, spatial location and urgency level. In practice, the environmental risk perception unit first constructs a smoke and fire detection model based on the improved YOLOv5 framework. It enhances the feature extraction capability for small-scale smoke and flames by introducing a CBAM attention mechanism and optimizes the detection effect of different smoke and fire patterns using an adaptive anchor box matching strategy. Then, it applies a Gaussian mixture model for background modeling, detects moving targets in the foreground using the frame difference method, and analyzes the target's trajectory using optical flow to identify illegal intrusion behavior at the construction area perimeter. Subsequently, it uses a U-Net semantic segmentation network to perform pixel-level segmentation of the construction material pile. By calculating the spatial relationship between the material pile outline and the preset safety boundary, it determines whether there is a risk of stacking beyond the boundary. Finally, it outputs the environmental risk type, spatial coordinates, risk level assessment, and disposal suggestions in real time, providing comprehensive environmental risk situational awareness for construction safety supervision.

[0048] Finally, the refined risk detection submodule extracts refined risk features of the construction scene based on preprocessed UAV imagery data and point cloud data, and obtains refined risk targets through computer vision and deep learning technologies. In this invention, the infrastructure monitoring submodule is used to acquire infrastructure displacement data based on BeiDou high-precision positioning data, through time series analysis and deformation recognition algorithms, and to monitor the structural safety status of critical infrastructure.

[0049] Specifically, the infrastructure monitoring submodule includes a displacement analysis unit, a deformation identification unit, and a stability assessment unit; The displacement analysis unit uses a time series decomposition method, decomposes the BeiDou three-dimensional coordinate sequence into trend, periodic and residual terms using the STL algorithm, identifies displacement abrupt change points based on sliding window variance analysis, and extracts vibration characteristics of different frequency bands by combining wavelet transform, thereby quantifying the overall displacement and vibration intensity of the infrastructure. In practice, the displacement analysis unit first preprocesses the original coordinate sequences of longitude, latitude, and elevation collected by BeiDou, eliminating gross errors and filling in reasonable missing values. Then, it applies the STL seasonal decomposition algorithm to decompose the coordinate sequence into long-term trend terms, periodic terms, and residual terms. The trend term reflects the cumulative displacement of the infrastructure, the periodic term reflects the periodic changes caused by environmental factors such as temperature, and the residual term includes random noise and instantaneous vibrations. Next, it uses a sliding window with a window size of 30 minutes to calculate the variance of the displacement sequence. When the variance value exceeds three times the standard deviation of the historical mean, it is marked as a displacement mutation point. Finally, it analyzes the residual sequence through wavelet transform to extract the vibration energy characteristics in the 0.1-10Hz frequency band and quantifies the response intensity of the infrastructure under dynamic actions such as wind load and mechanical vibration.

[0050] The deformation identification unit uses principal component analysis to perform feature dimensionality reduction on multi-point displacement sequences, identifies similar deformation patterns through K-means clustering, constructs a displacement transformation model based on rigid body kinematics principles, distinguishes deformation types such as settlement, slip, and torsion of infrastructure, and outputs deformation pattern classification and confidence level. In practice, the deformation recognition unit first performs principal component analysis on the displacement sequences of the four basic measuring points of the tower, extracting the top three principal component features with a contribution rate exceeding 85%. Then, it uses the K-means clustering algorithm to divide the historical deformation data into four typical pattern categories, corresponding to stable state, uniform settlement, differential settlement, and overall tilt, respectively. Next, it constructs a six-degree-of-freedom displacement transformation model based on the rigid body kinematics principle, and solves the translation vector and rotation matrix of the infrastructure using the least squares method. Finally, it calculates the rotational components around the X, Y, and Z axes through Euler angle decomposition and determines whether torsional deformation has occurred, outputting the recognition result containing the deformation type, deformation parameters, and a confidence level of over 90%.

[0051] The stability assessment unit combines displacement rate, deformation type and structural design parameters to establish a comprehensive assessment model based on fuzzy reasoning. It quantifies the safety status of each indicator through membership function, calculates the stability index using weighted average method, and triggers graded early warning when the index exceeds the threshold, and generates a structural safety diagnosis report. In practice, the stability assessment unit first constructs an evaluation system that includes indicators such as displacement rate, deformation type, and vibration intensity. For example, a displacement rate of ≤2 mm / day is set as a safe state, 2-5 mm / day as a warning state, and ≥5 mm / day as a dangerous state. Then, a triangular membership function is used to fuzzify each indicator, and the weight coefficients of each indicator are determined by expert evaluation, with displacement rate having a weight of 0.4, deformation type having a weight of 0.35, and vibration intensity having a weight of 0.25. Next, a weighted average method is applied to calculate the infrastructure stability index, with a value range of 0-1. Thresholds corresponding to four levels—stable, alert, warning, and dangerous—are set according to actual needs. When the stability index exceeds the warning threshold of 0.7, a structural safety diagnostic report containing the current state, development trend, risk level, and disposal suggestions is automatically generated, and the warning information is pushed through a visualization platform.

[0052] In this invention, the risk diagnosis and assessment module is used to comprehensively process the risk identification results output by the macro-risk identification submodule, the refined risk detection submodule, and the infrastructure monitoring submodule, and generate a comprehensive and optimized construction safety risk assessment result through multi-source information fusion and cross-validation technology.

[0053] Specifically, the risk diagnosis and assessment module includes a data association and registration unit, a conflict detection and resolution unit, and a comprehensive risk assessment unit; The data association and registration unit is used to establish the association mapping relationship of multi-source risk information by adopting the spatiotemporal consistency matching algorithm, through unified coordinate system transformation and temporal alignment, so as to realize the spatial registration of macro risk distribution, refined risk targets and infrastructure displacement data; In practice, the data association and registration unit first converts the macro-risk distribution heat map identified by satellite remote sensing, the refined risk targets detected by UAVs, and the infrastructure displacement data monitored by BeiDou positioning into the local coordinate system of the construction area. Then, based on the UTC time reference, a cubic spline interpolation algorithm is used to unify the data with different sampling frequencies to a common time axis of 1Hz. Next, through spatial buffer analysis and nearest neighbor matching algorithms, the association between the risk areas identified by satellite and the refined risk targets detected by UAVs is established. When the refined risk targets are located within the buffer zone set in the risk area, a strong association mapping is established. Finally, a spatiotemporal association matrix is ​​constructed to record the spatial location, timestamp, and association strength of each risk element, providing a foundation for subsequent fusion analysis.

[0054] The conflict detection and resolution unit is used to build a multi-source information fusion model based on DS evidence theory. It quantifies the credibility of each data source through a basic probability allocation function. When a conflict occurs in the monitoring results, a confidence weighting strategy is used to optimize the decision and improve the reliability of risk assessment. In practice, the conflict detection and resolution unit first sets initial credibility weights for the three data sources: satellite remote sensing, UAV monitoring, and BeiDou positioning. The weight for BeiDou positioning data can be set to 0.4, for UAV monitoring data to 0.35, and for satellite remote sensing data to 0.25. Then, the weight coefficients are dynamically adjusted based on the historical accuracy of each data source. When a monitoring conflict occurs (e.g., satellite identifies high risk while BeiDou shows normal displacement), the basic probability allocation of each proposition is calculated using DS evidence theory, and evidence fusion is performed using the Depst synthesis rule. For persistent monitoring conflicts, a multi-source verification mechanism is activated, and UAVs are dispatched to conduct key verification of the conflict area. Real-time video streams collected by front-end cameras are used as arbitration evidence to update the credibility weights of each data source.

[0055] The comprehensive risk assessment unit is used to construct a multi-level fuzzy comprehensive evaluation system. By weighted integration of multi-dimensional risk indicators and rule-based reasoning in the expert knowledge base, it generates hierarchical early warning information and risk assessment reports, and obtains a comprehensive and optimized construction safety risk assessment result. In practice, the comprehensive risk assessment unit first establishes an evaluation index system encompassing four dimensions: personnel safety, equipment safety, environmental safety, and structural safety, with several secondary indicators under each dimension. Then, the analytic hierarchy process (AHP) is used to determine the weights of each indicator, with structural safety weight set to 0.3, equipment safety weight to 0.3, personnel safety weight to 0.2, and environmental safety weight to 0.2. Next, the quantified values ​​of each indicator are converted into standardized risk levels using a fuzzy membership function, and a weighted average operator is applied to calculate the comprehensive risk index. When the risk index exceeds the warning threshold of 0.7, combined with emergency response rules from the expert knowledge base, a comprehensively optimized construction safety risk assessment result is automatically generated, including the risk location, risk type, risk level, impact range, and response recommendations, and pushed to relevant responsible personnel through a visualization platform.

[0056] Finally, the risk diagnosis and assessment module comprehensively processes the risk identification results generated from data from different sources to generate a comprehensive and optimized construction safety risk assessment result. The multi-source data cross-validation mechanism in the risk diagnosis and assessment module can be set according to the actual situation. It is a conventional technology in this field and will not be elaborated here.

[0057] The following example illustrates the multi-source data cross-validation mechanism of the risk diagnosis and assessment module: When the BeiDou positioning system detects a 10cm instantaneous displacement (exceeding the safety threshold of 5cm / h) on a power tower during strong winds: 1. The system automatically marks the displacement data as a suspected anomaly and initiates a multi-source verification mechanism; 2. Immediately access the latest image data from the satellite remote sensing system and verify whether surface displacement has occurred in the tower area through InSAR deformation analysis; 3. The system dispatches drones to collect high-definition images and lidar point cloud data of the tower, and verifies the actual position of the tower through visual ranging and point cloud comparison; 4. If both satellite and drone data show that the tower position is normal, it is determined that the BeiDou displacement data is abnormal due to strong wind interference. The system automatically uses the drone measurement results as the arbitration basis to avoid false alarms. At the same time, the abnormal case is recorded in the knowledge base to optimize the BeiDou data filtering algorithm under strong wind weather. Conversely, when dense fog causes the drone's visual monitoring to fail, the system automatically increases the weight of BeiDou positioning and satellite remote sensing data, and maintains risk assessment capabilities through displacement trend analysis and surface deformation monitoring, ensuring system reliability when a single data source fails.

[0058] Through the above mechanism, the risk diagnosis and assessment module realizes intelligent cross-validation and complementarity of multi-source data, which significantly improves the accuracy of risk assessment and the robustness of the system.

[0059] In this invention, the intelligent prediction and early warning module is used to predict the trend of future construction safety risks and generate a comprehensive prediction report based on the comprehensive and optimized risk assessment results, combined with historical data and predicted meteorological information.

[0060] The intelligent prediction and early warning module can predict future construction safety risks based on comprehensive risk assessment results, combined with historical data and forecast meteorological information, and provide predictive early warnings through time series analysis and machine learning algorithms. Specifically, the intelligent prediction and early warning module includes a risk trend prediction unit, a multi-step prediction and deduction unit, and a prediction result fusion unit; The risk trend prediction unit uses an LSTM neural network to build a time series prediction model. By analyzing the changing patterns of historical risk data and combining it with meteorological forecast information, it predicts the risk development trend in the next 24 hours. In practice, the risk trend prediction unit first preprocesses historical risk data, including missing value imputation, outlier correction, and data standardization. Then, it constructs an LSTM network with 128 hidden units, with an input sequence length of 24 hours, and outputs the predicted risk index value for the next 24 hours. The model training uses the mean squared error loss function, and the Adam optimizer is used for parameter optimization. Finally, parameters such as wind speed and rainfall from weather forecasts are used as external features input to the model to improve prediction accuracy. When the predicted risk index exceeds 0.8, a risk trend warning is generated in advance.

[0061] The multi-step prediction and simulation unit, based on physical models and data-driven methods, constructs a multi-step risk evolution prediction chain to simulate the propagation path and impact range of risks under different meteorological conditions; In practice, the multi-step prediction and simulation unit first establishes a directed graph model of risk propagation, defining the causal relationships and propagation weights among various risks. Then, it uses the Monte Carlo method to conduct stochastic simulations, setting the probability distributions of meteorological factors (wind speed follows a two-parameter Weibull distribution, temperature follows a normal distribution, and humidity follows a Beta distribution), and fitting the distribution parameters based on historical meteorological data. Considering the influence of meteorological factors such as wind speed, temperature, and humidity, 1000 risk evolution paths are generated. Next, the probability of occurrence of each risk node is calculated using the kernel density estimation method to identify high-risk propagation paths. Finally, the propagation weights are dynamically adjusted based on a graph neural network, outputting risk distribution prediction maps for the next 6 hours, 12 hours, and 24 hours.

[0062] The prediction result fusion unit uses the DS evidence theory to fuse multi-source prediction results and generates a comprehensive prediction report through confidence assessment and conflict resolution. In practice, the prediction result fusion unit first standardizes the risk index prediction value output by the risk trend prediction unit, the risk distribution prediction map output by the multi-step prediction extrapolation unit, and the expert experience prediction results. Then, it assigns initial confidence levels based on the historical accuracy of each prediction method, with a confidence level of 0.4 for the risk trend prediction unit, 0.35 for the multi-step prediction extrapolation unit, and 0.25 for the expert experience. When prediction results conflict, a weighted average method is used for fusion, and a sensitivity analysis is performed on the fusion results. Finally, a comprehensive prediction report containing the predicted value, confidence interval, risk type, and scope of impact is output.

[0063] Ultimately, the intelligent prediction and early warning module generates a comprehensive prediction report based on the optimized risk assessment results, combined with historical data and forecast meteorological information, to predict future construction safety risks.

[0064] In this invention, the early warning decision and response module is used to generate tiered early warning information and optimized response plans based on the current risk assessment results and future risk change trends. In this embodiment, the early warning decision and response module includes an early warning level assessment unit, a response plan generation unit, and an emergency response linkage unit.

[0065] The early warning level assessment unit, based on the fuzzy comprehensive evaluation method, combines the current risk index and predicted risk trends to determine the early warning level and the scope of issuance; In practice, the early warning level assessment unit first establishes an evaluation index system that includes three dimensions: risk severity, urgency, and scope of impact. Then, it uses the analytic hierarchy process (AHP) to determine the weights of each index, with risk severity weighted at 0.5, urgency weighted at 0.3, and scope of impact weighted at 0.2. Next, it uses a triangular membership function to fuzzify each index and applies a weighted average operator to calculate the comprehensive early warning index. Finally, based on the early warning index, the early warning is divided into three levels: red, orange, and yellow, and the corresponding early warning issuance scope and response time limit are determined.

[0066] The response plan generation unit automatically generates the optimal response plan based on case reasoning and rule engine technology, combined with the warning level and risk characteristics; In practice, the disposal plan generation unit first constructs a knowledge base containing historical disposal cases, with each case including risk scenarios, disposal measures, and effect evaluations. When a new warning is generated, the k-nearest neighbor algorithm is used to retrieve similar cases from the knowledge base and calculate the feature similarity between cases. Then, the rule engine matches the preset disposal rules to generate a preliminary disposal plan. Finally, the plan is adaptively adjusted based on the actual situation on site, and a detailed disposal plan including disposal steps, responsible departments, resource requirements, and completion deadlines is output.

[0067] An emergency response unit establishes a multi-departmental collaborative response mechanism, using an information platform to achieve rapid distribution of early warning information and full-process tracking of the handling process; In practice, the emergency response unit first establishes rules for distributing early warning information, determines the scope and method of notification based on the warning level, and requires all relevant personnel to be notified within 5 minutes for red warnings; then, it drives the handling process through a workflow engine, automatically assigns tasks to responsible departments and personnel, and sets processing time limits and assessment requirements; finally, it collects on-site handling information in real time through a mobile APP, automatically generates a handling progress report, and automatically urges overdue steps to be completed.

[0068] Finally, the early warning decision-making and response module generates tiered early warning information and optimized response plans based on the current risk assessment results and future risk change trends. In this invention, the visualization and feedback module is used to provide a human-computer interaction interface to display analysis results and early warning information; The visualization and feedback module can construct a panoramic view of the construction safety situation through digital twin technology, providing multi-dimensional information display and human-computer interaction functions; The visualization and feedback module can display setting information, including image information and data information, and provide human-computer interaction functions. It is a conventional technology in this field and will not be described in detail here.

[0069] D: Based on the trained construction safety analysis model, multi-source data is collected using an integrated air-space collaborative monitoring network, combined with a meteorological database, to conduct online monitoring, early warning, and handling of power distribution network construction safety.

[0070] Through the organic connection and iterative process of the above steps, this invention constructs a complete technical system from multi-source data acquisition in space and air, intelligent risk identification, integrated situation assessment, prediction and early warning to visualized decision-making. It achieves full-process and all-round coverage of power distribution network construction safety monitoring, significantly improves the accuracy of safety hazard identification and the timeliness of early warning, and provides strong technical support for the construction safety of power distribution network projects.

[0071] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0072] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for integrated air-space safety monitoring of power distribution network construction based on unmanned aerial vehicles (UAVs), characterized in that: Includes the following steps: A: Construct an integrated space-air collaborative monitoring network and use it to collect multi-source data from the power distribution network construction area. The multi-source data includes macroscopic change data of the construction area, refined scene data, three-dimensional spatial data, and infrastructure displacement data. B: Obtain historical and forecast meteorological information for the construction area and construct a meteorological database; C: Construct a construction safety analysis model; the construction safety analysis model includes a data acquisition and preprocessing module, a multi-source data fusion analysis module, a risk diagnosis and assessment module, an intelligent prediction and early warning module, an early warning decision and response module, and a visualization and feedback module; The system comprises the following modules: a data acquisition and preprocessing module, used to standardize and spatiotemporally align the input multi-source data, providing a data foundation for subsequent analysis; a multi-source data fusion and analysis module, used to deeply integrate macroscopic change data, refined scene data, 3D spatial data, infrastructure displacement data, and meteorological data of the construction area, thereby obtaining multi-source fused data and generating independent risk identification results based on data from different sources; a risk diagnosis and assessment module, used to comprehensively process the risk identification results generated from data from different sources, generating a comprehensively optimized construction safety risk assessment result; an intelligent prediction and early warning module, used to predict future construction safety risks based on the comprehensively optimized risk assessment result, combined with historical data and forecast meteorological information; an early warning decision and response module, used to generate tiered early warning information and optimized response plans based on the current risk assessment result and future risk change trends; and a visualization and feedback module, used to provide a human-computer interaction interface to display analysis results and early warning information. D: Based on the trained construction safety analysis model, multi-source data is collected using an integrated air-space collaborative monitoring network, combined with a meteorological database, to conduct online monitoring, early warning, and handling of power distribution network construction safety.

2. The safety monitoring method according to claim 1, characterized in that: The integrated air-space collaborative monitoring network includes a satellite remote sensing monitoring system, an unmanned aerial vehicle (UAV) near-ground monitoring system, and a BeiDou high-precision positioning monitoring system. The satellite remote sensing monitoring system is used to collect macroscopic change data of the construction area; the UAV near-ground monitoring system is used to collect detailed scene data and three-dimensional spatial data of the construction area; and the BeiDou high-precision positioning monitoring system is used to collect three-dimensional displacement data of critical infrastructure in the absolute geographic coordinate system.

3. The safety monitoring method according to claim 1, characterized in that: The data acquisition and preprocessing module includes an image processing submodule, a point cloud data processing submodule, a numerical data processing submodule, and a positioning data processing submodule; The image processing submodule is used to process satellite remote sensing image data, UAV visible light image data and infrared image data to obtain standardized image data; The point cloud data processing submodule is used to process the 3D point cloud data collected by the UAV LiDAR to obtain a structured point cloud feature dataset. The numerical data processing submodule is used to process numerical monitoring data collected by meteorological sensors or obtained from meteorological databases to obtain standardized numerical time-series datasets. The positioning data processing submodule is used to process the three-dimensional displacement data collected by the Beidou high-precision positioning monitoring system and obtain a standardized displacement time series dataset.

4. The safety monitoring method according to claim 1, characterized in that: The multi-source data fusion and analysis module includes a macro-risk identification submodule, a refined risk detection submodule, and an infrastructure monitoring submodule; The macro-risk identification submodule is used to extract features and recognize patterns from preprocessed satellite remote sensing data, and generate macro-risk identification results and risk areas based on satellite data. The refined risk detection submodule is used to extract refined risk features, i.e. refined risk targets, of construction scenarios based on pre-processed UAV imagery data and point cloud data, using computer vision and deep learning technologies. The infrastructure monitoring submodule is used to obtain infrastructure displacement data based on BeiDou high-precision positioning data, through time series analysis and deformation recognition algorithms.

5. The safety monitoring method according to claim 4, characterized in that: The macro-risk identification submodule includes a target detection unit, a change detection unit, and a regional risk assessment unit; The target detection unit is used to build a construction machinery recognition model based on the improved YOLOv5 deep learning framework. By introducing the CBAM attention mechanism and optimizing the anchor frame size, it can realize the automatic recognition and accurate positioning of construction machinery and output the machinery type, position coordinates and confidence level. The change detection unit adopts a semantic segmentation model based on the U-Net architecture. It extracts multi-scale features through an encoder-decoder structure, compares pixel-level differences in satellite images of different time phases, automatically identifies surface changes, and outputs vector boundaries of changed areas and change types. The regional risk assessment unit integrates target detection and change detection results to construct a construction activity intensity index. Based on the fuzzy comprehensive evaluation method, it classifies risk levels by incorporating multi-dimensional features including weighted fusion of mechanical equipment density, area of ​​change, and construction duration, and generates a macro-risk distribution heat map.

6. The safety monitoring method according to claim 4, characterized in that: The refined risk detection submodule includes a personnel safety supervision unit, an equipment safety monitoring unit, and an environmental risk perception unit; The personnel safety supervision unit uses a multi-target tracking algorithm combined with a classification network to continuously track construction workers, and judges the wearing status of protective equipment based on an attention mechanism. It outputs personnel trajectory, behavioral norms assessment and violation alarms in real time to obtain refined risk targets. The equipment safety monitoring unit, based on LiDAR point cloud spatial analysis and combined with point cloud nearest neighbor search, calculates the spatial distance between the moving parts of mechanical equipment and power lines in real time. When the distance value is lower than the preset threshold, it triggers a collision warning. At the same time, it identifies equipment overheating abnormalities through infrared temperature field analysis. The environmental risk perception unit constructs a smoke detection model based on the YOLO framework, enhances the detection capability of small targets through multi-scale feature fusion, extracts the smoke diffusion area by combining image segmentation algorithm, and identifies environmental anomalies in the construction area by using background modeling and frame difference method, outputting the risk type, spatial location and urgency level.

7. The safety monitoring method according to claim 4, characterized in that: The infrastructure monitoring submodule includes a displacement analysis unit, a deformation identification unit, and a stability assessment unit; The displacement analysis unit uses a time series decomposition method, decomposes the BeiDou three-dimensional coordinate sequence into trend, periodic and residual terms using the STL algorithm, identifies displacement abrupt change points based on sliding window variance analysis, and extracts vibration characteristics of different frequency bands by combining wavelet transform, thereby quantifying the overall displacement and vibration intensity of the infrastructure. The deformation identification unit uses principal component analysis to perform feature dimensionality reduction on multi-point displacement sequences, identifies similar deformation patterns through K-means clustering, constructs a displacement transformation model based on rigid body kinematics principles, distinguishes the deformation types of infrastructure, and outputs deformation pattern classification and confidence level. The stability assessment unit combines displacement rate, deformation type and structural design parameters to establish a comprehensive assessment model based on fuzzy reasoning. It quantifies the safety status of each indicator through membership function, calculates the stability index using weighted average method, and triggers graded early warning when the index exceeds the threshold, and generates a structural safety diagnosis report.

8. The safety monitoring method according to claim 1, characterized in that: The risk diagnosis and assessment module includes a data association and registration unit, a conflict detection and resolution unit, and a comprehensive risk assessment unit; The data association and registration unit is used to establish the association mapping relationship of multi-source risk information by adopting the spatiotemporal consistency matching algorithm, through unified coordinate system transformation and temporal alignment, so as to realize the spatial registration of macro risk distribution, refined risk targets and infrastructure displacement data; The conflict detection and resolution unit is used to build a multi-source information fusion model based on DS evidence theory. It quantifies the credibility of each data source through a basic probability allocation function. When a conflict occurs in the monitoring results, it uses confidence to optimize the decision and improve the reliability of risk assessment. The comprehensive risk assessment unit is used to construct a multi-level fuzzy comprehensive evaluation system. By weighted integration of multi-dimensional risk indicators and rule-based reasoning from the expert knowledge base, it generates graded early warning information and risk assessment reports, and obtains a comprehensive and optimized construction safety risk assessment result.

9. The safety monitoring method according to claim 2, characterized in that: The UAV near-ground monitoring system uses a multi-rotor UAV platform and is equipped with a high-definition visible light camera, an infrared thermal imager, and a lidar. The high-definition visible light camera captures visible light images of the construction scene; the infrared thermal imager generates thermal infrared images; and the lidar acquires the three-dimensional spatial coordinates of the target surface.

10. The safety monitoring method according to claim 2, characterized in that: The BeiDou high-precision positioning and monitoring system uses a high-precision BeiDou / GPS dual-mode receiver that supports carrier phase differential technology.

Citation Information

Cited By

  • Industrial unmanned aerial vehicle cluster intelligent scheduling management method based on multi-data analysis

    CN121616059A

  • Sky-ground safety collaborative management and control device based on open-source swan-Mongolia pumped storage power station during construction period

    CN121660858A

  • Natural reserve self-adaptive monitoring system and method based on sky-ground coordination

    CN121937905A