A multi-dimensional construction safety risk management and control method for high-altitude highways

Abstract

The application discloses a kind of multi-dimension construction safety risk management and control methods for high-altitude highway, specifically related to construction risk control field, for solving the problem of slope monitoring lag and passive safety management and control caused by complex terrain in the process of existing high-altitude highway construction;By collecting the aerial survey image and laser point cloud data of the construction slope area according to fixed cycle, a time sequence three-dimensional real scene model sequence is generated, natural feature points and artificial structure feature points are identified and tracked in the model sequence, the feature point displacement field is calculated based on the accuracy registration between multiple models, the risk deformation data existing in time and space correlation is identified by combining historical construction activity records and construction activities, the potential instability area is labeled by spatial clustering analysis of risk deformation data, and the prospective construction risk classification control signal is generated according to the spatial range of potential instability area and construction information data, to realize the dynamic prediction and classification response of high-altitude highway construction risk.

Description

Technical Field

[0001] This invention relates to the field of construction risk management technology, and more specifically, to a multi-dimensional construction safety risk management method for high-altitude highways. Background Technology

[0002] Highway construction in high-altitude areas faces various safety risks due to complex terrain, harsh climate, and variable geological structures, including slope instability, landslides, rockfalls, and freeze-thaw damage. These risks are often characterized by their suddenness, uneven spatial distribution, and significant temporal variations.

[0003] Current construction safety management primarily relies on manual inspections and static monitoring methods, such as fixed-point measurements, video observation, or local displacement sensor monitoring. These methods struggle to provide a comprehensive and dynamic overview of the entire construction area, especially in high-altitude regions where low air pressure, strong radiation, and large diurnal temperature variations make monitoring equipment susceptible to environmental interference, compromising data continuity and reliability. With the development of digital mapping technologies such as UAV aerial surveying, laser point cloud scanning, and 3D reconstruction, utilizing multi-source spatial data to construct temporal 3D reality models of the construction area and combining this with construction behavior records for multi-dimensional risk analysis is becoming a new trend.

[0004] However, how to achieve accurate registration of multi-phase models and tracking and identification of potential instability points in dynamic construction scenarios, and how to output forward-looking risk classification signals in combination with construction information data for construction control in potential instability areas, remains a key technical challenge in the field of safety management of high-altitude highway construction. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a multi-dimensional construction safety risk management method for high-altitude highways to solve the problems mentioned in the background art.

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

[0007] A multi-dimensional construction safety risk management method for high-altitude highways includes the following steps:

[0008] S1. Collect aerial survey images and laser point cloud data of the construction slope area at fixed intervals to generate a time-series 3D real scene model sequence corresponding to each collection time point;

[0009] S2. In the sequence of three-dimensional real-scene models, identify and track the natural feature points and artificial structure feature points on the surface of each phase of the model;

[0010] S3. By registering the accuracy between multiple models, calculate the three-dimensional coordinate change of each feature point at continuous acquisition time points, and construct a feature point displacement field covering the entire slope area.

[0011] S4. Based on historical construction activity records and the spatiotemporal characteristics of displacement field, identify feature points in the displacement field that are spatiotemporally related to construction activities, and define deformation data that exhibits continuous acceleration or spatial diffusion as risk deformation data.

[0012] S5. Screen the spatially clustered distribution areas of risk deformation data and mark them as potential instability areas;

[0013] S6. Based on the spatial range of potential instability zones, statistically analyze the corresponding construction information data to generate a forward-looking construction risk classification and control signal.

[0014] In a preferred embodiment, step S1, which involves collecting aerial survey images and laser point cloud data of the construction slope area at fixed intervals to generate a sequence of temporal 3D real-scene models corresponding to each collection time point, specifically includes:

[0015] Aerial survey imagery and laser point cloud data of the construction slope area are collected at a preset fixed period.

[0016] The aerial survey image data includes multi-angle overlapping images covering the entire slope area, and the laser point cloud data includes a set of three-dimensional spatial coordinate points evenly distributed in the slope area.

[0017] A high-resolution digital surface model is generated based on aerial survey image sequences. The laser point cloud data is registered and fused with the digital surface model. The coordinate system of the laser point cloud data is mapped to the coordinate system of the digital surface model to construct a three-dimensional model framework.

[0018] The surface texture information extracted from the aerial survey image data collected in each cycle is mapped onto the surface of the 3D model framework, and a time-series 3D real scene model sequence is generated through texture mapping.

[0019] In a preferred embodiment, step S2, in the sequence of three-dimensional real-world models, specifically includes identifying and tracking natural feature points and man-made structure feature points on the surface of each model, including:

[0020] On the surface of the 3D real scene model, prominent bedrock edges and landform turning points are selected as natural feature points, and concrete structure corners in completed projects are selected as artificial structure feature points.

[0021] A feature description vector is constructed based on the three-dimensional spatial coordinates of feature points and surface texture information. The feature description vector contains the geometric shape parameters and texture distribution parameters of the feature points.

[0022] In the sequence of 3D reality models, the spatial position of the feature point in the 3D reality model of the previous acquisition cycle is used as the center. Within the preset search radius, the point with the highest similarity of the feature description vector is searched to perform cross-period tracking and matching of feature points.

[0023] In a preferred embodiment, step S3, through precision registration between multiple models, calculates the three-dimensional coordinate change of each feature point at continuous acquisition time points, and constructs a feature point displacement field covering the entire slope area, specifically including:

[0024] Using the 3D real-scene model corresponding to the previous acquisition cycle as the spatial reference benchmark, calculate the spatial transformation parameter sequence of the feature points in the 3D real-scene model sequence that are tracked and matched across periods relative to the spatial reference benchmark.

[0025] Numerical statistical analysis is performed on the spatial transformation parameters to identify stable feature point groups that are concentrated in the numerical value of the transformation parameters, and prominent feature points whose deviation from the transformation parameters and stable feature point groups exceeds a set threshold are selected.

[0026] The coordinate sequence of prominent feature points is extracted based on transformation parameters in a unified coordinate system, wherein the unified coordinate system is the spatial coordinate system of the three-dimensional real scene model in the first acquisition cycle;

[0027] The planar displacement components and elevation displacement components of all prominent feature points in each acquisition cycle are obtained and arranged in chronological order to construct a feature point displacement field covering the entire slope area.

[0028] In a preferred embodiment, in step S4, based on historical construction activity records and the spatiotemporal characteristics of the displacement field, feature points in the displacement field that are spatiotemporally related to the construction activities are identified. Deformation data exhibiting continuous acceleration or spatial diffusion are defined as risk deformation data, specifically including:

[0029] The construction area and construction time recorded in the historical construction activity records are spatiotemporally superimposed with the spatial distribution and change time of prominent feature points in the displacement field for analysis.

[0030] Feature points whose displacement changes are synchronized with the start time of construction activities in time and overlap with the area of ​​construction activities in space are identified as having a spatiotemporal correlation with construction activities.

[0031] Displacement changes of feature points that are spatiotemporally related to construction activities are converted into deformation data, and deformation data that exhibits continuous acceleration or spatial diffusion over time are defined as risk deformation data.

[0032] In a preferred embodiment, the risk deformation data consists of a sequence of relative displacement vectors between the corresponding prominent feature points and the geometric centers of a pre-labeled set of adjacent feature points; the adjacent feature points are determined based on pre-defined engineering zoning boundaries or geological structural boundaries.

[0033] In a preferred embodiment, step S5, screening spatially clustered areas of risk deformation data and marking them as potential instability zones, specifically includes:

[0034] Spatial clustering analysis is performed on the risk deformation data. Based on the relative displacement vector direction corresponding to the deformation, feature points with similar displacement vector directions or spatial distribution in a radial convergence pattern are grouped into the same clustered distribution area.

[0035] Based on the spatial extent of the clustered distribution area and the spatial density of risk deformation data, the boundary range of the potential instability zone is delineated and marked.

[0036] In a preferred embodiment, the similarity determination condition for the displacement vector direction is that the angle between the displacement vector direction and the average direction of all displacement vectors is less than a set angle threshold, and the determination condition for the radial convergence pattern is that the directions of each displacement vector point to the same geometric center region.

[0037] In a preferred embodiment, step S6, generating a forward-looking construction risk classification and control signal based on the construction information data corresponding to the spatial range of the potential instability zone, specifically includes:

[0038] Acquire construction information data, which includes the real-time location distribution of construction personnel and machinery and equipment, as well as the static location information of key structures.

[0039] Spatial overlay analysis of the spatial boundary of the potential instability zone and construction information data is performed to count the quantity and type of various construction information data falling into the potential instability zone.

[0040] Based on the spatial extent of the potential instability zone and the value and quantity of affected construction information data, a set of risk assessment factors is constructed, and a comprehensive risk value is calculated.

[0041] Based on the predefined range of the comprehensive risk value, forward-looking construction control signals corresponding to different risk levels are generated.

[0042] The technical effects and advantages of this invention's multi-dimensional construction safety risk management method for high-altitude highways:

[0043] By integrating aerial survey imagery with laser point cloud data into a fusion model, full-area three-dimensional dynamic monitoring of high-altitude highway construction areas can maintain high-precision data acquisition and spatial continuity even in complex terrain and harsh weather conditions. Through cross-period tracking of feature points and displacement field construction, traditional static monitoring is transformed into dynamic time-series analysis, accurately depicting the evolution trends of slopes, structures, and surface deformation, significantly improving the foresight and precision of risk identification. Combined with spatiotemporal correlation analysis of construction activity records, risk assessment no longer relies on single terrain changes but is based on a comprehensive judgment of the multi-dimensional coupling relationship between construction disturbances and geological responses, effectively reducing false alarms and missed alarms. Spatial clustering of risk deformation data can automatically delineate potential instability zones, providing quantitative basis for construction organization adjustments, equipment deployment, and safety early warning. The resulting forward-looking construction risk classification signal enables real-time risk classification and dynamic response in actual engineering management, improving the scientific and proactive nature of construction safety decisions and overall enhancing the safety management level and operational continuity of high-altitude highway projects in complex geological environments. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of a multi-dimensional construction safety risk management method for high-altitude highways according to the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0046] Example 1, Figure 1 This invention presents a multi-dimensional construction safety risk management method for high-altitude highways, which includes the following steps:

[0047] S1. Collect aerial survey images and laser point cloud data of the construction slope area at fixed intervals to generate a time-series 3D real scene model sequence corresponding to each collection time point;

[0048] S2. In the sequence of three-dimensional real-scene models, identify and track the natural feature points and artificial structure feature points on the surface of each phase of the model;

[0049] S3. By registering the accuracy between multiple models, calculate the three-dimensional coordinate change of each feature point at continuous acquisition time points, and construct a feature point displacement field covering the entire slope area.

[0050] S4. Based on historical construction activity records and the spatiotemporal characteristics of displacement field, identify feature points in the displacement field that are spatiotemporally related to construction activities, and define deformation data that exhibits continuous acceleration or spatial diffusion as risk deformation data.

[0051] S5. Screen the spatially clustered distribution areas of risk deformation data and mark them as potential instability areas;

[0052] S6. Based on the spatial range of potential instability zones, statistically analyze the corresponding construction information data to generate a forward-looking construction risk classification and control signal.

[0053] In step S1, aerial survey images and laser point cloud data of the construction slope area are collected at fixed intervals to generate a time-series three-dimensional real scene model sequence corresponding to each collection time point.

[0054] A comprehensive 3D dynamic monitoring dataset covering the construction slope area was established using a fixed-cycle aerial survey and laser point cloud acquisition method. The acquisition cycle was determined based on a combination of construction progress and weather conditions, set every 7 days during stable seasons and extended to 15 days during periods of strong winds or heavy snow to ensure data continuity and weather adaptability. Aerial survey imagery data was acquired using UAVs equipped with high-resolution digital cameras. The flight path followed the outline of the slope area, with an overlap rate of no less than 80%, ensuring sufficient common viewing area between images from different angles to support subsequent stereo reconstruction and elevation fitting. Laser point cloud data was simultaneously acquired by an airborne laser scanner during the same flight. The scanning density was automatically adjusted according to the terrain slope, ensuring a point cloud density of more than 50 points per square meter in areas with drastic slope changes and maintaining approximately 10 points per square meter in gentler areas.

[0055] Based on the acquired aerial survey image sequence, a high-resolution digital surface model (DSM) is generated using a 3D reconstruction algorithm. First, the aerial survey images undergo relative orientation and illumination correction to eliminate illumination differences and distortions between different flights. Then, common points between images are extracted using image feature point matching, and the 3D coordinates of each feature point in space are recovered based on triangulation principles, forming dense point cloud data. This dense point cloud is then interpolated to generate the DSM. The model resolution is set according to the ground sampling distance of the images, controlled within a range of 5 to 10 centimeters, balancing detail representation and processing efficiency. After generating the DSM, the laser point cloud data is incorporated into a registration process. Coarse registration aligns the laser point cloud data to the spatial framework of the DSM, followed by fine registration using an iterative nearest-point algorithm to eliminate minor misalignments caused by differences in equipment attitude or flight trajectory. After registration, the coordinate system of the laser point cloud data is mapped to the coordinate system of the DSM, achieving spatial unification of the two types of data and constructing a 3D model framework that includes terrain geometry and surface structure.

[0056] To enable visualization and temporal comparability of the 3D spatial structure, surface texture information from aerial survey imagery is mapped onto the model surface. Surface texture information is extracted from aerial survey imagery acquired in each period, and brightness deviations caused by climate change or differences in solar altitude angle are eliminated through illumination correction and color equalization. Then, based on the spatial coordinate information of the model framework, the spatial position of each surface point is matched with its corresponding image pixel, and a continuous texture map is generated using bilinear interpolation to ensure that the surface color distribution matches the actual ground texture. During the mapping process, for overlapping areas with multiple views, a view weight fusion strategy is used for texture fusion, selecting image areas with higher verticality of the shooting angle to improve texture clarity and detail reproduction. After complete texture processing, a 3D real-scene model covering the entire slope area is obtained. The 3D real-scene models generated in each period are arranged in chronological order to form a temporal sequence of 3D real-scene models.

[0057] In S2, in the three-dimensional real-scene model sequence, natural feature points and artificial structure feature points on the surface of each model are identified and tracked.

[0058] The model surface undergoes initial selection and classification of feature points. Natural feature points are selected based on the salience of the landform, focusing on identifying locations on slopes with abrupt geometric changes, such as bedrock edges, fault intersections, gully inflections, and rock joint lines. Unstable locations prone to landslides, settlement, or collapses are prioritized. These locations exhibit significant curvature changes and elevation abrupt changes in spatial morphology, reflecting the stability and trends of the terrain. Feature points for man-made structures are selected primarily from completed structural entities, such as key connection points of elevated structures and tunnels, as well as corners, edge intersections, and surface connection nodes of concrete structures like retaining piles and anchor platforms. Unstable locations are prioritized based on historical accident records. To ensure the traceability of feature points across multiple model phases, each feature point must be spatially salient and have clear texture. Spatial salience is confirmed by a neighborhood elevation difference greater than 5 cm, and texture clarity is determined by a grayscale variation variance greater than a set threshold (e.g., 30). A feature point set covering both natural and man-made components is established through a combination of manual and automatic identification methods.

[0059] For each feature point, a feature description vector is established to describe its spatial and surface attributes. The feature description vector consists of two types of information: geometric shape parameters and surface texture distribution parameters. Geometric shape parameters characterize the spatial morphological features of the local region where the feature point is located, including local curvature, normal direction distribution, and elevation gradient changes. Specifically, this is obtained by statistically analyzing the consistency of surface normal direction and the range of elevation changes within a 20 cm diameter neighborhood around the feature point. Texture distribution parameters describe the image brightness and color characteristics within the feature point region. These are extracted by analyzing the texture map of the 3D model surface to obtain statistics such as average brightness, brightness variance, and main texture direction. To unify the data structure of different types of feature points, the feature description vectors of all feature points are normalized to ensure comparability across models from different acquisition periods.

[0060] Establish cross-period correspondences of feature points between 3D reality models from different acquisition cycles. Using the spatial location of a feature point in the current acquisition cycle within the model of the previous cycle as the center, a certain spatial search radius is set for similarity point searching. The search radius is determined based on the terrain change rate of the slope area and the acquisition cycle interval, defaulting to 0.2 to 0.5 meters to cover possible deformation ranges without introducing redundant matching. Within this search range, the similarity of feature description vectors between all candidate points in the current model and the target feature point is calculated. The similarity metric comprehensively considers geometric morphological differences and texture distribution differences. The geometric part uses a weighted average of normal direction deviation and elevation gradient deviation, while the texture part compares brightness difference and texture direction consistency. The candidate point with the highest overall similarity is confirmed as the corresponding point of that feature point in the current cycle, thus establishing a spatial tracking relationship between the two models. Performing this matching operation on all feature points yields complete cross-period tracking results, forming a dynamic correspondence sequence of feature points covering the entire construction slope area.

[0061] In S3, through precision registration between multiple models, the three-dimensional coordinate change of each feature point at continuous acquisition time points is calculated to construct a feature point displacement field covering the entire slope area.

[0062] Using the 3D reality model corresponding to the previous acquisition cycle as a spatial reference, all cross-cycle matched feature points in the current cycle model are mapped to the coordinate frame of this reference model. To achieve spatial unification between the two cycles, the spatial transformation parameters of each feature point relative to the reference model need to be calculated. These transformation parameters consist of three components: translation, rotation, and scale, which respectively characterize the model's relative position offset, pose change, and scale difference in 3D space. During calculation, cross-cycle tracking matched feature point pairs are used as input, and the overall spatial transformation relationship is determined through least-squares fitting. The relative spatial offset of each feature point is calculated individually to form a complete sequence of transformation parameters. Each parameter sequence represents the spatial change state of the feature point in the current acquisition cycle relative to the previous cycle.

[0063] Numerical statistical analysis is performed on the parameters to identify stable feature point groups in the model that are least affected by terrain or structure deformation. Specifically, the translation and rotation components of all feature points are statistically analyzed, and their average and standard deviations are calculated to form the overall distribution characteristics of the transformation parameters. A stable feature point group is defined as a set of points whose transformation parameter values ​​are concentrated near the average and whose deviations are less than a set threshold. This threshold is determined based on measurement accuracy and terrain stability. For example, when the positioning accuracy of the acquisition device is 2 cm, the translation deviation threshold is set to 3 cm, and the rotation deviation threshold is set to 0.5 degrees. After statistical analysis, points whose transformation parameters deviate from the stable feature point group by more than the set threshold are identified as prominent feature points. Prominent feature points reflect the deformation of the local surface or the displacement characteristics of structures, and their distribution is the core basis for subsequent deformation identification. In this way, interference caused by overall translation is reduced in complex terrain and construction disturbance environments, preserving more accurate information on local changes.

[0064] The spatial coordinate sequence of prominent feature points is extracted under a unified coordinate system, and their displacement changes in each acquisition cycle are calculated. The unified coordinate system is based on the coordinate system of the 3D real-scene model in the first acquisition cycle, ensuring that the data at all time points have a directly comparable spatial reference. The model coordinates of subsequent cycles are gradually mapped to this initial coordinate system through accumulated transformation parameters, realizing the spatial integration of time-series data. For each prominent feature point, its 3D coordinate changes in continuous acquisition cycles are extracted and decomposed into planar displacement components (along the horizontal plane) and elevation displacement components (along the vertical direction). Planar displacement reflects the horizontal movement trend of the ground surface or structure, and elevation displacement reflects the vertical uplift or settlement trend. The displacement data of all prominent feature points are arranged in chronological order to form a multi-time period displacement sequence, and spatial interpolation and continuous processing are performed to construct a feature point displacement field covering the entire slope area. The displacement field is in the form of a spatial grid.

[0065] In step S4, based on historical construction activity records and the spatiotemporal characteristics of the displacement field, feature points in the displacement field that are spatiotemporally related to construction activities are identified, and deformation data that exhibits continuous acceleration or spatial diffusion are defined as risk deformation data.

[0066] Historical construction activity records are used as external spatiotemporal reference data for analysis. These records include the spatial boundary coordinates of the construction area, start and end times, and corresponding operation types, such as blasting, excavation, support, and pouring. The spatial boundary information in the construction activity records is converted into a spatial format consistent with the coordinates of the 3D reality model, and a correspondence is established between the start and end times and the timestamps of each acquisition period. Subsequently, prominent feature points in the displacement field are superimposed and compared in both spatial and temporal dimensions. Spatial superposition is performed by determining whether the coordinates of the feature points fall within the boundaries of the construction area or the construction impact (such as the estimated impact range of blasting). Temporal superposition is determined by comparing whether the displacement change time of the feature points and the start time of the construction activity are within the same monitoring window. A synchronization time tolerance of two days is set, meaning that when the time of the displacement change of the feature points differs from the start time of the construction activity by no more than two days, it is considered to be time synchronized. After spatiotemporal superposition, if the spatial location of a feature point overlaps with the construction activity area and its displacement change time matches the construction time, then the feature point is identified as having a spatiotemporal association with the construction activity.

[0067] The displacement changes of feature points that are spatiotemporally related to construction activities are converted into deformation data. Based on the spatial coordinate sequence of feature points within a continuous acquisition period, the displacement change trend of feature points in the time dimension is calculated, and the displacement sequence at discrete time points is interpolated to form a continuous deformation curve. The vertical axis of the deformation curve represents the displacement amount, and the horizontal axis represents time. When the slope of the curve gradually increases over a continuous time period, it indicates a deformation state that accelerates continuously with time. This usually occurs during the support construction stage or the slope cutting stage, when the soil and rock structure undergoes rapid displacement due to stress redistribution. If the deformation curve remains stable in time but the displacement direction between adjacent feature points gradually diverges, it indicates that the deformation has spatial diffusion characteristics, which is common in the early signs of local instability such as surface cracking and slope toe heave. The deformation data generation process not only retains the displacement amplitude information of feature points but also incorporates their temporal evolution characteristics into the data structure to distinguish different types of deformation patterns in subsequent analysis. All deformation data sets that show an accelerating trend with time or a diffusion pattern in spatial distribution are defined as risk deformation data, thereby identifying high-risk construction impact areas.

[0068] Risk deformation data is organized as a relative displacement vector sequence, consisting of the spatial displacement relationship between the corresponding prominent feature points and the geometric centers of their adjacent feature point sets. Specifically, within the local area of ​​each prominent feature point, adjacent feature point sets are determined according to pre-defined engineering zoning boundaries or geological structure boundaries. Feature points located within the same geological unit or construction section, such as identical bedrock slopes or fault fracture zones, are considered. The displacement difference between the prominent feature point and the geometric center of its adjacent feature point set is calculated in each acquisition cycle, forming a set of relative displacement vectors with clearly defined directions and magnitudes. These vectors are arranged in chronological order to form a time-series vector sequence of risk deformation data. This vector sequence reflects the local relative deformation of feature points during overall surface movement, while also revealing the changing patterns of local stress transmission paths and displacement directions on the slope, visually presenting the local deformation propagation process caused by construction disturbance at different time stages.

[0069] In step S5, areas where risk deformation data is clustered in space are selected and marked as potential instability zones.

[0070] Spatial clustering analysis is performed on deformation data, using the spatial coordinates of feature points and their corresponding relative displacement vector directions as the main input information. All risk deformation feature points are projected onto a 3D spatial grid in a unified coordinate system, and a spatial index structure of feature points is constructed according to their spatial proximity. Then, a set of neighboring feature points is extracted within a preset search radius, centered on each feature point. The search radius is determined based on the slope size and sampling density; for example, it is set to 2 meters in areas with large slope elevation differences and 5 meters in gentle slope areas. For each neighborhood set, the distribution characteristics of the displacement vector directions of all feature points are calculated, and feature points with smaller directional changes are grouped into the same initial cluster group. To ensure the uniformity of directional similarity determination, a displacement vector direction similarity determination condition is set: if the angle between the displacement vector direction of any feature point and the average direction of all displacement vectors in that group is less than a threshold of 30 degrees, it is determined that it belongs to the same directional cluster group.

[0071] After obtaining preliminary clusters with similar orientations, the system further identifies characteristic regions exhibiting a radial convergence pattern to reflect potential concentrated deformation or instability trends. Specifically, this is achieved by analyzing the directional extensions of all displacement vectors within each preliminary cluster, calculating the spatial convergence trend of these extensions, and statistically analyzing the geometric characteristics of their convergence centers. When the directional extensions of most (e.g., over 70%) displacement vectors point to the same local geometric center region, and the radius of this region does not exceed 2 meters, the cluster is determined to exhibit a radial convergence pattern. Radial convergence typically corresponds to deformation characteristics such as surface fracturing, concentrated slippage, or localized collapse, and is an important basis for identifying potential instability zones. During the identification process, spatial angle statistics and center aggregation degree analysis automatically determine the convergence intensity of different clusters, and clusters exhibiting radial convergence characteristics are merged a second time to form more physically meaningful clustered distribution areas, thereby macroscopically describing the concentrated direction and range of slope deformation energy.

[0072] After completing directional clustering and convergence identification, the spatial characteristics of all clustered distribution areas are used to delineate boundaries. Boundary delineation uses the spatial distribution density of risk deformation data as the core indicator, employing a density threshold method to determine the outer extent of potential instability zones. Specifically, the clustered areas are divided into regular spatial grid cells, with the side length of each cell ranging from 0.5 meters to 1 meter based on the average spacing between feature points. The number of risk deformation feature points within each grid cell is counted; when the feature point density exceeds twice the average density of the entire region, the cell is marked as a high-density area. Then, centered on the high-density cell, the density is gradually expanded to areas where the density gradually decreases to the threshold edge, forming a continuous spatial contour line. This contour line represents the boundary range of the potential instability zone. Finally, the boundary range is superimposed onto the 3D reality model in vector form, along with the cluster number and convergence center location of each instability zone.

[0073] In step S6, construction information data corresponding to the spatial range of potential instability zones are statistically analyzed to generate a forward-looking construction risk classification and control signal.

[0074] Acquire current construction information data from the construction site. This data originates from on-site monitoring and management records, including the real-time location distribution of construction personnel, the operating location and status of machinery and equipment, and the static spatial coordinates of key structures (such as slope protection structures, drainage ditches, retaining walls, and slope protection piles). The real-time location of construction personnel is obtained through wearable positioning devices or on-site marker nodes. Each personnel's location is recorded in two-dimensional or three-dimensional coordinates, typically updated once per minute. The location information of machinery and equipment is collected synchronously through built-in positioning modules and operational status sensors to ensure that location information matches operational conditions. The static location information of key structures is provided by construction drawings or on-site survey data, with a relatively low update frequency, but maintaining consistent spatial accuracy. All data from these different sources are then uniformly converted to the same spatial coordinate system as the 3D reality model, forming a standardized construction information dataset.

[0075] Spatial overlay analysis is performed between the spatial boundary of the potential instability zone and construction information data. The overlay analysis uses the intersection relationship in three-dimensional space as the criterion to determine whether construction personnel, machinery, or critical structures are within the potential instability zone. Specific steps include: first, calculating the spatial envelope of the potential instability zone boundary; then, sequentially traversing each spatial record in the construction information dataset to determine whether its location coordinates fall within the envelope or the boundary region. When a location point is located within the potential instability zone, the record is included in the "affected data set." The affected data set is grouped by type, including three categories of data: personnel, equipment, and structures. For construction personnel, the number of personnel falling into the instability zone and their corresponding job assignments are counted; for machinery, the equipment type, tonnage, and operating status are recorded; for critical structures, their structural type and spatial orientation are identified.

[0076] Based on the spatial extent of the potential instability zone and the quantity and type of affected construction information data, a risk assessment factor set is constructed, and a comprehensive risk value is calculated accordingly. The risk assessment factor set includes three main dimensions: first, a spatial impact factor, reflecting the area, volume, and proportion of the potential instability zone relative to the construction area; second, a resource impact factor, used to quantify the number of affected personnel, equipment value, and the importance level of key structures; and third, a dynamic impact factor, used to reflect the activity level of changes in construction status, such as equipment operation density or personnel concentration. Each factor is assigned a weight based on on-site safety regulations and historical monitoring experience. The weights can be determined based on empirical data, such as a weight of 0.4 for the spatial impact factor, 0.35 for the resource impact factor, and 0.25 for the dynamic impact factor. After standardizing each factor, a weighted sum is obtained to obtain the comprehensive risk value of the potential instability zone. The comprehensive risk value is then mapped to a predefined risk level range, which is divided into four levels: low risk (0~0.25), medium risk (0.25~0.5), high risk (0.5~0.75), and extremely high risk (above 0.75). When the overall risk value falls into different ranges, the system automatically generates corresponding level of forward-looking construction control signals. These signals are linked to the on-site dispatch system and are used to indicate different levels of control measures, such as suspending construction, relocating equipment, or strengthening monitoring within the risk area, thereby achieving real-time graded response and forward-looking intervention to potential risks.

[0077] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0078] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0079] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0080] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0081] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0082] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0083] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0084] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0085] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-dimensional construction safety risk management method for high-altitude highways, characterized in that, Includes the following steps: S1. Collect aerial survey images and laser point cloud data of the construction slope area at fixed intervals to generate a time-series 3D real scene model sequence corresponding to each collection time point; S2. In the sequence of three-dimensional real-scene models, identify and track the natural feature points and artificial structure feature points on the surface of each phase of the model; S3. By registering the accuracy between multiple models, calculate the three-dimensional coordinate change of each feature point at continuous acquisition time points, and construct a feature point displacement field covering the entire slope area. S4. Based on historical construction activity records and the spatiotemporal characteristics of displacement field, identify feature points in the displacement field that are spatiotemporally related to construction activities, and define deformation data that exhibits continuous acceleration or spatial diffusion as risk deformation data. S5. Screen the spatially clustered distribution areas of risk deformation data and mark them as potential instability areas; S6. Based on the spatial range of potential instability zones, statistically analyze the corresponding construction information data to generate forward-looking construction risk classification and control signals; In step S2, identifying and tracking natural feature points and artificial structure feature points on the surface of each model in the three-dimensional real-scene model sequence specifically includes: On the surface of the 3D real scene model, prominent bedrock edges and landform turning points are selected as natural feature points, and concrete structure corners in completed projects are selected as artificial structure feature points. A feature description vector is constructed based on the three-dimensional spatial coordinates of feature points and surface texture information. The feature description vector contains the geometric shape parameters and texture distribution parameters of the feature points. In the sequence of 3D real scene models, the spatial position of the feature point in the 3D real scene model of the previous acquisition cycle is taken as the center, and the point with the highest similarity of the feature description vector is found within the preset search radius to perform cross-period tracking and matching of feature points. In step S5, screening spatially clustered areas of risk deformation data and marking them as potential instability zones specifically includes: Spatial clustering analysis is performed on the risk deformation data. Based on the relative displacement vector direction corresponding to the deformation, feature points with similar displacement vector directions or spatial distribution in a radial convergence pattern are grouped into the same clustered distribution area. Based on the spatial extent of the clustered distribution area and the spatial density of risk deformation data, the boundary range of the potential instability zone is delineated and marked.

2. The multi-dimensional construction safety risk management method for high-altitude highways according to claim 1, characterized in that, In step S1, the aerial survey images and laser point cloud data of the construction slope area are collected at fixed intervals to generate a time-series three-dimensional real-scene model sequence corresponding to each collection time point, specifically including: Aerial survey imagery and laser point cloud data of the construction slope area are collected at a preset fixed period. The aerial survey image data includes multi-angle overlapping images covering the entire slope area, and the laser point cloud data includes a set of three-dimensional spatial coordinate points evenly distributed in the slope area. A high-resolution digital surface model is generated based on aerial survey image sequences. The laser point cloud data is registered and fused with the digital surface model. The coordinate system of the laser point cloud data is mapped to the coordinate system of the digital surface model to construct a three-dimensional model framework. The surface texture information extracted from the aerial survey image data collected in each cycle is mapped onto the surface of the 3D model framework, and a time-series 3D real scene model sequence is generated through texture mapping.

3. The multi-dimensional construction safety risk management method for high-altitude highways according to claim 1, characterized in that, In step S3, through precision registration between multiple models, the three-dimensional coordinate change of each feature point at continuous acquisition time points is calculated to construct a feature point displacement field covering the entire slope area. Specifically, this includes: Using the 3D real-scene model corresponding to the previous acquisition cycle as the spatial reference benchmark, calculate the spatial transformation parameter sequence of the feature points in the 3D real-scene model sequence that are tracked and matched across periods relative to the spatial reference benchmark. Numerical statistical analysis is performed on the spatial transformation parameters to identify stable feature point groups that are concentrated in the numerical value of the transformation parameters, and prominent feature points whose deviation from the transformation parameters and stable feature point groups exceeds a set threshold are selected. The coordinate sequence of prominent feature points is extracted based on transformation parameters in a unified coordinate system, wherein the unified coordinate system is the spatial coordinate system of the three-dimensional real scene model in the first acquisition cycle; The planar displacement components and elevation displacement components of all prominent feature points in each acquisition cycle are obtained and arranged in chronological order to construct a feature point displacement field covering the entire slope area.

4. The multi-dimensional construction safety risk management method for high-altitude highways according to claim 1, characterized in that, In step S4, based on historical construction activity records and the spatiotemporal characteristics of the displacement field, feature points in the displacement field that are spatiotemporally related to construction activities are identified. Deformation data exhibiting continuous acceleration or spatial diffusion is defined as risk deformation data, specifically including: The construction area and construction time recorded in the historical construction activity records are spatiotemporally superimposed with the spatial distribution and change time of prominent feature points in the displacement field for analysis. Feature points whose displacement changes are synchronized with the start time of construction activities in time and overlap with the area of ​​construction activities in space are identified as having a spatiotemporal correlation with construction activities. Displacement changes of feature points that are spatiotemporally related to construction activities are converted into deformation data, and deformation data that exhibits continuous acceleration or spatial diffusion over time are defined as risk deformation data.

5. A multi-dimensional construction safety risk management method for high-altitude highways according to claim 4, characterized in that, The risk deformation data consists of a sequence of relative displacement vectors between the geometric centers of the corresponding prominent feature points and the pre-labeled set of adjacent feature points; the adjacent feature points are determined based on pre-defined engineering zoning boundaries or geological structural boundaries.

6. The multi-dimensional construction safety risk management method for high-altitude highways according to claim 1, characterized in that, The similarity criterion for displacement vector directions is that the angle between the displacement vector direction and the average direction of all displacement vectors is less than a set angle threshold, and the criterion for the radial convergence pattern is that the directions of all displacement vectors point to the same geometric center region.

7. The multi-dimensional construction safety risk management method for high-altitude highways according to claim 1, characterized in that, In step S6, generating a forward-looking construction risk classification and control signal based on the construction information data corresponding to the spatial range of the potential instability zone specifically includes: Acquire construction information data, which includes the real-time location distribution of construction personnel and machinery and equipment, as well as the static location information of key structures. Spatial overlay analysis of the spatial boundary of the potential instability zone and construction information data is performed to count the quantity and type of various construction information data falling into the potential instability zone. Based on the spatial extent of the potential instability zone and the value and quantity of affected construction information data, a set of risk assessment factors is constructed, and a comprehensive risk value is calculated. Based on the predefined range of the comprehensive risk value, forward-looking construction control signals corresponding to different risk levels are generated.

Images