Deep foundation pit intelligent safety monitoring method and system for intelligent construction site

Abstract

The invention discloses a deep foundation pit intelligent safety monitoring method for an intelligent construction site, and relates to the field of safety monitoring, and the method comprises the steps: collecting a plurality of pieces of foundation pit point cloud data of a target foundation pit; marking a point cloud vacant area; collecting regional point cloud data of the point cloud vacant region, and combining the regional point cloud data and all the foundation pit point cloud data to construct a foundation pit three-dimensional model; constructing a foundation pit map network, and determining a foundation pit key monitoring area according to the foundation pit map network; generating a foundation pit monitoring strategy of the target foundation pit; obtaining foundation pit monitoring data of the target foundation pit; updating the three-dimensional model of the foundation pit in real time; and risk evaluation is carried out on the target foundation pit, and a risk evaluation result is visually displayed through the foundation pit three-dimensional model. According to the invention, continuous risk monitoring and risk visual display of the foundation pit can be effectively realized.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of safety monitoring, and in particular to a method and system for intelligent safety monitoring of deep foundation pits for smart construction sites. Background Art

[0002] According to statistics on construction accidents in recent years, the probability of engineering accidents occurring in foundation pit projects is much higher than that in main projects, causing significant economic losses and serious consequences to society. Therefore, it is particularly important to carry out safety monitoring of deep foundation pits.

[0003] The existing technology for safety monitoring of deep foundation pits mainly relies on manual monitoring, which uses manual labor to regularly observe and record data at monitoring points, and then manually process and analyze the data. This method is not only inefficient, but also the manual analysis of data is easily affected by subjective factors, resulting in analysis errors, which has serious consequences. In addition, the inability to continuously monitor deep foundation pits manually will also affect the timeliness of the monitoring results, which may cause foundation pit safety accidents, and in extreme weather, monitoring work will be difficult to carry out. Some projects will also use automated instruments to continuously observe and record monitoring points, but still rely on manual data analysis. However, for some large foundation pits, there are many monitoring points that need to be monitored, and the monitoring data is complicated. Relying solely on manual data analysis is extremely wasteful of human resources, and because the monitoring results cannot intuitively display the monitoring location corresponding to the monitoring results, it is not conducive to subsequent risk elimination work. Summary of the Invention

[0004] The embodiments of the present application provide a method and system for intelligent safety monitoring of deep foundation pits for smart construction sites, which are used to solve the problem that the existing technology is difficult to achieve continuous risk monitoring and risk visualization of foundation pits.

[0005] To achieve the above objectives, the embodiments of the present application adopt the following technical solutions:

[0006] In a first aspect, a method for intelligent safety monitoring of deep foundation pits for smart construction sites is provided, the method comprising:

[0007] Use detection drones to collect multiple foundation pit point cloud data of the target foundation pit;

[0008] Position matching is performed on all the foundation pit point cloud data with the pre-acquired foundation pit plan construction drawing to obtain a point cloud position matching result, and point cloud vacant areas in the foundation pit plan construction drawing are marked according to the point cloud position matching result;

[0009] Controlling the unmanned detection vehicle to collect regional point cloud data of the point cloud vacancy area, and combining the regional point cloud data with all the foundation pit point cloud data to construct a three-dimensional foundation pit model of the target foundation pit;

[0010] Constructing a foundation pit map network according to pre-acquired foundation pit risk monitoring specifications, and determining a key monitoring area of ​​the target foundation pit according to the foundation pit map network;

[0011] Generating a foundation pit monitoring strategy for the target foundation pit by combining the foundation pit key monitoring area and the foundation pit three-dimensional model;

[0012] Arranging different types of monitoring instruments at various monitoring points of the target foundation pit according to the foundation pit monitoring strategy, and using the monitoring instruments to monitor the target foundation pit in real time to obtain foundation pit monitoring data of the target foundation pit;

[0013] Position matching is performed between the foundation pit monitoring points corresponding to the foundation pit monitoring data and the foundation pit three-dimensional model, and the foundation pit three-dimensional model is updated in real time using the foundation pit monitoring data according to the position matching result;

[0014] A risk assessment is performed on the target foundation pit based on the foundation pit monitoring data, and the risk assessment result is visualized through the foundation pit three-dimensional model.

[0015] Optionally, performing position matching on all the foundation pit point cloud data with a pre-acquired foundation pit plan construction drawing to obtain a point cloud position matching result, and marking point cloud vacant areas in the foundation pit plan construction drawing according to the point cloud position matching result comprises the following steps:

[0016] Preprocessing all the foundation pit point cloud data;

[0017] Perform preliminary perspective alignment on all the pre-processed foundation pit point cloud data using a feature matching algorithm;

[0018] aligning the coordinate systems of all the foundation pit point cloud data using an iterative optimization algorithm to obtain a plurality of standard foundation pit point cloud data;

[0019] Performing point cloud vacancy identification on all the standard foundation pit point cloud data to determine the locations of multiple point cloud vacancy areas;

[0020] The positions of all the missing areas of the point cloud are mapped to the pre-acquired foundation pit plan construction drawing to determine the missing areas of the point cloud.

[0021] Optionally, performing point cloud vacancy identification on all the standard foundation pit point cloud data to obtain multiple point cloud vacancy area locations includes the following steps:

[0022] For each of the standard foundation pit point cloud data, the standard foundation pit point cloud data is projected at multiple angles to obtain a plurality of multi-directional foundation pit point cloud data;

[0023] Gridding all the two-dimensional foundation pit point cloud data to obtain a plurality of grid areas;

[0024] Calculating the grid density in the grid area;

[0025] If there is a sparse grid area in all the grid areas where the grid density is less than a preset density threshold, analyzing the number of grids in all the sparse grid areas that are connected by edges;

[0026] If the number of grids is greater than a preset threshold, all the sparse grid areas are acquired and integrated to obtain a point cloud sparse area;

[0027] The center point coordinates of all the sparse areas of the point cloud are obtained, and the positions of all the vacant areas of the point cloud are determined according to the center point coordinates.

[0028] Optionally, controlling the unmanned detection vehicle to collect regional point cloud data of the point cloud missing area, and combining the regional point cloud data with all the foundation pit point cloud data to construct a three-dimensional foundation pit model of the target foundation pit includes the following steps:

[0029] Performing point cloud splicing on all the standard foundation pit point cloud data to obtain target foundation pit point cloud data;

[0030] generating an unmanned vehicle detection route for the detection unmanned vehicle by combining the point cloud vacant area and the foundation pit remote sensing information of the target foundation pit;

[0031] Controlling the detection unmanned vehicle to collect regional point cloud data of the point cloud vacant area according to the detection route of the unmanned vehicle;

[0032] Preprocessing the regional point cloud data;

[0033] Aligning the coordinate system of the pre-processed regional point cloud data with the target foundation pit point cloud data as a reference;

[0034] Filling the regional point cloud data with the completed coordinate system alignment into the target foundation pit point cloud data to obtain the complete foundation pit point cloud data of the target foundation pit;

[0035] Performing surface fitting on the complete foundation pit point cloud data to obtain multiple fitting surfaces;

[0036] The plurality of fitting curved surfaces are connected according to the edge point cloud coordinates of the fitting curved surfaces to obtain a three-dimensional foundation pit model of the target foundation pit.

[0037] Optionally, the step of generating a detection route for an unmanned vehicle by combining the point cloud missing area and the foundation pit remote sensing information comprises the following steps:

[0038] Generating the shortest detection route for the unmanned vehicle in the foundation pit plan construction drawing according to the vacant area of ​​the point cloud and using the shortest path algorithm;

[0039] Preprocessing the foundation pit remote sensing information;

[0040] Performing spectral analysis on all the pre-processed foundation pit remote sensing information, and calculating a water body index of the foundation pit remote sensing information based on the spectral analysis results;

[0041] Marking the area in the foundation pit planned construction drawing where the water body index is greater than a preset index threshold as a foundation pit water body area;

[0042] Position matching is performed on the target foundation pit point cloud data and the foundation pit planned construction drawing, and base point cloud data in the target foundation pit point cloud data is obtained according to the foundation pit planned construction drawing;

[0043] Performing surface fitting on the base point cloud data to obtain a base fitting surface of the base area of ​​the target foundation pit;

[0044] Calculating the distance differences between all the base point clouds and the base fitting surface in all the base point cloud data;

[0045] Screening out the base point clouds whose distance difference is greater than a preset difference threshold to obtain a plurality of base abnormal point clouds;

[0046] Fitting all the basement abnormal point clouds to obtain the foundation pit deposition area;

[0047] Analyzing whether there is any overlap between the shortest detection route and the foundation pit water area and the foundation pit sediment area; if no overlap exists, using the shortest detection route as the unmanned vehicle detection route of the detection unmanned vehicle;

[0048] If the overlapping phenomenon exists, the foundation pit water area and the foundation pit sedimentation area are used as constraints, and the shortest detection route is optimized according to the constraints to obtain the unmanned vehicle detection route of the detection unmanned vehicle.

[0049] Optionally, constructing a foundation pit map network according to a pre-acquired foundation pit risk monitoring specification, and acquiring a foundation pit key monitoring area of ​​the target foundation pit according to the foundation pit map network includes the following steps:

[0050] Perform semantic analysis on the pre-acquired foundation pit risk monitoring specifications to obtain semantic analysis results;

[0051] Extracting key foundation pit construction items and foundation pit monitoring data types in the foundation pit risk monitoring specification from the semantic analysis results, and obtaining semantic association relationships between the key foundation pit construction items and the foundation pit monitoring data types based on the semantic analysis results;

[0052] The key foundation pit construction projects and the types of foundation pit monitoring data are used as graph nodes, and the semantic association relationships are used as node edges between the graph nodes to obtain a foundation pit knowledge graph;

[0053] Analyze the topological relationship and graph knowledge of the foundation pit knowledge graph, obtain key graph nodes in the foundation pit knowledge graph according to the analysis results, and use the key construction items of the foundation pit corresponding to the key graph nodes as key monitoring locations;

[0054] Performing structural matching between the key monitoring position and the three-dimensional model of the foundation pit to obtain the key monitoring position of the foundation pit in the three-dimensional model of the foundation pit;

[0055] The key monitoring position of the foundation pit, the water area of ​​the foundation pit and the sedimentation area of ​​the foundation pit are used as the key monitoring areas of the target foundation pit.

[0056] Optionally, performing structural matching between the key monitoring position and the three-dimensional model of the foundation pit to obtain the key monitoring position of the foundation pit in the three-dimensional model of the foundation pit comprises the following steps:

[0057] Construct a three-dimensional structure recognition model based on a neural network model;

[0058] Using a pre-built foundation pit structure training set to train the three-dimensional structure recognition model;

[0059] Using the trained three-dimensional structure recognition model to perform model structure recognition on the three-dimensional foundation pit model to obtain the foundation pit model structure of the three-dimensional foundation pit model;

[0060] Performing semantic recognition on the foundation pit model structure and the key monitoring position, and performing semantic matching on the foundation pit model structure and the key monitoring position according to the semantic recognition result to obtain a semantic matching result;

[0061] According to the semantic matching result, the foundation pit model structure corresponding to the key monitoring position is used as the foundation pit key monitoring position in the foundation pit three-dimensional model.

[0062] Optionally, the foundation pit monitoring strategy for generating the target foundation pit by combining the foundation pit key monitoring area and the foundation pit three-dimensional model includes the following steps:

[0063] Calculating the projected area of ​​the key monitoring area of ​​the foundation pit on the horizontal plane according to the three-dimensional model of the foundation pit;

[0064] Generating the number of foundation pit monitoring points in the key monitoring area of ​​the foundation pit according to the projected area;

[0065] According to the number of the foundation pit monitoring points corresponding to the foundation pit key monitoring area, a corresponding number of foundation pit monitoring point positions are set in the foundation pit key monitoring area;

[0066] A foundation pit monitoring strategy for the target foundation pit is generated based on the number of foundation pit monitoring points and the locations of the foundation pit monitoring points.

[0067] Optionally, after visually displaying the risk assessment results through the foundation pit three-dimensional model, the following steps are further included:

[0068] Storing the foundation pit monitoring data of all the foundation pit monitoring points as historical foundation pit monitoring data;

[0069] Matching the foundation pit monitoring data with the foundation pit monitoring points, and marking the foundation pit monitoring points with missing information as abnormal foundation pit monitoring points;

[0070] Acquiring the historical foundation pit monitoring data corresponding to the abnormal foundation pit monitoring point, and predicting the monitoring data of the abnormal foundation pit monitoring point based on the corresponding historical foundation pit monitoring data to obtain predicted monitoring data;

[0071] The predicted monitoring data is used as the monitoring data of the abnormal foundation pit monitoring point for data filling to obtain the complete foundation pit monitoring data of the target foundation pit, and the risk assessment of the target foundation pit is performed based on the complete foundation pit monitoring data.

[0072] In a second aspect, the present application provides a deep foundation pit intelligent safety monitoring system for a smart construction site, characterized by comprising:

[0073] a memory configured to store instructions; and

[0074] The processor is configured to call the instructions from the memory and to implement a method for intelligent safety monitoring of deep foundation pits in smart construction sites according to any one of the first aspects when executing the instructions.

[0075] Through the above technical solution, a detection drone is used to collect the target foundation pit point cloud data, the foundation pit point cloud data is analyzed to determine the point cloud vacancy area, and the detection unmanned vehicle is used to collect data in the point cloud vacancy area, so that the point cloud data used to construct the foundation pit three-dimensional model is more complete and accurate. The constructed foundation pit three-dimensional model can also more accurately reflect the geometric structure of the target foundation pit, facilitating the subsequent generation of foundation pit monitoring strategies for the target foundation pit based on the foundation pit three-dimensional model and the visualization of risk assessment results. Combining foundation pit remote sensing information and foundation pit point cloud data to identify the foundation pit water area and foundation pit sedimentation area in the target foundation pit, a smoother route can be planned for the detection drone, allowing the detection unmanned vehicle to more quickly reach the point cloud vacancy area for detection tasks. The regional point cloud data obtained by the detection unmanned vehicle can then be used to fill the foundation pit point cloud data obtained by the detection drone, obtaining complete foundation pit point cloud data, and using the complete foundation pit point cloud data to construct a more accurate foundation pit three-dimensional model. In addition, building a foundation pit knowledge graph based on the foundation pit risk monitoring specifications can more objectively and accurately determine the key monitoring locations that need to be monitored, making the subsequent foundation pit monitoring data more reliable and comprehensive. In summary, this application provides a more comprehensive and accurate method for real-time monitoring of target mining areas, and can visualize the foundation pit risk assessment results, making it easier for foundation pit staff to conduct risk investigations.

[0076] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 A schematic diagram of a process flow of a deep foundation pit intelligent safety monitoring method for a smart construction site provided in an embodiment of the present application;

[0078] Figure 2 A schematic diagram of a process for marking vacant areas in a point cloud provided in an embodiment of the present application. DETAILED DESCRIPTION

[0079] To make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0080] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indications will also change accordingly.

[0081] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the fact that they can be implemented by ordinary technicians in this field. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0082] Figure 1 The following schematically shows a flow chart of a method for intelligent safety monitoring of deep foundation pits in a smart construction site according to an embodiment of the present application. Figure 1 As shown, the embodiment of the present application provides a method for intelligent safety monitoring of deep foundation pits for smart construction sites, which may include the following steps:

[0083] S101. Collect multiple foundation pit point cloud data of the target foundation pit using a detection drone.

[0084] In this embodiment, the detection drone is equipped with a laser radar module, which includes a laser transmitter and a receiver. The laser transmitter emits a beam of light pulses. When these light pulses encounter the surface of an object, they are reflected back and received by the receiver. The receiver measures the time from the emission to the reflection of the light pulse, and combined with the speed of light, the distance between the object and the sensor can be calculated. By recording multiple such measurement results, the foundation pit point cloud data of multiple different areas of the target foundation pit is obtained. In addition, the detection drone is also equipped with a remote sensing module, which includes visible light cameras, infrared cameras, radars, hyperspectral cameras and other equipment. While the detection drone obtains the foundation pit point cloud data of the target foundation pit, it uses infrared cameras, hyperspectral cameras and other equipment to obtain the foundation pit remote sensing information of the target foundation pit.

[0085] S102 , position matching is performed on all foundation pit point cloud data with the pre-acquired foundation pit plan construction drawing to obtain a point cloud position matching result, and point cloud vacant areas in the foundation pit plan construction drawing are marked according to the point cloud position matching result.

[0086] In this embodiment, the planned foundation pit construction drawings are pre-designed by relevant staff based on the target foundation pit's geographic location and surrounding environment before construction of the target foundation pit. The target foundation pit is also excavated according to the planned foundation pit construction drawings. All foundation pit point cloud data is subjected to denoising and outlier removal, followed by preliminary perspective alignment and coordinate system alignment. This is because the drone's flight angle may vary when collecting point cloud data from different locations during the acquisition of the target foundation pit, in order to more accurately capture the target pit's point cloud data. Therefore, perspective alignment and coordinate system alignment are necessary. Preliminary perspective alignment can be performed using feature matching algorithms, point cloud matching algorithms, etc., while coordinate system alignment can be performed using iterative optimization algorithms, Horn quaternion closure solvers, and Marta's iterative least squares solver. After completing the preliminary perspective alignment and coordinate system alignment, multiple standard foundation pit point cloud data sets are obtained. Using the standard foundation pit point cloud data, areas where the drone failed to capture point cloud data, i.e., areas with missing point cloud data, can be more accurately determined. Specifically, the standard foundation pit point cloud data is mapped to different directions around it. Based on the point cloud density of the standard foundation pit point cloud data in different directions around it, it is determined whether there are any point cloud gaps in the standard foundation pit point cloud data, as well as the locations of the point cloud gaps in different directions. Based on the locations of the point cloud gaps in different directions, the locations of the point cloud gap areas in the standard foundation pit point cloud data are determined. By mapping the locations of all point cloud gap areas to the foundation pit planned construction drawing, the locations of the point cloud gap areas in the foundation pit planned construction drawing can be obtained, namely the point cloud gap areas, which facilitates the subsequent secondary detection of the target foundation pit by the unmanned vehicle based on the foundation pit planned construction drawing.

[0087] S103 , controlling the detection unmanned vehicle to collect regional point cloud data of the point cloud vacant areas, and combining the regional point cloud data with all foundation pit point cloud data to construct a three-dimensional foundation pit model of the target foundation pit.

[0088] In this embodiment, a shortest path algorithm is used to generate the shortest detection route connecting all missing areas in the point cloud. Common shortest path algorithms include the Dijkstra algorithm, the Bellman-Ford algorithm, and the Floyd algorithm. Based on the remote sensing information of the foundation pit, the presence of waterlogged areas, i.e., water bodies, within the foundation pit is analyzed. Simultaneously, based on the target foundation pit point cloud data, the presence of sedimentary areas at the bottom of the target foundation pit is determined. Sedimentary areas include deep pits at the bottom of the foundation pit caused by manual excavation or pit collapse, as well as piles of soil created by excavation and raised areas caused by pit uplift. The target foundation pit point cloud data refers to the relatively complete point cloud obtained after the point cloud of all standard foundation pit point clouds is completed. If water bodies or sedimentary areas exist along the shortest detection route, the detection drone will be unable to travel smoothly, thereby affecting the detection mission of the unmanned vehicle. Therefore, the unmanned vehicle should avoid water bodies and sedimentary areas as much as possible when traveling to the missing areas in the point cloud. The specific method is to mark the foundation pit water area and foundation pit sedimentation area on the foundation pit construction plan, and analyze whether there is any overlap between the shortest detection route and the foundation pit water area and / or foundation pit sedimentation area. If there is overlap, it means that if the detection unmanned vehicle follows the shortest detection route, it will pass through the foundation pit settlement area and the foundation pit water area. This may make it difficult for the detection unmanned vehicle to successfully drive to the point cloud vacant area, thus affecting the detection task, so the route needs to be adjusted. The shortest detection route can be optimized using algorithms such as the gray wolf optimization algorithm, the ant colony algorithm, and dynamic programming. The optimized detection route is used as the unmanned vehicle detection route for the detection unmanned vehicle. If there is no overlap, it means that the detection unmanned vehicle will not pass through the foundation pit settlement area and the foundation pit water area when traveling along the optimized unmanned vehicle detection route, so there is no need to further optimize the detection route.

[0089] The unmanned detection vehicle is controlled to follow its detection route to all areas with missing point clouds, acquiring regional point cloud data. The regional point cloud data undergoes geometric correction, image denoising, and radiometric calibration. The point cloud data of the standard foundation pit point cloud is spliced ​​to obtain the target foundation pit point cloud data. The regional point cloud data is aligned with the target foundation pit point cloud data before being spliced ​​with the target foundation pit point cloud data to obtain the complete target foundation pit point cloud data. Surface fitting is performed on the complete foundation pit point cloud data to obtain multiple fitted surfaces. The coordinates of the edge points of these multiple fitted surfaces are obtained. Based on the edge point coordinates, all fitted surfaces are translated and rotated, aligning and connecting the edge points of the different fitted surfaces in the same coordinate system to obtain a 3D model of the target foundation pit.

[0090] The purpose of constructing the three-dimensional model of the foundation pit is to facilitate the subsequent design of the foundation pit monitoring strategy based on the actual structural characteristics of the target foundation pit point cloud. At the same time, it can also more intuitively display the foundation pit risk levels in different areas of the target foundation pit, making it easier for relevant staff of the target foundation pit to obtain risk information of the target foundation pit.

[0091] S104: construct a foundation pit map network according to the pre-acquired foundation pit risk monitoring specification, and determine the key monitoring area of ​​the target foundation pit according to the foundation pit map network.

[0092] In this embodiment, the foundation pit risk monitoring specification includes the monitoring equipment, monitoring methods, monitoring items, and monitoring data used in the foundation pit monitoring process. A semantic analysis of the foundation pit risk monitoring specification is performed to obtain key content within the specification, namely, key foundation pit construction items and types of foundation pit monitoring data, as well as the semantic associations between the two. Key foundation pit construction items primarily include foundation pit slope areas, foundation pit support structures, and surrounding buildings. Types of foundation pit monitoring data include groundwater level, horizontal displacement, vertical displacement, internal force, axial force, and inclination. Semantic associations refer to the correspondence between key foundation pit construction items and types of foundation pit monitoring data, such as the groundwater level at the bottom of the foundation pit, vertical displacement at the bottom of the foundation pit, internal force of foundation pit pillars, axial force of foundation pit anchor rods, horizontal displacement of the foundation pit slope, and crack width of the foundation pit slope. Key foundation pit construction items and types of foundation pit monitoring data are used as graph nodes, and semantic associations are used as node edges between the graph nodes to obtain a foundation pit knowledge graph. According to the knowledge graph, the node degree and node edge weight of each graph node are obtained, and the key graph nodes in the knowledge graph are screened out based on the node degree and node edge weight. The key construction projects of the foundation pit corresponding to the key graph nodes are used as key monitoring positions. The names of the various structures in the foundation pit three-dimensional model, i.e., the foundation pit model structure, are identified using a pre-built three-dimensional structure recognition model. The foundation pit model structure is semantically matched with the foundation pit model structure, and the foundation pit model structure corresponding to the key monitoring position is used as the foundation pit key monitoring position in the foundation pit three-dimensional model based on the semantic matching results. This step is mainly to obtain the positions that need to be monitored in the foundation pit monitoring process in a more scientific and objective manner, such as foundation pit slopes, foundation pit pillars, etc., and then find the foundation pit model structures corresponding to the foundation pit slopes and foundation pit pillars in the foundation pit three-dimensional model, so as to facilitate the subsequent installation of various monitoring instruments in the target foundation pit according to the foundation pit three-dimensional model.

[0093] S105. Generate a foundation pit monitoring strategy for the target foundation pit by combining the foundation pit key monitoring area and the foundation pit three-dimensional model.

[0094] In this embodiment, the projection area of ​​the key monitoring area of ​​the foundation pit on the horizontal plane is calculated based on the three-dimensional model of the foundation pit, and different numbers of foundation pit monitoring points are generated according to the different projection areas. Specifically, the area with a large projection area generates more foundation pit monitoring points, and vice versa. In addition, for the key monitoring areas of the foundation pit with a small horizontal projection area such as foundation pit pillars and soil anchor rods, their vertical projection areas will be additionally calculated, and the horizontal projection areas will be used as an auxiliary to generate the corresponding number of foundation pit monitoring points. According to the number of foundation pit monitoring points, a corresponding number of foundation pit monitoring point positions are generated (which can be evenly distributed), and the foundation pit monitoring strategy of the target foundation pit is generated in combination with the number of foundation pit monitoring points and the foundation pit monitoring point positions. Subsequently, the relevant staff of the target foundation pit will arrange different types of monitoring instruments according to the foundation pit monitoring strategy to achieve all-round and high-precision monitoring of the target foundation pit.

[0095] S106. Arrange different types of monitoring instruments at each foundation pit monitoring point of the target foundation pit according to the foundation pit monitoring strategy, use the monitoring instruments to perform real-time monitoring of the target foundation pit, and obtain foundation pit monitoring data of the target foundation pit.

[0096] In this embodiment, the foundation pit monitoring strategy includes the number and location of foundation pit monitoring points corresponding to different key monitoring areas of the target pit, as well as the types of foundation pit monitoring data corresponding to these key monitoring areas. Based on the foundation pit monitoring strategy, different types of monitoring instruments are deployed at the foundation pit monitoring points within each key monitoring area of ​​the target pit. The type of monitoring instrument is selected based on the type of foundation pit monitoring data. For example, an inclinometer is required to monitor the inclination of foundation pit pillars, and a water level gauge is required to monitor the groundwater level at the bottom of the pit. The monitoring instruments installed in each key monitoring area of ​​the target pit perform comprehensive, real-time monitoring of the target pit. The soil pressure, inclination, and water level data obtained from the monitoring are the target pit monitoring data.

[0097] S107: Positionally match the foundation pit monitoring points corresponding to the foundation pit monitoring data with the foundation pit three-dimensional model, and update the foundation pit three-dimensional model in real time using the foundation pit monitoring data according to the position matching result.

[0098] In this embodiment, the foundation pit monitoring points corresponding to the foundation pit monitoring data are matched with the foundation pit three-dimensional model according to the positions of the foundation pit monitoring points to obtain the structural positions corresponding to the foundation pit monitoring data on the foundation pit three-dimensional model. The foundation pit monitoring data is mapped to the foundation pit three-dimensional model according to the structural positions, and the foundation pit monitoring data is associated with the corresponding foundation pit structure in the foundation pit three-dimensional model using digital twin technology. The foundation pit three-dimensional model is used as a carrier so that the foundation pit monitoring data can be visualized in the foundation pit three-dimensional model, and the foundation pit three-dimensional model is updated in real time according to the foundation pit monitoring data acquired in real time. Data twin technology is a technology that collects, transmits, processes and analyzes data, performs modeling and simulation, and constructs a digital model that is highly similar to the physical entity in a virtual space, and realizes the monitoring, analysis and optimization of the physical entity through real-time data exchange and analysis.

[0099] S108. Conduct risk assessment on the target foundation pit based on the foundation pit monitoring data, and visualize the risk assessment results through a three-dimensional model of the foundation pit.

[0100] In this embodiment, the risk level of the foundation pit monitoring data is divided by setting different thresholds. For example, if the cumulative displacement value of the foundation pit slope cracks is greater than 10 mm, the risk level of the foundation pit slope is level three. The risk assessment results are visualized through the foundation pit three-dimensional model. Specifically, after the level division is completed, the model color of the corresponding position of the foundation pit three-dimensional model is changed according to the level division result. For example, when the risk level of the foundation pit slope is level three, the area around the foundation pit monitoring point corresponding to the foundation pit slope turns red, turns orange when the risk level is level two, and turns yellow when the risk level is level one. The risk level is the most serious when it is level three. At this time, the foundation pit staff needs to be immediately arranged to go to the location corresponding to the target foundation pit to eliminate the risk. The foundation pit staff can directly determine the location of the area with a risk level of level three in the target foundation pit based on the foundation pit three-dimensional model, which is convenient for the foundation pit staff to check the risk in time.

[0101] In one embodiment, Figure 2 As shown, all foundation pit point cloud data are matched with the pre-acquired foundation pit plan construction drawing to obtain the point cloud position matching result. The point cloud vacant area in the foundation pit plan construction drawing is marked according to the point cloud position matching result, including the following steps:

[0102] S201, pre-processing all foundation pit point cloud data;

[0103] S202, performing preliminary perspective alignment on all pre-processed foundation pit point cloud data using a feature matching algorithm;

[0104] S203, aligning the coordinate systems of all foundation pit point cloud data using an iterative optimization algorithm to obtain multiple standard foundation pit point cloud data;

[0105] S204, performing point cloud vacancy identification on all standard foundation pit point cloud data to determine the locations of multiple point cloud vacancy areas;

[0106] S205: Map the positions of all the missing areas of the point cloud to the pre-acquired foundation pit plan construction drawing to determine the missing areas of the point cloud.

[0107] In this embodiment, all foundation pit point cloud data are subjected to denoising and outlier removal, and then all pre-processed foundation pit point cloud data are subjected to preliminary perspective alignment and coordinate system alignment. This is because in the process of collecting target foundation pits, in order to more accurately collect point cloud data of the target foundation pit, the flight angle of the detection drone may be different when collecting point cloud data at different positions, so perspective alignment and coordinate system alignment are required. Feature matching algorithms, point cloud matching algorithms, etc. can be used for preliminary perspective alignment. The steps of using feature matching algorithms to perform preliminary perspective alignment include: using point cloud key point monitoring algorithms to extract key point clouds from all foundation pit point cloud data, using the ISS algorithm to analyze the local geometric structure of points to select points with significant geometric features in different foundation pit point cloud data as key point clouds, calculating features such as curvature and normal vectors of key point clouds, matching different foundation pit point cloud data according to features such as curvature and normal vectors of key point clouds, and completing preliminary perspective alignment. The steps of aligning the coordinate systems of all foundation pit point cloud data using an iterative optimization algorithm include calculating the positional error between key point clouds after preliminary perspective alignment, calculating the change matrix between key point clouds of different foundation pit point cloud data based on the positional error, and then aligning all foundation pit point cloud data again to obtain multiple standard foundation pit point cloud data. An iterative optimization algorithm is a type of algorithm that gradually approaches the optimal solution through an iterative process. It can minimize or maximize an objective function by continuously updating the model parameters to achieve the optimal solution. Using standard foundation pit point cloud data, areas where the detection drone failed to collect point cloud data, i.e., point cloud gap areas, can be more accurately determined. Specifically, the standard foundation pit point cloud data is mapped to different surrounding directions. Based on the point cloud density of the standard foundation pit point cloud data in different surrounding directions, whether there are point cloud gaps in the standard foundation pit point cloud data and the locations of the point cloud gaps in different directions are determined. Based on the locations of the point cloud gaps in different directions, the locations of the point cloud gap areas in the standard foundation pit point cloud data are determined. The missing areas in the foundation pit are caused by the detection drone being unable to obtain point cloud data for certain areas due to obstruction by tilted foundation pit pillars and other structures during the detection process of the target foundation pit. The detection unmanned vehicle needs to collect point cloud data for these areas again to obtain regional point cloud data. By mapping the locations of all missing areas in the point cloud to the foundation pit plan construction drawing, the location of the missing areas in the point cloud can be determined, i.e., the missing areas in the point cloud. This facilitates the subsequent secondary detection of the target foundation pit by the detection unmanned vehicle based on the foundation pit plan construction drawing.

[0108] In one embodiment, performing point cloud vacancy identification on all standard foundation pit point cloud data to obtain multiple point cloud vacancy area locations includes the following steps:

[0109] For each standard foundation pit point cloud data, the standard foundation pit point cloud data is projected at multiple angles to obtain multiple multi-directional foundation pit point cloud data;

[0110] Gridding is performed on all two-dimensional foundation pit point cloud data to obtain multiple grid areas;

[0111] Calculate the mesh density in the mesh area;

[0112] If there are sparse grid areas with a grid density less than a preset density threshold in all grid areas, the number of grids in all sparse grid areas that have edge connections is analyzed;

[0113] If the number of grids is greater than the preset threshold, all sparse grid areas are obtained and integrated to obtain the point cloud sparse area;

[0114] Get the center point coordinates of all sparse point cloud areas, and determine the locations of all missing point cloud areas based on the center point coordinates.

[0115] In this embodiment, each standard foundation pit point cloud data is projected at multiple angles to obtain multiple multi-directional foundation pit point cloud data. The multiple angles can be the positive direction of the X-axis, the negative direction of the X-axis, the positive direction of the Y-axis, the negative direction of the Y-axis, the positive direction of the Z-axis, and the negative direction of the Z-axis in the three-dimensional coordinate system. Specifically, the maximum and minimum values ​​of the coordinates on the X-axis, Y-axis, and Z-axis in each standard foundation pit point cloud data are found, and the standard foundation pit point cloud data are mapped according to the maximum and minimum values ​​of the coordinates. For example, the maximum value of the X-axis is 9, the minimum value of the X-axis is -6, the three-dimensional coordinates of the foundation pit point cloud are (1, 4, 7), the three-dimensional coordinates after mapping to the positive direction of the X-axis are (9, 4, 7), and the three-dimensional coordinates after mapping to the negative direction of the X-axis are (-6, 4, 7). The multi-directional foundation pit point cloud data is gridded (can be evenly divided) to obtain multiple grid areas, and the number of point clouds in each grid area is calculated. The number of point clouds is divided by the area of ​​the grid area to obtain the grid density. The grids with a grid density greater than the preset density threshold are screened out to obtain sparse grid areas. The sparse grid areas with edge connections are used as adjacent grids. The number of adjacent grids in each sparse grid area is calculated. If the number of grids is greater than the preset number threshold, it means that there is a point cloud vacant area in that direction. Therefore, all sparse grid areas with a number of grids greater than the number threshold are extracted and integrated to obtain a point cloud sparse area. The point cloud sparse area is the point cloud vacant area in the corresponding direction. After obtaining the point cloud sparse areas in all different directions, the coordinates of the center points of all point cloud sparse areas are obtained, and the positions of all point cloud vacant areas are determined based on the coordinates of multiple point cloud center points. For example, a standard foundation pit point cloud data has two point cloud center point coordinates of (6, 8, 2) and (-7, 7, 3). Then the position of the point cloud vacancy area in the standard foundation pit point cloud data is between the range of (6, -7) in the X-axis direction, between the range of (8, 7) in the Y-axis direction, and between the range of (2, 3) in the Z-axis direction.

[0116] In one embodiment, controlling the unmanned detection vehicle to collect regional point cloud data of a point cloud-missing area, and combining the regional point cloud data with all foundation pit point cloud data to construct a three-dimensional foundation pit model of the target foundation pit includes the following steps:

[0117] All standard foundation pit point cloud data are stitched together to obtain the target foundation pit point cloud data;

[0118] Combining the missing areas in the point cloud with the remote sensing information of the target foundation pit to generate a detection route for the unmanned vehicle;

[0119] Control the unmanned vehicle to collect regional point cloud data of the point cloud vacant areas according to the unmanned vehicle detection route;

[0120] Preprocess regional point cloud data;

[0121] Align the coordinate system of the pre-processed regional point cloud data with the target foundation pit point cloud data as the benchmark;

[0122] Fill the regional point cloud data after the coordinate system alignment into the target foundation pit point cloud data to obtain the complete foundation pit point cloud data of the target foundation pit;

[0123] Perform surface fitting on the complete foundation pit point cloud data to obtain multiple fitting surfaces;

[0124] Multiple fitting surfaces are connected according to the edge point cloud coordinates of the fitting surfaces to obtain a three-dimensional model of the target foundation pit.

[0125] In this embodiment, the foundation pit remote sensing information of different areas of the target foundation pit is first spliced ​​according to the foundation pit plan construction drawing. Commonly used point cloud splicing methods include feature matching method, global optimization method and iterative closest point method (ICP). Taking the iterative closest point method as an example, multiple standard foundation pit point cloud data are selected as reference point cloud data, and other standard foundation pit point cloud data are used as target point cloud data. Multiple target point clouds in the target point cloud are selected, and the nearest point cloud is found as the corresponding point for all target point clouds in the reference point cloud data. The transformation parameters such as translation and rotation of all target point clouds are estimated according to the corresponding coordinates, and the target point cloud data is transformed according to the transformation parameters. After repeating the above steps many times, the point cloud splicing is completed.

[0126] A shortest path algorithm is used to generate the shortest detection route connecting all missing areas in the point cloud. Based on the foundation pit remote sensing information, the presence of waterlogged areas, also known as pit water bodies, is analyzed. Simultaneously, based on the target foundation pit point cloud data, the presence of sedimentation areas at the bottom of the target foundation pit is determined. The shortest detection route is analyzed for overlap with the pit water and sedimentation areas. If overlap exists, the unmanned vehicle will pass through the pit settlement and water areas if it follows the shortest detection route. This may hinder the vehicle's ability to reach the missing areas, impacting the detection mission and necessitating route adjustments. The shortest detection route can be optimized using algorithms such as the Gray Wolf Optimization Algorithm, the Ant Colony Algorithm, and dynamic programming. The resulting optimized detection route is then used as the unmanned vehicle's detection route. If no overlap exists, the unmanned vehicle will not pass through the pit settlement and water areas by following the shortest detection route. Therefore, optimization of the shortest detection route is unnecessary and the shortest detection route can be used directly.

[0127] The unmanned detection vehicle is controlled to follow the unmanned vehicle detection route to all the vacant areas of the point cloud to obtain regional point cloud data. The regional point cloud data is processed by denoising and removing outliers. Common denoising and outlier removal methods include radius filtering and Gaussian filtering, which can effectively remove noise points and outliers in the regional point cloud data. After preprocessing, the regional point cloud data is aligned with the target foundation pit point cloud data as the reference. Methods that can be used for coordinate system alignment include the Umeyama (uniform scaling transformation) algorithm, Horn quaternion closure solution algorithm, and iterative least squares solution. Taking the Umeyama algorithm as an example, the least squares method is used to calculate the positional relationship between the regional point cloud data and the target foundation pit point cloud data, thereby obtaining the rotation and translation matrices between the two. The regional point cloud data is rotated and translated according to the rotation and translation matrices to complete the coordinate system alignment. After the coordinate system alignment is completed, the regional point cloud data is filled into the target foundation pit point cloud data. This step is equivalent to performing point cloud splicing between the regional point cloud data and the target foundation pit point cloud data. Methods such as feature matching, global optimization, and iterative closest point (ICP) can also be used to obtain the complete foundation pit point cloud data of the target foundation pit. Surface fitting is performed on the complete foundation pit point cloud data to obtain multiple fitting surfaces. Polynomial regression, spline interpolation, and triangulation algorithms can also be used to obtain the edge point coordinates of multiple fitting surfaces. All fitting surfaces are translated and rotated according to the edge point coordinates so that the edge points of different fitting surfaces are aligned and connected in the same coordinate system, thereby obtaining a three-dimensional foundation pit model of the target foundation pit.

[0128] In one embodiment, generating a detection route for an unmanned vehicle by combining point cloud vacant areas and foundation pit remote sensing information includes the following steps:

[0129] Based on the missing areas in the point cloud and using the shortest path algorithm, the shortest detection route for the unmanned vehicle is generated in the foundation pit construction plan;

[0130] Pre-process foundation pit remote sensing information;

[0131] Perform spectral analysis on all pre-processed foundation pit remote sensing information, and calculate the water index of the foundation pit remote sensing information based on the spectral analysis results;

[0132] Mark the area in the foundation pit plan construction drawing where the water body index is greater than the preset index threshold as the foundation pit water body area;

[0133] Position matching is performed between the target foundation pit point cloud data and the foundation pit planned construction drawing, and the base point cloud data in the target foundation pit point cloud data is obtained according to the foundation pit planned construction drawing;

[0134] Perform surface fitting on the base point cloud data to obtain the base fitting surface of the base area of ​​the target foundation pit;

[0135] Calculate the distance difference between all base point clouds and the base fitting surface in all base point cloud data;

[0136] The base point clouds with distance differences greater than a preset difference threshold are screened out to obtain multiple base abnormal point clouds;

[0137] Fit all basement abnormal point clouds to obtain the foundation pit deposition area;

[0138] Analyze whether there is any overlap between the shortest detection route and the foundation pit water area and foundation pit sediment area. If there is no overlap, use the shortest detection route as the unmanned vehicle detection route;

[0139] If there is an overlap, the foundation pit water area and the foundation pit sedimentation area are used as constraints, and the shortest detection route is optimized according to the constraints to obtain the unmanned vehicle detection route.

[0140] In this embodiment, a shortest path algorithm is used to generate the shortest detection route connecting all the vacant areas of the point cloud. Commonly used shortest path algorithms include Dijkstra's algorithm, Bellman-Ford's algorithm, and Floyd's algorithm. Taking the Dijkstra algorithm as an example, its main steps are: first mark the starting point as a visited node, calculate the distance between the visited node and other unvisited nodes, mark the unvisited node with the shortest distance as a visited node, continue to calculate the distance between the node just marked as visited and other unvisited nodes, repeat the above steps until all nodes are marked as visited nodes, and obtain the shortest detection route. In order to facilitate the generation of the shortest detection route, the vacant areas of the point cloud will be marked in the foundation pit plan construction drawing in advance, and then the shortest detection route will be generated based on the foundation pit plan construction drawing.

[0141] Based on the remote sensing information of the foundation pit, the presence of areas of accumulated water, also known as pit water bodies, is analyzed. Deep pits formed by artificial excavation or ground subsidence can accumulate in the pit if precipitation occurs in the area where the target pit is located. These water-filled pits are known as pit water bodies. If the unmanned detection vehicle passes through a pit water body while traveling to an area missing from the point cloud, the accumulated water may damage the unmanned detection vehicle. Alternatively, if the pit is deep, the unmanned detection vehicle may become trapped and unable to complete the detection mission. Furthermore, if the drainage system of the target foundation pit malfunctions, this can lead to accumulation and seepage of water in the target pit. Water accumulation or seepage at the bottom of the pit can weaken the foundation strength and cause instability. Passing through an area of ​​accumulated water or seepage can cause ground collapse in that area, potentially preventing the detection robot from successfully completing its detection mission. Since these areas of accumulated water cannot be identified using the pit point cloud data, remote sensing information from the pit is required for identification.

[0142] First, remote sensing information from different areas of the target pit is spliced ​​together according to the planned construction drawings. The complete remote sensing information then undergoes geometric correction, image denoising, and radiometric calibration. Geometric correction is the process of eliminating or correcting geometric errors in remote sensing images. These errors are caused by factors such as material deformation, lens distortion, atmospheric refraction, and topographical fluctuations. Image denoising enhances the clarity of remote sensing information. Radiometric calibration converts the brightness grayscale values ​​of remote sensing information into absolute radiometric brightness values. These preprocessing steps facilitate the subsequent accurate identification of water bodies within the target pit. After preprocessing, the remote sensing information undergoes spectral analysis, and the water index of the information is calculated based on the spectral analysis results. The main principle is that when visible light, near-infrared light and other light hit the surface of the water body, reflection, scattering and absorption phenomena will occur. For example, the reflectivity of the blue-green band in the visible light band is higher than that of the red band, while the reflection energy of the near-infrared and short-wave infrared parts is less. By analyzing the reflectivity of different bands, the water body index of the target foundation pit can be obtained. The specific calculation formula is NDWI = (G-NI R) / (G+NI R), where G is the reflectivity of the green band, NIR is the reflectivity of the near-infrared band, and NDWI refers to the water body index, also called the normalized difference water body index. The area with a water body index greater than zero is screened out to obtain the foundation pit water body area of ​​the target foundation pit, and the screened foundation pit water body area is marked on the foundation pit plan construction drawing to facilitate subsequent optimization of the shortest inspection route.

[0143] Due to reasons such as high groundwater levels and unreasonable geotechnical engineering design, the bottom of the target foundation pit may collapse, while insufficient support burial depth, support strength, and rigidity may cause the foundation pit to bulge. If the collapse or bulge is severe, it may affect the driving of the detection unmanned vehicle. In addition, if there are artificially excavated deep pits and piled soil at the bottom of the foundation pit, it may also affect the driving of the detection unmanned vehicle. Therefore, it is necessary to use the foundation pit point cloud data to identify abnormal areas where the above abnormal conditions occur. The above abnormal areas are the foundation pit sedimentation areas. Specifically, based on the foundation pit plan construction drawing as the scale benchmark, the three-dimensional coordinates of the point cloud in the target foundation pit point cloud data are proportionally scaled, and the scaled three-dimensional coordinates of the point cloud are converted into two-dimensional coordinates of the point cloud. For example, if the three-dimensional coordinates of the point cloud are (a, b, c), the converted two-dimensional coordinates of the point cloud are (a, b). The obtained two-dimensional coordinates of the point cloud are matched with the two-dimensional coordinates of the base area in the foundation pit plan construction drawing to obtain the foundation pit point cloud data of the base area, that is, the base point cloud data. The base area refers to the area at the bottom of the foundation pit. This is because the presence of supporting columns, beams, and other structures surrounding the foundation pit walls can hinder the identification of the pit bottom's sedimentation area, for example, misidentifying supporting columns as pit sedimentation areas. Surface fitting is performed on the base point cloud data to obtain a base fitting surface for the base area. Polynomial regression, spline interpolation, and triangulation algorithms can be used to construct a mathematical surface from the point cloud, making this surface as close as possible to all point clouds. This surface is then obtained for the base fitting surface at the pit bottom. Taking the triangulation algorithm as an example, the base point clouds in the base point cloud data are first connected into triangles to construct a triangular mesh. Adjacent triangular meshes are stitched together to form a continuous surface. The size, shape, and distribution of the triangles are then adjusted to ensure a smooth surface without obvious defects. Finally, the stitched and optimized triangular mesh is used to obtain the base fitting surface. The vertical distance (distance difference) between all base point clouds in the base point cloud data and the base fitting surface is calculated. Base point clouds with distance differences greater than a preset difference threshold are screened out and considered base anomaly point clouds. The large difference in distance indicates that the corresponding base point cloud is not a point cloud on a plane, but a point cloud of a deep pit or a pile at the bottom of the foundation pit. After screening out all abnormal base point clouds and fitting, the area with a deep pit or bulge at the bottom of the foundation pit can be obtained, that is, the foundation pit deposition area.

[0144] The pit settlement area and the pit water area are marked on the pit plan construction drawing. The regional coordinates of the pit settlement area and the pit water area in the pit plan construction drawing are obtained. The regional coordinates are analyzed to see if there is any overlap with the route coordinates of the shortest detection route in the pit plan construction drawing. If there is overlap, it means that if the detection unmanned vehicle follows the shortest detection route, it will pass through the pit settlement area and the pit water area. This may make it difficult for the detection vehicle to successfully drive to the point cloud gap area, thus affecting the detection task. Therefore, the route needs to be adjusted. The shortest detection route can be optimized using algorithms such as the gray wolf optimization algorithm, the ant colony algorithm, and dynamic programming. Taking the gray wolf optimization algorithm as an example, multiple primary detection routes are randomly generated based on the detection start and end points in the shortest detection route and the location of the point cloud gap area in the pit plan construction drawing. All primary detection routes are encoded and the maximum number of iterations of the gray wolf optimization algorithm is set. Then, the pit settlement area and the pit water area are used as constraints of the gray wolf optimization algorithm to update all primary detection routes iteratively until the maximum number of iterations is reached and the optimal population is output. The fitness of the detection routes included in the optimal population is calculated based on a pre-set fitness function. The detection route with the highest fitness is used as the detection route for the unmanned vehicle. If there is no overlap, it means that the detection vehicle is following the shortest detection route and will not pass through the foundation pit settlement area or the foundation pit water area. Therefore, there is no need to optimize the shortest detection route and it can be directly used as the detection route for the unmanned vehicle.

[0145] In one embodiment, constructing a foundation pit map network according to pre-acquired foundation pit risk monitoring specifications, and obtaining key monitoring areas of a target foundation pit according to the foundation pit map network includes the following steps:

[0146] Perform semantic analysis on the pre-acquired foundation pit risk monitoring specifications to obtain semantic analysis results;

[0147] Extract key foundation pit construction items and foundation pit monitoring data types in the foundation pit risk monitoring specification from the semantic analysis results, and obtain the semantic association relationship between key foundation pit construction items and foundation pit monitoring data types based on the semantic analysis results;

[0148] The key construction projects and types of foundation pit monitoring data are used as graph nodes, and the semantic association relationships are used as node edges between graph nodes to obtain the foundation pit knowledge graph.

[0149] Analyze the topological relationship and graph knowledge of the foundation pit knowledge graph, obtain the key graph nodes in the foundation pit knowledge graph based on the analysis results, and use the key construction projects of the foundation pit corresponding to the key graph nodes as key monitoring locations;

[0150] Perform structural matching between the key monitoring positions and the three-dimensional model of the foundation pit to obtain the key monitoring positions of the foundation pit in the three-dimensional model;

[0151] The key monitoring positions of the foundation pit, the water area of ​​the foundation pit and the sedimentation area of ​​the foundation pit are taken as the key monitoring areas of the target foundation pit.

[0152] In this embodiment, the foundation pit risk monitoring specification includes the monitoring equipment, monitoring methods, monitoring items, monitoring data, etc. in the foundation pit monitoring process. The foundation pit risk monitoring specification is semantically analyzed to obtain the key contents in the foundation pit risk monitoring specification, namely, the foundation pit key construction items and foundation pit monitoring data types, as well as the semantic association relationship between the two. The foundation pit key construction items mainly include the foundation pit slope area, foundation pit support structure and buildings around the foundation pit. The foundation pit monitoring data types include groundwater level, horizontal displacement, vertical displacement, internal force, axial force, inclination, etc. The semantic association relationship refers to the correspondence between the foundation pit key construction items and the foundation pit monitoring data types, such as the groundwater level at the bottom of the foundation pit, the vertical displacement of the bottom of the foundation pit, the internal force of the foundation pit pillar, the axial force of the foundation pit anchor, the horizontal displacement of the foundation pit slope and the width of the foundation pit slope crack.

[0153] Methods that can be used for semantic analysis include support vector machines, latent semantic analysis models (LSA), Transformer models (converter models), word embedding models, and BERT models (pre-trained natural language processing models). Taking the BERT model as an example, the foundation pit risk monitoring specifications are input into the pre-trained BERT model for semantic recognition, and the text of the foundation pit risk monitoring specifications is classified according to the semantic recognition results. The monitoring items and monitoring data are filtered out to obtain the key text containing the monitoring items and monitoring data. The text content containing the monitoring items and monitoring data in the key text is semantically recognized to obtain the semantic association relationship between the monitoring items and the monitoring data. Then, the core keywords in the key text are extracted using algorithms such as the TF-IDF (term frequency-inverse document frequency) algorithm and the keyword extraction algorithm. Specifically, key construction projects and types of foundation pit monitoring data are analyzed. Using the TF-IDF algorithm as an example, the frequency of each word in the text and the inverse document frequency of each word in the entire corpus are calculated. The TF-IDF value of each word is then calculated based on its frequency and inverse document frequency. A higher value indicates a more important word in the document. Words with TF-IDF values ​​greater than a preset TF-IDF threshold are then filtered out to identify key construction projects and types of foundation pit monitoring data. Semantic associations between key construction projects and types of foundation pit monitoring data are then extracted from the semantic associations between monitoring projects and monitoring data. BERT is based on the Transformer model, which consists of two parts: an encoder and a decoder. BERT uses only the encoder, leveraging a self-attention mechanism and a feedforward neural network to process the input sequence. BERT employs a bidirectional training approach, leveraging both left- and right-hand information of the input sequence during pre-training to generate deep, bidirectional language representations.

[0154] The foundation pit knowledge graph is obtained by taking the key construction projects of the foundation pit and the types of foundation pit monitoring data as graph nodes, and the semantic association relationship between the key construction projects of the foundation pit and the types of foundation pit monitoring data as the node edges between the graph nodes. In the library and information science community, the knowledge graph is called knowledge domain visualization or knowledge domain mapping map. It is a series of various graphics that show the development process and structural relationship of knowledge. It uses visualization technology to describe knowledge resources and their carriers, and to mine, analyze, construct, draw and display knowledge and their mutual connections. The foundation pit knowledge graph can intuitively display the connection relationship between the key construction projects of the foundation pit and the types of foundation pit monitoring data. For example, the foundation pit slope area is connected with the horizontal displacement, and the foundation pit pillar is connected with the inclination.

[0155] Topological relationships refer to the relationships between spatial data that satisfy the principles of topological geometry. These relationships primarily manifest as connections between key excavation construction projects and types of excavation monitoring data. Graph knowledge refers to the attributes of graph nodes within the knowledge graph. Based on topological relationships and graph knowledge, the degree centrality, eigenvector centrality, betweenness centrality, and closeness centrality of graph nodes corresponding to all key excavation construction projects are calculated. Degree centrality is determined by node degree, which refers to the number of edges connected to a graph node. Eigenvector centrality is determined by the node degrees of its neighbors. A higher degree indicates a higher eigenvector centrality for the graph node, indicating a more important graph node. Betweenness centrality measures the extent to which a graph node lies on the shortest path between other pairs of nodes. Graph nodes with high betweenness centrality play a key role in information transmission and resource flow, indicating that graph nodes with high betweenness centrality are more important. Closeness centrality is based on the sum of the shortest paths from the graph node to all other graph nodes in the network. The smaller the sum of the shortest paths, the higher the closeness centrality of the graph node, which also means the graph node is more important. The degree centrality, eigenvector centrality, betweenness centrality, and closeness centrality of all graph nodes are calculated. Graph nodes of high importance are screened based on these centrality characteristics. A centerline threshold can be set to filter out graph nodes with a centerline threshold greater than the preset threshold, resulting in key graph nodes. Key construction projects in the foundation pit corresponding to key graph nodes are designated as key monitoring locations. Key graph nodes in the knowledge graph can also be screened based on node degree and node edge weights, with the key construction projects in the foundation pit corresponding to these key graph nodes designated as key monitoring locations. A pre-built 3D structure recognition model is then used to identify the names of the various structures in the foundation pit 3D model, namely the foundation pit model structures. Semantic matching is then performed on the foundation pit model structures. Based on the semantic matching results, the foundation pit model structures corresponding to the key monitoring locations are designated as key monitoring locations in the foundation pit 3D model.

[0156] Finally, the key monitoring positions of the foundation pit, the water body area of ​​the foundation pit, and the sedimentation area of ​​the foundation pit are taken as the key monitoring areas of the target foundation pit. The key monitoring positions are the areas that need to be monitored according to the foundation pit risk monitoring specifications, and are set according to the general situation of the foundation pit, while the water body area and the sedimentation area are the areas that need to be monitored based on the foundation pit point cloud data and the remote sensing information of the foundation pit, and are set according to the specific situation of the target foundation pit. This is because the water body area of ​​the foundation pit is an area where seepage or accumulation of water exists, and the seepage or accumulation of water in the target foundation pit may be due to insufficient groundwater drainage or the lack of drainage ditches before excavation of the foundation pit, which may affect the stability of the foundation pit and may even cause earthwork collapse, resulting in safety accidents such as ground collapse. In addition, the sedimentation area of ​​the foundation pit may be due to collapse or uplift at the bottom of the foundation pit, so it also needs to be monitored.

[0157] In one embodiment, performing structural matching between the key monitoring positions and the three-dimensional model of the foundation pit to obtain the key monitoring positions of the foundation pit in the three-dimensional model of the foundation pit includes the following steps:

[0158] Construct a three-dimensional structure recognition model based on a neural network model;

[0159] Use the pre-built foundation pit structure training set to train the 3D structure recognition model;

[0160] Using the trained three-dimensional structure recognition model, the model structure of the foundation pit three-dimensional model is recognized to obtain the foundation pit model structure of the foundation pit three-dimensional model;

[0161] Perform semantic recognition on the foundation pit model structure and the key monitoring position, and perform semantic matching on the foundation pit model structure and the key monitoring position according to the semantic recognition result to obtain a semantic matching result;

[0162] According to the semantic matching results, the foundation pit model structure corresponding to the key monitoring position is used as the foundation pit key monitoring position in the foundation pit three-dimensional model.

[0163] In this embodiment, a fully convolutional neural network (FCN) is used as the core architecture. This model effectively preserves the spatial structure of the input data and is particularly suitable for processing the complex structures of three-dimensional foundation pit models. The model structure consists of an encoder and a decoder. The encoder consists of multiple 3D convolutional layers, batch normalization layers, and ReLU activation functions to extract multi-scale features of the foundation pit model. The decoder uses transposed convolutional layers for upsampling, restoring the spatial resolution of the feature map. To improve model performance, skip connections are added to fuse the low-level features of the encoder with the high-level features of the decoder. The final layer uses a softmax activation function to output the probability of each voxel belonging to a different structural category. The loss function uses cross-entropy, and the model is trained using the Adam optimizer. The learning rate is initially set to 0.001 and a learning rate decay strategy is used. In addition, a spatial attention module is introduced to enhance the model's ability to recognize key structures. The training set should contain a diverse set of 3D foundation pit models and their corresponding annotated data, covering a variety of pit types, sizes, and complexities. Data augmentation techniques such as random rotation, scaling, mirror flipping, and noise addition are applied to improve the model's generalization ability. To handle large 3D data, a block training strategy was adopted, splitting the large foundation pit model into small, overlapping blocks for training. Batch gradient descent was used for training, with each batch containing 8-16 samples. A cosine annealing strategy was used for the learning rate, and L2 regularization was introduced to prevent overfitting. Cross-validation was used to evaluate model performance, with the dataset divided into training, validation, and test sets. To improve training efficiency, transfer learning was employed: the model was pre-trained on a large-scale, general 3D dataset and then fine-tuned on the specific foundation pit dataset.

[0164] Next, the 3D foundation pit model to be analyzed is preprocessed to the same format and resolution as the training data. For large foundation pit models, a sliding window technique is used for block processing. The window size matches the block size used during model training, and the stride can be set to half the window size to ensure sufficient coverage. To address boundary issues, padding techniques such as mirror padding can be used. For overlapping regions, predictions are merged using a weighted averaging strategy. To improve recognition accuracy, test-time augmentation techniques are used to perform multiple transformations on the same input and then combine these predictions. Conditional random fields can be used to further refine the segmentation results, taking into account spatial consistency. Finally, morphological operations are applied to post-process the results to remove noise and smooth boundaries. Semantic representations are then generated for the foundation pit model structure and key monitoring locations. Word embedding techniques such as Word2Vec or GloVe are used to convert descriptions of structural components and monitoring points into dense vectors. To capture more complex semantic information, a pre-trained BERT model can be used for feature extraction. The semantic recognition process uses these representations to calculate similarity, with common methods including cosine similarity and dot-product attention. A multi-head attention mechanism is introduced to capture correlations at different semantic levels. The semantic matching process can be modeled as a bipartite graph matching problem, using the Hungarian algorithm to find the optimal global match. To handle complex semantic relationships, graph neural networks can be used to model the relationships between structures, and then a graph matching algorithm can be used for matching. Introducing a domain knowledge base can enhance semantic understanding, for example, using a dictionary of foundation pit engineering terminology for entity linking.

[0165] Based on the semantic matching results from the previous step, a mapping function is constructed from the set of model structures to the set of monitoring locations. For each matching pair, the specific coordinates of the structure are located in the 3D model. To accurately locate the structure, a nearest neighbor search algorithm is used to find the most suitable point on the structure surface. Given the physical limitations of the monitoring equipment, constraints such as minimum spacing are applied. For complex structures, a Monte Carlo method can be used to generate multiple candidate points, and the optimal one is selected. To optimize the overall monitoring effect, a greedy algorithm is used to gradually select monitoring points, each time choosing the point that maximizes monitoring coverage. An expert system is introduced to integrate domain knowledge to adjust and verify the rationality of monitoring locations. For visualization and interaction, the selected monitoring locations are marked in the 3D model, using different colors or symbols to represent different types of monitoring points. Finally, a detailed monitoring plan report is generated, including the exact coordinates, monitoring type, installation requirements, and other information for each monitoring point. This integrated approach allows the precise location of key monitoring locations in the 3D foundation pit model, achieving intelligent and visual monitoring layout.

[0166] In one embodiment, a foundation pit monitoring strategy for generating a target foundation pit by combining key monitoring areas of the foundation pit and a three-dimensional foundation pit model includes the following steps:

[0167] Calculate the projected area of ​​the key monitoring area of ​​the foundation pit on the horizontal plane based on the three-dimensional model of the foundation pit;

[0168] Generate the number of foundation pit monitoring points in the key monitoring area of ​​the foundation pit based on the projected area;

[0169] According to the number of foundation pit monitoring points corresponding to the key monitoring area of ​​the foundation pit, a corresponding number of foundation pit monitoring point positions are set in the key monitoring area of ​​the foundation pit;

[0170] The foundation pit monitoring strategy for the target foundation pit is generated by combining the number and location of foundation pit monitoring points.

[0171] In this embodiment, the projection area of ​​the key monitoring area of ​​the foundation pit on the horizontal plane is calculated based on the three-dimensional model of the foundation pit, and different numbers of foundation pit monitoring points are generated according to the different projection areas. Specifically, the area with a large projection area generates more foundation pit monitoring points, and vice versa. In addition, for the key monitoring areas of the foundation pit with a small horizontal projection area such as foundation pit pillars and soil anchor rods, their vertical projection areas will be additionally calculated, and the horizontal projection areas will be used as an auxiliary to generate the corresponding number of foundation pit monitoring points. According to the number of foundation pit monitoring points, a corresponding number of foundation pit monitoring point positions are generated (which can be evenly distributed), and the foundation pit monitoring strategy of the target foundation pit is generated in combination with the number of foundation pit monitoring points and the foundation pit monitoring point positions. Subsequently, the relevant staff of the target foundation pit will arrange different types of monitoring instruments according to the foundation pit monitoring strategy to achieve all-round and high-precision monitoring of the target foundation pit.

[0172] In one embodiment, after visually displaying the risk assessment results through the three-dimensional model of the foundation pit, the following steps are further included:

[0173] Storing the foundation pit monitoring data of all foundation pit monitoring points as historical foundation pit monitoring data;

[0174] Match the foundation pit monitoring data with the foundation pit monitoring points, and mark the foundation pit monitoring points with missing information as abnormal foundation pit monitoring points;

[0175] Obtaining historical foundation pit monitoring data corresponding to the abnormal foundation pit monitoring point, and predicting the monitoring data of the abnormal foundation pit monitoring point based on the corresponding historical foundation pit monitoring data to obtain predicted monitoring data;

[0176] The predicted monitoring data is used as the monitoring data of the abnormal foundation pit monitoring point for data filling to obtain the complete foundation pit monitoring data of the target foundation pit, and the risk assessment of the target foundation pit is carried out based on the complete foundation pit monitoring data.

[0177] In this embodiment, the foundation pit monitoring data of all foundation pit monitoring points obtained from monitoring within a period of time are stored as historical foundation pit monitoring data. The foundation pit monitoring data obtained from each monitoring is matched with the corresponding foundation pit monitoring point. If there is a foundation pit monitoring point that does not have corresponding foundation pit monitoring data, it means that there is an information gap at the foundation pit monitoring point, which may be due to a failure of the monitoring instrument at the foundation pit monitoring point. In order to ensure that the target foundation pit can be monitored in all directions, it is necessary to fill in the information of the foundation pit monitoring point with missing information. Specifically, the foundation pit monitoring point with missing information is marked as an abnormal foundation pit monitoring point, and the historical foundation pit monitoring data corresponding to the abnormal foundation pit monitoring point in the pre-stored historical foundation pit monitoring data is extracted to obtain the historical foundation pit monitoring data of the abnormal foundation pit monitoring point. The historical foundation pit monitoring data is subjected to time series analysis to obtain the changing trend of the historical foundation pit monitoring data of the abnormal foundation pit monitoring point, and the predicted monitoring data of the abnormal foundation pit monitoring point is predicted based on the changing trend. The predicted monitoring data is used as the monitoring data of abnormal foundation pit monitoring points to fill in the information of the foundation pit monitoring data with missing information, to obtain complete foundation pit monitoring data, and to conduct risk assessment on the target foundation pit based on the complete foundation pit monitoring data.

[0178] In addition, when an abnormal foundation pit monitoring point is marked, the foundation pit staff will immediately obtain the abnormal monitoring point location of the abnormal foundation pit monitoring point in the target foundation pit based on the foundation pit three-dimensional model, and go to the abnormal monitoring point location to check the monitoring instrument. If the monitoring instrument is found to be faulty, it will be repaired or replaced immediately.

[0179] The present application also discloses a deep foundation pit intelligent safety monitoring system for a smart construction site, which is characterized by comprising:

[0180] a memory configured to store instructions; and

[0181] A processor is configured to call instructions from a memory and implement any of the above-mentioned methods for intelligent safety monitoring of deep foundation pits in smart construction sites when executing the instructions.

[0182] Among them, the processor can adopt a central processing unit (CPU). Of course, according to actual usage, other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be adopted. The general-purpose processor can adopt a microprocessor or any conventional processor, etc., and this application does not impose any restrictions on this.

[0183] Among them, the memory can be an internal storage unit of a computer device, such as a hard disk or memory of a computer device, or an external storage device of a computer device, such as a plug-in hard disk, smart memory card (SMC), secure digital card (SD) or flash memory card (FC) equipped on the computer device. In addition, the memory can also be a combination of an internal storage unit and an external storage device of a computer device. The memory is used to store computer programs and other programs and data required by the computer device. The memory can also be used to temporarily store data that has been output or is to be output. This application does not impose any restrictions on this.

[0184] An embodiment of the present application also provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute the above-mentioned method for intelligent safety monitoring of deep foundation pits for smart construction sites.

[0185] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0186] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0187] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0188] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0189] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0190] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0191] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transit media), such as modulated data signals and carrier waves.

[0192] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0193] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included within the scope of the claims of the present application.

Claims

1. A deep foundation pit intelligent safety monitoring method for a smart construction site, characterized in that: The method comprises the steps of: Use detection drones to collect multiple foundation pit point cloud data of the target foundation pit; Position matching is performed on all the foundation pit point cloud data with the pre-acquired foundation pit plan construction drawing to obtain a point cloud position matching result, and a point cloud vacant area in the foundation pit plan construction drawing is marked according to the point cloud position matching result; Controlling the detection unmanned vehicle to collect regional point cloud data of the point cloud vacant area, and combining the regional point cloud data with all the foundation pit point cloud data to construct a three-dimensional foundation pit model of the target foundation pit; Constructing a foundation pit map network according to the pre-acquired foundation pit risk monitoring specification, and determining the foundation pit key monitoring area of ​​the target foundation pit according to the foundation pit map network; Generating a foundation pit monitoring strategy for the target foundation pit in combination with the foundation pit key monitoring area and the foundation pit three-dimensional model; Arranging different types of monitoring instruments at each foundation pit monitoring point of the target foundation pit according to the foundation pit monitoring strategy, and using the monitoring instruments to perform real-time monitoring of the target foundation pit to obtain foundation pit monitoring data of the target foundation pit; Position matching is performed between the foundation pit monitoring point corresponding to the foundation pit monitoring data and the foundation pit three-dimensional model, and the foundation pit three-dimensional model is updated in real time using the foundation pit monitoring data according to the position matching result; A risk assessment is performed on the target foundation pit according to the foundation pit monitoring data, and the risk assessment result is visualized through the foundation pit three-dimensional model.

2. The method according to claim 1, characterized in that The step of matching all the foundation pit point cloud data with the pre-acquired foundation pit plan construction drawing to obtain a point cloud position matching result, and marking the point cloud vacant area in the foundation pit plan construction drawing according to the point cloud position matching result comprises the following steps: Preprocessing all the foundation pit point cloud data; All the pre-processed foundation pit point cloud data are preliminarily aligned using a feature matching algorithm; Using an iterative optimization algorithm to align the coordinate systems of all the foundation pit point cloud data to obtain a plurality of standard foundation pit point cloud data; Performing point cloud vacancy identification on all the standard foundation pit point cloud data to determine the locations of multiple point cloud vacancy areas; The positions of all the missing areas of the point cloud are mapped to the pre-acquired foundation pit plan construction drawing to determine the missing areas of the point cloud.

3. The method according to claim 2, characterized in that The step of performing point cloud vacancy identification on all the standard foundation pit point cloud data to obtain multiple point cloud vacancy area locations comprises the following steps: For each of the standard foundation pit point cloud data, the standard foundation pit point cloud data is projected at multiple angles to obtain a plurality of multi-directional foundation pit point cloud data; Gridding all the two-dimensional foundation pit point cloud data to obtain a plurality of grid areas; Calculating and obtaining the grid density in the grid area; If there is a sparse grid area whose grid density is less than a preset density threshold among all the grid areas, analyzing the number of grids in the sparse grid area that are connected by edges among all the sparse grid areas; If the number of grids is greater than a preset number threshold, all the sparse grid areas are acquired and integrated to obtain a point cloud sparse area; The center point coordinates of all the sparse areas of the point cloud are obtained, and the positions of all the vacant areas of the point cloud are determined according to the center point coordinates.

4. The method according to claim 3, characterized in that The controlling the detection unmanned vehicle to collect the regional point cloud data of the point cloud vacant area, and combining the regional point cloud data with all the foundation pit point cloud data to construct the three-dimensional model of the target foundation pit includes the following steps: All the standard foundation pit point cloud data are stitched together to obtain target foundation pit point cloud data; Combining the point cloud missing area with the foundation pit remote sensing information of the target foundation pit to generate an unmanned vehicle detection route for the detection unmanned vehicle; Control the detection unmanned vehicle to collect regional point cloud data of the point cloud missing area according to the detection route of the unmanned vehicle; Preprocessing the regional point cloud data; Aligning the coordinate system of the pre-processed regional point cloud data with the target foundation pit point cloud data as a reference; Filling the regional point cloud data with the coordinate system aligned into the target foundation pit point cloud data to obtain the complete foundation pit point cloud data of the target foundation pit; Performing surface fitting on the complete foundation pit point cloud data to obtain multiple fitting surfaces; A plurality of the fitting curved surfaces are connected according to the edge point cloud coordinates of the fitting curved surfaces to obtain a three-dimensional foundation pit model of the target foundation pit.

5. The method according to claim 4, characterized in that The step of generating a detection route for an unmanned vehicle by combining the point cloud vacant area and the foundation pit remote sensing information comprises the following steps: Generate the shortest detection route for the detection unmanned vehicle in the foundation pit planned construction drawing according to the vacant area of ​​the point cloud and using the shortest path algorithm; Preprocessing the foundation pit remote sensing information; Performing spectral analysis on all the pre-processed foundation pit remote sensing information, and calculating the water body index of the foundation pit remote sensing information according to the spectral analysis result; Marking the area in the foundation pit planned construction drawing where the water body index is greater than a preset index threshold as a foundation pit water body area; Position matching is performed on the target foundation pit point cloud data and the foundation pit planned construction drawing, and base point cloud data in the target foundation pit point cloud data is obtained according to the foundation pit planned construction drawing; Performing surface fitting on the base point cloud data to obtain a base fitting surface of the base area of ​​the target foundation pit; Calculating the distance difference between all the base point clouds and the base fitting surface in all the base point cloud data; Screening out the base point clouds whose distance difference is greater than a preset difference threshold to obtain a plurality of base abnormal point clouds; Fitting all the abnormal point clouds of the basement to obtain the foundation pit deposition area; Analyze whether there is any overlap between the shortest detection route and the foundation pit water area and the foundation pit sediment area. If there is no overlap, use the shortest detection route as the unmanned vehicle detection route of the detection unmanned vehicle. If the overlapping phenomenon exists, the foundation pit water area and the foundation pit sedimentation area are used as constraints, and the shortest detection route is optimized according to the constraints to obtain the unmanned vehicle detection route of the detection unmanned vehicle.

6. The method according to claim 5, characterized in that The step of constructing a foundation pit map network according to the pre-acquired foundation pit risk monitoring specification and acquiring the foundation pit key monitoring area of ​​the target foundation pit according to the foundation pit map network comprises the following steps: Perform semantic analysis on the pre-acquired foundation pit risk monitoring specifications to obtain semantic analysis results; Extracting key foundation pit construction items and foundation pit monitoring data types in the foundation pit risk monitoring specification from the semantic analysis results, and obtaining semantic association relationships between the key foundation pit construction items and the foundation pit monitoring data types according to the semantic analysis results; The key construction items of the foundation pit and the types of foundation pit monitoring data are used as graph nodes, and the semantic association relationships are used as node edges between the graph nodes to obtain a foundation pit knowledge graph; Analyze the topological relationship and graph knowledge of the foundation pit knowledge graph, obtain key graph nodes in the foundation pit knowledge graph according to the analysis results, and use the key construction items of the foundation pit corresponding to the key graph nodes as key monitoring positions; Performing structural matching between the key monitoring position and the three-dimensional model of the foundation pit to obtain the key monitoring position of the foundation pit in the three-dimensional model of the foundation pit; The key monitoring position of the foundation pit, the water area of ​​the foundation pit and the sedimentation area of ​​the foundation pit are taken as the key monitoring areas of the foundation pit of the target foundation pit.

7. The method according to claim 6, characterized in that The structural matching of the key monitoring position with the foundation pit three-dimensional model to obtain the key monitoring position of the foundation pit in the foundation pit three-dimensional model comprises the following steps: Construct a three-dimensional structure recognition model based on a neural network model; Using a pre-constructed foundation pit structure training set to train the three-dimensional structure recognition model; Using the trained three-dimensional structure recognition model to perform model structure recognition on the foundation pit three-dimensional model to obtain the foundation pit model structure of the foundation pit three-dimensional model; Performing semantic recognition on the foundation pit model structure and the key monitoring position, and performing semantic matching on the foundation pit model structure and the key monitoring position according to the semantic recognition result to obtain a semantic matching result; According to the semantic matching result, the foundation pit model structure corresponding to the key monitoring position is used as the foundation pit key monitoring position in the foundation pit three-dimensional model.

8. The method according to claim 1, characterized in that The foundation pit monitoring strategy for generating the target foundation pit by combining the foundation pit key monitoring area and the foundation pit three-dimensional model comprises the following steps: Calculating the projection area of ​​the key monitoring area of ​​the foundation pit on the horizontal plane according to the three-dimensional model of the foundation pit; Generating the number of foundation pit monitoring points in the key monitoring area of ​​the foundation pit according to the projection area; According to the number of the foundation pit monitoring points corresponding to the foundation pit key monitoring area, a corresponding number of foundation pit monitoring point positions are set in the foundation pit key monitoring area; A foundation pit monitoring strategy for the target foundation pit is generated based on the number of foundation pit monitoring points and the locations of the foundation pit monitoring points.

9. The method according to claim 1, characterized in that: After visually displaying the risk assessment results through the foundation pit three-dimensional model, the following steps are also included: Storing the foundation pit monitoring data of all the foundation pit monitoring points as historical foundation pit monitoring data; Matching the foundation pit monitoring data with the foundation pit monitoring points, and marking the foundation pit monitoring points with missing information as abnormal foundation pit monitoring points; Acquire the historical foundation pit monitoring data corresponding to the abnormal foundation pit monitoring point, and predict the monitoring data of the abnormal foundation pit monitoring point according to the corresponding historical foundation pit monitoring data to obtain predicted monitoring data; The predicted monitoring data is used as the monitoring data of the abnormal foundation pit monitoring point for data filling to obtain the complete foundation pit monitoring data of the target foundation pit, and the risk assessment of the target foundation pit is performed based on the complete foundation pit monitoring data.

10. A deep foundation pit intelligent safety monitoring system for smart construction sites, characterized in that: include: a memory configured to store instructions; as well as A processor is configured to call the instructions from the memory and to implement a method for intelligent safety monitoring of deep foundation pits for smart construction sites according to any one of claims 1 to 9 when executing the instructions.

Images