Intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing

Abstract

The present invention relates to the field of construction engineering technology, and specifically to an intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing, comprising the following steps: S1: collecting point cloud data of geometric shapes, sizes and position information of various parts of the steel structure; S2: registering and fusing point cloud data from different perspectives to form a three-dimensional point cloud model; S3: extracting geometric features of the steel structure from the three-dimensional point cloud model, and establishing a three-dimensional model of the steel structure; S4: comparing the three-dimensional model of the steel structure established in S3 with the construction drawings to identify geometric differences; S5: analyzing the construction deviation report to identify potential quality problems in construction. The present invention, by combining drone inspection, point cloud processing and graph neural network algorithms, can efficiently and accurately identify quality problems in steel structure construction and assess safety risks, significantly improving the automation and intelligence level of construction quality monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of building engineering, and in particular to an intelligent monitoring method for steel structure construction based on unmanned aerial vehicle inspection and point cloud processing. Background Art

[0002] With the rapid development of the construction industry, steel structure, as a widely used structural form in modern buildings, has a construction quality that directly affects the safety and stability of buildings. Traditional steel structure construction quality monitoring methods usually rely on manual inspections and manual measurements, which are inefficient and easily affected by human factors. In addition, steel structure construction sites often have complex environments and long construction cycles, and quality problems (such as structural deformation, cracks or dislocations) are usually difficult to detect in time during the construction process. Even if problems are discovered through traditional detection methods, the response speed to solve the problems is relatively slow, resulting in potential safety hazards not being dealt with in a timely manner, increasing construction risks, and may even endanger the safety of the entire building.

[0003] At present, with the development of drone technology, lidar and computer vision technology, the monitoring of steel structure construction quality is gradually transforming towards automation and intelligence. Drones equipped with lidar and high-definition cameras can quickly obtain a large amount of point cloud data from the construction site, and with the help of computer processing technology, they can generate detailed three-dimensional models, thereby automating the detection of construction deviations. However, in the existing technology, there are still great challenges in how to efficiently and accurately use point cloud data to extract the geometric features of steel structures, identify potential quality problems, and combine construction environment data for intelligent analysis. Summary of the invention

[0004] Based on the above objectives, the present invention provides an intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing.

[0005] The intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing includes the following steps:

[0006] S1: Use the laser radar and high-definition camera carried by the drone to inspect the steel structure construction site and collect point cloud data of the geometric shape, size and position information of each part of the steel structure;

[0007] S2: De-noise and standardize the point cloud data collected by S1, and use the iterative closest point algorithm to register and fuse the point cloud data from different perspectives to form a three-dimensional point cloud model;

[0008] S3: Extract the geometric features of the steel structure from the 3D point cloud model, including the spatial coordinates, shape and size information of the steel beams, steel columns and connection points, and establish a 3D model of the steel structure;

[0009] S4: Compare the three-dimensional model of the steel structure established in S3 with the construction drawings, identify the geometric differences between the actually constructed steel structure and the design drawings, and generate a construction deviation report;

[0010] S5: Combined with the construction environment data, the graph neural network algorithm is used to analyze the construction deviation report generated in S4 to identify potential quality problems in construction, evaluate the safety risks brought by quality problems, and generate a construction risk assessment report.

[0011] Optionally, the S1 specifically includes:

[0012] S11: Based on the specific layout and monitoring requirements of the steel structure construction site, the flight path of the drone is pre-planned to ensure coverage of all steel structure construction areas;

[0013] S12: The laser radar carried by the drone scans according to the predetermined flight path, collects the distance and reflection intensity data of each part of the steel structure in real time, and generates point cloud data containing at least 100,000 points per square meter;

[0014] S13: Use the high-definition camera on the drone to capture high-resolution images of the construction site and obtain the texture and color information of the steel structure surface;

[0015] S14: Time synchronization processing is performed on the lidar data collected in S12 and S13 and the high-definition camera image to ensure accurate data alignment.

[0016] Optionally, the S2 specifically includes:

[0017] S21: Apply a statistical outlier removal algorithm to the point cloud data collected in S1, and remove noise points in the point cloud data by setting a distance threshold to ensure the accuracy of the remaining data;

[0018] S22: performing scale normalization processing on the denoised point cloud data, unifying the coordinate range of the point cloud data, and eliminating the scale difference between data collected by different sensors;

[0019] S23: Using the iterative closest point algorithm, the standardized point cloud data from different viewpoints are registered, and the distance error between point clouds is minimized through an iterative optimization process;

[0020] S24: Based on the registered point cloud data, the multi-view point cloud data are fused using a weighted average method to generate a three-dimensional point cloud model.

[0021] Optionally, the S23 specifically includes:

[0022] S231: Select a set of reference point cloud data and initialize a transformation matrix for representing an initial alignment relationship between the source point cloud data and the target point cloud data;

[0023] S232: for each point in the source point cloud data, find the corresponding point in the target point cloud data that is closest to it, and specifically determine the best matching relationship by calculating the distance between each pair of points;

[0024] S233: Calculate the optimal rigid transformation matrix from the source point cloud data to the target point cloud data using the least square method;

[0025] S234: Apply the calculated optimal transformation matrix to the source point cloud data to update its position; and repeat steps S232 and S233 to gradually optimize the registration result until the error between the point clouds is reduced to below a predetermined threshold.

[0026] Optionally, the S24 specifically includes:

[0027] S241: Calculate weighting factors for each point cloud dataset , the calculation formula is: ,in, Indicates The weighting factor of the point cloud dataset of each viewpoint; Indicates Quality scoring of point cloud datasets from different viewpoints; Indicates Quality scoring of point cloud datasets from different viewpoints; Represents the total number of point cloud datasets from all viewpoints;

[0028] S242: For each point Coordinates in the point cloud dataset , according to the weighting factor Perform weighted calculation to obtain weighted coordinates , the formulas are: ; ; ;in, Respectively represent From a perspective The X, Y, and Z coordinates of a point in three-dimensional space; Respectively represent the weighted From a perspective The X, Y, and Z coordinates of a point;

[0029] S243: Weighted coordinates in the weighted point cloud dataset of all viewpoints Accumulate and generate a unified 3D point cloud model , the expression is: ,in, Represents the final generated three-dimensional point cloud model; Indicates A weighted point cloud dataset with 3000 points of view.

[0030] Optionally, the S3 specifically includes:

[0031] S31: Calculate the normal vector of each point based on the neighborhood point set of each point in the three-dimensional point cloud model through a normal vector calculation method, and filter out the point set belonging to the steel beam, steel column and connection point in the point cloud data according to the direction of the normal vector and the point cloud density characteristics;

[0032] S32: Based on the point set selected in S31, a clustering-based algorithm is used to classify the steel beams, steel columns and connection points into different geometric entities;

[0033] S33: Based on S32, the spatial coordinates, shapes and size information of the steel beams, steel columns and connection points are extracted respectively;

[0034] S34: establishing a three-dimensional steel structure model according to the geometric feature information of the steel structure extracted in S33;

[0035] The three-dimensional steel structure model expression is: ,in, For the A three-dimensional representation of geometric features, for all geometric features.

[0036] Optionally, the S4 specifically includes:

[0037] S41: Obtaining geometric information of the steel structure in the design drawing, including spatial coordinates, shape and size information of steel beams, steel columns and connection points; and establishing a three-dimensional model of the design drawing by extracting geometric features in the drawing data;

[0038] S42: Based on the least squares method, a rigid registration algorithm is used to spatially align the three-dimensional model of the steel structure established in S3 and the three-dimensional model of the design drawing extracted in S41;

[0039] S43: Calculate the spatial distance between each corresponding point of the registered three-dimensional model of the steel structure and the three-dimensional model of the design drawing. If the spatial distance exceeds a set threshold, the point is considered to be a deviation point.

[0040] S44: All calculated deviation point data are aggregated to form a construction deviation report, including the specific location of the deviation point, the deviation value and the deviation type.

[0041] Optionally, the S5 specifically includes:

[0042] S51: Obtain construction environment data, including meteorological information, soil conditions, construction equipment conditions, and real-time dynamic data of the construction site, including temperature and humidity;

[0043] S52: Standardize different data sources to ensure consistent data dimensions;

[0044] S53: integrating the deviation point information in the construction deviation report with the construction environment data to construct a multidimensional data graph, wherein each node represents a construction element, and the characteristics of the node include the spatial position of the corresponding element, geometric differences, and correlation characteristics with the construction environment data;

[0045] S54: Using a graph neural network algorithm, training the multidimensional data graph to identify potential quality problems, including structural deformation, cracks, or dislocations;

[0046] S55: Evaluate the impact of quality problems on construction safety based on the severity of the quality problems and construction environment conditions;

[0047] S56: Generate a construction risk assessment report based on the identified quality issues and assessed safety risks, including the nature, scope of impact and risk level of each potential quality issue.

[0048] Optionally, the S54 specifically includes:

[0049] S541: feature encoding is performed on each node in the multidimensional data graph, wherein each node represents an element in the steel structure, and the node features include spatial position, geometric difference, deviation amount, and features related to the construction environment data;

[0050] S542: Use the graph convolution network to perform graph convolution operations, combine the edge information between nodes, update the node features, and thus capture the local and global dependencies between nodes;

[0051] S543: After completing the multi-layer graph convolution, the final features of each node are input into a fully connected network for classification to identify potential quality issues including structural deformation, cracks or misalignment.

[0052] Optionally, the S55 specifically includes:

[0053] S551: Based on the quality issues identified in S54, for each issue, calculate its weighted risk value, the formula is: ,in, For quality issues The risk value, For quality issues The weighting coefficient of For quality issues the severity of the disease;

[0054] S552: Combine construction environment data to calculate the influencing factors of the construction environment ; The formula is: ,in, is the construction environment influencing factor, For the construction environment Item data, For the The weight coefficient of the construction environment data, is the number of environmental data items;

[0055] S553: ​​Combine the quality problem risk value and construction environment impact factors, and use the weighted average method to calculate the total risk value , represents the overall construction safety risk, and the formula is: ,in, is the overall construction safety risk value, is the number of quality issues, For quality issues The risk value, is the construction environment influencing factor;

[0056] S554: Based on the total risk value , the construction risk assessment is divided into different safety levels, which are judged by setting thresholds:

[0057] like If it is higher than the first threshold, the construction risk is high and safety measures need to be taken immediately;

[0058] like If it is below the first threshold but above the second threshold, the construction risk is medium and regular inspection is required;

[0059] like Below the second threshold, the construction risk is low and no additional safety measures are needed immediately.

[0060] Beneficial effects of the present invention:

[0061] The present invention uses laser radar and drones equipped with high-definition cameras to conduct inspections, which can efficiently collect data such as the geometric shape, size and position of steel structures, and use an iterative closest point algorithm to accurately align and fuse point cloud data from different perspectives to generate a high-quality three-dimensional point cloud model; this automated and efficient data collection and processing method significantly improves the accuracy and efficiency of monitoring, and solves the problems of inefficiency and large errors in traditional manual inspections.

[0062] The present invention, by combining construction environment data and utilizing a graph neural network algorithm to intelligently analyze construction deviation reports, can identify potential quality problems in construction in real time and assess the impact of quality problems on construction safety based on the severity of the problems and construction environment conditions; it can calculate construction risk values ​​using a weighted average method, providing a scientific basis for safety management during the construction process. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0064] Figure 1 A schematic diagram of an intelligent monitoring method for steel structure construction according to an embodiment of the present invention;

[0065] Figure 2 The figure is a schematic diagram of a process of forming a three-dimensional point cloud model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0066] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0067] It should be noted that the references to "one embodiment", "embodiment", "exemplary embodiments", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0068] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0069] like Figure 1-Figure 2As shown, the intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing includes the following steps:

[0070] S1: Use the laser radar and high-definition camera carried by the drone to inspect the steel structure construction site and collect point cloud data of the geometric shape, size and position information of each part of the steel structure;

[0071] S2: De-noise and standardize the point cloud data collected by S1, and use the iterative closest point algorithm to register and fuse the point cloud data from different perspectives to form a three-dimensional point cloud model to ensure the spatial consistency of the point cloud data;

[0072] S3: Extract the geometric features of the steel structure from the 3D point cloud model, including the spatial coordinates, shape and size information of the steel beams, steel columns and connection points, and establish a 3D model of the steel structure;

[0073] S4: Compare the three-dimensional model of the steel structure established in S3 with the construction drawings, identify the geometric differences between the actually constructed steel structure and the design drawings, and generate a construction deviation report;

[0074] S5: Combined with the construction environment data, the graph neural network (GNN) algorithm is used to analyze the construction deviation report generated in S4 to identify potential quality problems in construction, including structural deformation, cracks or misalignment. At the same time, the safety risks brought by quality problems are evaluated and a construction risk assessment report is generated.

[0075] S1 specifically includes:

[0076] S11: Based on the specific layout and monitoring requirements of the steel structure construction site, the flight path of the drone is pre-planned to ensure coverage of all steel structure construction areas;

[0077] S12: The laser radar carried by the drone scans according to the predetermined flight path, collects the distance and reflection intensity data of each part of the steel structure in real time, and generates point cloud data containing at least 100,000 points per square meter;

[0078] S13: Use the high-definition camera on the drone to capture high-resolution images of the construction site and obtain the texture and color information of the steel structure surface;

[0079] S14: Perform time synchronization processing on the lidar data collected in S12 and S13 and the high-definition camera images to ensure accurate data alignment, thereby providing an accurate basis for subsequent data fusion.

[0080] S2 specifically includes:

[0081] S21: Apply a statistical outlier removal algorithm to the point cloud data collected in S1, and remove noise points in the point cloud data by setting a distance threshold to ensure the accuracy of the remaining data;

[0082] S22: performing scale normalization processing on the denoised point cloud data, unifying the coordinate range of the point cloud data, eliminating the scale differences between data collected by different sensors, and ensuring data consistency;

[0083] S23: Iterative Closest Point (ICP) algorithm is used to register standardized point cloud data from different viewpoints. Through an iterative optimization process, the distance error between point clouds is minimized to achieve accurate alignment of multi-viewpoint cloud data.

[0084] S24: Based on the registered point cloud data, the multi-view point cloud data is fused using the weighted average method to generate a three-dimensional point cloud model to ensure the integrity and high precision of the model.

[0085] S23 specifically includes:

[0086] S231: selecting a set of reference point cloud data and initializing a transformation matrix for representing an initial alignment relationship between the source point cloud data and the target point cloud data. The initial transformation matrix is ​​usually set to a unit matrix to achieve preliminary alignment of the point cloud data.

[0087] S232: for each point in the source point cloud data, find the corresponding point in the target point cloud data that is closest to it, and specifically determine the best matching relationship by calculating the distance between each pair of points to ensure accurate pairing of the point cloud data;

[0088] S233: using the least squares method, calculating the optimal rigid transformation matrix from the source point cloud data to the target point cloud data; this step aims to minimize the total distance error between the source point cloud data after transformation and the target point cloud data, thereby achieving high-precision alignment;

[0089] S234: Apply the calculated optimal transformation matrix to the source point cloud data to update its position; and repeat steps S232 and S233 to gradually optimize the registration results until the error between the point clouds is reduced to below a predetermined threshold, thereby ensuring the convergence and accuracy of the registration process.

[0090] The specific steps of the above iterative closest point algorithm are as follows:

[0091] Initial registration step: Selecting a reference point cloud dataset , and initialize the transformation matrix ,in , is the initial rotation matrix, is the initial translation vector, and the initial values ​​are set to the unit matrix and the zero vector respectively;

[0092] Closest point matching step: For the source point cloud dataset Every point in , in the target point cloud dataset Find the nearest point , the calculation formula is: ,in, represents the Euclidean distance, It is the result of calculating the Euclidean distance between two points, and the two points with the shortest Euclidean distance are considered as matching relationships;

[0093] Optimal transformation matrix calculation steps: Calculate the rotation matrix using the least squares method and translation vectors , to minimize the following objective function: ,in, is the rotation matrix, is the translation vector, is the total number of points;

[0094] Iterative update step: Apply the optimized transformation matrix Source point cloud dataset Transform to obtain the transformed point cloud dataset ; Then update the reference point cloud dataset for , and Assign to ; Repeat the above steps until the change in the transformation matrix meets the following convergence conditions: ,in, is a predetermined convergence threshold; through the above steps, the standardized point cloud data from different perspectives can be accurately registered, providing a reliable data basis for subsequent geometric feature extraction and construction deviation analysis.

[0095] S24 specifically includes:

[0096] S241: Calculate weighting factors for each point cloud dataset , the weighting factor is determined based on the acquisition quality of the point cloud data and the importance of the viewing angle, and the calculation formula is: ,in, Indicates The weighting factor of the point cloud dataset of each viewpoint; Indicates Quality scoring of point cloud datasets from different viewpoints; Indicates Quality scoring of point cloud datasets from different viewpoints; Represents the total number of point cloud datasets from all viewpoints;

[0097] S242: For each point Coordinates in the point cloud dataset , according to the weighting factor Perform weighted calculation to obtain weighted coordinates , the formulas are: ; ; ;in, Respectively represent From a perspective The X, Y, and Z coordinates of a point in three-dimensional space; Respectively represent the weighted From a perspective The X, Y, and Z coordinates of a point;

[0098] S243: Weighted coordinates in the weighted point cloud dataset of all viewpoints Accumulate and generate a unified 3D point cloud model , the expression is: ,in, Represents the final generated 3D point cloud model, which contains high-precision point cloud data after fusion of all view angles; Indicates The above weighted method enables high-quality perspective data to obtain a larger weight during data fusion, thereby playing a more important role in the final generated 3D point cloud model and improving the accuracy and reliability of the entire model.

[0099] S3 specifically includes:

[0100] S31: The normal vector calculation method is used to calculate the normal vector of each point based on the neighborhood point set of each point in the three-dimensional point cloud model, and the point set belonging to the steel beam, steel column and connection point in the point cloud data is screened out according to the direction of the normal vector and the point cloud density characteristics; the calculation formula of the normal vector is: ,in, For point The normal vector of For point The set of points in the neighborhood; For point The coordinates of Neighborhood points The coordinates of For point The size of the neighborhood point set;

[0101] S32: Based on the point set selected in S31, a clustering-based algorithm is used to classify the steel beams, steel columns and connection points into different geometric entities;

[0102] Specifically, the least squares method is used to fit the steel beam point set to a straight line, and the formula is: ,in, is a point on the line; is the starting point of the straight line; is the direction vector of the line; is a parameter used to control the position of the point on the line;

[0103] The cylinder fitting algorithm is used to fit the steel column point set, and the radius and height of the cylinder are optimized by the least squares method. The formula is: ,in, For point The coordinates of is the center coordinate of the cylinder; is the radius of the cylinder;

[0104] Use a distance-based clustering algorithm (such as DBSCAN) to cluster the set of connection points; the formula is:

[0105] ,

[0106] in, are two points in the point set; is the distance between two points; ;

[0107] S33: Based on S32, the spatial coordinates, shape and size information of the steel beam, steel column and connection point are extracted respectively, wherein the spatial coordinates of the steel beam are obtained by the straight line fitting parameters, the shape is a straight line segment, and the size is the length and cross-sectional size of the steel beam; the spatial coordinates of the steel column are obtained by the cylinder fitting parameters, the shape is a cylinder, and the size is the height and diameter of the steel column; the spatial coordinates of the connection point are obtained by the point clustering algorithm, the shape is a point, and the size is the coordinates of the connection point; for example, the formula for calculating the length of the steel beam is: ,in, is the length of the steel beam; and are the spatial coordinates of the starting and ending points of the steel beam; the calculation formula for the steel column diameter is: ,in, is the diameter of the steel column; is the radius obtained by cylinder fitting;

[0108] S34: establishing a three-dimensional steel structure model according to the geometric feature information of the steel structure extracted in S33, wherein the steel beams, steel columns and connection points in the model are accurately located and constructed according to their spatial coordinates and size information;

[0109] The three-dimensional steel structure model expression is: ,in, For the A three-dimensional representation of a geometric feature (beam, column or connection point), For all geometric features; through the above technical scheme, the geometric features of the steel structure are extracted from the 3D point cloud model, including the spatial coordinates, shape and size information of the steel beams, steel columns and connection points, and the geometric characteristics of each structure are accurately extracted by using straight line fitting, cylinder fitting and clustering methods; these steps ensure the accuracy and completeness of the 3D model of the steel structure, and thus provide high-precision data support for subsequent construction deviation analysis, quality inspection, etc.

[0110] S4 specifically includes:

[0111] S41: Acquire geometric information of the steel structure in the design drawing, including spatial coordinates, shape and size information of steel beams, steel columns and connection points; and establish a three-dimensional model of the design drawing by extracting geometric features in the drawing data; in this step, firstly use the drawing parsing algorithm to extract the geometric information of the steel structure in the construction drawing, and the extracted content includes the length of the steel beam, the diameter and height of the steel column, the spatial coordinates of the connection point, etc., to form a digital three-dimensional model of the design drawing;

[0112] S42: Based on the least squares method, the rigid registration algorithm is used to spatially align the three-dimensional model of the steel structure established in S3 and the three-dimensional model of the design drawing extracted in S41. The formula is: ,in, is the minimum error after registration; is the rotation matrix; is the translation vector; For the design drawing 3D model The coordinates of the points; The first The coordinates of the points; is the total number of points in the point cloud;

[0113] S43: Calculate the spatial distance between each corresponding point of the registered three-dimensional model of the steel structure and the three-dimensional model of the design drawing. If the spatial distance exceeds a set threshold, the point is considered to be a deviation point.

[0114] S44: All calculated deviation point data are aggregated to form a construction deviation report, including the specific location of the deviation point, deviation value and deviation type (such as displacement, angular error, etc.); through the above steps, the geometric differences between the actually constructed steel structure and the design drawing can be accurately compared, and an accurate construction deviation report can be generated; this method ensures the precise alignment between models through the rigid registration algorithm of the least squares method, and uses spatial distance calculation to identify deviation points, ensuring high-precision detection of deviations; the generated construction deviation report can record the deviation location and size in detail, providing accurate data support for subsequent quality control and safety inspections.

[0115] S5 specifically includes:

[0116] S51: Acquire construction environment data, including meteorological information, soil conditions, construction equipment conditions, and real-time dynamic data of the construction site, including temperature and humidity. In this step, the construction environment data is collected through sensors, meteorological stations, equipment monitoring systems and other equipment to ensure the real-time and accuracy of the data.

[0117] S52: Standardize different data sources to ensure consistent data dimensions for subsequent analysis;

[0118] S53: The deviation point information in the construction deviation report is integrated with the construction environment data to construct a multidimensional data graph, in which each node represents a construction element (such as a steel beam, steel column, etc.), and the characteristics of the node include the spatial position of the corresponding element, geometric differences, and correlation characteristics with the construction environment data;

[0119] S54: Use the graph neural network algorithm to train the multidimensional data graph to identify potential quality problems, including structural deformation, cracks or dislocations. In this step, the graph neural network learns the relationship between nodes by transmitting information between nodes, thereby identifying possible structural quality problems. The graph neural network model uses graph convolution operations to update the features of each node and finally identify potential construction problems.

[0120] S55: Evaluate the impact of quality problems on construction safety based on the severity of the quality problems and construction environment conditions;

[0121] S56: Generate a construction risk assessment report based on the identified quality problems and assessed safety risks, including the nature, scope of impact and risk level of each potential quality problem; through the above technical solution, it is possible to combine construction environment data and use the graph neural network algorithm to conduct in-depth analysis of the construction deviation report, identify potential quality problems such as structural deformation, cracks or dislocation, and evaluate the possible safety risks they may bring, providing an important basis for construction safety management and effectively improving the safety and intelligence level of the construction process.

[0122] S54 specifically includes:

[0123] S541: Feature encoding is performed on each node in the multidimensional data graph, where each node represents an element in the steel structure, and the node features include spatial position, geometric difference, deviation, and features related to the construction environment data; the node feature encoding process of the graph neural network is used to convert the node's geometric data and environmental data into vector representations, so that each node can embed rich information in the graph, which is convenient for subsequent learning;

[0124] S542: Use the graph convolution network (GCN) to perform graph convolution operations, combine the edge information between nodes (that is, the relationship between different nodes), update the node features, and thus capture the local and global dependencies between nodes; during the graph convolution process, the feature vector of each node is updated by weighted average with adjacent nodes; the formula is as follows: ,in, For the Layer Node feature vector, For Node The set of neighbor nodes of For Node and nodes The edge weights between and Respectively The weight matrix and bias term of the layer, is the activation function (such as ReLU);

[0125] S543: After completing the multi-layer graph convolution, the final features of each node are input into a fully connected network (FCN) for classification to identify potential quality problems, including structural deformation, cracks or misalignment. Through the above technical solution, the graph neural network algorithm can be used to deeply analyze the construction deviation data and automatically identify potential quality problems (such as structural deformation, cracks or misalignment).

[0126] S55 specifically includes:

[0127] S551: Based on the quality issues identified in S54, for each issue, calculate its weighted risk value, the formula is: ,in, For quality issues The risk value indicates the potential impact of the problem on construction safety. For quality issues The weighting coefficient of , which indicates the relative importance of the problem in the construction process, For quality issues The severity of the problem is given by the regression output value, reflecting its urgency and impact range;

[0128] S552: Combine construction environment data (such as temperature, humidity, vibration, etc.) to calculate the influencing factors of the construction environment , which indicates the amplification or alleviation effect of the construction environment on quality problems; the formula is: ,in, is the construction environment influencing factor, For the construction environment Items of data (such as temperature, humidity, etc.), For the The weight coefficient of the construction environment data, is the number of environmental data items;

[0129] S553: ​​Combine the quality problem risk value and construction environment impact factors, and use the weighted average method to calculate the total risk value , represents the overall construction safety risk, and the formula is: ,in, is the overall construction safety risk value, is the number of quality issues, For quality issues The risk value, is the construction environment influencing factor;

[0130] S554: Based on the total risk value , the construction risk assessment is divided into different safety levels, which are judged by setting thresholds:

[0131] like If it is higher than the first threshold, the construction risk is high and safety measures need to be taken immediately;

[0132] like If it is below the first threshold but above the second threshold, the construction risk is medium and regular inspection is required;

[0133] like If it is lower than the second threshold, the construction risk is low and no additional safety measures need to be taken immediately. The above technical solution can comprehensively consider the severity of quality problems and construction environment conditions to accurately assess the safety risks in the construction process. The weighted average method is used to comprehensively calculate the risk value, combined with actual environmental data to ensure that the risk assessment results are more accurate and reliable. At the same time, a hierarchical management method is used to provide more effective decision-making support for safety management in the construction process, which helps to prevent potential safety hazards and ensure construction safety.

[0134] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0135] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. Intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing, characterized in that: The following steps are involved: S1: Use the laser radar and high-definition camera carried by the drone to inspect the steel structure construction site and collect point cloud data of the geometric shape, size and position information of each part of the steel structure; S2: De-noise and standardize the point cloud data collected by S1, and use the iterative closest point algorithm to register and fuse the point cloud data from different perspectives to form a three-dimensional point cloud model; The S2 specifically includes: S21: Apply a statistical outlier removal algorithm to the point cloud data collected in S1, and remove noise points in the point cloud data by setting a distance threshold to ensure the accuracy of the remaining data; S22: performing scale normalization processing on the denoised point cloud data, unifying the coordinate range of the point cloud data, and eliminating the scale difference between data collected by different sensors; S23: Using the iterative closest point algorithm, the standardized point cloud data from different viewpoints are registered, and the distance error between point clouds is minimized through an iterative optimization process; S24: Based on the registered point cloud data, the multi-view point cloud data are fused using a weighted average method to generate a three-dimensional point cloud model; The S24 specifically includes: S241: Calculate weighting factors for each point cloud dataset , the calculation formula is: ,in, Indicates The weighting factor of the point cloud dataset of each viewpoint; Indicates Quality scoring of point cloud datasets from different viewpoints; Indicates Quality scoring of point cloud datasets from different viewpoints; Represents the total number of point cloud datasets from all viewpoints; S242: For each point Coordinates in the point cloud dataset , according to the weighting factor Perform weighted calculation to obtain weighted coordinates , the formulas are: ; ; ;in, Respectively represent From a perspective The X, Y, and Z coordinates of a point in three-dimensional space; Respectively represent the weighted From a perspective The X, Y, and Z coordinates of a point; S243: Weighted coordinates in the weighted point cloud dataset of all viewpoints Accumulate and generate a unified 3D point cloud model , the expression is: ,in, Represents the final generated three-dimensional point cloud model; Indicates A weighted point cloud dataset with 1000 points of view; S3: Extract the geometric features of the steel structure from the 3D point cloud model, including the spatial coordinates, shape and size information of the steel beams, steel columns and connection points, and establish a 3D model of the steel structure; S4: Compare the three-dimensional model of the steel structure established in S3 with the construction drawings, identify the geometric differences between the actually constructed steel structure and the design drawings, and generate a construction deviation report; S5: Combined with the construction environment data, the graph neural network algorithm is used to analyze the construction deviation report generated in S4, identify potential quality problems in construction, evaluate the safety risks caused by quality problems, and generate a construction risk assessment report; The S5 specifically includes: S51: Obtain construction environment data, including meteorological information, soil conditions, construction equipment conditions, and real-time dynamic data of the construction site, including temperature and humidity; S52: Standardize different data sources to ensure consistent data dimensions; S53: integrating the deviation point information in the construction deviation report with the construction environment data to construct a multidimensional data graph, wherein each node represents a construction element, and the characteristics of the node include the spatial position of the corresponding element, geometric differences, and correlation characteristics with the construction environment data; S54: Using a graph neural network algorithm, training the multidimensional data graph to identify potential quality problems, including structural deformation, cracks, or dislocations; S55: Evaluate the impact of quality problems on construction safety based on the severity of the quality problems and construction environment conditions; S56: Generate a construction risk assessment report based on the identified quality issues and assessed safety risks, including the nature, scope of impact and risk level of each potential quality issue.

2. The intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing according to claim 1 is characterized in that: The S1 specifically includes: S11: Based on the specific layout and monitoring requirements of the steel structure construction site, the flight path of the drone is pre-planned to ensure coverage of all steel structure construction areas; S12: The laser radar carried by the drone scans according to the predetermined flight path, collects the distance and reflection intensity data of each part of the steel structure in real time, and generates point cloud data containing at least 100,000 points per square meter; S13: Use the high-definition camera on the drone to capture high-resolution images of the construction site and obtain the texture and color information of the steel structure surface; S14: Time synchronization processing is performed on the lidar data collected in S12 and S13 and the high-definition camera image to ensure accurate data alignment.

3. The intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing according to claim 1 is characterized in that: The S23 specifically includes: S231: Select a set of reference point cloud data and initialize a transformation matrix for representing an initial alignment relationship between the source point cloud data and the target point cloud data; S232: for each point in the source point cloud data, find the corresponding point in the target point cloud data that is closest to it, and specifically determine the best matching relationship by calculating the distance between each pair of points; S233: Calculate the optimal rigid transformation matrix from the source point cloud data to the target point cloud data using the least square method; S234: Apply the calculated optimal transformation matrix to the source point cloud data to update its position; and repeat steps S232 and S233 to gradually optimize the registration result until the error between the point clouds is reduced to below a predetermined threshold.

4. The intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing according to claim 1 is characterized in that: The S3 specifically includes: S31: Calculate the normal vector of each point based on the neighborhood point set of each point in the three-dimensional point cloud model through a normal vector calculation method, and filter out the point set belonging to the steel beam, steel column and connection point in the point cloud data according to the direction of the normal vector and the point cloud density characteristics; S32: Based on the point set selected in S31, a clustering-based algorithm is used to classify the steel beams, steel columns and connection points into different geometric entities; S33: Based on S32, the spatial coordinates, shapes and size information of the steel beams, steel columns and connection points are extracted respectively; S34: establishing a three-dimensional steel structure model according to the geometric feature information of the steel structure extracted in S33; The three-dimensional steel structure model expression is: ,in, For the A three-dimensional representation of geometric features, for all geometric features.

5. The intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing according to claim 1 is characterized in that: The S4 specifically includes: S41: Obtaining geometric information of the steel structure in the design drawing, including spatial coordinates, shape and size information of steel beams, steel columns and connection points; and establishing a three-dimensional model of the design drawing by extracting geometric features in the drawing data; S42: Based on the least squares method, a rigid registration algorithm is used to spatially align the three-dimensional model of the steel structure established in S3 and the three-dimensional model of the design drawing extracted in S41; S43: Calculate the spatial distance between each corresponding point of the registered three-dimensional model of the steel structure and the three-dimensional model of the design drawing. If the spatial distance exceeds a set threshold, the point is considered to be a deviation point. S44: All calculated deviation point data are aggregated to form a construction deviation report, including the specific location of the deviation point, the deviation value and the deviation type.

6. The intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing according to claim 1 is characterized in that: The S54 specifically includes: S541: feature encoding is performed on each node in the multidimensional data graph, wherein each node represents an element in the steel structure, and the node features include spatial position, geometric difference, deviation amount, and features related to the construction environment data; S542: Use the graph convolution network to perform graph convolution operations, combine the edge information between nodes, update the node features, and thus capture the local and global dependencies between nodes; S543: After completing the multi-layer graph convolution, the final features of each node are input into a fully connected network for classification to identify potential quality issues including structural deformation, cracks or misalignment.

7. The intelligent monitoring method for steel structure construction based on drone inspection and point cloud processing according to claim 6 is characterized in that: The S55 specifically includes: S551: Based on the quality issues identified in S54, for each issue, calculate its weighted risk value, the formula is: ,in, For quality issues The risk value, For quality issues The weighting coefficient of For quality issues the severity of the disease; S552: Combine construction environment data to calculate the influencing factors of the construction environment ; The formula is: ,in, is the construction environment influencing factor, For the construction environment Item data, For the The weight coefficient of the construction environment data, is the number of environmental data items; S553: ​​Combine the quality problem risk value and construction environment impact factors, and use the weighted average method to calculate the total risk value , represents the overall construction safety risk, and the formula is: ,in, is the overall construction safety risk value, is the number of quality issues, For quality issues The risk value, is the construction environment influencing factor; S554: Based on the total risk value , the construction risk assessment is divided into different safety levels, which are judged by setting thresholds: like If it is higher than the first threshold, the construction risk is high and safety measures need to be taken immediately; like If it is below the first threshold but above the second threshold, the construction risk is medium and regular inspection is required; like Below the second threshold, the construction risk is low and no additional safety measures are needed immediately.

Images