Building construction measurement method and system, electronic equipment and storage medium

Abstract

The invention provides a building construction measurement method and system, electronic equipment and a storage medium, and relates to the technical field of building construction measuration.The building construction measurement method comprises the steps that steel structure nodes are shot at multiple angles through image acquisition equipment, and original images are fused to obtain a target image containing a weld joint contour and a bolt spacing; calling a twinning neural network to compare the twinning neural network with geometric features of corresponding nodes of a preset BIM model to obtain weld joint and bolt distance difference information; a phased array microwave radar scans the surface of the component, the distance is measured to form multiple sets of data, a three-dimensional point cloud grid is generated through three-dimensional reconstruction, and the three-dimensional point cloud grid is compared with theoretical coordinates of corresponding nodes of the BIM model to obtain space coordinate deviation; and associating the three types of data to obtain a comprehensive deviation, fusing the point cloud to generate an installation quality evaluation report, collecting data through images and radars, combining a twin neural network and a three-dimensional reconstruction technology, comparing actual characteristics of a steel structure with a BIM model, and integrating deviation information to generate a quality evaluation report, thereby realizing comprehensive measurement and evaluation of construction quality.

Description

Technical Field

[0001] This application relates to the technical field of building construction surveying, and in particular to a building construction surveying method, system, electronic device and storage medium. Background Technology

[0002] In building construction, the installation accuracy of steel structure components directly affects the safety and stability of the project. In particular, the weld quality, bolt spacing, and overall spatial position deviation at joints require precise measurement for real-time control. During construction, it is necessary to simultaneously acquire the microscopic features of the joints (such as weld contours and bolt spacing) and the macroscopic spatial coordinates of the components, and compare them with the pre-set BIM model to quantify deviations and provide data support for installation adjustments. This requires measurement methods with multi-dimensional, high-precision, and efficient integrated analysis capabilities.

[0003] Currently, to address the aforementioned needs, the industry commonly uses a measurement solution based on 3D laser scanning. This involves scanning the entire steel structure component using a laser scanner, generating point cloud data, extracting surface features of the component, comparing the coordinates with the BIM model using a point cloud matching algorithm, outputting spatial position deviation data, and supplementing microscopic feature information such as welds and bolts by manually identifying point cloud details.

[0004] However, the existing solution has obvious limitations. Although 3D laser scanning can obtain macroscopic spatial coordinates, it is not accurate enough in capturing microscopic features such as the edge lines of the weld contour and the subtle values ​​of bolt spacing. It is prone to losing details due to the limitation of point cloud resolution. Moreover, its data processing relies on manual intervention in the identification of microscopic features. It lacks intelligent correlation analysis capabilities when comparing with the BIM model, making it difficult to achieve efficient integration of microscopic features and macroscopic coordinates. This results in low efficiency and insufficient completeness in deviation quantification, and cannot meet the needs of synchronous and accurate measurement of multi-dimensional data in construction scenarios. Summary of the Invention

[0005] The purpose of this application is to provide a construction surveying method, system, electronic device, and storage medium to solve the problem of low accuracy in construction surveying in the prior art.

[0006] To solve the above-mentioned technical problems, in a first aspect, this application provides a construction surveying method, comprising: During the construction process, the joints of steel structure components are photographed from multiple angles using image acquisition equipment to obtain original images from multiple angles. All original images are then fused to obtain a target image that includes the weld contour and bolt spacing at the joint. The twin neural network is invoked to compare the weld contour and bolt spacing in the target image with the geometric features of the corresponding node in the preset BIM model to obtain weld contour difference information and bolt spacing difference information. The entire surface of the steel structure component is scanned from multiple angles using a phased array microwave radar array. During the scanning process, the distance between the radar in the phased array microwave radar array and the entire surface is measured to form multiple sets of distance data. The multiple sets of distance data are combined using 3D reconstruction technology to generate a 3D point cloud mesh. The actual spatial coordinates in the 3D point cloud mesh are compared with the theoretical spatial coordinates of the corresponding node positions in the preset BIM model to obtain spatial coordinate deviation data. The spatial coordinate deviation data, the weld contour difference information, and the bolt spacing difference information are correlated to obtain comprehensive deviation data. The comprehensive deviation data is then fused with the three-dimensional point cloud mesh to generate a comprehensive evaluation report on the installation quality of the steel structure components.

[0007] Optionally, the step of calling the Siamese neural network to compare the weld contour and bolt spacing in the target image with the geometric features of the corresponding node in the preset BIM model to obtain weld contour difference information and bolt spacing difference information includes the following steps: The first sub-network of the twin neural network performs layer-by-layer feature processing on the edge line information of the weld contour and the standard edge line information in the preset BIM model to obtain the first local feature and the second local feature. The difference calculation is performed on the coordinate intermediate feature and the angle intermediate feature of each point in the first local feature and the second local feature to obtain the first feature vector. The second sub-network of the twin neural network performs layer-by-layer feature processing on the numerical information of the bolt spacing and the numerical information of the standard bolt spacing in the preset BIM model to obtain the first spacing length feature and the second spacing length feature. The difference between the first spacing length feature and the second spacing length feature is calculated to output the second feature vector. Based on the output layer of the twin neural network, the weld contour difference information is determined based on the first feature vector, and the bolt spacing difference information is determined based on the second feature vector.

[0008] Optionally, the step of performing layer-by-layer feature processing on the edge line information of the weld contour and the standard edge line information in the preset BIM model through the first sub-network of the twin neural network to obtain the first local feature and the second local feature includes the following steps: The recognition layer of the first sub-network identifies feature points whose curvature values ​​exceed a preset curvature threshold on the edge lines in the edge line information and the edge lines in the standard edge line information, and treats the edge lines between adjacent feature points as continuous line segments. Through the feature extraction layer of the first sub-network, point positions are selected on the continuous line segment at preset intervals, and the horizontal and vertical coordinate values ​​of all the point positions in the two-dimensional plane are recorded to form coordinate intermediate features. Through the computation layer of the first sub-network, the angle between the line connecting the two endpoints on the continuous line segment and the horizontal reference direction is calculated, as well as the angle change of the line connecting adjacent points on the continuous line segment, forming an angle-type intermediate feature. Through the feature association layer of the first sub-network, the coordinate-class intermediate features and the angle-class intermediate features of the edge line information of the weld contour are associated to obtain the first local feature, and the coordinate-class intermediate features and the angle-class intermediate features of the standard edge line information are associated to obtain the second local feature.

[0009] Optionally, the step of performing layer-by-layer feature processing on the numerical information of the bolt spacing and the numerical information of the standard bolt spacing in the preset BIM model through the second sub-network of the Siamese neural network to obtain the first spacing length feature and the second spacing length feature includes the following steps: The identification layer of the second sub-network identifies the position information in the numerical information of the bolt spacing and the position information in the numerical information of the standard bolt spacing, respectively. Based on the position information, two adjacent bolts are paired to form a spacing unit, wherein the spacing unit contains the position identifier of two adjacent bolts and the corresponding spacing length value. Through the first association layer of the second sub-network, the position identifiers corresponding to two adjacent bolts in each spacing unit are associated with the corresponding spacing length values ​​to form a numerical intermediate feature; Through the computation layer of the second sub-network, the spacing units are arranged according to the bolt arrangement order on the steel structure component to form a bolt arrangement sequence. The proportional relationship between the consecutive spacing length values ​​within the bolt arrangement sequence is calculated to form intermediate features of the relationship class. Through the second association layer of the second sub-network, the numerical class intermediate features of the bolt spacing numerical information and the standard bolt spacing numerical information are associated with the relation class intermediate features to obtain the first spacing length feature and the second spacing length feature.

[0010] Optionally, determining the weld contour difference information based on the first feature vector and the bolt spacing difference information based on the second feature vector includes the following steps: Calculate the difference in horizontal coordinate values ​​and the difference in vertical coordinate values ​​between each actual point in the first local feature of the first feature vector and the corresponding standard point in the second local feature to obtain the coordinate value difference; Calculate the angle difference between the included angle value in the first local feature and the included angle value in the second local feature, and the difference between the angle change values ​​between corresponding adjacent points in the first local feature and the second local feature, to obtain the angle difference. The coordinate value differences and angle value differences are associated in the order of continuous line segments and integrated into weld contour difference information that includes the positional deviation and direction deviation of each continuous line segment. Calculate the difference between the first spacing length feature and the second spacing length feature in the second feature vector to obtain the basic spacing difference. Based on the continuous spacing units in the bolt arrangement sequence, calculate the difference between the proportional relationship in the first spacing length feature and the proportional relationship in the second spacing length feature to obtain the proportional difference. The basic spacing difference and the proportional difference are associated according to the order of the bolt arrangement sequence, and integrated into bolt spacing difference information that includes the length deviation of each spacing unit and the sequence proportional deviation.

[0011] Optionally, the step of combining the multiple sets of distance data using 3D reconstruction technology to generate a 3D point cloud mesh, and comparing the actual spatial coordinates in the 3D point cloud mesh with the theoretical spatial coordinates of the corresponding node positions in the preset BIM model to obtain spatial coordinate deviation data includes the following steps: By using 3D reconstruction technology, the distance information of any node position in the multiple sets of distance data collected at different angles is associated to obtain the actual spatial coordinates of the node position in 3D space. The actual spatial coordinates of all the node positions are integrated to form a 3D point set. The adjacent node positions in the 3D point set are connected to form a 3D point cloud mesh. The actual spatial coordinates of each node position in the three-dimensional point cloud mesh in three-dimensional space are compared with the theoretical spatial coordinates of the corresponding node position in the preset BIM model. The numerical differences between the actual spatial coordinates and the theoretical spatial coordinates of the node position in three axes are calculated respectively. The numerical differences in the three axes are integrated to generate the spatial coordinate deviation data of the node position.

[0012] Optionally, the step of associating the spatial coordinate deviation data, the weld contour difference information, and the bolt spacing difference information to obtain comprehensive deviation data, and then fusing the comprehensive deviation data with the three-dimensional point cloud mesh to generate a comprehensive evaluation report on the installation quality of the steel structure component, includes the following steps: The spatial coordinate deviation, weld contour difference, and bolt spacing difference at the same node position of the steel structure component are integrated to form comprehensive deviation data that characterizes the overall deviation value and distribution of the steel structure component. The comprehensive deviation data is associated with the actual spatial coordinates of the corresponding node positions in the three-dimensional point cloud mesh to form a three-dimensional model containing deviation information. Based on the three-dimensional model, the comprehensive deviation data within the same installation area of ​​the steel structure component are classified and organized according to the overall installation area. The overall deviation value and distribution of all comprehensive deviation data within each installation area are recorded to form a comprehensive evaluation report on the installation quality of the steel structure component.

[0013] Secondly, this application provides a construction surveying system, comprising: The acquisition module is used to capture images of the joints of steel structure components from multiple angles during the construction process using image acquisition equipment, obtain original images from multiple angles, and fuse all the original images to obtain a target image containing the weld contour and bolt spacing at the joint. The first comparison module is used to call the twin neural network to compare the weld contour and bolt spacing in the target image with the geometric features of the corresponding node in the preset BIM model, so as to obtain weld contour difference information and bolt spacing difference information. The measurement module is used to perform multi-angle scanning of the overall surface of the steel structure component using a phased array microwave radar array, and to measure the distance between the radar in the phased array microwave radar array and the overall surface during the scanning process, thereby generating multiple sets of distance data. The second comparison module is used to combine the multiple sets of distance data using three-dimensional reconstruction technology to generate a three-dimensional point cloud mesh, and compare the actual spatial coordinates in the three-dimensional point cloud mesh with the theoretical spatial coordinates of the corresponding node positions in the preset BIM model to obtain spatial coordinate deviation data. The fusion module is used to associate the spatial coordinate deviation data, the weld contour difference information, and the bolt spacing difference information to obtain comprehensive deviation data, and then fuse the comprehensive deviation data with the three-dimensional point cloud mesh to generate a comprehensive evaluation report on the installation quality of the steel structure components.

[0014] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the steps of a construction surveying method as described in the first aspect above.

[0015] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of a construction surveying method as described in the first aspect above.

[0016] This application provides a method for construction surveying. During construction, image acquisition equipment is used to capture multi-angle images of the joints of steel structure components, obtaining original images from multiple angles. These original images are then fused to obtain a target image containing the weld contour and bolt spacing at the joints. A twin neural network is used to compare the weld contour and bolt spacing in the target image with the geometric features of the corresponding joints in a pre-defined BIM model, obtaining weld contour difference information and bolt spacing difference information. A phased array microwave radar is then used to perform multi-angle scanning of the entire surface of the steel structure component. During the scanning process... The distance between the radar in the phased array microwave radar array and the overall surface is measured to form multiple sets of distance data. Three-dimensional reconstruction technology is used to combine these multiple sets of distance data to generate a three-dimensional point cloud mesh. The actual spatial coordinates in the three-dimensional point cloud mesh are compared with the theoretical spatial coordinates of the corresponding node positions in the preset BIM model to obtain spatial coordinate deviation data. The spatial coordinate deviation data, weld contour difference information, and bolt spacing difference information are correlated to obtain comprehensive deviation data. The comprehensive deviation data is then fused with the three-dimensional point cloud mesh to generate a comprehensive evaluation report on the installation quality of the steel structure components.

[0017] The technical solution of this application has the following beneficial effects: This application achieves multi-dimensional collaboration in steel structure construction measurement through a complete process from image acquisition and fusion, twin neural network comparison, radar scanning, 3D reconstruction to data association and fusion: multi-angle image fusion ensures that microscopic features such as weld contours and bolt spacing at nodes are fully captured; the twin neural network intelligently compares microscopic features with the BIM model, accurately outputting the difference information of welds and bolt spacing; phased array microwave radar scanning combined with 3D reconstruction obtains the overall macroscopic spatial coordinates of the components and quantifies the deviation from the BIM model; finally, the microscopic and macroscopic deviation data are correlated and fused with 3D point clouds to generate a comprehensive evaluation report. This not only achieves synchronous measurement of node details and overall space, but also improves the completeness and accuracy of deviation quantification through multi-source data integration, providing efficient and accurate technical support for comprehensive control of steel structure installation quality.

[0018] Furthermore, this application processes the weld contour edge line information and standard information layer by layer through the first sub-network. First, feature points are identified and continuous line segments are determined. Intermediate features of coordinate and angle classes are extracted and associated to obtain the first and second local features. Then, the corresponding coordinate and angle differences are calculated and integrated into weld contour difference information. Through the second sub-network, the bolt spacing numerical information and standard information are processed layer by layer. First, position information is identified to form spacing units. The numerical value and position identifier are associated and the proportional relationship of the arrangement sequence is calculated to obtain the first and second spacing length features. Then, the basic spacing and proportional differences are calculated and integrated into bolt spacing difference information. Finally, the two types of difference information are determined based on the feature vectors output by the two sub-networks.

[0019] This application achieves precise capture of the coordinate and angular features of weld contour edge lines and the length and proportion of bolt spacing through hierarchical feature extraction and association using a twin neural network. By performing targeted difference calculations, the micro-features are compared with the corresponding features of the preset BIM model in a fine-grained manner, ultimately obtaining weld contour difference information including position and orientation deviations and bolt spacing difference information including length and proportion deviations. This provides comprehensive and accurate micro-feature deviation data support for the quality assessment of steel structure nodes.

[0020] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A schematic flowchart of a construction surveying method provided in this application embodiment; Figure 2 A scene diagram illustrating a construction surveying method provided in an embodiment of this application; Figure 3 A structural schematic diagram of a building construction surveying system provided in this application embodiment; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0023] Existing measurement solutions based on 3D laser scanning can acquire the macroscopic spatial coordinates of components in steel structure construction measurement, but their accuracy in capturing microscopic features such as the edge details of weld contours and the minute values ​​of bolt spacing is limited. Insufficient scanning resolution often leads to the loss of details. At the same time, data processing requires manual assistance to identify microscopic features, and the comparison with BIM models lacks intelligent integration capabilities. It is difficult to efficiently correlate microscopic feature deviations with macroscopic spatial coordinate deviations, resulting in insufficient completeness and efficiency of measurement results. This fails to meet the needs of synchronous and accurate control of multi-dimensional data during construction.

[0024] To address the aforementioned issues, this application proposes a construction surveying method. This method involves capturing images of key nodes from multiple angles using image acquisition equipment and fusing the images to accurately capture microscopic features such as weld contours and bolt spacing. A twin neural network (an intelligent comparison model) is then used to automatically compare these features with corresponding parts in the BIM model, outputting the differences in microscopic features. Simultaneously, a phased-array microwave radar scans the component surface, and combined with 3D reconstruction to generate a 3D model. Macroscopic spatial coordinates are obtained and compared with the BIM model to determine coordinate deviations. Finally, the microscopic and macroscopic deviation data are correlated and fused into a comprehensive 3D model, generating a quality assessment report. This solution, through image focusing on microscopic details, radar capturing macroscopic coordinates, automatic intelligent model comparison, and multi-source data fusion and correlation, not only solves the problems of inaccurate microscopic feature capture and reliance on manual labor in existing methods but also achieves efficient integration of microscopic and macroscopic deviations, improving the accuracy, completeness, and efficiency of the measurement.

[0025] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] The core of this application is to provide a construction surveying method, the flowchart of one specific implementation of which is shown below. Figure 1 As shown, the method includes: S101. During the construction process, the joints of the steel structure components are photographed from multiple angles using image acquisition equipment to obtain original three-dimensional images from multiple angles. All original three-dimensional images are then fused to obtain a target image containing the weld contour and bolt spacing at the joints.

[0027] In the above scheme, image acquisition equipment refers to equipment used to capture images; the node of a steel structure component refers to the connection part of different components in the steel structure, which usually has welds and bolts; multi-angle shooting refers to shooting the same node from different directions and angles; multi-angle original images refer to the initial unprocessed images obtained by shooting from multiple angles; fusion refers to integrating multiple original images to eliminate overlapping information and retain key details; target image refers to the image obtained after fusion that fully presents the details of the node; weld contour refers to the shape and lines of the weld edge; bolt spacing refers to the distance between two adjacent bolts at the node.

[0028] In this application example, step S101 first determines the shooting object and equipment. In building construction, the steel structure component nodes that need to be measured (such as the connection nodes between columns and beams) are identified, and suitable image acquisition equipment (such as an industrial camera with anti-backlight function) is selected to ensure that the equipment resolution meets the requirements for capturing the weld contour and bolt spacing details. For example, for the weld (width about 5-10mm) and bolt (diameter about 12-20mm) at the node, a camera with a resolution of not less than 5 million pixels is selected to ensure clear details.

[0029] Next, multi-angle shooting is conducted around the node, taking pictures from different angles (such as front, left 45°, right 45°, top view, and bottom view, usually 4-8 angles are selected) to ensure that the image from each angle covers different areas of the node, and that there is a certain overlap between adjacent angles (overlap rate of about 30%-50%), providing a matching basis for subsequent fusion. For example, for node A (the connection between the column and the beam), pictures are taken from 5 angles: front, left 30°, right 30°, top view, and bottom view. Each image contains part of the weld and bolt, and about 40% of the area overlaps in adjacent images.

[0030] Finally, the original images are fused to obtain the target image. Image stitching techniques (such as SIFT-based feature point matching algorithms) are used to process the original images from multiple angles. First, feature points (such as the inflection points of the weld edge and the corner points of the bolt edges) are extracted from each image. The positional relationship between the images is determined by feature point matching. Then, pixel-level fusion is performed on the overlapping areas (to eliminate stitching traces). Finally, a target image that completely covers the node area is generated. This image needs to clearly show the complete outline of the weld and the relative positions of all bolts (i.e., bolt spacing). For example, for the 5 original images of node A, 10-15 feature points of the weld and 8-12 feature points of the bolts are extracted from each image. After matching, the image stitching order is determined. In the target image obtained after fusion, the edge lines of the weld are continuous and unbroken, the positions of all bolts are clearly distinguishable, and the distance between adjacent bolts can be directly observed.

[0031] In practical application, during the steel structure construction phase of a building project, construction personnel needed to acquire images of the connection node (node ​​M) between steel columns and beams before measurement. An E-brand industrial camera (6-megapixel resolution) was selected as the image acquisition device. Images were taken from five angles around node M: front, 45° left, 45° right, overhead, and bottom. One image was captured from each angle, resulting in five original images (the imaging area of ​​node M occupies approximately 60% of each image, with approximately 45% overlap between adjacent angles). Subsequently, image processing software was used to fuse the five original images: first, edge feature points of the weld (62 extracted in total) and corner feature points of the bolts (45 extracted in total) were identified in each image. The feature points were matched to determine the splicing coordinates between images. Then, the pixels in the overlapping areas were smoothed to finally generate the target image. This target image fully presents the entirety of node M, with a clear and continuous weld outline, clearly defined positions of the eight bolts, and easily distinguishable spacing between adjacent bolts.

[0032] The above-mentioned S101 overall solution ensures that all key details of the node are captured by multi-angle shooting, avoiding information omissions caused by shooting from a single angle; by using image fusion technology to integrate multiple original images into a complete target image, the perspective limitations of multi-angle shooting are eliminated, and the details of welds and bolts are clearly presented in the same image, providing a comprehensive and accurate image foundation for subsequent feature comparison using twin neural networks, ensuring that the subsequent comparison process can be carried out based on complete visual information.

[0033] S102. The twin neural network is invoked to compare the weld contour and bolt spacing in the target image with the geometric features of the corresponding node in the preset BIM model to obtain weld contour difference information and bolt spacing difference information. The corresponding node refers to the model node in the preset BIM model that corresponds to the actual node of the steel structure component captured in the target image. That is, the target image captures a specific node of a steel structure component, while the preset BIM model pre-designs a complete model of the component, in which the model node completely matches the actual captured node in terms of position and function.

[0034] Optionally, step S102, which involves calling a Siamese neural network to compare the weld contour and bolt spacing in the target image with the geometric features of the corresponding nodes in a preset BIM model to obtain weld contour difference information and bolt spacing difference information, includes the following steps: Step 1021: The edge line information of the weld contour and the standard edge line information in the preset BIM model are processed layer by layer through the first sub-network of the twin neural network to obtain the first local feature and the second local feature. The difference calculation is performed on the coordinate intermediate feature and the angle intermediate feature of each point in the first local feature and the second local feature to obtain the first feature vector.

[0035] Step 1021 may specifically include the following process: Through the recognition layer of the first sub-network, feature points whose curvature values ​​exceed a preset curvature threshold are identified on the edge lines in the edge line information and the edge lines in the standard edge line information, respectively; the edge lines between adjacent feature points are treated as continuous line segments; through the feature extraction layer of the first sub-network, point positions are selected on the continuous line segments at preset intervals, and the horizontal and vertical coordinate values ​​of all point positions in the two-dimensional plane are recorded to form coordinate-type intermediate features; through the calculation layer of the first sub-network, the angle between the line connecting the two endpoints on the continuous line segment and the horizontal reference direction, as well as the angle change value of the line connecting adjacent point positions on the continuous line segment, are calculated to form angle-type intermediate features; through the feature association layer of the first sub-network, the coordinate-type intermediate features of the weld contour edge line information are associated with the angle-type intermediate features to obtain a first local feature, and the coordinate-type intermediate features of the standard edge line information are associated with the angle-type intermediate features to obtain a second local feature.

[0036] The aforementioned intermediate features of coordinate type specifically refer to the intermediate data formed by systematically recording the horizontal and vertical coordinate values ​​of multiple points selected at preset intervals on a continuous line segment. These data intuitively reflect the spatial position and morphological details of the continuous line segment and serve as the basis for subsequent feature comparison. For example, if three points are selected on a continuous line segment with coordinates (x1, y1), (x2, y2), and (x3, y3), the set of these coordinates constitutes the intermediate features of the line segment. The horizontal reference direction refers to the unified horizontal reference direction used when processing target images and preset BIM models. It is usually a fixed, globally universal horizontal direction, such as the bottom horizontal line of the target image (e.g., the edge of a component that is kept horizontal during shooting) or the global horizontal coordinate axis in the preset BIM model (e.g., the positive X-axis direction in the model) as the reference. All angle calculations are performed with reference to this direction.

[0037] Step 1022: The numerical information of the bolt spacing and the numerical information of the standard bolt spacing in the preset BIM model are processed layer by layer through the second sub-network of the twin neural network to obtain the first spacing length feature and the second spacing length feature. The difference between the first spacing length feature and the second spacing length feature is calculated to output the second feature vector. The spacing length value is the core component of the numerical information and is the key data describing the bolt spacing. However, the numerical information is a broader concept. In addition to including the spacing length value of adjacent bolts (such as 100mm), it may also include auxiliary numerical information related to the spacing (such as the bolt number corresponding to the spacing, the position number of the spacing in the arrangement sequence, etc.). The spacing length value only refers to the actual distance between two adjacent bolts and is a direct quantification of the physical length of the spacing.

[0038] Step 1022 may specifically include the following process: Through the identification layer of the second sub-network, the position information in the numerical information of the bolt spacing and the position information in the numerical information of the standard bolt spacing are identified respectively. Based on the position information, adjacent bolts are paired to form a spacing unit, wherein the spacing unit contains the position identifiers of the two adjacent bolts and the corresponding spacing length values; through the first association layer of the second sub-network, the position identifiers corresponding to the two adjacent bolts in each spacing unit are associated with the corresponding spacing length values ​​to form a numerical intermediate feature; through the calculation layer of the second sub-network, the spacing units are arranged according to the bolt arrangement order on the steel structure component to form a bolt arrangement sequence, and the proportional relationship between consecutive spacing length values ​​within the bolt arrangement sequence is calculated to form a relational intermediate feature; through the second association layer of the second sub-network, the numerical intermediate features of the bolt spacing and the standard bolt spacing are associated with the relational intermediate feature to obtain a first spacing length feature and a second spacing length feature.

[0039] Step 1023: Based on the output layer of the twin neural network, determine the weld contour difference information based on the first feature vector, and determine the bolt spacing difference information based on the second feature vector.

[0040] Step 1023 may specifically include the following processes: calculating the difference in horizontal coordinate values ​​and the difference in vertical coordinate values ​​between each actual point in the first local feature of the first feature vector and the corresponding standard point in the second local feature, to obtain coordinate value differences; calculating the difference in angle values ​​between the included angle values ​​in the first local feature and the included angle values ​​in the second local feature, and the difference in the angle change values ​​between corresponding adjacent points in the first local feature and the second local feature, to obtain angle value differences; associating the coordinate value differences and the angle value differences according to the order of continuous line segments, and integrating them into weld contour difference information containing the positional deviation and direction deviation of each continuous line segment; calculating the difference between the first spacing length feature and the second spacing length feature in the second feature vector to obtain basic spacing differences; calculating the difference between the proportional relationship in the first spacing length feature and the proportional relationship in the second spacing length feature based on continuous spacing units in the bolt arrangement sequence, to obtain proportional differences; associating the basic spacing differences and the proportional differences according to the order of the bolt arrangement sequence, and integrating them into bolt spacing difference information containing the length deviation of each spacing unit and the sequence proportional deviation.

[0041] The aforementioned coordinate numerical differences refer to the horizontal coordinate differences (e.g., Δx = x actual - x standard) and vertical coordinate differences (e.g., Δy = y actual - y standard) between actual points on continuous line segments in the target image and their corresponding standard points in the preset BIM model. These can be understood as specific, local numerical differences. Positional deviation, on the other hand, reflects the overall spatial deviation of the entire continuous line segment from the standard line segment after integrating these coordinate numerical differences in the order of the continuous line segments (e.g., the entire line segment shifts to the left or upward). Positional deviation is calculated based on coordinate numerical differences, but these differences are the numerical results of local points; positional deviation is a comprehensive description of the overall positional deviation of the line segment. Angle numerical differences include the angle difference between the line connecting the two endpoints of the continuous line segment and the horizontal reference direction (e.g., Δθ = θ actual - θ standard), and the angle change difference between the lines connecting adjacent points (e.g., Δα = α actual - α standard). These are specific angle numerical differences. Direction deviation, after integrating these angle numerical differences, reflects the deviation of the entire continuous line segment from the standard line segment in its extension direction (direction). The overall situation (such as the line segment tilting upwards at an excessive angle) is considered. The direction deviation is calculated based on the difference in angle values, but the difference in angle values ​​is a local angle result. The direction deviation is a comprehensive description of the overall deviation of the line segment's direction. The spacing unit length deviation refers to the difference between the actual measured bolt spacing length value in each spacing unit (formed by two adjacent bolts, including the position identifiers of these two bolts and the corresponding spacing length value) and the standard length value of the corresponding spacing unit in the preset BIM model. It directly reflects the degree of deviation between the actual length and the standard length of a single pair of adjacent bolts. The sequence ratio deviation refers to the difference between the ratio between the actual length values ​​of multiple consecutive spacing units in the bolt arrangement sequence (multiple spacing units arranged according to the bolt arrangement order on the steel structure component) and the ratio between the actual length values ​​of multiple consecutive spacing units in the preset BIM model. It reflects the degree of deviation of the overall arrangement ratio of multiple adjacent bolts from the standard ratio, reflecting the overall coordination deviation of the bolt arrangement.

[0042] In the above scheme, a twin neural network refers to a model containing two sub-networks with identical structures, used to compare similar features; the target image refers to an image containing the weld contour and bolt spacing of the node; the weld contour refers to the shape of the edge line of the weld; the bolt spacing refers to the distance between two adjacent bolts; the preset BIM model refers to a pre-built building information model containing standard geometric features of the node; geometric features refer to feature data describing shape and position in the model; the first sub-network refers to the feature used to process weld contour-related features; the second sub-network refers to the feature used to process bolt spacing-related features; the first local feature and the second local feature refer to the feature data of the weld contour in the target image and the BIM model, respectively; the first spacing length feature and the second spacing length feature refer to the feature data of the bolt spacing in the target image and the BIM model, respectively; the feature vector is a vector form that integrates feature data; the weld contour difference information refers to the deviation data between the actual weld and the standard feature; the bolt spacing difference information refers to the deviation data between the actual bolt spacing and the standard feature.

[0043] In this application example, the weld contour features are processed through step 1021. The first step is to perform recognition layer processing. The recognition layer of the first sub-network uses an edge detection algorithm to extract the edge line information of the weld in the target image and the standard edge line information in the BIM model. The curvature calculation identifies feature points whose bending degree exceeds a preset threshold (such as a preset bending degree threshold of 30°, where the curvature calculation formula is the bending degree of the curve at a certain point, which is simplified here to be determined as bending if the angle formed by connecting three adjacent points is less than 30°). The straight line segments between adjacent feature points are determined as continuous line segments. For example, the weld edge of node M has 4 feature points (P1, P2, P3, P4), where the curvature of the line segments between P1 and P2, P2 and P3, and P3 and P4 is less than 30°, thus forming 3 continuous line segments (S1: P1-P2, S2: P2-P3, S3: P3-P4). The second step is to perform feature extraction layer processing. The feature extraction layer selects point positions at preset intervals on each continuous line segment and records the X-axis and Y-axis coordinates (in millimeters) of each point, forming coordinate-type intermediate features. For example, the coordinates of 10 points on S1 are (x1, y1), (x2, y2)...(x10, y10), where x1=100, y1=200; x2=105, y2=200, etc. These coordinates constitute the coordinate-type intermediate features of S1. The third step is to perform computational layer processing, using the formula θ=arctan[(y2-y1) / (x2-x1)] to calculate the angle between the line connecting the two endpoints of each continuous line segment and the horizontal reference direction (where (x1, y1) and (x2, y2) are the coordinates of the two endpoints of the line segment), and to calculate the angle change Δθ=|θafter-θbefore| between adjacent points (where θbefore is the angle between the line connecting the first two points and the horizontal line, and θafter is the angle between the line connecting the last two points and the horizontal line), forming the angle-type intermediate features. For example, if the two endpoints of S1 are P1(100, 200) and P2(150, 250), then y2-y1=50, x2-x1=50, θ=arctan(50 / 50)=45°; the θ-before the line connecting the adjacent points (x1, y1) and (x2, y2) on S1 is 45°, and the θ-after the line connecting (x2, y2) and (x3, y3) is 47°, then Δθ=|47°-45°|=2°; the fourth step is to perform feature association layer and difference calculation, special The feature association layer associates the coordinate-type intermediate features and angle-type intermediate features of each continuous line segment in point order to obtain the first local feature (target image weld feature) and the second local feature (BIM model standard weld feature). Then, it calculates the coordinate difference (Δx = x actual - x standard, Δy = y actual - y standard) and angle difference (Δθ angle = θ actual - θ standard, Δθ change = Δθ actual - Δθ standard) of the corresponding points of the two, and integrates these differences in line segment order into the first feature vector.For example, if the actual coordinates of a point in S1 are (102, 201) and the standard coordinates are (100, 200), then Δx = 2 and Δy = 1; if the actual included angle of S1 is 45° and the standard included angle is 43°, then the included angle Δθ = 2°. These data, in sequence, form part of the content of the first feature vector.

[0044] Next, the bolt spacing features are processed in step 1022. The first step is the recognition layer processing, and the recognition layer of the second sub-network uses a target detection algorithm to identify the position information of the bolts in the target image and the standard position information of the bolts in the BIM model. Based on the principle of proximity (e.g., two bolts with a distance of less than 100mm are considered adjacent), adjacent bolts are paired to form spacing units (format: "position identifier: spacing length"). The spacing length is determined by the formula... Calculate, where, Let (x1, y1) and (x2, y2) be the spacing lengths, and (x1, y1) and (x2, y2) be the center coordinates of adjacent bolts. For example, if node M has 8 bolts (B1-B8), after identification, 7 spacing units are formed. The spacing L between B1 (200, 300) and B2 (250, 300) is 50mm, and this unit is denoted as "B1-B2: 50mm". The second step is to perform the first association layer processing, which maps the position identifier of each spacing unit to the spacing length (e.g., "B1-B2" corresponds to "50mm"), forming a numerical intermediate feature. For example, the intermediate features of the numerical class of 7 units are "B1-B2: 50mm; B2-B3: 52mm; ...; B7-B8: 49mm"; the third step is to process the calculation layer. The calculation layer arranges the spacing units into a bolt arrangement sequence according to the bolt arrangement order (e.g., arranged in order from smallest to largest X-axis coordinate, B1 to B8 in sequence). The length ratio of consecutive spacing units in the calculation sequence is R = L_before / L_after (where L_before is the spacing length of the previous unit and L_after is the spacing length of the next unit), forming the intermediate features of the relation class. For example, the ratio of B1-B2 (50mm) to B2-B3 (52mm) in the sequence is R=50 / 52≈0.96, and the ratio of B2-B3 to B3-B4 (51mm) is R=52 / 51≈1.02, etc.; the fourth step is to perform the second association layer and difference calculation. The second association layer associates the intermediate features of the numerical class with the intermediate features of the relation class of each spacing unit (such as "B1-B2: 50mm, with a ratio of 0.96 with B2-B3") to obtain the first spacing length feature (target image bolt feature) and the second spacing length feature (BIM model standard bolt feature); then the spacing difference (ΔL=Lactual-LStandard) and the ratio difference (ΔR=Ractual-RStandard) of the two are calculated and integrated into the second feature vector. For example, if the standard spacing between B1 and B2 is 52mm, then ΔL = 50 - 52 = -2mm; if the standard ratio is 0.98, then ΔR = 0.96 - 0.98 = -0.02. These data, arranged in sequence, form part of the second feature vector.

[0045] Finally, difference information is generated through step 1023. The output layer receives the first feature vector and the second feature vector. Based on the first feature vector, the coordinate numerical differences (such as Δx and Δy of S1, Δx and Δy of S2, etc.) and angle numerical differences (such as the included angle and change of Δθ of S1, the included angle and change of Δθ of S2, etc.) of each continuous line segment are associated in the order of the line segments and integrated into weld contour difference information containing the positional deviation (reflected by Δx and Δy) and direction deviation (reflected by the included angle and change of Δθ) of each line segment. For example, the positional deviation of S1 is Δx=2mm, Δy=1mm, and the direction deviation is Δθ angle=2°, Δθ change=0.5°. These data, after association, become part of the weld contour difference information. At the same time, based on the second feature vector, the basic spacing difference ΔL and the proportional difference ΔR of each spacing unit are associated in the order of the bolt arrangement sequence and integrated into bolt spacing difference information containing the length deviation of each unit (reflected by ΔL) and the sequence proportional deviation (reflected by ΔR). For example, the length deviation of B1-B2 is -2mm and the ratio deviation is -0.02. These data, when correlated, become part of the bolt spacing difference information.

[0046] In practical applications, taking steel structure node M (the connection between a column and a beam) as an example in a certain building project, the first sub-network identifies four feature points (P1-P4) on the weld edge, forming three continuous line segments (S1-S3). S1 is 50mm long, and 10 points are selected at 5mm intervals, with coordinates (100, 200), (105, 200)...(150, 200). The included angle θ between the two ends of S1 is calculated using the formula θ=arctan[(y2-y1) / (x2-x1)], which is 0° (horizontal line segment). The angle change Δθ between adjacent points is 0°. In the BIM model, the standard coordinates of S1 are (100, 200), (105, 199)...(150, 199), with a standard included angle θ=0° and Δθ=0°. After association, the coordinate differences between the first and second local features are calculated as Δx = 0 mm and Δy = 0-1 mm, and the angle difference Δθ = 0° and Δθ change = 0°, which are integrated into the first feature vector. The second sub-network identifies the 8 bolts (B1-B8) of node M, arranged in X-axis order as B1 (200, 300) to B8 (550, 300), and uses the formula... The calculated spacing units are: B1-B2=50mm, B2-B3=52mm, B3-B4=51mm, B4-B5=50mm, B5-B6=53mm, B6-B7=51mm, B7-B8=49mm; the continuous unit ratios are 50 / 52≈0.96, 52 / 51≈1.02, 51 / 50≈1.02, 50 / 53≈0.94, 53 / 51≈1.04, 51 / 49≈1.04. In the BIM model, the standard spacing is 52mm, 52mm, 50mm, 50mm, 52mm, 52mm, 50mm, and the standard scale is 1.00, 1.04, 1.00, 1.00, 1.00, 1.04. After association, the calculated ΔL values ​​are -2mm, 0mm, 1mm, 0mm, 1mm, -1mm, and -1mm, respectively, and the ΔR values ​​are -0.04, -0.02, 0.02, -0.06, 0.04, and -0.00, respectively, which are integrated into the second feature vector. The output layer associates the coordinate difference and angle difference of the first feature vector in the order of S1-S3 to obtain the weld contour difference information (e.g., the position deviation of S1 Δy = 0-1mm, and the direction has no deviation). The ΔL and ΔR of the second feature vector are associated according to the bolt sequence to obtain the bolt spacing difference information (e.g., the length deviation of B1-B2 is -2mm, and the scale deviation is -0.04).

[0047] The aforementioned S102 overall solution, through hierarchical processing of a twin neural network, achieves accurate extraction and comparison of weld contour and bolt spacing features. Specifically, the detailed processing of the coordinate and angular features of the weld contour and the length and proportion features of the bolt spacing ensures the comprehensiveness of the difference information; the automated association and calculation of sub-networks reduces manual intervention and improves comparison efficiency; the final weld contour and bolt spacing difference information clearly reflects the deviation between the actual construction and the BIM model, providing detailed and reliable micro-feature deviation data for subsequent quality assessment.

[0048] S103. The phased array microwave radar array is used to scan the overall surface of the steel structure component from multiple angles. During the scanning process, the distance between the radar in the phased array microwave radar array and the overall surface is measured to form multiple sets of distance data.

[0049] In the above scheme, a phased array microwave radar array refers to a device composed of multiple microwave radar units arranged in a certain manner, and the beam direction can be adjusted by controlling the phase of microwave transmission / reception of each unit; the overall surface of the steel structure component refers to all the exposed outer surfaces of the component; multi-angle scanning refers to scanning the overall surface from different azimuths and angles; multiple sets of distance data refer to the set of distance values ​​measured by different scanning angles and different radar units, and each set of data includes the measurement location identifier and the corresponding distance value.

[0050] In this application example, step S103 first determines the scanning range and parameters. Based on the dimensions of the steel structure component, the scanning coverage range of the phased array microwave radar array is set to ensure that the entire surface of the component is covered. At the same time, the scanning angle interval, microwave signal frequency, and sampling frequency are set. For example, for a 3m long steel column component, the scanning angles are set to 0° (front), 30°, 60°...330°, a total of 12 angles, with the 16 elements of the radar array working synchronously at each angle.

[0051] Next, multi-angle scanning is performed. The phased array microwave radar array uses beamforming technology to control the phase difference of each radar unit, causing the microwave beam to point at a preset angle. Microwave signals are then emitted sequentially from each set angle onto the entire surface of the component. The signals are reflected after contacting the surface, and the radar array receives the reflected signals. By calculating the signal propagation time t, the formula is used... Where d is the distance between the radar and the surface, and c is the speed of light. For example, at a 0° angle, 16 radar units transmit signals towards the front of the steel column. After receiving the reflected signals, the distance values ​​are calculated to be 1.2m, 1.21m, ..., 1.23m. The measurement results of all radar units at each scanning angle are recorded in the format of "angle-radar unit number-distance value". The data from all angles are integrated to form multiple sets of distance data. For example, 12 angles × 16 units yield 192 sets of data. Each set of data, such as "30°-unit 5-1.18m", indicates that the distance measured by the 5th radar unit at a 30° angle is 1.18m. These data will be used for subsequent 3D reconstruction.

[0052] In practical applications, taking steel beam component C (6m long, 0.8m high, and 0.3m wide) as an example in the construction of a steel structure factory, a phased array microwave radar array consisting of 20 radar units was selected. The scanning angles were set to 0°, 45°, 90°, 135°, 180°, 225°, 270°, and 315°, a total of 8 angles. The microwave frequency was 24GHz, and the sampling frequency was 50 times per second. During scanning, the radar array sequentially emits microwave signals from eight angles onto the entire surface of steel beam C (including the upper and lower flanges, web, and connecting nodes). At each angle, 20 elements are measured simultaneously: the distance measured at 0° angle (directly facing the web) ranges from 1.5 to 1.55 m, the distance measured at 45° angle (side to the flange) ranges from 1.3 to 1.33 m, and so on. All data are recorded in the format of "angle-element number-distance", ultimately forming 8×20=160 sets of distance data, such as "0°-element 3-1.52 m" and "45°-element 10-1.31 m". These data will be used to generate a three-dimensional point cloud mesh in the future.

[0053] The aforementioned S103 overall scheme achieves comprehensive coverage measurement of the entire surface of the steel structure component through multi-angle scanning of the phased array microwave radar array, avoiding the information blind spots of single-angle scanning; it utilizes the characteristics of microwave signals and beamforming technology to ensure the accuracy of distance measurement at different locations; the resulting multiple sets of distance data contain the spatial position information of each point on the component surface, providing complete and reliable original data support for the subsequent generation of accurate three-dimensional point cloud mesh using three-dimensional reconstruction technology, thus ensuring the comprehensiveness and accuracy of macroscopic spatial coordinate measurement.

[0054] S104. The multiple sets of distance data are combined using three-dimensional reconstruction technology to generate a three-dimensional point cloud mesh. The actual spatial coordinates in the three-dimensional point cloud mesh are compared with the theoretical spatial coordinates of the corresponding node positions in the preset BIM model to obtain spatial coordinate deviation data.

[0055] Optionally, in step S104, three-dimensional reconstruction technology is used to combine the multiple sets of distance data to generate a three-dimensional point cloud mesh. The actual spatial coordinates in the three-dimensional point cloud mesh are compared with the theoretical spatial coordinates of the corresponding node positions in the preset BIM model to obtain spatial coordinate deviation data, including the following steps: Step 1041: Using 3D reconstruction technology, the distance information of any node position in the multiple sets of distance data collected at different angles is associated to obtain the actual spatial coordinates of the node position in 3D space. The actual spatial coordinates of all the node positions are integrated to form a 3D point set. The adjacent node positions in the 3D point set are connected to form a 3D point cloud mesh.

[0056] Step 1042: Compare the actual spatial coordinates of each node position in the three-dimensional point cloud mesh with the theoretical spatial coordinates of the corresponding node position in the preset BIM model. Calculate the numerical differences between the actual spatial coordinates and the theoretical spatial coordinates of the node position in the three axes respectively, and integrate the numerical differences in the three axes to generate spatial coordinate deviation data of the node position.

[0057] In the above scheme, 3D reconstruction technology refers to the technology of constructing a 3D model of an object by integrating multi-source spatial data; the 3D point cloud mesh is a mesh-like model formed by connecting a large number of 3D spatial points according to their positional relationships, where each point contains 3D coordinate information; node position refers to the key connection point on the steel structure component; actual spatial coordinates are the coordinates of the node position in the 3D point cloud mesh in 3D space; the theoretical spatial coordinates of the corresponding node position in the preset BIM model are the standard 3D coordinates of the node pre-designed in the model; spatial coordinate deviation data is the set of differences between the actual spatial coordinates and the theoretical spatial coordinates in three axes, reflecting the degree of spatial deviation of the node position.

[0058] In this application example, firstly, a three-dimensional point cloud mesh is generated through step 1041. Multiple sets of distance data are associated using the principle of triangulation. For each node position, the distance value is extracted from the distance data at different scanning angles. The three-dimensional coordinates are then calculated using a formula based on the scanning angle parameters. , , Where d is the distance value, The horizontal rotation angle, For the vertical tilt angle, the actual spatial coordinates of all node positions are integrated into a three-dimensional point set, and then adjacent node positions are connected by a point cloud stitching algorithm (such as the ICP algorithm) to form a three-dimensional point cloud mesh.

[0059] Secondly, by calculating the spatial coordinate deviation data in step 1042, the actual spatial coordinates of each node position are extracted from the 3D point cloud mesh, and the theoretical spatial coordinates of the corresponding nodes are obtained from the preset BIM model. The numerical differences of the three axes are then calculated respectively. , , These three differences are then integrated into the spatial coordinate deviation data of the node.

[0060] In practical applications, taking the node group (including nodes M, N, O, etc.) of steel structure component D in Project A as an example, 3D reconstruction technology was used to process 160 sets of distance data. Taking node N as an example, its two key data sets in 8 scanning angles are: angle 1, where the horizontal rotation angle is... Vertical tilt angle Measuring distance Through formula , , Calculate the three-dimensional coordinates , , Angle 2, where the horizontal rotation angle is 2. Vertical tilt angle Measuring distance Calculated , , After integrating data from eight angles and optimizing using a fusion algorithm, the actual spatial coordinates of node N are (2.30, 1.10, 0.50). Similarly, the coordinates of all nodes are obtained and integrated into a 3D point set. Adjacent nodes are connected using the ICP algorithm (intervals less than 0.5m are considered adjacent) to form a 3D point cloud mesh. The actual coordinates of node N (2.30, 1.10, 0.50) are extracted and compared with its theoretical coordinates (2.32, 1.08, 0.50) in the preset BIM model to calculate... , , The spatial coordinate deviation data of node N is integrated, and the same operation is performed on nodes M, O, etc., to finally obtain the spatial coordinate deviation data of all nodes.

[0061] The aforementioned S104 overall scheme uses 3D reconstruction technology to transform multiple sets of distance data into a 3D point cloud mesh, realizing the visualization of the spatial form of steel structure components. By comparing actual spatial coordinates with theoretical coordinates, the deviation of each node position in three axes is accurately quantified, and the resulting spatial coordinate deviation data reflects the overall spatial installation accuracy of the components. The entire process transforms scattered distance data into a system 3D model and deviation information, providing key data support in the macroscopic spatial dimension for subsequent integration of microscopic feature differences and generation of a comprehensive quality assessment report.

[0062] S105. The spatial coordinate deviation data, the weld contour difference information, and the bolt spacing difference information are correlated to obtain comprehensive deviation data. The comprehensive deviation data is then fused with the three-dimensional point cloud mesh to generate a comprehensive evaluation report on the installation quality of the steel structure components.

[0063] Optionally, step S105, which involves associating the spatial coordinate deviation data, the weld contour difference information, and the bolt spacing difference information to obtain comprehensive deviation data, and then fusing the comprehensive deviation data with the three-dimensional point cloud mesh to generate a comprehensive evaluation report on the installation quality of the steel structure component, includes the following steps: Step 1051: Integrate the spatial coordinate deviation, weld contour difference, and bolt spacing difference at the same node position of the steel structure component to form comprehensive deviation data that characterizes the overall deviation value and distribution of the steel structure component.

[0064] Step 1052: Associate the comprehensive deviation data with the actual spatial coordinates of the corresponding node positions in the three-dimensional point cloud mesh to form a three-dimensional model containing deviation information. The deviation information specifically refers to the three types of specific deviation content contained in the comprehensive deviation data, namely, the differences between the same node position and the preset standard in three dimensions: spatial coordinates, weld contour, and bolt spacing.

[0065] Step 1053: Based on the three-dimensional model, according to the overall installation area of ​​the steel structure component, classify and organize the comprehensive deviation data within the same installation area, and record the overall deviation value and distribution of all the comprehensive deviation data within each installation area to form a comprehensive evaluation report on the installation quality of the steel structure component.

[0066] In the above scheme, spatial coordinate deviation data refers to the difference between the actual coordinates and theoretical coordinates of the node position in three-dimensional space; weld contour difference information refers to the position and direction deviation of the actual edge of the weld from the standard edge; bolt spacing difference information refers to the length and proportion deviation of the actual bolt spacing from the standard spacing; comprehensive deviation data is an overall deviation record formed by integrating the spatial coordinate deviation of the same node, weld contour difference, and bolt spacing difference; three-dimensional point cloud mesh is a mesh model formed by connecting the three-dimensional coordinates of nodes; corresponding node position refers to the specific node in the three-dimensional point cloud mesh associated with the deviation data; installation area refers to the construction area divided according to the component structure, such as top, middle, and bottom; the comprehensive installation quality assessment report is a summary document that records the deviation distribution of each area and the overall quality.

[0067] In this application example, the comprehensive deviation data is integrated through step 1051. A data association algorithm (such as node ID matching) is used to bind the three types of deviation data at the same node location (such as spatial coordinate deviation, weld contour difference, and bolt spacing difference of node N are all associated with the "node N" identifier). The data is integrated in the format of "node ID-spatial coordinate deviation-weld contour difference-bolt spacing difference" to form comprehensive deviation data. For example, the comprehensive deviation data of node M is "node M: spatial coordinate deviation (-0.02m, 0.02m, 0m); weld contour difference (position deviation 1-2mm, direction deviation 1-3°); bolt spacing difference (length deviation -2-+1mm, proportional deviation -0.03-+0.01)".

[0068] Step 1052 generates a 3D model containing deviation information. Using coordinate matching technology, the node ID in the comprehensive deviation data is associated with the actual spatial coordinates of the corresponding node in the 3D point cloud mesh (e.g., the ID of node M corresponds to the node with coordinates (1.18m, 0.32m, 0m) in the 3D point cloud mesh). This makes the 3D coordinates of each node accompanied by its comprehensive deviation data, forming a visualized 3D model (the magnitude of the deviation can be indicated by color, such as red indicating a larger deviation). For example, the comprehensive deviation data is labeled next to the coordinates of node M in the 3D model. Clicking on the node allows you to view detailed spatial, weld, and bolt deviations.

[0069] Step 1053 generates an evaluation report, dividing the installation area according to the component structure (e.g., dividing steel columns into top, middle, and bottom areas). A region clustering algorithm is used to categorize the comprehensive deviation data of nodes within the same area. The report then statistically analyzes the deviation distribution in each area (e.g., the number of nodes with significant deviations, and the main types of deviations). Combined with engineering quality standards, the report describes the quality status of each area, ultimately forming a report that includes a summary of deviations in each area and an overall quality assessment. For example, the main deviation in the top area is bolt spacing deviation, while in the middle area it is mainly spatial coordinate deviation. The report records these deviations and provides adjustment suggestions for each area.

[0070] In practical applications, during the steel structure construction of Project A, when inspecting the installation quality of node N (the connection node between the steel column and the steel beam) of component D, the coordinates of the two endpoints P1 (100mm, 200mm) and P2 (150mm, 250mm) of the continuous line segment S1 at the weld edge are obtained using image acquisition equipment. The formula is then used... The actual included angle was calculated. The standard included angle of this line segment in the preset BIM model is 43°, therefore the angle deviation is... On S1, select adjacent points A (105mm, 200mm) and B (110mm, 201mm). The calculated angle between the lines connecting the two points is approximately 11.3°, and the angle between the lines connecting B and C (115mm, 201mm) is approximately 11.8°. The actual angle variation is 0.5°. Since the standard angle variation at this location in the BIM model is 0°, the angle variation deviation is 0.5°. Bolts B1 (200mm, 300mm) and B2 (250mm, 300mm) are identified through target detection using the formula... The calculated actual spacing is 50mm, while the standard spacing in the BIM model is 52mm, resulting in a length deviation of 50-52=-2mm; the bolt arrangement at node N is B1-B2-B3, and the measured actual spacing between B2-B3 is 52mm, which is in proportion to the actual scale. In the BIM model, the standard spacing between B1 and B2 is 52mm, and the standard spacing between B2 and B3 is 53mm, at the standard scale. The proportional deviation is 0.96-0.98=-0.02. These data are integrated into the comprehensive deviation data of node N, and after being associated with the three-dimensional point cloud mesh, they are included in the comprehensive evaluation report of the installation quality of component D.

[0071] The aforementioned S105 overall solution integrates three types of deviation data to form comprehensive deviation data, thereby establishing a correlation between microscopic feature deviations and macroscopic spatial deviations and avoiding the biased assessment caused by data dispersion. By fusing the comprehensive deviation data with a three-dimensional point cloud mesh, the deviation information is visualized, facilitating an intuitive understanding of the spatial distribution of deviations. Data is organized by installation area and reports are generated, clearly presenting the quality status of each area and the overall installation accuracy, providing a comprehensive and accurate basis for construction adjustments, and ensuring the systematicness and reliability of steel structure installation quality assessment.

[0072] The following is a complete example for steps S101 to S105. Figure 2As shown, firstly, in the steel structure construction of Project A, a 6-megapixel industrial camera (resolution 3072×2048) was used to take pictures of the steel column and steel beam connection node M from five angles: front, left 45°, right 45°, top view, and bottom view. Adjacent images overlapped by 40% (e.g., the overlap area between the left 45° image and the front image accounted for 40% of their respective areas). After importing the five original images, the SIFT algorithm was used to extract weld feature points (12 per image, such as P1(100, 200) as the weld inflection point) and bolt feature points (10 per image, such as the center of B1(200, 300)). After matching, the images were stitched together, and the pixels in the overlapping area were weighted and averaged (the weights were distributed according to the distance from the edge) to generate the target image, which clearly showed the three weld contours (S1-S3) and the positions of the eight bolts (B1-B8) of node M.

[0073] Next, the generated target image is input into the Siamese neural network. The first sub-network extracts the coordinates of the two endpoints of weld S1, P1 (100mm, 200mm) and P2 (150mm, 250mm), and uses the formula... Calculate the actual included angle, where , ,have to Compared with the standard angle of 43° for S1 in the BIM model, the angle difference is... Ten points are taken on S1 at 5mm intervals. The difference between the actual coordinates (110mm, 201mm) and the standard coordinates (108mm, 200mm) of the third point is... , This information is integrated into weld contour difference information. The second sub-network identifies bolts B1 (200mm, 300mm) and B2 (250mm, 300mm) using a formula. Calculate the actual spacing, where , ,have to The difference from the standard spacing of 52mm in the BIM model is The ratio of B1-B2 to B2-B3 (actual 52mm) The difference between this and the standard ratio 52 / 53≈0.98 is This information is integrated into bolt spacing difference information.

[0074] Then, a 20-element phased array radar was used to scan the steel column where node M is located, setting 8 angles (0°, 45°…315°) and a microwave frequency of 24GHz (wavelength 12.5mm); at a 30° angle… Next, the 5th unit transmits a signal to the surface of node M, and the reception time is... Using formula Calculate the distance Similarly, 160 sets of data were measured for 8 angles × 20 units (e.g., “30° - unit 5 - 1.18m”), which will be used for the three-dimensional reconstruction in step 4.

[0075] Next, the distance data was processed, and the distance to node M was set to 1.18m at a 30° angle and 1.22m at a 60° angle, combined with... , When calculating 30° using the formula , , After optimizing data from eight angles, the actual coordinates of node M are (2.30m, 1.10m, 0.50m). All node coordinates are integrated to form a point set, and adjacent points (with a spacing <0.5m) are connected using the ICP algorithm to generate a 3D point cloud mesh. This mesh is then compared with the theoretical coordinates of node M in the BIM model (2.32m, 1.08m, 0.50m). , This refers to spatial coordinate deviation data.

[0076] Finally, integrating the bolt spacing difference information and spatial coordinate deviation data, the comprehensive deviation of node M is "space (-0.02m, 0.02m, 0m); weld". ;bolt The value is associated with its actual coordinates (2.30m, 1.10m, 0.50m) and is marked in yellow (medium deviation) in the 3D model. After dividing the area into regions, the bolt deviation and spatial deviation of the top region (including nodes M and N) are statistically analyzed, and a report is finally generated. It is recommended to prioritize adjusting the top bolt spacing and the spatial position of the middle nodes. All data are used to evaluate the node installation quality.

[0077] Figure 3 This is a structural schematic diagram of a specific embodiment of a construction surveying system provided in this application, with reference to... Figure 3 The system may include: The acquisition module 31 is used to take multi-angle pictures of the node parts of the steel structure components through image acquisition equipment during the construction process, obtain multi-angle original images, and fuse all the original images to obtain a target image containing the weld contour and bolt spacing at the node parts.

[0078] The first comparison module 32 is used to call a twin neural network to compare the weld contour and bolt spacing in the target image with the geometric features of the corresponding node in the preset BIM model, so as to obtain weld contour difference information and bolt spacing difference information.

[0079] The measurement module 33 is used to perform multi-angle scanning of the overall surface of the steel structure component using a phased array microwave radar array, and to measure the distance between the radar in the phased array microwave radar array and the overall surface during the scanning process, thereby generating multiple sets of distance data.

[0080] The second comparison module 34 is used to combine the multiple sets of distance data using three-dimensional reconstruction technology to generate a three-dimensional point cloud mesh, and compare the actual spatial coordinates in the three-dimensional point cloud mesh with the theoretical spatial coordinates of the corresponding node positions in the preset BIM model to obtain spatial coordinate deviation data.

[0081] The fusion module 35 is used to associate the spatial coordinate deviation data, the weld contour difference information, and the bolt spacing difference information to obtain comprehensive deviation data, and to fuse the comprehensive deviation data with the three-dimensional point cloud mesh to generate a comprehensive evaluation report on the installation quality of the steel structure components.

[0082] This application provides a construction surveying system for implementing the aforementioned construction surveying method. Therefore, the specific implementation of the construction surveying system can be found in the embodiment section of the construction surveying method above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0083] like Figure 4 Furthermore, this application also provides an electronic device, comprising: a memory 41 for storing a computer program; and a processor 42 for executing the computer program to implement the steps of any of the above-described construction surveying methods.

[0084] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described construction surveying methods.

[0085] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0086] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described embodiments of the building construction measurement method.

[0087] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0088] The foregoing has provided a detailed description of a construction surveying method, system, electronic device, and storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for measuring construction, characterized in that, include: During the construction process, the joints of steel structure components are photographed from multiple angles using image acquisition equipment to obtain original images from multiple angles. All original images are then fused to obtain a target image that includes the weld contour and bolt spacing at the joint. The twin neural network is invoked to compare the weld contour and bolt spacing in the target image with the geometric features of the corresponding node in the preset BIM model to obtain weld contour difference information and bolt spacing difference information. The entire surface of the steel structure component is scanned from multiple angles using a phased array microwave radar array. During the scanning process, the distance between the radar in the phased array microwave radar array and the entire surface is measured to form multiple sets of distance data. The multiple sets of distance data are combined using 3D reconstruction technology to generate a 3D point cloud mesh. The actual spatial coordinates in the 3D point cloud mesh are compared with the theoretical spatial coordinates of the corresponding node positions in the preset BIM model to obtain spatial coordinate deviation data. The spatial coordinate deviation data, the weld contour difference information, and the bolt spacing difference information are correlated to obtain comprehensive deviation data. The comprehensive deviation data is then fused with the three-dimensional point cloud mesh to generate a comprehensive evaluation report on the installation quality of the steel structure components.

2. The method according to claim 1, characterized in that, The step of calling a Siamese neural network to compare the weld contour and bolt spacing in the target image with the geometric features of the corresponding nodes in the preset BIM model to obtain weld contour difference information and bolt spacing difference information includes the following steps: The first sub-network of the twin neural network performs layer-by-layer feature processing on the edge line information of the weld contour and the standard edge line information in the preset BIM model to obtain the first local feature and the second local feature. The difference calculation is performed on the coordinate intermediate feature and the angle intermediate feature of each point in the first local feature and the second local feature to obtain the first feature vector. The second sub-network of the twin neural network performs layer-by-layer feature processing on the numerical information of the bolt spacing and the numerical information of the standard bolt spacing in the preset BIM model to obtain the first spacing length feature and the second spacing length feature. The difference between the first spacing length feature and the second spacing length feature is calculated to output the second feature vector. Based on the output layer of the twin neural network, the weld contour difference information is determined based on the first feature vector, and the bolt spacing difference information is determined based on the second feature vector.

3. The method according to claim 2, characterized in that, The step of performing layer-by-layer feature processing on the edge line information of the weld contour and the standard edge line information in the preset BIM model through the first sub-network of the twin neural network to obtain the first local feature and the second local feature includes the following steps: The recognition layer of the first sub-network identifies feature points whose curvature values ​​exceed a preset curvature threshold on the edge lines in the edge line information and the edge lines in the standard edge line information, and treats the edge lines between adjacent feature points as continuous line segments. Through the feature extraction layer of the first sub-network, point positions are selected on the continuous line segment at preset intervals, and the horizontal and vertical coordinate values ​​of all the point positions in the two-dimensional plane are recorded to form coordinate intermediate features. Through the computation layer of the first sub-network, the angle between the line connecting the two endpoints on the continuous line segment and the horizontal reference direction is calculated, as well as the angle change of the line connecting adjacent points on the continuous line segment, forming an angle-type intermediate feature. Through the feature association layer of the first sub-network, the coordinate-class intermediate features and the angle-class intermediate features of the edge line information of the weld contour are associated to obtain the first local feature, and the coordinate-class intermediate features and the angle-class intermediate features of the standard edge line information are associated to obtain the second local feature.

4. The method according to claim 2, characterized in that, The step of performing layer-by-layer feature processing on the numerical information of the bolt spacing and the numerical information of the standard bolt spacing in the preset BIM model through the second sub-network of the twin neural network to obtain the first spacing length feature and the second spacing length feature includes the following steps: The identification layer of the second sub-network identifies the position information in the numerical information of the bolt spacing and the position information in the numerical information of the standard bolt spacing, respectively. Based on the position information, two adjacent bolts are paired to form a spacing unit, wherein the spacing unit contains the position identifier of two adjacent bolts and the corresponding spacing length value. Through the first association layer of the second sub-network, the position identifiers corresponding to two adjacent bolts in each spacing unit are associated with the corresponding spacing length values ​​to form a numerical intermediate feature; Through the computation layer of the second sub-network, the spacing units are arranged according to the bolt arrangement order on the steel structure component to form a bolt arrangement sequence. The proportional relationship between the consecutive spacing length values ​​within the bolt arrangement sequence is calculated to form intermediate features of the relationship class. Through the second association layer of the second sub-network, the numerical class intermediate features of the bolt spacing numerical information and the standard bolt spacing numerical information are associated with the relation class intermediate features to obtain the first spacing length feature and the second spacing length feature.

5. The method according to claim 2, characterized in that, The process of determining the weld contour difference information based on the first feature vector and the bolt spacing difference information based on the second feature vector includes the following steps: Calculate the difference in horizontal coordinate values ​​and the difference in vertical coordinate values ​​between each actual point in the first local feature of the first feature vector and the corresponding standard point in the second local feature to obtain the coordinate value difference; Calculate the angle difference between the included angle value in the first local feature and the included angle value in the second local feature, and the difference between the angle change values ​​between corresponding adjacent points in the first local feature and the second local feature, to obtain the angle difference. The coordinate value differences and angle value differences are associated in the order of continuous line segments and integrated into weld contour difference information that includes the positional deviation and direction deviation of each continuous line segment. Calculate the difference between the first spacing length feature and the second spacing length feature in the second feature vector to obtain the basic spacing difference. Based on the continuous spacing units in the bolt arrangement sequence, calculate the difference between the proportional relationship in the first spacing length feature and the proportional relationship in the second spacing length feature to obtain the proportional difference. The basic spacing difference and the proportional difference are associated according to the order of the bolt arrangement sequence, and integrated into bolt spacing difference information that includes the length deviation of each spacing unit and the sequence proportional deviation.

6. The method according to claim 1, characterized in that, The process involves combining multiple sets of distance data using 3D reconstruction technology to generate a 3D point cloud mesh. The actual spatial coordinates in the 3D point cloud mesh are then compared with the theoretical spatial coordinates of the corresponding node positions in the preset BIM model to obtain spatial coordinate deviation data. This includes the following steps: By using 3D reconstruction technology, the distance information of any node position in the multiple sets of distance data collected at different angles is associated to obtain the actual spatial coordinates of the node position in 3D space. The actual spatial coordinates of all the node positions are integrated to form a 3D point set. The adjacent node positions in the 3D point set are connected to form a 3D point cloud mesh. The actual spatial coordinates of each node position in the three-dimensional point cloud mesh in three-dimensional space are compared with the theoretical spatial coordinates of the corresponding node position in the preset BIM model. The numerical differences between the actual spatial coordinates and the theoretical spatial coordinates of the node position in three axes are calculated respectively. The numerical differences in the three axes are integrated to generate the spatial coordinate deviation data of the node position.

7. The method according to claim 1, characterized in that, The process of associating the spatial coordinate deviation data, the weld contour difference information, and the bolt spacing difference information to obtain comprehensive deviation data, and then fusing the comprehensive deviation data with the three-dimensional point cloud mesh to generate a comprehensive evaluation report on the installation quality of the steel structure components, includes the following steps: The spatial coordinate deviation, weld contour difference, and bolt spacing difference at the same node position of the steel structure component are integrated to form comprehensive deviation data that characterizes the overall deviation value and distribution of the steel structure component. The comprehensive deviation data is associated with the actual spatial coordinates of the corresponding node positions in the three-dimensional point cloud mesh to form a three-dimensional model containing deviation information. Based on the three-dimensional model, the comprehensive deviation data within the same installation area of ​​the steel structure component are classified and organized according to the overall installation area. The overall deviation value and distribution of all comprehensive deviation data within each installation area are recorded to form a comprehensive evaluation report on the installation quality of the steel structure component.

8. A construction surveying system, characterized in that, include: The acquisition module is used to capture images of the joints of steel structure components from multiple angles during the construction process using image acquisition equipment, obtain original images from multiple angles, and fuse all the original images to obtain a target image containing the weld contour and bolt spacing at the joint. The first comparison module is used to call the twin neural network to compare the weld contour and bolt spacing in the target image with the geometric features of the corresponding node in the preset BIM model, so as to obtain weld contour difference information and bolt spacing difference information. The measurement module is used to perform multi-angle scanning of the overall surface of the steel structure component using a phased array microwave radar array, and to measure the distance between the radar in the phased array microwave radar array and the overall surface during the scanning process, thereby generating multiple sets of distance data. The second comparison module is used to combine the multiple sets of distance data using three-dimensional reconstruction technology to generate a three-dimensional point cloud mesh, and compare the actual spatial coordinates in the three-dimensional point cloud mesh with the theoretical spatial coordinates of the corresponding node positions in the preset BIM model to obtain spatial coordinate deviation data. The fusion module is used to associate the spatial coordinate deviation data, the weld contour difference information, and the bolt spacing difference information to obtain comprehensive deviation data, and then fuse the comprehensive deviation data with the three-dimensional point cloud mesh to generate a comprehensive evaluation report on the installation quality of the steel structure components.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of a construction surveying method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables a construction surveying method as described in any one of claims 1 to 7.

Images