Scaffold engineering compliance intelligent evaluation method based on three-dimensional reconstruction technology

Abstract

The invention discloses a scaffold engineering compliance intelligent evaluation method based on a three-dimensional reconstruction technology. According to the scheme, an unmanned aerial vehicle aerial photography sampling plan is made according to scaffold engineering characteristics, a machine vision model is trained to detect scaffold key components, and the distribution compliance of the scaffold system key components is evaluated according to the types and the number of the key components. A scaffold three-dimensional point cloud model is generated by adopting an incremental SfM and a multi-view stereo matching algorithm MVS based on an aerial image, a point cloud model of each rod piece is obtained through segmentation, and a central characteristic line equation of the rod piece is further extracted through a cylinder fitting algorithm. According to relevant specifications and standards of scaffolds, a scaffold engineering compliance set is summarized; and calculating the spatial relationship of the characteristic line equation, and comparing the scaffold compliance set to evaluate the construction compliance of the scaffold engineering of the project under construction. According to the method, the unmanned aerial vehicle, intelligent image recognition, three-dimensional reconstruction and feature line extraction technologies are comprehensively used, and autonomous and efficient evaluation of the safety compliance of the floor type fastener scaffold is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of building construction detection and engineering compliance, and particularly relates to a scaffold engineering compliance intelligent evaluation method based on three-dimensional reconstruction technology. BACKGROUND

[0002] Scaffolds play a crucial role in building construction. Instability of scaffolds caused by non-standard erection can lead to serious personnel casualties and property loss accidents. Checking the compliance of scaffold erection is not only the basis for protecting the safety of workers, but also an important link in maintaining the smooth progress of the whole project. Traditional scaffold compliance evaluation methods mainly rely on manual inspection and on-site measurement, which is risky, inefficient and subjective, and cannot meet the requirements of modern building construction.

[0003] In recent years, image recognition technology and unmanned aerial vehicle technology have provided a new way for the compliance evaluation of scaffold engineering. Image recognition technology can automatically analyze image data obtained by unmanned aerial vehicles, recognize and detect key components and erection details of scaffolds, and improve the automation level of scaffold engineering compliance evaluation. On the other hand, through high-precision and wide-coverage shooting by unmanned aerial vehicles, the data collection efficiency and accuracy are greatly improved. However, the single picture taken by the unmanned aerial vehicle has limited perspective, which makes it difficult to evaluate the overall structure compliance of the scaffold. By splicing a large number of single-view pictures to form the outer facade picture of the scaffold, the perspective limitation problem can be solved. However, this method requires the image acquisition device and the photographed surface to be perpendicular during shooting, which is difficult and time-consuming, and it is difficult to avoid the distortion problem of the scaffold components after splicing. In addition, the picture acquisition process is easily affected by factors such as light, shadow and reflection, which further affects the evaluation effect.

[0004] Two-dimensional pictures of scaffolds lack depth and spatial information, and some components are easily affected by angle and obstruction, so it is difficult to conduct comprehensive and accurate compliance evaluation of scaffolds through image recognition methods. Three-dimensional reconstruction technology generates accurate spatial models, avoiding the problems of lack of depth information, component distortion and high acquisition difficulty of two-dimensional scaffold images, and provides the possibility for accurate evaluation of scaffold compliance.

[0005] 3D reconstruction technology is categorized into active and passive methods based on the sensors used. Active 3D reconstruction requires sensors to actively detect depth information. Commonly used sensors include LiDAR, structured light sensors, and ToF cameras (Time-of-Flight cameras). Active sensors can provide high-precision depth information and detailed 3D structures, do not rely on natural light, and have strong environmental adaptability. However, their hardware costs are typically high, operation is complex, and they can produce significant data errors in complex environments such as highly reflective surfaces or multipath interference. Passive 3D reconstruction usually only requires conventional cameras, has low equipment costs, is easy to deploy, and is applicable to a wide range of scenarios, making it popular among researchers.

[0006] Passive 3D reconstruction can be divided into monocular depth estimation, binocular stereo matching, and multi-view methods, depending on the number of input views. Figure Three 3D reconstruction. Monocular depth estimation, relying on only a single image to infer the 3D shape of objects, suffers from inconsistent accuracy, failing to accurately reflect the complexities of scaffolding engineering scenarios. Binocular stereo matching infers the 3D structure of a scene by calculating the disparity between two images from different viewpoints, achieving higher reconstruction accuracy than monocular depth estimation. However, this method has limited field of view, is susceptible to occlusion, and suffers from inaccurate reconstruction of certain areas. Multi-view... Figure Three 3D reconstruction requires acquiring multiple images from different perspectives and reconstructing the 3D structure of a scene by matching feature points between these views. Common techniques include SfM (Structure from Motion) and MVS (Multi-view Stereo). Because multiple views provide more geometric information from various perspectives, it is possible to generate highly detailed, accurate, and large-scale 3D models of scenes. Images matched using multi-view stereo do not require correction, resulting in low image acquisition costs.

[0007] In summary, this invention proposes an intelligent evaluation method for the compliance of scaffolding projects based on 3D reconstruction technology. This method utilizes UAV oblique photography, simplifying the image acquisition process. Building upon image recognition technology for intelligent identification of key scaffolding components, it employs multi-view... Figure Three The 3D model generated by the 3D reconstruction technology provides detailed and accurate structural information for scaffolding projects, overcoming the problem of some scaffolding components being obscured, and resulting in a more accurate and comprehensive evaluation of the compliance of scaffolding projects. Summary of the Invention

[0008] In view of this, the purpose of this invention is to propose an intelligent evaluation method for the compliance of scaffolding projects based on three-dimensional reconstruction technology, which comprehensively utilizes UAV technology, intelligent image recognition technology, and multi-view technology. Figure ThreeThe three-dimensional reconstruction technology and the feature line extraction technology have reliable implementation, flexible application, and more accurate and comprehensive evaluation.

[0009] To achieve the above technical purposes, the technical scheme adopted by the present application is:

[0010] A scaffold engineering compliance intelligent evaluation method based on three-dimensional reconstruction technology, comprising:

[0011] Step 1) According to the engineering characteristics of the area where the scaffold engineering is located, a parameter plan for unmanned aerial vehicle sampling is formulated, and the scaffold engineering is sampled by using the unmanned aerial vehicle aerial photography mode to generate aerial photography pictures;

[0012] Step 2) The aerial photography pictures are preprocessed and then input into a trained machine vision model to identify the scaffold key components in the aerial photography pictures, and then the categories and quantities of the scaffold key components are counted and used to evaluate the scaffold system component distribution rationality;

[0013] Step 3) The aerial photography pictures containing scaffold components are reversely generated into a scaffold three-dimensional point cloud model by using an incremental SfM and a multi-view stereo matching algorithm MVS;

[0014] Step 4) The scaffold three-dimensional point cloud model is segmented by using a density clustering algorithm DBSCAN combined with a distance and angle judgment method between adjacent components to obtain a point cloud model of each rod of the scaffold;

[0015] Step 5) The point cloud model of the rod is filtered by using a cylindrical fitting algorithm combined with a RANSAC algorithm to extract the rod center feature line;

[0016] Step 6) According to the corresponding technical specifications and technical standards of the scaffold, a scaffold engineering compliance set is induced;

[0017] Step 7) The rod center line spatial equation is calculated, the scaffold compliance requirements are expressed in terms of the spatial relationship of the rod feature line, and the scaffold engineering construction compliance is evaluated in combination with the scaffold engineering compliance set.

[0018] As a possible implementation manner, further, in step 1) of the present scheme, the engineering characteristics of the area where the scaffold engineering is located include the scaffold facade characteristics and / or the site restriction conditions of the site where the scaffold is located.

[0019] As a possible implementation manner, further, the parameter plan for unmanned aerial vehicle sampling in the present scheme includes the path of aerial photography of the unmanned aerial vehicle and the aerial photography parameters; the aerial photography parameters include one or more of the heading, the side image overlap degree, the camera inclination angle, the photography distance, and the light intensity selection.

[0020] As a possible implementation manner, further, in step 2) of the present scheme, the preprocessing of the aerial photography pictures includes:

[0021] The aerial images meeting the quality requirements are screened out, and image enhancement, denoising, smoothing and geometric transformation processing are performed on the aerial images.

[0022] As a possible implementation, further, the machine vision model trained by the scheme is a YOLOv8 model, and the training method comprises:

[0023] The scaffold pictures are collected, classified and labeled according to categories, and then divided into a training set, a verification set and a test set. The training set is input into the YOLOv8 model for training, and the verification set and the test set are used to verify and test the trained YOLOv8 model. Then, after the test and verification meet the requirements, the converged YOLOv8 model is output as a category identifier, which is used to classify the scaffold components in the aerial pictures.

[0024] As a preferred implementation option, preferably, the category identifier described in the scheme is used to identify one or more components of the wall connecting piece, the sweeping rod, the connecting fastener and the scissors support in the key components of the scaffold in the aerial picture.

[0025] As a possible implementation, further, in step 3) of the scheme, the scaffold three-dimensional point cloud model refers to the point cloud model of the aerial scene obtained by using incremental SfM and MVS technology to reversely reconstruct the picture set processed in step 2), then carrying out voxel downsampling and filtering denoising operation, and satisfying the whole coordinate transformation that the axis of the bottom layer vertical rod is parallel to the Z-axis direction of the whole coordinate system.

[0026] As a possible implementation, further, in step 5) of the scheme, the rod center feature line is a spatial straight line segment passing through the center points of both ends of each point cloud rod in the scaffold three-dimensional point cloud model in the whole coordinate system; it further comprises: according to the three-dimensional coordinate relationship of the rod end point and the scaffold construction feature, the rod is numbered, wherein the rod center feature line has semantic information.

[0027] As a possible implementation, further, in step 6) of the scheme, the scaffold engineering compliance set is formed by systematically summarizing key indicators affecting the safety structure of the scaffold in the current national scaffold safety technical specifications and standards, which includes horizontal and vertical horizontal rods, sweeping rods, vertical rods, scissors supports, and safety inspection items related to main and auxiliary nodes of horizontal and vertical horizontal rods on the same layer.

[0028] As a possible implementation, further, in step 7) of the scheme, the scaffold engineering construction compliance evaluation comprises:

[0029] The key indicators of scaffold safety structure are transformed into geometric relationships of the center feature lines of the members. When the axis of the bottom upright is parallel to the Z-axis of the overall coordinate system, the geometric relationships of the center feature lines of the members include:

[0030] 1) Verticality deviation, horizontal deviation, overhang length, and maximum z-coordinate value of the center feature line of a single vertical or horizontal member;

[0031] 2) The included angle, shortest distance, and x / y / z coordinate difference between the center feature lines of two vertical or horizontal poles;

[0032] 3) The angle between the center feature line of the inclined scissor brace and the XOY plane, the number of intersection points between the center feature line of the scissor brace and the center feature line of the upright in the orthographic projection plane, the difference in x / y coordinates between the endpoints of the center feature lines of two intersecting scissor brace members, and the difference in x / y coordinates between the endpoints of the center feature lines of two adjacent but non-intersecting scissor brace members.

[0033] Based on the above, this solution also proposes a scaffolding project compliance detection method, which applies the aforementioned intelligent evaluation method for scaffolding project compliance based on three-dimensional reconstruction technology.

[0034] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art: This solution overcomes the traditional method's reliance on manual on-site inspection, avoiding the problems of limited operating range, high risk of high-altitude operations, low efficiency, and certain subjectivity inherent in the traditional method; This solution effectively solves the problems of missing depth information, distortion of some components, susceptibility to occlusion, and high acquisition difficulty caused by vertical shooting when directly applying image recognition; In addition, this solution also helps to obtain more comprehensive and accurate compliance evaluation results for scaffolding projects, detect potential safety hazards in scaffolding projects early, and provide early warning and reference for technicians and safety managers. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a flowchart illustrating the workflow of an intelligent evaluation method for compliance of scaffolding projects based on three-dimensional reconstruction technology, as described in this invention.

[0037] Figure 2 This is a schematic diagram of the fastener-type scaffolding structure according to an embodiment of the present invention;

[0038] Figure 3 L-shaped scaffold for the embodiment of the present application;

[0039] Figure 4 scaffold key component image detection effect drawing for the embodiment of the present application;

[0040] Figure 5 scaffold three-dimensional reconstruction point cloud model drawing for the embodiment of the present application;

[0041] Figure 6 scaffold point cloud model filtering and voxel downsampling processing effect drawing for the embodiment of the present application;

[0042] Figure 7 scaffold rod point cloud segmentation effect drawing for the embodiment of the present application;

[0043] Figure 8 scaffold feature line model drawing for the embodiment of the present application. DETAILED DESCRIPTION

[0044] The present application will be further described below in conjunction with the drawings and embodiments. It is particularly pointed out that the following embodiments are only used to illustrate the present application, but do not limit the scope of the present application. Similarly, the following embodiments are only part of the embodiments of the present application, not all embodiments, and all other embodiments obtained by those skilled in the art without creative labor are within the scope of the present application.

[0045] The scaffold mentioned in the embodiment scheme is a fastener type scaffold; it is a stable support system connected by fasteners of different types of rods, used to provide a temporary work platform. Among them, the fastener type scaffold includes vertical rods, horizontal rods, longitudinal horizontal rods, scissors, fasteners, wall connecting pieces and other basic components.

[0046] In combination with Figure 1 shown, the embodiment scheme is a scaffold engineering compliance intelligent evaluation method based on three-dimensional reconstruction technology, which includes:

[0047] Step 1) According to the engineering characteristics of the area where the scaffold engineering is located, the parameter plan of the unmanned aerial vehicle equipment sampling is made, the unmanned aerial vehicle aerial photography is used to sample the scaffold engineering, and the aerial photography picture is generated;

[0048] Step 2) The aerial photography picture is preprocessed, and then input into the trained machine vision model to identify the scaffold key components in the aerial photography picture, and then count the types and quantities of the scaffold key components, which are used to evaluate the rationality of the scaffold system component distribution;

[0049] Step 3) The aerial photography picture containing the scaffold components is reversely generated into a three-dimensional point cloud model of the scaffold by using an incremental SfM and a multi-view stereo matching algorithm MVS.

[0050] Step 4) The three-dimensional point cloud model of the scaffold is segmented by using the density clustering algorithm DBSCAN combined with the distance and angle judgment method between adjacent components to obtain the point cloud model of each rod of the scaffold;

[0051] Step 5) The point cloud model of the rod is filtered by using the cylindrical fitting algorithm combined with the RANSAC algorithm to extract the rod center feature line;

[0052] Step 6) According to the corresponding technical specifications and technical standards of the scaffold, the scaffold engineering compliance set is summarized;

[0053] Step 7) The spatial equation of the rod center line is calculated, the spatial relationship of the rod feature line is expressed, the scaffold compliance requirements are expressed, and the scaffold engineering compliance set is combined to evaluate the scaffold engineering construction compliance.

[0054] In combination Figure 2 As shown in the scheme, the engineering features of the scaffold engineering area include the scaffold facade features and / or the site restriction conditions of the scaffold site. Among them, the aerial sampling plan of the scaffold generally includes sampling objects, sampling methods, sampling equipment and sampling routes. The sampling objects need to clearly indicate the geometric parameters such as the height, width, complexity and density of the on-site scaffold engineering; the sampling method specifies the sampling height, angle, distance, frequency and other parameter conditions; the sampling equipment mainly considers the resolution, focal length, imaging rate of the image sensor, and the flight speed, ceiling height, endurance time, wind resistance level, RTK positioning accuracy and other parameters of the aircraft; the sampling route considers the heading and side image overlap, and the construction site restriction conditions, including space obstacles, air control requirements and climate conditions and other factors.

[0055] In terms of parameter planning, as a possible implementation manner, further, the parameter plan of the unmanned aerial vehicle equipment sampling includes the path of the unmanned aerial vehicle for aerial photography and the aerial photography parameters; the aerial photography parameters include one or more of the heading, side image overlap, camera tilt angle, photography distance, and light intensity selection.

[0056] In step 2) of the scheme, the aerial picture is preprocessed, including: screening out aerial images meeting the quality requirements, and performing image enhancement, denoising, smoothing and geometric transformation processing on the aerial images.

[0057] The collected scaffold images are screened, and the screening criteria include image quality indicators such as resolution, clarity, and exposure, so that the retained images can clearly present the details of the scaffold structure and key components. In order to facilitate data reuse and model training, the screened images are subjected to image enhancement, denoising, smoothing, and geometric transformation, etc. to expand the number of scaffold image samples. The processed image dataset of each component is divided into training set, validation set and test set according to the proportion (or the existing network data can be preprocessed into training data for model training), and the YOLOv8 software is used to train the scaffold basic component intelligent recognition model, which is used to identify the basic components in the aerial scaffold image, automatically count the number of each type of component, and evaluate the rationality of the scaffold component distribution.

[0058] In the model training, as a possible implementation, further, the trained machine vision model is a YOLOv8 model, and the training method includes:

[0059] The scaffold pictures are collected, classified and labeled according to categories, and then divided into training set, validation set and test set. The training set is input into the YOLOv8 model for training, and the trained YOLOv8 model is verified and tested by using the validation set and the test set. Then, after the test and verification meet the requirements, the converged YOLOv8 model is output as a category recognizer, which is used to classify the scaffold components in the aerial picture.

[0060] The category recognizer described in the present solution is used to identify one or more components of the scaffold key components in the aerial picture, such as wall connecting parts, sweeping rods, connecting fasteners, and scissors supports.

[0061] As a possible implementation, further, in step 3) of the present solution, the scaffold three-dimensional point cloud model refers to a point cloud model of the aerial scene generated by using incremental SfM and MVS technology to reversely three-dimensionally reconstruct the picture set processed in step 2), then performing voxel downsampling and filtering denoising operations, and performing overall coordinate transformation to satisfy that the axis of the bottom layer vertical rod is parallel to the Z-axis direction of the overall coordinate system.

[0062] As an example, the present solution uses incremental SfM technology to generate a scaffold three-dimensional point cloud model, and the specific steps include:

[0063] 1) Use SIFT image processing algorithm to extract feature points in the aerial scaffold image, so that the feature points are rich enough and uniformly distributed in the entire image area;

[0064] 2) Through the nearest neighbor search and RANSAC algorithm, the feature points of the images under different angles are matched to ensure that the points of the same object under different angles are correctly associated;

[0065] 3) Based on the matched feature points, the camera pose and internal parameters corresponding to each image are calculated to generate a sparse point cloud model;

[0066] 4) On the basis of the sparse point cloud, a more dense three-dimensional point cloud model is generated by using MVS technology.

[0067] The scheme can use statistical filtering (SOR) combined with radius filtering (ROR) method to remove noise from the point cloud model generated by SfM technology, and remove invalid points and isolated points. According to the resolution requirement, the voxel downsampling operation is performed, and the simplified point cloud model is constructed through the processes of voxel grid size definition, point cloud space three-dimensional grid division, and voxel average method sampling, etc., to reduce the point cloud redundancy.

[0068] For step 4), the scheme clusters the point cloud model by using the density clustering algorithm DBSCAN to identify the point cloud of each steel pipe; further, the clustered point cloud region is segmented to ensure the integrity of the steel pipe region; the connection points between the steel pipes and the adjacent components are automatically segmented by the distance and angle between the points to extract the clear point cloud of the steel pipe component.

[0069] The cylindrical fitting algorithm is applied to each steel pipe point cloud to calculate the axial parameters of the cylinder; during fitting, the RANSAC algorithm is used to filter outliers to ensure the fitting accuracy; the center line of each cylindrical steel pipe is extracted to form a feature line model of the three-dimensional point cloud model.

[0070] As a possible implementation, further, in step 5) of the scheme, the rod center feature line is a spatial straight line segment passing through the center points of both ends of each point cloud rod in the whole coordinate system of the scaffold three-dimensional point cloud model; it further comprises: according to the three-dimensional coordinate relationship of the rod end points and the scaffold construction features, the rod is numbered.

[0071] The whole coordinate transformation of the scaffold point cloud model is performed, the center line of a certain vertical rod at the bottom layer is selected, the rotation axis is found and the direction angle between the global coordinate system Z axis is calculated, the rotation matrix is constructed, the point cloud model is rotated to satisfy the selected vertical rod axis parallel to the global coordinate system Z axis. Further, the point cloud model is translated to satisfy the coincidence of the bottom end point of the vertical rod center line with the global coordinate system origin.

[0072] The scheme compares the feature line model with the geometric size of the actual scaffold structure to evaluate the three-dimensional reconstruction accuracy, and the evaluation formula is shown in the following formulas (1) and (2):

[0073]

[0074] In the formula, each parameter refers to: R avg : average scale; M i : program measured size; S i : actual size of the scaffold structure; n: number of measurement points; E avg : average error.

[0075] Accurate positioning and classification of the feature line model is performed, and semantic information is assigned to it, and the specific steps are as follows:

[0076] 1) Bottom pole axis reference: select the bottom pole axis parallel to the Z axis direction through the global coordinate system origin as the reference;

[0077] 2) Angle classification of rod: other straight line segments are classified into different categories according to the included angle with the vertical rod, and the search tolerance of angle classification is 10 degrees.

[0078] a) 0 vertical rod: rod with an included angle of 0°±10° or 180°±10° with the reference vertical rod;

[0079] b) Horizontal and vertical horizontal rods and horizontal and vertical sweeping rods: rod with an included angle of 90°±10° with the vertical rod, and the horizontal and vertical horizontal rods and sweeping rods are distinguished according to the length;

[0080] c) Scissors support inclined rod: rod forming other included angles with the vertical rod.

[0081] 3) Layered processing: according to the z-axis value of the center point coordinates of the straight line segment equation, the rod of each layer is processed.

[0082] 4) Naming and numbering: finally, according to the rod spacing and step distance, the rods in the feature line model are named and numbered to facilitate accurate positioning and management.

[0083] As a possible implementation, further, in step 6) of the scheme, the scaffold engineering compliance set is formed by systematically summarizing the key indicators that affect the safety structure of the scaffold in the current national scaffold safety technical specifications and standards.

[0084] Specifically, in the aspect of summarizing the scaffold engineering compliance set, the scheme induces the compliance set of the fastener type scaffold structure according to the current national fastener type scaffold safety technical specifications, including "Uniform Standard for Safety Technology of Scaffolding for Building Construction" (GB51210-2016), "Standard for Internal Scaffolding for Building Construction", "Safety Technical Code for Steel Tubular Scaffolding with Couplers for Building Construction" (JGJ130-2011), and "Safety Technical Code for High Altitude Operation in Building Construction" (JGJ80-2016), and induces the compliance set of the fastener type scaffold structure, covering the key safety evaluation indicators of quantifiable components such as horizontal rods, vertical rods, scissors support inclined rods, and main and auxiliary nodes of horizontal and vertical rods in the same layer, as follows:

[0085] 1) Horizontal bar construction requirements

[0086] a) The longitudinal horizontal bar level tolerance deviation value Δh≤±20mm;

[0087] b) The longitudinal horizontal bar step distance Δs≤1800mm;

[0088] c) The longitudinal horizontal bar spacing Δg≤1800mm;

[0089] d) The transverse longitudinal horizontal bar overhanging length is not less than l≤15mm;

[0090] e) The height of the scaffold sweeping bar is not more than h≤200mm;

[0091] f) The scaffold transverse sweeping bar must be lower than the longitudinal sweeping bar;

[0092] 2) Vertical bar construction requirements;

[0093] a) The vertical bar allowable longitudinal distance and erection height meet the following table requirements.

[0094] Table 1 Scaffold allowable vertical bar longitudinal distance and erection height

[0095]

[0096] a) The vertical bar longitudinal distance tolerance deviation value δ v ≤±20mm;

[0097] b) The erection height H satisfies the overall verticality tolerance deviation δ H ≤

[0098] ±100mm;

[0099] 3) Construction requirements of scissors support

[0100] a) The scissors support diagonal bar and the ground inclination angle α satisfies 45°≤α≤60°;

[0101] b) The number of spans and the width of the scissors support meet the following table 2:

[0102] Table 2 Number of spans and width of scissors support diagonal bar

[0103]

[0104] c) When the scaffold erection height H≤24m, the net distance d of the scissors support in the scaffold is ≤15m;

[0105] 4) Safety check of same layer transverse and longitudinal horizontal bar main and auxiliary nodes The maximum spacing of the transverse uneven bar at the non-main node of the scaffold working layer is not more than 1 / 2 of the longitudinal distance.

[0106] As a possible implementation, further, in step 7) of the scheme, the scaffold engineering construction compliance evaluation includes:

[0107] The key indicators of the scaffold safety structure are converted into the geometric relationship of the center characteristic line of the rod, and when the axis of the bottom vertical rod is parallel to the Z-axis direction of the overall coordinate system, the geometric relationship of the center characteristic line of the rod includes:

[0108] 1) The verticality deviation, horizontal deviation value, overhanging length, and maximum z coordinate value of the center characteristic line of a single vertical rod or horizontal rod;

[0109] 2) The included angle, shortest distance, and x / y / z coordinate difference between the center characteristic lines of two vertical rods or horizontal rods;

[0110] 3) The included angle between the center characteristic line of the diagonal scissors strut and the XOY plane, the number of intersection points of the center characteristic line of the scissors strut and the center characteristic line of the vertical rod in the orthographic plane, the x / y coordinate difference between the end points of two intersecting scissors strut center characteristic lines, and the x / y coordinate difference between the end points of two adjacent but non-intersecting scissors strut center characteristic lines.

[0111] Wherein, the solution of the spatial relationship of the characteristic line includes the following two implementation steps:

[0112] 1) Based on the center points (x1, y1, z1) and (x2, y2, z2) of the two end sections of the steel pipe, the spatial line segment equation of the corresponding steel pipe characteristic line in the overall coordinate system is solved, assuming that x1≤x2, y1≤y2, z1≤z2, then the center characteristic line spatial equation of the steel pipe is:

[0113]

[0114] 2) Based on the characteristic line spatial equation, the spatial relationship of two steel pipes is solved, including the included angle, the shortest distance, the projection relationship on the coordinate plane, and the characteristic line end point. For two steel pipes with the following characteristic line equations:

[0115]

[0116] a) Included angle solution: calculate the included angle of the direction vectors of the characteristic lines of the two steel pipes,

[0117]

[0118] wherein: is the direction vector on the straight line L1;

[0119] is the direction vector on the straight line L2.

[0120] b) Shortest distance solution: the shortest distance between the two characteristic lines is solved by using the distance formula between spatial straight lines, and the formula is as follows:

[0121]

[0122] wherein: is a point on the straight line L1; is a point on the straight line L2.

[0123] c) Projection intersection and perpendicular distance: Project the feature line to the coordinate plane, solve the intersection and perpendicular distance in the plane by geometric projection, further analyze the verticality deviation, horizontal deviation, overhanging length, maximum z coordinate value of a single vertical pole or horizontal pole, the x / y / z coordinate difference value of the center feature line of two vertical poles or horizontal poles, the angle between the diagonal bracing diagonal and the ground and its intersection or adjacent relationship with the vertical pole.

[0124] The corresponding relationship between the construction compliance requirements of the fastener type scaffold and the spatial relationship of the pole feature line is as follows:

[0125] 1) Safety inspection of transverse / longitudinal horizontal poles and sweeping poles

[0126] a) Levelness: the levelness deviation of the transverse / longitudinal horizontal poles is the z coordinate difference value of the two end points of the longitudinal horizontal pole on the same layer;

[0127] b) Step or spacing: the step of the longitudinal horizontal pole is the shortest distance between the adjacent longitudinal horizontal feature lines of the upper and lower two layers; the spacing of the longitudinal horizontal pole is the shortest distance between the adjacent longitudinal horizontal feature line segments on the same layer;

[0128] c) Overhanging length: the distance from the projection point of the spatial intersection point of the edge transverse / longitudinal horizontal pole and the vertical pole on the feature line of the former to the end point thereof;

[0129] d) Sweeping pole height: the distance from the projection point of the spatial intersection point of the transverse / longitudinal sweeping pole and the vertical pole on the feature line of the latter to the XOY plane;

[0130] e) Relative height of transverse and longitudinal horizontal poles: whether the center z coordinate of the transverse sweeping pole feature line is less than the center z coordinate of the longitudinal sweeping pole;

[0131] f) Passing through the vertical pole: the X coordinate of the longitudinal horizontal pole or the sweeping pole is less than the vertical pole.

[0132] 2) Safety inspection of vertical poles

[0133] a) Erection height: the maximum z coordinate value among the end points of all vertical pole feature lines;

[0134] b) Settlement value: the average change value of the z coordinate value of the midpoint of all sweeping pole feature lines at different times;

[0135] c) Longitudinal deviation: Calculate the difference in x or y coordinates of two adjacent vertical poles with the same z coordinate (the difference in z coordinate is less than 0.05), compare it with the design value, and determine whether the deviation value meets the specifications.

[0136] d) Verticality deviation: The distance between the projection points of the two ends of the feature line of each pole in the XOY plane.

[0137] 3) Safety inspection of scissor bracing

[0138] a) Inclination angle: The angle between the scissor brace feature line and the XOY plane;

[0139] b) Number of spans and width: The number of intersections between the projections of each scissor brace and the upright on the XOY plane is the number of spans of the scissor brace, and the length of the projection is the actual width of the scissor brace;

[0140] c) Spacing: The shortest distance between the projection points of the diagonal ends of two adjacent scissor braces in the XOY plane.

[0141] 4) Safety inspection of main and secondary nodes of horizontal and longitudinal bars on the same floor

[0142] Spacing: The shortest distance between the uprights along the Y-axis is the longitudinal spacing of the uprights; the intersection with the longitudinal horizontal bar is the main node; the transverse bars within a certain range of the main node are transverse horizontal bars, and the rest are transverse horizontal bars. The shortest distance between adjacent transverse horizontal bars is called the spacing, and the maximum spacing does not exceed 1 / 2 of the longitudinal spacing.

[0143] Based on the above, this solution also proposes a scaffolding project compliance detection method, which applies the aforementioned intelligent evaluation method for scaffolding project compliance based on three-dimensional reconstruction technology.

[0144] The following is a specific example to further illustrate this solution:

[0145] Taking an L-shaped right-angle, three-span, coupler-type scaffolding as an example, it uses Φ48.3×3.6mm steel pipes, with an overall length × width × height of 3m × 3m × 2m. See the attached diagram for details. Figure 3 As shown.

[0146] The data acquisition device used was a DJI mini4 Pro, with camera parameters of 1 / 1.3-inch CMOS, image size of 10.3mm*7.7mm, 48 million effective pixels, 82.1° field of view, and 7mm focal length.

[0147] The flight strategy employed a 360° surround method, with flight altitudes of 3m, 2m, and 1m for filming. The camera angle was perpendicular to the shooting surface, and the overlap was set to 60% and 80%. Flight times were chosen in the morning (500-600W / m²). 2), 12:00 noon (800-1000 W / m 2 The data acquisition parameters are shown in Table 3.

[0148] Table 3 Data acquisition parameter table

[0149]

[0150] Collecting pictures of fastener scaffolding, including scissors, wall components, and sweeping rods, about 700, through data enhancement, including cropping, brightness adjustment, blur processing and noise addition, etc. Methods, expand the data to 5295, divide according to the training set and validation set 4:1.

[0151] Based on the nano model in the YOLOv8 instance segmentation pre-training model, after training and parameter adjustment, the information extraction of the key features of the scaffold is realized, and the edge contour information of each type of key component in each picture can be extracted, as shown in the accompanying Figure 4 .

[0152] Based on image information, unmanned aerial vehicle flight data and camera data, using SfM technology and MVS technology for scaffold three-dimensional reconstruction, as shown in the accompanying Figure 5 .

[0153] Using statistical filtering (SOR) combined with radius filtering (ROR) method to filter out noise points, and performing voxel downsampling operation, to obtain a three-dimensional point cloud model containing only scaffold components, as shown in the accompanying Figure 6 .

[0154] Through the density clustering algorithm DBSCAN combined with the distance and angle between points to extract the steel pipe point cloud, as shown in the accompanying Figure 7 .

[0155] Extract the centerline of each cylindrical steel pipe to obtain the feature line model. After rotation and translation, the spatial equation of part of the feature line model is shown in the accompanying Figure 8 and Table 4 below,

[0156] Table 4 Scaffold test line model data (part)

[0157]

[0158]

[0159] Evaluate the deviation of the feature line model from the L-shaped scaffold structure size, indicating that the deviation range is within 0-3.2 cm, and the average error is 1.15 cm.

[0160] According to the comparison of the scaffold structure compliance set and the spatial relationship of the feature line model, the construction compliance judgment result of the L-shaped scaffold is obtained (see Table 5).

[0161] Table 5 Scaffold test model construction normative determination (part)

[0162]

[0163] The above merely describes some embodiments of the present application, and does not limit the protection scope of the present application, and any equivalent device or equivalent process transformation using the content of the present application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for intelligent evaluation of scaffold engineering compliance based on three-dimensional reconstruction technology, characterized in that, It comprises: Step 1) According to the engineering characteristics of the area where the scaffold project is located, the parameter plan of unmanned aerial vehicle equipment sampling is made, the scaffold project is sampled by unmanned aerial vehicle aerial photography, and aerial photography pictures are generated; Step 2) The aerial photography pictures are preprocessed and then input into the trained machine vision model to identify the key components of the scaffold in the aerial photography pictures, and then the categories and quantities of the key components of the scaffold are counted and used to evaluate the rationality of the component distribution of the scaffold system; Step 3) The aerial photography pictures containing the scaffold components are used to generate the scaffold three-dimensional point cloud model in reverse by using the incremental SfM and multi-view stereo matching algorithm MVS; Step 4) The scaffold three-dimensional point cloud model is segmented by using the density clustering algorithm DBSCAN combined with the distance and angle judgment method between adjacent components to obtain the point cloud model of each rod of the scaffold; Step 5) The point cloud model of the rod is used to extract the rod center feature line by using the cylindrical fitting algorithm combined with the RANSAC algorithm to filter outliers; Step 6) According to the corresponding technical specifications and standards of the scaffold, the scaffold engineering compliance set is summarized; Step 7) The spatial equation of the rod center line is calculated, the spatial relationship of the rod feature line is used to express the scaffold compliance requirements, and the scaffold engineering compliance set is combined to evaluate the construction compliance of the scaffold engineering.

2. The method of claim 1, wherein the method further comprises: In step 1), the engineering characteristics of the area where the scaffold project is located include the facade features of the scaffold and / or the site restrictions of the site where the scaffold is located; The parameter plan of unmanned aerial vehicle equipment sampling includes the path and aerial photography parameters of the unmanned aerial vehicle aerial photography; the aerial photography parameters include one or more of the heading, the side image overlap, the camera tilt angle, the photography distance, and the light intensity selection.

3. The method of claim 1, wherein the method further comprises: In step 2), the preprocessing of the aerial photography pictures includes: Filtering out aerial photography images that meet the quality requirements, and performing image enhancement, denoising, smoothing and geometric transformation processing on them.

4. The intelligent evaluation method of scaffold engineering compliance based on three-dimensional reconstruction technology according to claim 1, characterized in that, The trained machine vision model is a YOLOv8 model, and its training method includes: Collecting scaffold pictures, classifying and labeling them by category, and then dividing them into a training set, a validation set and a test set, training the YOLOv8 model by inputting the training set into the model, verifying and testing the trained YOLOv8 model using the validation set and the test set, and then outputting the converged YOLOv8 model as a category identifier for classifying the key components of the scaffold in the aerial photography pictures after the test and verification meet the requirements.

5. The intelligent evaluation method of scaffold engineering compliance based on three-dimensional reconstruction technology according to claim 4, characterized in that, The category identifier is used to identify one or more components of the scaffold key components in the aerial photography pictures, such as wall connecting parts, floor sweeping rods, connecting fasteners and scissors supports.

6. The method for intelligent evaluation of scaffold engineering compliance based on three-dimensional reconstruction technology according to claim 1, wherein, In step 3), the scaffold three-dimensional point cloud model refers to the point cloud model of the aerial photography scene reconstructed in reverse by using incremental SfM and MVS technology on the picture set processed in step 2), and then performing voxel downsampling and filtering denoising operations, and after the overall coordinate transformation, the scaffold three-dimensional point cloud model satisfies that the axis of the bottom vertical rod is parallel to the Z-axis direction of the overall coordinate system.

7. The method of claim 1, wherein the method further comprises: In step 5), the rod center feature line is a spatial straight line segment passing through the center points of the two ends of each point cloud rod in the scaffold three-dimensional point cloud model in the overall coordinate system. It also includes: according to the relationship between the three-dimensional coordinates of the end of the rod and the characteristics of the scaffold construction, the rod is numbered.

8. The method of claim 1, wherein the method further comprises: In step 6), the scaffold engineering compliance set is formed by systematically summarizing the key indicators affecting the safety of scaffold construction in the current national scaffold safety technical specifications and standards, including horizontal and vertical horizontal rods, floor rods, vertical rods, scissors, and safety check items related to main and auxiliary nodes of horizontal and vertical horizontal rods on the same layer.

9. The method for intelligent evaluation of the scaffold engineering compliance based on the three-dimensional reconstruction technology according to claim 1, characterized in that, In step 7), the scaffold engineering construction compliance evaluation includes: The key indicators of scaffold safety construction are converted into the geometric relationship of rod center characteristic line. When the axis of the vertical rod on the bottom layer is parallel to the Z-axis direction of the overall coordinate system, the geometric relationship of the rod center characteristic line includes: 1) The verticality deviation, horizontal deviation value, overhanging length, and maximum z coordinate value of the single vertical rod or horizontal rod rod center characteristic line; 2) The angle between the center characteristic lines of two vertical rods or horizontal rods, the shortest distance, and the x / y / z coordinate difference value; 3) The angle between the center characteristic line of the inclined scissors and the XOY plane, the number of intersection points of the center characteristic line of the scissors in the orthographic plane and the center characteristic line of the vertical rod, the x / y coordinate difference value of the two intersecting scissors rod center characteristic line endpoints, and the x / y coordinate difference value of the two adjacent but non-intersecting scissors rod center characteristic line endpoints.

10. A method for detecting the compliance of scaffolding projects, characterized in that, The application has the scaffold engineering compliance intelligent evaluation method based on three-dimensional reconstruction technology according to any one of claims 1-9.

Images