Board identification method based on geometric figure range discrimination

Abstract

The invention relates to the technical field of plate recognition, in particular to a plate recognition method based on geometric figure range discrimination. The method mainly comprises the following steps: acquiring geometric data of a plate by using an image acquisition device, a laser scanner and a BIM model, wherein the geometric data comprises boundary contour, surface morphology, spatial position and size information; extracting boundaries and key geometric feature points of the plates through a graph cut algorithm; generating a three-dimensional geometric model by adopting a multi-scale shape analysis algorithm; performing geometric range discrimination by using a Markov random field algorithm, and identifying an abnormal region; and comparing the BIM model design data through an iterative nearest point algorithm to generate an identification report. According to the invention, accurate identification and range discrimination of the building structural slab are realized, and the monitoring and management efficiency of the construction quality is effectively improved.

Description

Technical Field

[0001] The invention relates to the technical field of plate recognition, and in particular to a plate recognition method based on geometric figure range discrimination. Background Art

[0002] In building structure design, plates are widely used in various buildings as key structural components for bearing and transferring loads. Plates in building structures are usually made of materials such as concrete and steel, with various shapes and sizes, and have their own geometric characteristics according to different load-bearing requirements and structural arrangements. However, with the complexity of modern building design, especially in high-rise buildings, special-shaped buildings and complex frame structures, the geometric shape and layout of plates have become more complex. At present, there are still the following problems: most traditional plate recognition methods rely on manual visual inspection or simple mechanical recognition methods, which are not only inefficient but also have high error rates; traditional methods are mostly used for planes or simple two-dimensional structures, and cannot effectively recognize and process complex three-dimensional structural plates. There is a lack of refined three-dimensional geometric model generation and analysis methods, resulting in low recognition accuracy in practical applications; when processing complex building plates, the existing technology is easily affected by changes in geometric form and spatial position, resulting in inaccurate range discrimination results or poor stability, which cannot meet the needs of high-precision recognition. Summary of the invention

[0003] To solve the above problems, the present invention provides a plate recognition method based on geometric range discrimination, which solves the problem of low efficiency and high error of the plate recognition method when the geometric shape and spatial position change in complex building design, and effectively generates and analyzes complex three-dimensional geometric models, ensuring the accuracy and stability of the range discrimination results to meet the needs of high-precision recognition, thereby realizing accurate recognition and range discrimination of building structure plates.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] A plate recognition method based on geometric figure range discrimination comprises the following steps:

[0006] S1: using an image acquisition device, a laser scanner and a BIM model to obtain geometric data of a building structure plate; the geometric data includes boundary contours, surface morphology, spatial position and size information of the building structure plate;

[0007] S2: Based on the geometric data, a graph cut algorithm is used to extract the boundary, obtain the contour line and boundary feature points of the building structure plate, and identify key feature points, including geometric inflection points and curve features of the building structure plate;

[0008] S3: Analyze the key feature points using a multi-scale shape analysis algorithm to generate a three-dimensional geometric model of the building structure plate;

[0009] S4: Based on the three-dimensional geometric model, the Markov random field algorithm is used to construct a geometric figure range discrimination model for range discrimination and generate range discrimination results;

[0010] S5: Based on the range identification result, the range identification result identified based on the iterative closest point algorithm is compared and verified with the design data in the BIM model, and an identification report of the building structure plate is generated.

[0011] Furthermore, the building structural panels include concrete panels, steel structural panels, aluminum panels and composite structural panels.

[0012] Furthermore, the step S2 includes the following steps:

[0013] S21: Based on the geometric data, collecting point cloud data of the building structure plate by means of a laser scanner, and performing a preprocessing operation on the point cloud data; the point cloud data includes three-dimensional coordinate point information of the building structure plate;

[0014] S22: extracting the boundary contour of the pre-processed point cloud data by using a graph cut algorithm, and combining the boundary data of the BIM model to determine the initial boundary of the building structure plate;

[0015] S23: Based on the initial boundary and in combination with the material characteristics of the building structure plate, the initial boundary is optimized using a normal vector smoothing algorithm to identify key feature points; the key feature points include geometric inflection points, straight line edges, curve segments, and curvature change feature points of the building structure plate.

[0016] Furthermore, the step S23 includes the following steps:

[0017] According to the geometric data and material properties of the building structure plate, its surface normal vector is calculated, and a preliminary normal vector field is constructed based on the distribution of the normal vector;

[0018] Performing weighted smoothing processing on the normal vector field by using a weighted average method to generate an optimized normal vector field;

[0019] The optimized normal vector field is used to identify the key feature points of the building structure plate based on the feature extraction algorithm based on curvature change analysis and differential geometry.

[0020] Furthermore, the step S3 includes the following steps:

[0021] S31: constructing a preliminary geometric outline of the building structure plate based on the key feature points;

[0022] S32: Analyze the key feature points using a multi-scale shape analysis algorithm, and generate a three-dimensional geometric model based on the preliminary geometric outline of the building structure plate;

[0023] S33: Performing surface refinement processing on the generated three-dimensional geometric model, eliminating irregularities using a smooth surface fitting algorithm, optimizing boundary accuracy, and obtaining an optimized three-dimensional geometric model.

[0024] Further, the step S4 includes the following steps:

[0025] S41: Based on the three-dimensional geometric model, construct a preliminary geometric range of the building structure plate, and define geometric boundary conditions and restriction conditions of the building structure plate;

[0026] S42: constructing a geometric figure range discrimination model based on the geometric boundary conditions of the building structure plate through a Markov random field algorithm, and modeling and optimizing the relationship between adjacent geometric units in the model; the nodes in the geometric figure range discrimination model represent geometric boundary points and feature points, and the edges represent the mutual geometric relationship between these points;

[0027] S43: Optimizing the Markov random field model using an energy function, calculating the energy value of each geometric unit, and identifying abnormal areas or areas that do not meet the design specifications within the geometric range;

[0028] S44: Generate geometric range discrimination results based on the optimized geometric figure range discrimination model, and output range discrimination information of the building structure plate.

[0029] Furthermore, the formula of the geometric figure range discrimination model is as follows:

[0030]

[0031] Where E(X) represents the total energy of the geometric range discrimination model, which is used to judge whether the geometric range of the building structure plate meets the specifications; V represents the node set of the geometric figure, that is, the key feature points on the building structure plate; E represents the edge set of the geometric figure, that is, the relationship between the key feature points; N i Represents the normal direction of the building structure plate surface at the feature point i; represents the normal direction of the building structure plate at the characteristic point i under ideal conditions; x i and x j represents the position vector of nodes i and j, that is, the three-dimensional coordinates of two adjacent key feature points on the building structure plate; α represents the weight parameter of the normal vector difference; β represents the weight parameter of the geometric relationship between adjacent feature points.

[0032] Furthermore, step S5 includes the following steps:

[0033] S51: Based on the geometric range discrimination result, the boundary feature information of the building structure plate is extracted, and the identified geometric feature points and boundary lines are preliminarily aligned with the design data in the BIM model;

[0034] S52: using an iterative closest point algorithm to accurately align the recognition results, and optimizing the registration accuracy of the geometric model of the building structure plate and the BIM model design data through multiple iterations;

[0035] S53: Based on the alignment result, identifying the deviation area between the geometric model and the design data, analyzing the nature and source of the deviation, and determining whether the building structure plate has a geometric deviation exceeding the allowable tolerance;

[0036] S54: Generate an identification report of the building structure plate, the report content including geometric deviation information, spatial position of the deviation area, deviation range and degree of conformity with the BIM model design data.

[0037] Furthermore, the formula of the iterative closest point algorithm is as follows:

[0038]

[0039] Among them, T k+1 represents the transformation matrix after the k+1th iteration; T represents the rigid body transformation matrix, that is, the rotation and translation operations are performed on the geometric data of the building structure plate collected by the laser scanner; p i represents the three-dimensional coordinates of the i-th feature point on the building structure plate collected by the laser scanner; q i Indicates the BIM model with p i The three-dimensional coordinates of the nearest design point corresponding to the feature point; N represents the total number of feature point pairs; ||T(p i )-q i || 2 represents the squared Euclidean distance.

[0040] The beneficial effects of the present invention are:

[0041] The present invention can accurately obtain the geometric data of the building structure plate through the combination of image acquisition equipment, laser scanner and BIM model, realize the accurate description of the plate's outline, surface morphology, spatial position and size, and provide high-precision basic data for subsequent identification. Using advanced computer vision and mathematical models such as graph cut algorithm, Markov random field algorithm, multi-scale shape analysis algorithm, etc., the boundaries, outlines and feature points of the building structure plate can be automatically identified, reducing manual intervention, and improving recognition efficiency and intelligence level. Through multi-scale shape analysis of key feature points, a three-dimensional geometric model is generated, making the recognition process more three-dimensional and detailed, and the morphology and position of the building structure plate can be accurately presented, providing strong support for subsequent comparison. The geometric figure range discrimination model constructed by the Markov random field algorithm can accurately discriminate the range of the building structure plate, ensuring that the recognition result is more accurate in spatial range. The iterative closest point algorithm is used to compare and verify the identified geometric model with the design data in the BIM model to ensure the accuracy and reliability of the recognition result, effectively reduce the error, and finally generate a reliable recognition report to provide data support for engineering construction and monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a flow chart of a plate recognition method based on geometric figure range discrimination according to the present invention.

[0043] Figure 2 It is a flowchart diagram of step S4 provided in one embodiment of the present invention.

[0044] Figure 3 It is a flowchart diagram of step S5 provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0045] See also Figure 1-3 As shown, the present invention relates to a plate recognition method based on geometric figure range discrimination.

[0046] Example

[0047] A plate recognition method based on geometric figure range discrimination comprises the following steps:

[0048] S1: Use image acquisition equipment, laser scanners and BIM models to obtain geometric data of building structural panels; the geometric data includes boundary contours, surface morphology, spatial positions and size information of the building structural panels; the building structural panels include concrete panels, steel structural panels, aluminum panels and composite structural panels.

[0049] Specifically, select image acquisition equipment suitable for the construction site environment, such as high-resolution digital cameras, drone photography equipment, or stereo vision cameras. Depending on the specific environment and conditions of the building structure board, an industrial-grade camera with dustproof and waterproof functions can be selected. For example, using drones for high-altitude and large-scale photography is particularly suitable for large building structures such as steel structure boards and concrete slabs. For indoor structural panels (such as composite structural panels or steel structure panels), a stereo camera mounted on a track can be used for close-up photography.

[0050] Select a laser scanner that is suitable for the current size and position of the building structure plate. For example, using a 3D laser scanner (such as a LiDAR device) for all-round scanning can provide millimeter-level accuracy, which is suitable for obtaining geometric data of larger or complex structural plates. For scenes that require high-precision geometric information, use a high-precision static laser scanner. The device can be fixed in a specific position to scan the boundary, surface, and spatial position information of the plate multiple times to obtain complete 3D point cloud data. If the building structure plate is large or the environment is complex, a dynamic laser scanner (such as a portable laser scanning device or a LiDAR system mounted on a drone) can also be used. The device can be mobile for collection, and a large-scale scan covers the entire building structure plate.

[0051] Extract BIM (Building Information Modeling) data directly from the design phase of a building project. The BIM model contains design information of the building panels, such as size, boundary shape, thickness, material, and installation location. By accessing BIM software (such as Revit, AutoCAD, or other architectural design tools), export the geometric data related to the building structural panels. These geometric data include two-dimensional plane information (such as the boundary lines of the panels) and three-dimensional solid geometry (such as the surface model and spatial position of the panels).

[0052] Integrate the geometric data obtained from image acquisition equipment, laser scanners and BIM models. Use spatial coordinate system 1 to align image data and laser scanning point cloud data to the same coordinate system. Use data fusion technology to fuse data from three sources into a complete three-dimensional geometric model. Image data provides texture and color information, laser scanning provides accurate spatial position information, and BIM models provide design references. In order to ensure the accuracy of the data, use calibration plates or control points with known positions on the ground to calibrate the image acquisition equipment and laser scanners to reduce data deviations caused by equipment differences or environmental factors.

[0053] S2: Based on the geometric data, a graph cut algorithm is used to extract the boundary, obtain the contour line and boundary feature points of the building structure plate, and identify key feature points, including geometric inflection points and curve features of the building structure plate;

[0054] Wherein, the step S2 comprises the following steps:

[0055] S21: Based on the geometric data, collecting point cloud data of the building structure plate by means of a laser scanner, and performing a preprocessing operation on the point cloud data; the point cloud data includes three-dimensional coordinate point information of the building structure plate;

[0056] S22: extracting the boundary contour of the pre-processed point cloud data by using a graph cut algorithm, and combining the boundary data of the BIM model to determine the initial boundary of the building structure plate;

[0057] Specifically, the preprocessed 3D point cloud data is projected onto a 2D plane to form a 2D outline of the building structure board. This is because the boundaries of most building boards can be effectively extracted by their plane outlines. The projected data is discretized to create an adjacency graph containing all points, where each node represents a 3D point and the edges represent the spatial relationship between nodes.

[0058] The graph cut algorithm is applied to the two-dimensional projection of the point cloud to extract the boundary contour. The algorithm constructs the point cloud data into a weighted graph model, regards the boundary of the building board as a secant, and determines the best boundary contour by optimizing the secant weight. The weight value in the graph cut algorithm is related to the characteristics of the point cloud data, especially based on the distance between points, the difference in surface normal vectors, and the boundary information of the BIM model. Through the minimum cut method, the secant with the minimum cost is found in the graph model, and the secant is the initial boundary of the building structure plate. Through iterative solution, the position of the secant is gradually optimized. The initial boundary extracted by the graph cut algorithm is compared with the design boundary in the BIM model to ensure that the extracted boundary is basically consistent with the design model. If there is a deviation between the initial boundary and the boundary of the BIM model, the spatial position of the point cloud data is adjusted by matching, and the position and shape of the initial boundary are further corrected.

[0059] S23: Based on the initial boundary and in combination with the material characteristics of the building structure plate, the initial boundary is optimized using a normal vector smoothing algorithm to identify key feature points; the key feature points include geometric inflection points, straight line edges, curve segments, and curvature change feature points of the building structure plate.

[0060] The step S23 comprises the following steps:

[0061] According to the geometric data and material properties of the building structure plate, its surface normal vector is calculated, and a preliminary normal vector field is constructed based on the distribution of the normal vector;

[0062] Specifically, the surface normal vector of each point is calculated based on the three-dimensional geometric data (point cloud) and material properties of the building structure board. The normal vector is a vector that represents the direction of the surface, and is usually generated by fitting a plane in the neighborhood or by a triangular mesh. A preliminary normal vector field is constructed based on the normal vectors of all surface points of the building board. The normal vector field can be regarded as the normal vector distribution of each point on the surface of the building board, which can describe the geometric morphology of the board surface. For the mutation parts in the normal vector field (such as edges or inflection points), special marking is required so that these geometric changes can be accurately captured in the subsequent feature point recognition process.

[0063] Performing weighted smoothing processing on the normal vector field by using a weighted average method to generate an optimized normal vector field;

[0064] Specifically, when smoothing the normal vector field, the weighted average method is used to weight the normal vector of each point to smooth its local fluctuations and remove noise. In the weighted average process, the contribution of each point in the neighborhood to the normal vector of the target point is related to its distance and geometric position. The closer the point is, the greater the weight is, and it is inversely proportional to the change of the normal vector of the point.

[0065] The formula for weighted average of normal vectors is as follows:

[0066]

[0067] Among them, n′(p) represents the optimized normal vector of point p. This is the final normal vector of point p after weighted smoothing, indicating the direction of the surface of the point; p represents a sampling point in the currently processed building board point cloud. This point p is a point in the three-dimensional point cloud, and its three-dimensional coordinates are (x p ,y p , z p ), located at a certain position on the surface of the building structure plate; N(p) represents the neighborhood of point p; w(p, q) represents the weight between point p and the neighborhood point q. The weight value depends on the distance between points p and q, and the similarity of their normal vectors. For the point cloud data of the building structure plate, if the distance between p and q is close or the direction of the normal vector is similar, the weight is larger; n(q) represents the normal vector of point q.

[0068] The optimized normal vector field is used to identify the key feature points of the building structure plate based on the feature extraction algorithm based on curvature change analysis and differential geometry.

[0069] Specifically, in the optimized normal vector field, the curvature is estimated by calculating the rate of change of the neighborhood normal vector of each point. Based on the principle of differential geometry, the curvature formula is used to calculate the change of the normal vector in the neighborhood and identify the area on the surface where significant curvature changes occur. Commonly used curvature calculation methods include neighborhood surface fitting or finite difference method.

[0070] The identification of feature points is as follows:

[0071] Geometric inflection point identification: By analyzing the curvature change in the normal vector field, the geometric inflection points of the building structure are identified. Geometric inflection points are usually the locations where the normal vector changes suddenly, and correspond to corner points and turning points on the boundary.

[0072] Straight edge recognition: In areas where the normal vector changes slowly, the straight edges of the building board are identified by linearly fitting the normal vector. These edges are usually manifested as areas with consistent normal vector directions and zero or very small curvature.

[0073] Curve segment recognition: A curve segment is an area where the surface normal vector changes gradually along the curve. The curvature is relatively smooth but significantly different from the straight line edge. By analyzing the curvature change rate of the normal vector, the boundary of the curve segment can be identified.

[0074] Identification of curvature mutation points: In the normal vector field, curvature mutation points refer to areas where the curvature change rate is large, such as the extreme points and turning points of the curve. These mutation points usually correspond to important geometric features of the building board.

[0075] S3: Analyze the key feature points using a multi-scale shape analysis algorithm to generate a three-dimensional geometric model of the building structure plate;

[0076] Wherein, the step S3 comprises the following steps:

[0077] S31: constructing a preliminary geometric outline of the building structure plate based on the key feature points;

[0078] Specifically, the preliminary geometric outline constructed from the key feature points is used, that is, these feature points are combined into the preliminary outline of the building board through a connection algorithm. The process includes straight line fitting and curve fitting to ensure that the connection lines between the feature points can accurately reflect the overall outline of the building board.

[0079] Line fitting: For straight edges, a least squares fitting algorithm is used to generate an optimal straight line that connects all feature points on the edge. This process eliminates the influence of noise and ensures smooth edges.

[0080] Curve fitting: For curve segments, use polynomial curve fitting or spline interpolation algorithms to connect the characteristic points of the curve section to generate a smooth curve profile. Use interpolation methods such as B-spline or Catmull-Rom spline to ensure that the generated curve is accurate near the inflection points and curvature change points.

[0081] The fitted contour lines are projected into three-dimensional space to form a preliminary geometric framework of the building structural plate. The framework shows the main structural contours of the plate, including boundary lines, straight lines, curve segments, and corner points. This geometric information provides the basis for subsequent model generation. For building panels with complex shapes (such as curved panels or composite panels with concave and convex changes), this contour framework can preliminarily represent the general shape of the plate, but does not yet contain surface details.

[0082] S32: Analyze the key feature points using a multi-scale shape analysis algorithm, and generate a three-dimensional geometric model based on the preliminary geometric outline of the building structure plate;

[0083] Specifically, a preliminary coarse-scale 3D geometric model is generated based on the previous geometric contour. At this stage, the overall shape of the building board is mainly fitted through key feature points and contour lines. This coarse model describes the approximate position, size and boundary shape of the building board. Detailed information is gradually added, especially the local optimization of the surface part. When processing the surface part, the surface fitting algorithm (such as Bézier surface, NURBS surface, etc.) is used to optimize the local geometric shape to ensure the smooth transition of the surface and the accuracy of the boundary. At this stage, the parts with large curvature changes are analyzed in detail to ensure the accuracy of the curve segments and surface parts. The model is optimized layer by layer by using the Gaussian pyramid algorithm by processing the geometric shape in layers. The geometric information of each layer is further enhanced in detail in the next level of processing. At each scale level, combined with the material properties of the building board (such as the roughness of the concrete board surface and the smoothness of the aluminum board), the curvature change and normal vector field near the key points are accurately calculated and refined to ensure that the generated 3D model can reflect the material characteristics of the actual building structure board. Based on the above multi-scale shape analysis results, combined with the preliminary geometric outline of the building board, the final three-dimensional geometric model is generated. The model should include all key features such as the board's boundaries, surface curves, inflection points, and curvature change points, and should be highly consistent with the actual building structure's geometric form.

[0084] S33: Performing surface refinement processing on the generated three-dimensional geometric model, eliminating irregularities using a smooth surface fitting algorithm, optimizing boundary accuracy, and obtaining an optimized three-dimensional geometric model.

[0085] S4: Based on the three-dimensional geometric model, the Markov random field algorithm is used to construct a geometric figure range discrimination model for range discrimination and generate range discrimination results;

[0086] Wherein, the step S4 comprises the following steps:

[0087] S41: Based on the three-dimensional geometric model, construct a preliminary geometric range of the building structure plate, and define geometric boundary conditions and restriction conditions of the building structure plate;

[0088] S42: constructing a geometric figure range discrimination model based on the geometric boundary conditions of the building structure plate through a Markov random field algorithm, and modeling and optimizing the relationship between adjacent geometric units in the model; the nodes in the geometric figure range discrimination model represent geometric boundary points and feature points, and the edges represent the mutual geometric relationship between these points;

[0089] Specifically, the three-dimensional geometric model of the building structure plate is represented as a Markov random field (MRF) model. MRF describes the relationship between geometric units through the nodes and edges of the graph. Nodes represent geometric boundary points and feature points, and edges represent the mutual geometric relationship between these points. In this model, each geometric feature point of the building structure plate (such as inflection points, boundary points, curve segments, etc.) is represented as a node. These nodes are connected by edges, and the weight of the edge represents the spatial relationship between the geometric feature points (such as distance, angle, curvature change between adjacent points, etc.).

[0090] The relationships between adjacent geometric cells are modeled by edges in the Markov random field. These relationships include:

[0091] Geometric distance: The Euclidean distance between two adjacent geometric points. This is used to indicate the spatial proximity between geometric feature points. The closer the distance, the higher the weight of the edge, indicating a stronger connection between the two points.

[0092] Normal vector relationship: The difference between the normal vectors of two geometric feature points. By calculating the change in the normal vectors of adjacent geometric units, the curvature change of the building structure plate can be described. If the normal vector difference is large, it may indicate an inflection point or an area with large curvature.

[0093] Curvature Change: Identify if a building panel surface has significant curvature by comparing the curvature of adjacent geometric points. This is especially important when dealing with curved panels and composite panels with concave and convex structures.

[0094] In order to improve the accuracy of the geometric range discrimination model, an optimization algorithm is used to adjust the edge weights in the model. By minimizing the distance difference, normal vector difference and curvature change between adjacent points, the relationship between adjacent geometric units in the model can accurately reflect the geometry of the real building board.

[0095] S43: Optimizing the Markov random field model using an energy function, calculating the energy value of each geometric unit, and identifying abnormal areas or areas that do not meet the design specifications within the geometric range;

[0096] It should be noted that the results of energy optimization can effectively identify areas in the building panels that do not meet the geometric specifications. The nodes in these areas will show high energy values ​​during the optimization process, indicating that the geometric characteristics of this area are significantly different from other parts, and there may be design or construction errors. Abnormal areas may include areas with edges outside the design range, uneven surfaces, or areas with uneven thickness.

[0097] S44: Generate geometric range discrimination results based on the optimized geometric figure range discrimination model, and output range discrimination information of the building structure plate.

[0098] Furthermore, the formula of the geometric figure range discrimination model is as follows:

[0099]

[0100] Where E(X) represents the total energy of the geometric range discrimination model, which is used to judge whether the geometric range of the building structure plate meets the specifications; V represents the node set of the geometric figure, that is, the key feature points on the building structure plate; E represents the edge set of the geometric figure, that is, the relationship between the key feature points; N i Represents the normal direction of the building structure plate surface at the feature point i; represents the normal direction of the building structure plate at the characteristic point i under ideal conditions; x i and x j represents the position vector of nodes i and j, that is, the three-dimensional coordinates of two adjacent key feature points on the building structure plate; α represents the weight parameter of the normal vector difference, which is used to control the contribution of the normal vector deviation to the total energy; β represents the weight parameter of the geometric relationship between adjacent feature points, which controls the influence of the distance deviation between feature points on the total energy.

[0101] S5: Based on the range identification result, the range identification result identified based on the iterative closest point algorithm is compared and verified with the design data in the BIM model, and an identification report of the building structure plate is generated;

[0102] Wherein, the step S5 comprises the following steps:

[0103] S51: Based on the geometric range discrimination result, the boundary feature information of the building structure plate is extracted, and the identified geometric feature points and boundary lines are preliminarily aligned with the design data in the BIM model;

[0104] S52: using an iterative closest point algorithm to accurately align the recognition results, and optimizing the registration accuracy of the geometric model of the building structure plate and the BIM model design data through multiple iterations;

[0105] Specifically, the iterative closest point algorithm (ICP) is used to further accurately align the geometric model with the BIM model. The ICP algorithm finds the closest point pair between two sets of point cloud data (geometric model and design data) and minimizes the distance between the two in each iteration. The goal of the ICP algorithm is to accurately match the geometric range discrimination result with the design data in the BIM model in space through multiple iterations.

[0106] The specific steps of the ICP algorithm include: for each feature point in the geometric model, find the closest point in the BIM design data to form a point pair. Based on the distance between the point pairs, calculate the rigid transformation parameters of the geometric model, including translation and rotation. Through the least squares method, calculate the optimal rigid transformation to minimize the distance between the two groups of points. Apply the calculated transformation to the geometric model to update the position and posture of the model. Repeat the above steps to gradually reduce the distance between the two groups of points until the set error threshold or iteration limit is reached. After each iteration, the alignment effect is evaluated by calculating the root mean square error of the two sets of data. When the error is less than the predetermined threshold, it means that the alignment result meets the accuracy requirements.

[0107] S53: Based on the alignment result, identifying the deviation area between the geometric model and the design data, analyzing the nature and source of the deviation, and determining whether the building structure plate has a geometric deviation exceeding the allowable tolerance;

[0108] It should be noted that based on the ICP alignment results, the deviation between each point in the geometric model and the corresponding point in the BIM design data is calculated. Deviations include differences in geometric parameters such as plane position, thickness, and surface curvature. Use the deviation analysis tool to identify areas in the geometric model that deviate greatly from the design data. Common deviation areas include boundary line misalignment, uneven thickness, abnormal surface undulation, etc.

[0109] The analysis of the nature of the deviation is as follows:

[0110] Geometric deviation classification: Geometric deviations are classified into different types, such as:

[0111] Boundary Deviation: The boundary of the building panel does not match the designed boundary, such as misaligned or bent edges.

[0112] Thickness deviation: The actual thickness does not match the designed thickness, which may cause structural performance problems.

[0113] Surface Topography Deviation: Deviation in surface flatness that may affect surface smoothness or strength.

[0114] Calculation of deviation size: For each deviation area, calculate the specific value of the deviation, including distance deviation (mm), curvature deviation, etc., to ensure that the degree of each deviation area can be quantified.

[0115] According to the tolerance requirements in the design specifications and BIM model, determine whether each deviation area exceeds the allowable tolerance range. Areas that exceed the tolerance range will be marked as unqualified areas and need to be processed or corrected. If the deviation is within the allowable range, the geometric model can be considered to meet the design requirements.

[0116] S54: Generate an identification report of the building structure plate, the report content including geometric deviation information, spatial position of the deviation area, deviation range and degree of conformity with the BIM model design data.

[0117] Furthermore, the formula of the iterative closest point algorithm is as follows:

[0118]

[0119] Among them, T k+1 represents the transformation matrix after the k+1th iteration; T represents the rigid body transformation matrix, that is, the rotation and translation operations are performed on the geometric data of the building structure plate collected by the laser scanner; p i represents the three-dimensional coordinates of the i-th feature point on the building structure plate collected by the laser scanner; q i Indicates the BIM model with p i The three-dimensional coordinates of the nearest design point corresponding to the feature point; N represents the total number of feature point pairs; ||T(p i )-q i || 2 Represents the square of the Euclidean distance, which is used to measure the distance between the characteristic point of the i-th building structure plate after transformation T and the nearest design point q in the BIM model. i The geometric deviation between .

[0120] In summary, the present invention can comprehensively utilize the advantages of different devices to ensure the comprehensiveness and accuracy of geometric information through the fusion of geometric data of image acquisition equipment, laser scanners and BIM models. The use of graph cut algorithm and normal vector smoothing algorithm can efficiently extract the boundaries of building structure plates and identify key geometric feature points, such as inflection points, straight line edges and curve segments. This process combines the design data of the BIM model to effectively optimize the boundary position and ensure high-precision restoration of geometric features.

[0121] The present invention uses a multi-scale shape analysis algorithm to optimize the geometric model layer by layer, which can accurately restore the complex building board morphology, and is particularly suitable for boards with large surface curvature or concave-convex structures. At the same time, the refinement of the curvature and normal vectors ensures the smooth transition and boundary accuracy of the model. The range discrimination based on the Markov random field model can automatically identify abnormal areas within the geometric range, especially the geometric deviations that may occur during the construction process. The accuracy of the discrimination results is guaranteed by energy optimization and modeling of the spatial relationship of geometric units.

[0122] The invention uses the iterative closest point algorithm (ICP) to accurately align the geometric model with the BIM design data, and can identify the deviation between the geometric model and the design specification. The identification report finally generated records the geometric deviation and its location in detail, providing an important basis for subsequent quality control and correction.

[0123] The above implementation modes are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A plate recognition method based on geometric figure range discrimination, characterized in that: The following steps are involved: S1: using an image acquisition device, a laser scanner and a BIM model to obtain geometric data of a building structure plate; the geometric data includes boundary contours, surface morphology, spatial position and size information of the building structure plate; S2: Based on the geometric data, a graph cut algorithm is used to extract the boundary, obtain the contour line and boundary feature points of the building structure plate, and identify key feature points, including geometric inflection points and curve features of the building structure plate; S3: Analyze the key feature points using a multi-scale shape analysis algorithm to generate a three-dimensional geometric model of the building structure plate; S4: Based on the three-dimensional geometric model, the Markov random field algorithm is used to construct a geometric figure range discrimination model for range discrimination and generate range discrimination results; S5: Based on the range identification result, the range identification result identified based on the iterative closest point algorithm is compared and verified with the design data in the BIM model, and an identification report of the building structure plate is generated.

2. A plate recognition method based on geometric figure range discrimination according to claim 1, characterized in that: The building structural panels include concrete panels, steel structural panels, aluminum panels and composite structural panels.

3. A plate recognition method based on geometric figure range discrimination according to claim 1, characterized in that: The step S2 comprises the following steps: S21: Based on the geometric data, collecting point cloud data of the building structure plate by means of a laser scanner, and performing a preprocessing operation on the point cloud data; the point cloud data includes three-dimensional coordinate point information of the building structure plate; S22: extracting the boundary contour of the pre-processed point cloud data by using a graph cut algorithm, and combining the boundary data of the BIM model to determine the initial boundary of the building structure plate; S23: Based on the initial boundary and in combination with the material characteristics of the building structure plate, the initial boundary is optimized using a normal vector smoothing algorithm to identify key feature points; the key feature points include geometric inflection points, straight line edges, curve segments, and curvature change feature points of the building structure plate.

4. A plate recognition method based on geometric figure range discrimination according to claim 3, characterized in that: The step S23 comprises the following steps: According to the geometric data and material properties of the building structure plate, its surface normal vector is calculated, and a preliminary normal vector field is constructed based on the distribution of the normal vector; Performing weighted smoothing processing on the normal vector field by using a weighted average method to generate an optimized normal vector field; The optimized normal vector field is used to identify the key feature points of the building structure plate based on the feature extraction algorithm based on curvature change analysis and differential geometry.

5. The plate recognition method based on geometric figure range discrimination according to claim 1 is characterized in that: The step S3 comprises the following steps: S31: constructing a preliminary geometric outline of the building structure plate based on the key feature points; S32: Analyze the key feature points using a multi-scale shape analysis algorithm, and generate a three-dimensional geometric model based on the preliminary geometric outline of the building structure plate; S33: Performing surface refinement processing on the generated three-dimensional geometric model, eliminating irregularities using a smooth surface fitting algorithm, optimizing boundary accuracy, and obtaining an optimized three-dimensional geometric model.

6. A plate recognition method based on geometric figure range discrimination according to claim 1, characterized in that: The step S4 comprises the following steps: S41: Based on the three-dimensional geometric model, construct a preliminary geometric range of the building structure plate, and define geometric boundary conditions and restriction conditions of the building structure plate; S42: constructing a geometric figure range discrimination model based on the geometric boundary conditions of the building structure plate through a Markov random field algorithm, and modeling and optimizing the relationship between adjacent geometric units in the model; the nodes in the geometric figure range discrimination model represent geometric boundary points and feature points, and the edges represent the mutual geometric relationship between these points; S43: Optimizing the Markov random field model using an energy function, calculating the energy value of each geometric unit, and identifying abnormal areas or areas that do not meet the design specifications within the geometric range; S44: Generate geometric range discrimination results based on the optimized geometric figure range discrimination model, and output range discrimination information of the building structure plate.

7. A plate recognition method based on geometric figure range discrimination according to claim 6, characterized in that: The formula of the geometric figure range discrimination model is as follows: Where E(X) represents the total energy of the geometric range discrimination model, which is used to judge whether the geometric range of the building structure plate meets the specifications; V represents the node set of the geometric figure, that is, the key feature points on the building structure plate; E represents the edge set of the geometric figure, that is, the relationship between the key feature points; N i Represents the normal direction of the building structure plate surface at the feature point i; represents the normal direction of the building structure plate at the characteristic point i under ideal conditions; x i and x j represents the position vector of nodes i and j, that is, the three-dimensional coordinates of two adjacent key feature points on the building structure plate; α represents the weight parameter of the normal vector difference; β represents the weight parameter of the geometric relationship between adjacent feature points.

8. A plate recognition method based on geometric figure range discrimination according to claim 1, characterized in that: The step S5 comprises the following steps: S51: Based on the geometric range discrimination result, the boundary feature information of the building structure plate is extracted, and the identified geometric feature points and boundary lines are preliminarily aligned with the design data in the BIM model; S52: The iterative closest point algorithm is used to accurately align the recognition results, and the registration accuracy between the geometric model of the building structure plate and the BIM model design data is optimized through multiple iterations; S53: Based on the alignment results, identify the deviation area between the geometric model and the design data, analyze the nature and source of the deviation, and determine whether the building structure plate has a geometric deviation that exceeds the allowable tolerance; S54: Generate an identification report of the building structure plate, the report content includes geometric deviation information, spatial position of the deviation area, deviation range and degree of conformity with BIM model design data.

9. A plate recognition method based on geometric figure range discrimination according to claim 8, characterized in that: The formula of the iterative closest point algorithm is as follows: Among them, T k+1 represents the transformation matrix after the k+1th iteration; T represents the rigid body transformation matrix, that is, the rotation and translation operations are performed on the geometric data of the building structure plate collected by the laser scanner; p i represents the three-dimensional coordinates of the i-th feature point on the building structure plate collected by the laser scanner; q i Indicates the BIM model with p i The three-dimensional coordinates of the nearest design point corresponding to the feature point; N represents the total number of feature point pairs; ||T(p i )-q i || 2 represents the squared Euclidean distance.

Images