A prefabricated wallboard mold size measurement method and system based on point cloud data

Abstract

The application provides a prefabricated wallboard mold size measurement method and system based on point cloud data, and relates to the technical field of precise three-dimensional measurement of prefabricated building components. By collecting original three-dimensional point clouds, pure mold point clouds are obtained after preprocessing and cutting; three-dimensional reference point clouds are generated based on the vertices of the mold bottom surface, and complete mold boundaries are extracted in combination with three-dimensional nearest neighbor search; no two-dimensional projection transformation is performed throughout, and RANSAC algorithm is directly used to fit spatial straight lines to the original three-dimensional boundary point clouds, and three-dimensional space intersection points are solved simultaneously and mapped to obtain mold corner points carrying complete XYZ coordinates; based on the three-dimensional corner points, the lengths of the four sides of the mold, the perpendicularity of the side walls, the flatness of the forming surface and the overall warping degree are calculated synchronously; finally, the measured point clouds are compared with the standard design point clouds of the mold, a deviation heat map is generated, and a traceable quality inspection report is output, realizing integrated mold size and deformation detection, improving measurement accuracy, and adapting to prefabricated component factory batch online quality inspection scenes.

Description

Technical Field

[0001] This invention relates to the field of precision three-dimensional measurement technology for prefabricated components of prefabricated buildings, and in particular to a method and system for measuring the dimensions of prefabricated wall panel molds based on point cloud data. Background Technology

[0002] In building construction, the dimensional inspection of precast wall panel molds is a crucial step in quality control. Traditional manual steel ruler measurement methods are highly random, require at least two people to complete, and are inefficient.

[0003] Existing technologies suffer from the following drawbacks: 1. Generating point clouds through photography and registering them with the design model to output a dimension report relies on 2D image reconstruction. This results in insufficient Z-axis accuracy of the point cloud and a lack of independent 3D straight-line fitting logic. Instead, it forcibly projects 3D boundary points onto a 2D plane for dimension estimation, directly leading to the permanent loss of height (Z-axis) information at the mold corners and failing to accurately reflect the verticality and spatial warping of the mold sidewalls. 2. Deep learning-based wall point cloud detection uses 3D laser acquisition combined with PointNet boundary extraction, but its core processing still reduces all point cloud data to a 2D plane before performing straight-line fitting. This only outputs the plane's length and width dimensions, without designing specific 3D spatial analysis logic for the cavity, corner wear, and sidewall deformation of precast wall panel molds. 3. Existing methods largely rely on a single AABB bounding box and cannot adapt to tilted molds. They lack a global scale calibration step based on coded targets, resulting in large point cloud scale deviations and failing to meet the industrial requirements for high-precision closed-loop quality inspection of prefabricated molds. Summary of the Invention

[0004] The main objective of this invention is to provide a method and system for measuring the dimensions of precast wall panel molds based on point cloud data, which solves the problems of lost Z coordinates in two-dimensional projection, distorted boundary extraction, large point cloud scale deviation, low measurement accuracy, and lack of automated out-of-tolerance marking and quality inspection report output.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for measuring the dimensions of prefabricated wall panel molds based on point cloud data, comprising the following steps: S1: Encoded targets are deployed throughout the precast wall panel mold, and a structured light camera is used to scan the mold area to collect the original 3D point cloud and perform preprocessing. S2: Segment and trim the mold point cloud containing only the precast wall panel mold from the preprocessed 3D point cloud; S3: Obtain the AABB bounding box of the mold point cloud, select the four vertices of the bottom surface of the bounding box to construct four sets of starting vectors, and generate a 3D reference point cloud with a fixed step size; use 3D nearest neighbor search to match the mold point cloud and extract the complete 3D boundary point cloud of the bottom surface of the mold. S4: Based directly on the original 3D boundary point cloud that is not projected onto the 2D plane, the RANSAC random sampling consensus algorithm is used to fit four spatial lines and obtain the 3D line equations; the 3D center point of the mold point cloud is calculated, and the four spatial lines are sorted based on the angle between the vector from the sampling point to the center point and the X-axis; the 3D spatial intersection points of adjacent spatial lines after sorting are solved in sequence, and the nearest neighbor 3D point of each intersection point is retrieved in the original mold point cloud as the 3D corner points of the mold carrying complete X / Y / Z coordinates, and the four corner points are sorted according to the clockwise rule of rectangles; S5: Based on the complete three-dimensional coordinates of the sorted three-dimensional corner points, calculate the three-dimensional Euclidean distance between adjacent corner points to obtain the actual length of the four sides of the mold; simultaneously construct a spatial reference plane based on the four corner points, solve all spatial geometric parameters of the mold sidewall perpendicularity, forming surface flatness, and overall diagonal warping, and output the complete size measurement and deformation detection results of the mold.

[0006] In the preferred embodiment, the preprocessing includes: performing downsampling and outlier denoising preprocessing on the original point cloud, and performing point cloud scale calibration based on target coordinates.

[0007] In the preferred scheme, the four sets of starting point pairs in step S3 are combinations of AABB bottom face vertices; each set of starting point pairs generates a continuous three-dimensional reference point cloud by iteratively generating the two-point three-dimensional vectors as directions and setting a fixed millimeter-level step size.

[0008] In the preferred embodiment, the Z-axis height information of the point cloud is preserved throughout the three-dimensional RANSAC straight-line fitting process in step S4, and the XY plane projection transformation is not performed.

[0009] In the preferred scheme, in step S4, the solution order for the intersection points of adjacent spatial lines is as follows: from the intersection point of the first and second lines to the intersection point of the second and third lines, to the intersection point of the third and fourth lines, and then to the intersection point of the fourth and first lines, thus forming four three-dimensional intersection points on the bottom surface of the mold rectangle.

[0010] In the preferred embodiment, the side lengths of the four sides in step S5 are calculated using the three-dimensional Euclidean distance formula between adjacent three-dimensional corner points; the perpendicularity is solved by the angle between the three-dimensional vectors of adjacent sides; the flatness is calculated based on the average point cloud deviation of the least squares plane of the four corner points; and the warping is determined based on the difference in the three-dimensional height difference of the mold diagonal.

[0011] The preferred solution includes the S6 deviation quality inspection step: the three-dimensional corner points and geometric parameters obtained from this measurement are registered and compared with the standard design point cloud of the precast wall panel mold to generate a full-domain dimensional deviation heat map, automatically mark the out-of-tolerance areas and output a traceable quality inspection report.

[0012] Secondly, the present invention provides a precast wall panel mold size measurement system based on point cloud data, used to implement the aforementioned precast wall panel mold size measurement method based on point cloud data, including: The point cloud acquisition and preprocessing module is used to deploy coded targets across the entire area of ​​the precast wall panel mold, use a structured light camera to scan the mold area, acquire the original 3D point cloud, and perform preprocessing. The point cloud trimming module is used to segment and trim the mold point cloud, which contains only the prefabricated wall panel mold, from the pre-processed 3D point cloud. The 3D boundary extraction module is used to obtain the AABB bounding box of the mold point cloud, select the four vertices of the bottom surface of the bounding box to construct four sets of starting vectors, and generate a 3D reference point cloud with a fixed step size; the 3D nearest neighbor search is used to match the mold point cloud and extract the complete 3D boundary point cloud of the bottom surface of the mold. The 3D line fitting and corner point solving module is used to directly fit four spatial lines and obtain the 3D line equations based on the original 3D boundary point cloud that is not projected onto the 2D plane using the RANSAC random sampling consensus algorithm. It calculates the 3D center point of the mold point cloud and sorts the four spatial lines based on the angle between the vector from the sampling point to the center point and the X-axis. After sequentially solving the system of adjacent spatial lines, it solves the 3D spatial intersection point. It retrieves the nearest neighbor 3D point of each intersection point in the original mold point cloud as the 3D corner point of the mold carrying complete X / Y / Z coordinates, and completes the sorting of the four corner points according to the clockwise rectangular rule. The multi-parameter dimensional deformation calculation module is used to calculate the three-dimensional Euclidean distance between adjacent corner points based on the complete three-dimensional coordinates of the sorted three-dimensional corner points to obtain the actual length of the four sides of the mold; simultaneously, a spatial reference plane is constructed based on the four corner points to solve all spatial geometric parameters such as the perpendicularity of the mold sidewall, the flatness of the forming surface, and the overall diagonal warp, and output the complete size measurement and deformation detection results of the mold.

[0013] The preferred solution also includes a quality inspection comparison output module, which is used to register and compare the three-dimensional corner points and geometric parameters obtained from this measurement with the standard design point cloud of the precast wall panel mold, generate a full-domain dimensional deviation heat map, automatically mark the out-of-tolerance areas, and output a traceable quality inspection report.

[0014] This invention provides a method for measuring the dimensions of precast wall panel molds based on point cloud data, comprising: setting up coded targets, acquiring original point clouds with a structured light camera, and preprocessing them; segmenting and retaining only the 3D point cloud of the mold area; switching between AABB / OBB dual-mode bounding boxes, generating a 3D reference point cloud from the bottom vertices, and extracting the complete 3D boundary through nearest neighbor retrieval; directly fitting the boundary lines in the original 3D space, solving for the 3D intersection points, and mapping to obtain 3D corner points with complete XYZ coordinates; based on the complete 3D coordinates of the sorted 3D corner points, calculating the 3D Euclidean distance between adjacent corner points to obtain the actual length of the four sides of the mold, solving for all spatial geometric parameters of the mold sidewall verticality, forming surface flatness, and overall diagonal warpage, and outputting the complete size measurement and deformation detection results of the mold; retaining the XYZ 3D coordinates of the point cloud throughout the process, directly completing line fitting and intersection point solving in 3D space, fundamentally solving the information distortion problem caused by the above-mentioned 2D projection, reducing size measurement errors, improving scene adaptability and automation, and improving measurement accuracy. Attached Figure Description

[0015] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the measurement method of the present invention; Figure 2 This is a diagram of the mold and the AABB enclosure of the present invention; Figure 3 This is a boundary diagram of the AABB enclosing frame of the mold of the present invention; Figure 4 This is a diagram of the boundary lines and intersection points of the AABB bounding box of this invention; Figure 5 This is a comparison diagram of the AABB / OBB dual bounding box switching of the present invention; Figure 6 This is the three-dimensional reference point cloud and boundary point cloud extraction image of the present invention; Figure 7 This invention relates to three-dimensional spatial line fitting and three-dimensional corner point positioning. Detailed Implementation

[0016] Example 1 like Figure 1-7 As shown, a method for measuring the dimensions of precast wall panel molds based on point cloud data includes the following steps: S1 Target Calibration and Point Cloud Acquisition: Encoded targets are deployed throughout the precast wall panel mold. A structured light camera is used to scan the mold area to acquire the original 3D point cloud. The original point cloud is then preprocessed by downsampling and outlier denoising. The point cloud scale is calibrated based on the target coordinates. S2 point cloud trimming: Segment and trim the mold point cloud containing only the prefabricated wall panel mold from the preprocessed 3D point cloud; S3 3D Boundary Point Cloud Extraction: Obtain the AABB bounding box of the mold point cloud, select the four vertices of the bottom surface of the bounding box to construct four sets of starting vectors, and generate a 3D reference point cloud with a fixed step size; use 3D nearest neighbor search to match the mold point cloud and extract the complete 3D boundary point cloud of the bottom surface of the mold. S4 Unprojected 3D Spatial Line Fitting and 3D Corner Point Solving: Based directly on the original 3D boundary point cloud that is not projected onto the 2D plane, the RANSAC random sampling consensus algorithm is used to fit four spatial lines and obtain the 3D line equations; the 3D center point of the mold point cloud is calculated, and the four spatial lines are sorted based on the angle between the vector from the sampling point to the center point and the X-axis; the 3D spatial intersection points of adjacent spatial lines after sorting are solved in sequence, and the nearest neighbor 3D point of each intersection point is retrieved in the original mold point cloud as the 3D corner points of the mold carrying complete X / Y / Z coordinates, and the four corner points are sorted according to the rectangular clockwise rule; S5 Multidimensional Dimension and Spatial Form and Position Parameter Calculation: Based on the complete three-dimensional coordinates of the sorted three-dimensional corner points, the three-dimensional Euclidean distance between adjacent corner points is calculated to obtain the actual length of the four sides of the mold; Simultaneously, a spatial reference plane is constructed based on the four corner points to solve all spatial geometric parameters such as the perpendicularity of the mold sidewall, the flatness of the forming surface, and the overall diagonal warp, and output the complete size measurement and deformation detection results of the mold.

[0017] S6 Deviation Inspection Steps: Register the three-dimensional corner points and geometric parameters obtained from this measurement with the standard design point cloud of the precast wall panel mold (using ICP or rigid registration based on a target), generate a global dimensional deviation heat map, automatically mark out-of-tolerance areas, and output a traceable quality inspection report.

[0018] In this embodiment, target calibration preprocessing, point cloud clipping, AABB 3D boundary extraction, projection-free 3D RANSAC spatial line fitting, 3D corner point solving, simultaneous calculation of dimensions, perpendicularity, flatness, and warpage, and deviation quality inspection are performed. This preserves the integrity of the mold's 3D spatial information, reduces structured light camera imaging distortion and point cloud scale errors, improves measurement accuracy, reduces manual intervention, improves computational efficiency, and enhances scene adaptability.

[0019] The following steps will be explained in detail.

[0020] Step S1: Target calibration and point cloud acquisition Encoded targets were deployed across the entire precast wall panel mold: checkerboard targets were placed at the four corners and the center of the mold. A structured light camera was used to scan the mold area and acquire the original 3D point cloud. The original point cloud was downsampled with a voxel size of 5mm, and outlier denoising was performed, such as statistical filtering, with the number of neighboring points k=15 and the standard deviation factor set to 2.0. Based on the known world coordinates of the coded targets, the point cloud was scaled and aligned to eliminate camera distortion and installation errors.

[0021] Step S2: Point Cloud Cropping Using the target coordinates or prior position information of the mold area, the mold point cloud containing only the precast wall panel mold is segmented and cropped from the preprocessed 3D point cloud, and background noise and interference points on the support platform are removed.

[0022] Step S3: Extraction of 3D Boundary Point Cloud like Figure 2-3 As shown, the four pairs of starting points are combinations of the bottom vertices of an AABB matrix. , , , Each starting point pair uses a two-point 3D vector as its direction, and iteratively generates a continuous 3D reference point cloud with a fixed millimeter-level step size, specifically: Step S31: Extracting bounding box vertices Obtain the AABB (Axial Bounding Box) bounding box of the mold point cloud, and take 4 out of the 8 vertices of the bounding box:

[0023] Step S32: Reference Point Cloud Generation Set starting point pair , , , Generate a reference point cloud based on the starting point pair, using point pairs For example, the reference point for cloud computing is as follows: Each group starts at the first vertex and ends at the second vertex. For example: calculate , The vector, denoted as . .by Starting point As the endpoint, according to Generate a reference point cloud for the step size, i.e. until the finish line is reached. The reference point sets generated from the four sets of point pairs are merged into a three-dimensional reference point cloud. P .

[0024] Step S33: Nearest Neighbor Search for Boundary Points Using the nearest neighbor search algorithm, the nearest neighbor of each point in P in the mold point cloud is retrieved, and this set is denoted as Q, which is the boundary point set. This set is the complete 3D boundary point cloud of the bottom surface of the mold, containing Z-axis height information.

[0025] like Figure 6 As shown, in this embodiment, the boundary extraction accuracy is improved based on the boundary extraction range of the mold bottom surface contour constraint, and it is adapted to the standard bottom surface structure of rectangular steel molds; no manual annotation of mold edge lines is required, which reduces the complexity of the algorithm.

[0026] Step S4: Line Fitting and Corner Point Solving in Unprojected 3D Space This step preserves the Z-coordinates of the point cloud throughout, directly performing line fitting and intersection point calculation in 3D space to avoid the loss of height information caused by 2D projection. The specific sub-steps are as follows: S41: Three-dimensional spatial linear fitting Based on the boundary point set Q, the RANSAC (Random Sample Consensus) algorithm is used to directly fit four spatial lines. The RANSAC parameters are set as follows: 1000 iterations and an interior point distance threshold of 2 mm. The fitting process employs a sequential strategy: after the first spatial line is fitted using RANSAC, its interior points are recorded and removed from the boundary point set Q; the RANSAC fitting is repeated for the remaining boundary points, sequentially extracting the second, third, and fourth spatial lines, until all four lines are fitted. This strategy effectively avoids the RANSAC algorithm from repeatedly fitting the same side, ensuring that all four edges are uniquely extracted.

[0027] Each fitting yields the direction vector of a spatial straight line. and over the point The equation of the straight line is expressed as (where t is a real number), resulting in four spatial lines. The fitting process does not perform any two-dimensional projection, thus fully preserving the XYZ three-dimensional information of the point cloud.

[0028] S42: Spatial Linear Sort Calculate the 3D center point of the mold point cloud N is the number of points in the point cloud of the mold; arbitrarily select a point on each spatial straight line. Construct vector Calculate the vector relative to the positive X-axis of the world coordinate system (unit vector). The included angle .according to Sort the lines from smallest to largest (0°~360°) to obtain the sorted linear sequence. These correspond to the front, right, back, and left sides of the bottom surface of the mold, respectively.

[0029] S43: Solving for intersection points in three-dimensional space Following the ordered sequence of lines, solve for the spatial intersection points by solving the equations of adjacent lines one by one. The order of solving for the spatial intersection points of adjacent lines is: intersection of the first and second lines → intersection of the second and third lines → intersection of the third and fourth lines → intersection of the fourth and first lines, thus forming the four three-dimensional intersection points of the base of the mold rectangle, as follows. Figure 4 As shown, specifically: intersection United and ; intersection United and ; intersection United and ; intersection United and .

[0030] The method for finding the intersection point of two lines is as follows: Let the two lines be ( ) and ( ). , ,make Construct a system of linear equations and solve it. The coordinates of the three-dimensional intersection point are obtained. Due to point cloud noise and fitting error, the four lines may not be completely coplanar. The nearest point in the least squares sense is used as the approximate intersection point.

[0031] The method for finding the intersection point of two lines is as follows: Let the two lines be: , ,make Construct a system of linear equations to solve for the parameters. , The coordinates of the three-dimensional intersection point are obtained. Due to point cloud noise and fitting error, the four lines may not be completely coplanar (i.e., they are skew lines). In this case, the closest pair of points on each of the two lines is calculated, and the midpoint of the line connecting the two points is taken as the approximate three-dimensional intersection point to minimize the impact of point cloud noise and fitting error on the localization of the intersection point.

[0032] S44: 3D Corner Mapping and Sorting like Figure 7 As shown, in the original mold point cloud, for each intersection point Perform a nearest neighbor search to find the nearest 3D point as the corner point of the mold. This ensures that the corner points actually fall on the mold's point cloud, preserving their complete XYZ coordinates. The four corner points are then... The order is that the rectangles are arranged clockwise, completing the corner point sorting. The four spatial lines after sorting are denoted as follows: (Front side) (Right side) (Rear side) (Left side), its adjacent intersection points ( , , , ) are respectively mapped to mold corner points Click the four corners The order is that the rectangles are arranged clockwise, thus completing the corner point sorting.

[0033] This embodiment eliminates the calculation errors caused by projection distortion and coordinate compression, and reduces the three-dimensional coordinate errors of corner points.

[0034] In this embodiment, the three-dimensional intersection points have an inherent height dimension, which improves the accuracy of corner positioning in the real space of the mold, increases the degree of algorithm automation, reduces the corner sorting error rate, and improves measurement accuracy.

[0035] Step S5: Calculate multi-dimensional dimensions and spatial form and position parameters. Based on the complete coordinates of the four three-dimensional corner points, calculate the mold's dimensional parameters and deformation parameters simultaneously.

[0036] In the preferred scheme, the side lengths in step S5 are... , The perpendicularity is calculated by the three-dimensional Euclidean distance formula between adjacent three-dimensional corner points; the flatness is calculated by the angle between the three-dimensional vectors of adjacent sides; the flatness is calculated based on the average point cloud deviation of the least square plane of the four corner points; and the warpage is determined based on the difference in the three-dimensional height difference of the mold diagonal.

[0037] Specifically: S51: Calculation of the side lengths of four sides Sort by corner points, and calculate the lengths of the four sides in sequence. The side length is arrive distance, The side length is arrive distance, The side length is arrive distance, The side length is arrive The distance, i.e., the three-dimensional Euclidean distance between adjacent corner points calculated according to the corner point sorting, is expressed as: , , , .in, Output the actual lengths of the four sides of the mold.

[0038] S52: Calculation of sidewall verticality Perpendicularity is defined as the angular deviation between adjacent sides and the direction of the normal to the bottom surface. (The last part, "using the side...", appears to be an unrelated fragment and is left untranslated.) For example, its direction vector is The normal vector of the bottom surface is (Normalization). Perpendicularity Angle The ideal value is 90°, and the actual deviation is... Calculate the perpendicularity deviation of each of the four sides in sequence.

[0039] S53: Calculation of Flatness of Molded Surface Based on the three-dimensional coordinates of the four corner points, a spatial reference plane is fitted using the least squares method. Calculate the distance from each corner point to the plane. The flatness is calculated by taking the difference between the maximum and minimum values ​​of the four distances. The smaller the value, the smoother the molded surface.

[0040] S54: Overall Warpage Calculation Warpage is used to measure whether the mold twists diagonally. Calculate the three-dimensional height difference between the two diagonals: Diagonal 1 is... Its Z-axis elevation difference Diagonal 2 is Its Z-axis elevation difference Warpage This value reflects the degree of overall torsional deformation of the mold.

[0041] This embodiment improves the accuracy of tilting mold dimension measurement, avoids manual secondary conversion, and improves calculation efficiency.

[0042] In the preferred solution, such as Figure 5 As shown, when the mold scanning posture is tilted and not parallel to the world coordinate axis, an OBB directed bounding box is used to replace the AABB bounding box to regenerate the reference point cloud, in order to eliminate the boundary extraction error caused by the tilt. The specific judgment logic is as follows: calculate the principal axis direction of the mold point cloud (obtain the first principal component direction through principal component analysis PCA). If the absolute value of the angle between this direction and the X-axis or Y-axis of the world coordinate system is greater than a preset angle threshold (preferably 15°), it is determined to be a tilted posture, and the OBB bounding box is automatically switched to generate the subsequent reference point cloud.

[0043] In this embodiment, when the mold tilt exceeds the threshold, the OBB directed bounding box is switched to replace the AABB, which reduces the boundary extraction error under tilting conditions and improves the adaptability of unmanned automated detection scenarios.

[0044] This embodiment illustrates a specific scenario: Acquisition equipment: Industrial structured light 3D camera; Calibration targets: Checkerboard coded targets, placed at the four corners and center of the mold; Measured object: Standard PC precast wall panel steel mold, size 3000mm×2800mm.

[0045] S1: Downsampling sphere radius r=5mm; Outlier point denoising minimum neighborhood threshold k=15; Target global calibration corrects point cloud scale error; S2: Perform point cloud cropping: S3: Reference point cloud step size ε=10mm; Mold tilt judgment threshold: Automatically switch OBB bounding box when the angle between the mold spindle and the world X-axis is >15°; S4 3D RANSAC parameters: 1000 iterations, interior point distance threshold of 2mm; spatial line sorting uses the angle between the vector and the X-axis from 0 to 360°. S5 calculation example: Side length: Three-dimensional Euclidean distance between two points; Verticality: The angle is calculated by the dot product of the side vectors. If the angle is greater than 92°, the side wall tilt is considered to be out of tolerance. Flatness: The least squares plane at the four corners, a single point deviation > 0.3mm is considered an unqualified forming surface; Warpage: The mold is judged to be warped if the height difference in the Z direction of the diagonal is greater than 0.5mm.

[0046] S6 Deviation Inspection and Report Output: Standard mold design point cloud built-in database, deviations > standard thresholds are automatically marked in red on the heat map icon.

[0047] The three-dimensional corner points, edges, and geometric parameters obtained from this measurement are registered with the built-in standard mold design point cloud (using ICP or target-based rigid registration). The standard mold design point cloud is pre-built in the following way: the CAD three-dimensional model of the prefabricated wall panel mold (step or igs format) is imported, and the theoretical point cloud is generated by voxel downsampling (sampling interval preferably 2mm) and stored in the local database. At the same time, metadata such as mold model, design size, and tolerance grade are associated for automatic retrieval during subsequent comparison.

[0048] After registration, the point-by-point deviation between the mold point cloud and the design point cloud is calculated, generating a global dimensional deviation heatmap. Areas where deviations exceed preset thresholds (e.g., side length deviation > 2mm, perpendicularity > 2°, flatness > 0.3mm, warpage > 0.5mm) are automatically marked. These tolerance thresholds are comprehensively set based on the "Technical Specification for Production of Precast Concrete Components for Prefabricated Buildings" (JG / T 565-2018) and the factory's internal quality control specifications. In specific implementation, these thresholds can be customized according to the mold's precision level and customer requirements. A traceable quality inspection report (PDF or Excel format) containing all measured values, deviation values, and out-of-tolerance criteria is output.

[0049] In this embodiment, automatic switching of OBB bounding boxes is achieved, reducing boundary extraction errors.

[0050] Example 2 To further illustrate with reference to Embodiment 1, a prefabricated wall panel mold size measurement system based on point cloud data is provided for implementing the measurement method of Embodiment 1, comprising: The point cloud acquisition and preprocessing module is equipped with a structured light camera and a coded target recognition unit. It is used to deploy coded targets throughout the precast wall panel mold, scan the mold area with the structured light camera to acquire the original 3D point cloud, and perform preprocessing.

[0051] The point cloud trimming module is used to segment pure mold point cloud data from the preprocessed point cloud.

[0052] The 3D boundary extraction module includes a bounding box calculation unit, a reference point cloud generation unit, and a 3D nearest neighbor retrieval unit. It is used to obtain the AABB bounding box of the mold point cloud, select four vertices of the bottom surface of the bounding box to construct four sets of starting vectors, and generate a 3D reference point cloud with a fixed step size. The 3D nearest neighbor search is used to match the mold point cloud and extract the complete 3D boundary point cloud of the bottom surface of the mold.

[0053] The 3D line fitting and corner point solving module incorporates a 3D RANSAC fitting unit, a spatial line sorting unit, a 3D intersection point solving unit, and a corner point mapping unit. It directly fits four spatial lines based on the original 3D boundary point cloud not projected onto a 2D plane using the RANSAC random sampling consensus algorithm to obtain the 3D line equations. It calculates the 3D center point of the mold point cloud and sorts the four spatial lines based on the angle between the vector from the line sampling point to the center point and the X-axis. After sequentially solving the sorted system of adjacent spatial lines, it solves the 3D spatial intersection points. It then retrieves the nearest neighbor 3D point of each intersection point in the original mold point cloud as the 3D corner points of the mold carrying complete X / Y / Z coordinates, and completes the sorting of the four corner points according to a rectangular clockwise rule.

[0054] The multi-parameter dimensional deformation calculation module is used to simultaneously calculate the side length, perpendicularity, flatness, and warpage of the four sides based on the three-dimensional corner coordinates.

[0055] The quality inspection comparison output module is used to match the point cloud of the standard mold design and generate a deviation heat map and quality inspection report.

[0056] In the preferred embodiment, the bounding box calculation unit has a built-in AABB / OBB dual-mode switching subunit, which automatically switches the bounding box type according to the mold placement posture.

[0057] In the preferred embodiment, this embodiment provides a method for measuring the dimensions of prefabricated wall panel molds based on point cloud data. The working process, working details and technical effects can be found in Embodiment 1, and will not be repeated here.

[0058] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for measuring the dimensions of precast wall panel molds based on point cloud data, characterized in that, Includes the following steps: S1: Encoded targets are deployed throughout the precast wall panel mold, and a structured light camera is used to scan the mold area to collect the original 3D point cloud and perform preprocessing. S2: Segment and trim the mold point cloud containing only the precast wall panel mold from the preprocessed 3D point cloud; S3: Obtain the AABB bounding box of the mold point cloud, select the four vertices of the bottom surface of the bounding box to construct four sets of starting vectors, and generate a 3D reference point cloud with a fixed step size; use 3D nearest neighbor search to match the mold point cloud and extract the complete 3D boundary point cloud of the bottom surface of the mold. S4: Based directly on the original 3D boundary point cloud that is not projected onto the 2D plane, the RANSAC random sampling consensus algorithm is used to fit four spatial lines and obtain the 3D line equations; Calculate the 3D center point of the mold point cloud. Sort four spatial lines based on the angle between the vector from the sampling point to the center point and the X-axis. Solve the 3D spatial intersection point of adjacent spatial lines after sorting. In the original mold point cloud, retrieve the nearest neighbor 3D point of each intersection point as the 3D corner point of the mold carrying complete X / Y / Z coordinates. Sort the four corner points according to the clockwise rule of rectangle. S5: Based on the complete three-dimensional coordinates of the sorted three-dimensional corner points, calculate the three-dimensional Euclidean distance between adjacent corner points to obtain the actual length of the four sides of the mold; simultaneously construct a spatial reference plane based on the four corner points, solve all spatial geometric parameters of the mold sidewall perpendicularity, forming surface flatness, and overall diagonal warping, and output the complete size measurement and deformation detection results of the mold.

2. The method for measuring the dimensions of precast wall panel molds based on point cloud data according to claim 1, characterized in that, The preprocessing includes: performing downsampling and outlier denoising preprocessing on the original point cloud, and performing point cloud scale calibration based on target coordinates.

3. The method for measuring the dimensions of precast wall panel molds based on point cloud data according to claim 1, characterized in that, In step S3, the four sets of starting point pairs are combinations of AABB bottom face vertices; each set of starting point pairs uses the two-point three-dimensional vectors as directions and sets a fixed millimeter-level step size to iteratively generate a continuous three-dimensional reference point cloud.

4. The method for measuring the dimensions of precast wall panel molds based on point cloud data according to claim 1, characterized in that, In step S4, the Z-axis height information of the point cloud is preserved throughout the 3D RANSAC straight line fitting process, and the XY plane projection transformation is not performed.

5. The method for measuring the dimensions of precast wall panel molds based on point cloud data according to claim 1, characterized in that, In step S4, the solution order for the intersection points of adjacent spatial lines is as follows: from the intersection point of the first and second lines to the intersection point of the second and third lines, to the intersection point of the third and fourth lines, and then to the intersection point of the fourth and first lines, thus forming four three-dimensional intersection points on the bottom surface of the mold rectangle.

6. The method for measuring the dimensions of precast wall panel molds based on point cloud data according to claim 1, characterized in that, In step S5, the side lengths of the four sides are calculated using the three-dimensional Euclidean distance formula between adjacent three-dimensional corner points; the perpendicularity is solved by the angle between the three-dimensional vectors of adjacent sides; the flatness is calculated based on the average point cloud deviation of the least squares plane of the four corner points; and the warping is determined based on the difference in the three-dimensional height difference of the mold diagonal.

7. The method for measuring the dimensions of precast wall panel molds based on point cloud data according to claim 1, characterized in that, The S6 deviation quality inspection step includes: registering and comparing the three-dimensional corner points and geometric parameters obtained from this measurement with the standard design point cloud of the precast wall panel mold, generating a full-domain dimensional deviation heat map, automatically marking out-of-tolerance areas and outputting a traceable quality inspection report.

8. The method for measuring the dimensions of precast wall panel molds based on point cloud data according to claim 1, characterized in that, When the mold scanning posture is tilted and not parallel to the world coordinate axis, an OBB oriented bounding box is used to replace the AABB bounding box to regenerate the reference point cloud, eliminating the boundary extraction error caused by the tilt.

9. A precast wall panel mold size measurement system based on point cloud data, used to implement the measurement method according to any one of claims 1-8, characterized in that, include: The point cloud acquisition and preprocessing module is used to deploy coded targets across the entire area of ​​the precast wall panel mold, use a structured light camera to scan the mold area, acquire the original 3D point cloud, and perform preprocessing. The point cloud trimming module is used to segment and trim the mold point cloud, which contains only the prefabricated wall panel mold, from the pre-processed 3D point cloud. The 3D boundary extraction module is used to obtain the AABB bounding box of the mold point cloud, select the four vertices of the bottom surface of the bounding box to construct four sets of starting vectors, and generate a 3D reference point cloud with a fixed step size; the 3D nearest neighbor search is used to match the mold point cloud and extract the complete 3D boundary point cloud of the bottom surface of the mold. The 3D line fitting and corner point solving module is used to directly fit four spatial lines and obtain the 3D line equations based on the original 3D boundary point cloud that is not projected onto the 2D plane. Calculate the 3D center point of the mold point cloud. Sort four spatial lines based on the angle between the vector from the sampling point to the center point and the X-axis. Solve the 3D spatial intersection point of adjacent spatial lines after sorting. In the original mold point cloud, retrieve the nearest neighbor 3D point of each intersection point as the 3D corner point of the mold carrying complete X / Y / Z coordinates. Sort the four corner points according to the clockwise rule of rectangle. The multi-parameter dimensional deformation calculation module is used to calculate the three-dimensional Euclidean distance between adjacent corner points based on the complete three-dimensional coordinates of the sorted three-dimensional corner points to obtain the actual length of the four sides of the mold; simultaneously, a spatial reference plane is constructed based on the four corner points to solve all spatial geometric parameters such as the perpendicularity of the mold sidewall, the flatness of the forming surface, and the overall diagonal warp, and output the complete size measurement and deformation detection results of the mold.

10. The precast wall panel mold size measurement system based on point cloud data according to claim 9, characterized in that, It also includes a quality inspection comparison output module, which is used to register and compare the three-dimensional corner points and geometric parameters obtained from this measurement with the standard design point cloud of the precast wall panel mold, generate a full-domain dimensional deviation heat map, automatically mark the out-of-tolerance areas, and output a traceable quality inspection report.

Images