A prefabricated building hole occupying tube position detection method based on three-dimensional point cloud

Abstract

This invention discloses a method for detecting the location of prefabricated building hole occupants based on 3D point clouds, belonging to the field of building engineering quality inspection. This invention acquires a full-domain point cloud of the concealed area through 3D laser scanning. Addressing the problem of traditional cylindrical fitting being susceptible to noise interference, the RANSAC cylindrical fitting model is optimized with elevation priors. Combined with the vertical protrusion geometric representation of the pre-reserved hole pipe, accurate target extraction is achieved. An improved boundary extraction algorithm is used to obtain the mold edge, and finally, the positioning deviation from the center of the hole pipe to the mold edge is automatically calculated. This invention achieves non-contact, automated, and high-precision detection, effectively overcoming the influence of steel reinforcement skeleton occlusion and point cloud noise. The detection process is digitized, and the results are traceable, significantly improving the efficiency and quality control level of concealed works acceptance. The detection accuracy can reach within ±0.5mm, and the detection time for a single area is ≤3 minutes. It is suitable for concealed acceptance scenarios in construction sites with dense steel reinforcement, low light, and clutter, meeting the digital, standardized, and intelligent requirements of concealed works acceptance in prefabricated buildings.

Description

Technical Field

[0001] This invention relates to the field of building engineering quality inspection technology, specifically to a method for detecting the location of occupant pipes in prefabricated buildings based on three-dimensional point clouds. Background Technology

[0002] In the construction of prefabricated concrete buildings and the production process of precast components, reserved perforated pipes are the core reserved structures for electromechanical installation and pipeline connection. Their positioning accuracy directly determines the subsequent installation accuracy of electromechanical equipment, the pipeline connection effect, and the overall forming quality of the building structure. The distance from the center of the perforated pipe to the edge of the mold and the positioning deviation are mandatory inspection items during the concealed acceptance stage, which directly affect the building's functionality and structural safety. Currently, the industry generally uses tools such as manual tape measures, calipers, and laser rangefinders for random inspection, which has many obvious defects. Not only is the measurement accuracy insufficient, with manual reading errors typically ranging from ±2mm to ±5mm, which cannot meet the ±1mm acceptance requirements of high-precision precast components, but there is also the problem of unmeasurable concealed parts. The perforated pipe is surrounded by reinforcing bars, stirrups, and embedded parts, making it difficult to accurately measure its center position and distance from the edge of the mold, thus creating a blind spot for inspection.

[0003] Meanwhile, manual inspection is inefficient. The inspection of a single reserved hole tube requires repeated adjustments to the measurement angle, recording of data, and verification and calibration. A single inspection area (containing 3-5 hole tubes) takes 15-30 minutes, which is difficult to adapt to batch rapid acceptance. In addition, the data relies on paper records, with no original data retained. The acceptance conclusions are highly subjective and prone to disputes. Furthermore, there is a lack of visualization methods, which cannot intuitively show the positioning deviation of the hole tubes, which is not conducive to problem location, rectification, and re-inspection.

[0004] Existing point cloud-based dimensional inspection technologies are mostly focused on the inspection of precast components, steel structures, and civil engineering structures. They lack dedicated algorithmic processes for concealed acceptance scenarios, cylindrical structures with reserved holes, and mold reference edge extraction. These technologies generally suffer from problems such as sensitivity to on-site noise, susceptibility to burrs / noise interference in cylindrical fitting, instability in mold boundary extraction, mismatch between dimensional calculation logic and acceptance specifications, and inability to automatically output standard results for positioning deviations. Specifically, traditional RANSAC cylindrical fitting does not incorporate the vertical protrusion features of reserved holes, resulting in strong sampling randomness, a high misfit rate, and a tendency to misidentify rebar and debris point clouds as holes. Mold boundary extraction relies solely on a single normal vector, which is prone to edge offset in weak boundary and low-contrast scenarios, leading to significant distance calculation errors. Therefore, developing a fully automated, high-precision, and visually traceable method for detecting the location of reserved holes based on 3D point clouds, oriented towards concealed acceptance, has become an urgent need for quality control in prefabricated buildings. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies, such as low accuracy of manual measurement, inaccessibility of concealed parts, low efficiency, and lack of data traceability, as well as the poor fitting stability, inaccurate boundary extraction, and weak anti-interference ability of conventional point cloud detection methods. This invention provides a method for detecting the location of reserved hole pipes based on optimized RANSAC cylinder fitting and improved boundary extraction algorithm, which realizes non-contact, automated, and high-precision positioning and detection of reserved hole pipes, and meets the digital, standardized, and intelligent requirements of concealed acceptance.

[0006] The technical solution of this invention is as follows:

[0007] A method for detecting the location of occupiers in prefabricated buildings based on 3D point clouds includes the following steps:

[0008] (1) Obtain three-dimensional point cloud data of the concealed acceptance area using a three-dimensional laser scanning device;

[0009] (2) Preprocess the original point cloud to achieve noise filtering, outlier removal and dataset optimization;

[0010] (3) Based on the vertical protrusion geometric representation of the pre-embedded sleeve, the target feature point cloud is extracted by elevation domain segmentation;

[0011] (4) The optimized RANSAC cylindrical fitting model is used to fit the candidate point cloud to the surface, and the three-dimensional coordinates of the center of the reserved hole tube are obtained by combining the circular contour analysis.

[0012] (5) The mold region is segmented from the point cloud by an improved boundary feature extraction algorithm, and the reference edge line of the mold is extracted and fitted;

[0013] (6) Calculate the shortest distance from the center of the hole tube to the edge of the mold through spatial geometric operations to obtain the measured positioning value;

[0014] (7) Compare the measured values ​​with the design values, and output the deviation results and acceptance judgment conclusions;

[0015] Furthermore, the preprocessing method in step (2) is as follows: statistical filtering is used to remove point cloud noise points, the k-neighborhood value is 32, data simplification is achieved by voxel downsampling, the voxel grid size is set to 0.8mm, and computational redundancy is reduced while preserving the geometric features of the target component, thereby improving the efficiency of subsequent processing.

[0016] Furthermore, the optimization of the RANSAC cylindrical fitting model in step (4) is as follows: On the traditional RANSAC random sampling fitting framework, a prior constraint mechanism for mold elevation is introduced. Based on the vertical protruding structural features of the pre-reserved hole pipe in prefabricated buildings, the vertically protruding area where the point cloud elevation is higher than the mold reference plane is defined as the core priority sampling interval. The elevation range of the sampling point cloud is strictly limited to the interval of 35mm protruding from the mold plane. Non-target point cloud interference such as the mold reference plane and exposed steel bars are directly eliminated through pre-filtering of elevation threshold, thus avoiding invalid iterations from the sampling source. At the same time, the number of model iterations is adaptively set to 1800. Secondly, balancing fitting efficiency and global optimal solution search capability, the interior point distance threshold is finely set to 0.2mm to meet the high-precision fitting requirements of mold hole tubes, significantly improving the stability of cylinder fitting and the accuracy of size recognition under complex point cloud noise. On this basis, a cylinder topology consistency verification rule is added. By verifying three types of topological features of the point cloud, namely axial continuity, radial cross-sectional roundness, and tubular radial wall thickness consistency, non-tubular stray interference points and local protrusion noise points are automatically eliminated. The fitting model is strictly constrained to only match the standard cylindrical topology structure of the reserved hole tube, eliminating misfitting of non-target structures.

[0017] Furthermore, the improved boundary feature extraction algorithm in step (5) is as follows: combining the difference of normal vectors and the curvature analysis of the neighborhood, first calculate the angle difference between the normal vector of the single point cloud of the mold plane and the normal vector of the overall reference plane of the mold, screen out the candidate boundary points of the normal vector mutation, and then perform a second verification by combining the curvature change rate in the k neighborhood of the point cloud. The boundary points are accurately locked by the dual threshold linkage discrimination. For the actual engineering scenario of weak reflection, low contrast and local sparseness of the edge point cloud of the mold, the neighborhood point cloud weighted smoothing processing and boundary continuity constraint are introduced to correct the boundary breakage, false detection and missed detection caused by single point noise. It breaks through the limitation of the traditional algorithm in the failure of boundary extraction under low quality point cloud, and greatly improves the robustness and completeness of the extraction of mold edge features under complex working conditions.

[0018] Furthermore, step (6) uses the Euclidean distance from a spatial point to a straight line for calculation with an accuracy of ≤ ±0.1 mm. The result is then projected onto the mold plane to form an in-plane positioning deviation, eliminating the influence of spatial elevation differences on positioning accuracy and ensuring that the deviation calculation is consistent with the engineering acceptance specifications.

[0019] Furthermore, in step (4), the least squares method is used to optimize the fitting of the circular contour analysis, and the point cloud of the top contour of the cylinder is fitted and optimized. The fitting error is ≤ ±0.1mm, which improves the accuracy of the center calculation and ensures the accuracy of the hole tube center positioning.

[0020] Using the above method, full-area point cloud acquisition is performed: an industrial-grade 3D laser scanner is used to conduct a full-area non-contact multi-station scan of the area containing reserved holes, pipes, steel reinforcement frames, and molds under concealed acceptance conditions to obtain complete original 3D point cloud data, which is then transmitted to the processing terminal; the scanning distance is set to 1500mm, and the point cloud density is not less than 80 points / cm². 2 This ensures that the detailed features of the perforated tube are clearly identifiable.

[0021] Point cloud preprocessing: Statistical filtering is used to remove noise and outliers from the original point cloud. Each point and its k-neighborhood are traversed (k can be between 16 and 32, with 32 being the best). The average distance is calculated, and discrete noise points with a distance exceeding the threshold are removed. Data redundancy is reduced by voxel downsampling. The voxel grid size is set to 0.8 mm to improve computational efficiency while preserving the geometric features of the components, and to ensure the stability of subsequent fitting and extraction.

[0022] Elevation domain feature segmentation: Based on the geometric representation of the pre-embedded sleeve protruding vertically from the mold plane, the point cloud is statistically analyzed for elevation distribution and segmented by threshold. The elevation segmentation threshold is set to 35mm. Candidate point clouds of the vertically protruding area are selected, and interference from the lower steel skeleton, ground, debris and other planar areas is eliminated to obtain a set of candidate point clouds containing only the pre-reserved hole pipe.

[0023] Optimize the RANSAC cylindrical fitting model to solve the center point: In view of the technical pain points of the traditional RANSAC algorithm in complex point cloud environment, which is easily affected by non-target points, has low fitting accuracy and is prone to misfitting, two core innovations are optimized to build a high-precision cylindrical fitting model: First, a prior constraint mechanism for mold elevation is introduced. Based on the candidate point cloud of the vertical protruding area obtained by segmentation in step (3), the sampling range is strictly limited to the core area of ​​35mm protruding from the mold plane. Non-target point clouds such as the mold reference plane and exposed steel bars are filtered from the sampling source to avoid invalid iteration and greatly improve the fitting efficiency. Second, a new cylindrical topology consistency verification rule is added. By verifying the three core topological features of the point cloud, namely the axial continuity, radial cross-section roundness and tubular radial wall thickness consistency, non-tubular stray interference points and local protrusion noise points are automatically eliminated. The fitting model is strictly constrained to only match the standard cylindrical topology structure of the reserved hole pipe, and misfitting of non-target structures is eliminated. Meanwhile, the adaptive model iteration count is set to 1800 times, balancing fitting efficiency and global optimal solution search capability. The interior point distance threshold is finely set to 0.2mm to meet the high-precision fitting requirements of the mold hole tube. After fitting, the top contour circle of the hole tube is optimized and fitted again using the least squares method, and the fitting error is strictly controlled within ≤±0.1mm. Finally, the accurate three-dimensional coordinates of the center of the reserved hole tube are obtained, which significantly improves the stability, accuracy and precision of cylinder fitting under complex working conditions.

[0024] An improved boundary extraction algorithm fits mold edges: Addressing the shortcomings of traditional boundary extraction algorithms in practical engineering scenarios such as weak reflection, low contrast, and local sparsity of mold edge point clouds, which easily lead to boundary breakage, false detections, and missed detections, resulting in insufficient robustness, an improved boundary extraction algorithm with dual-condition joint judgment is adopted to achieve stable and complete extraction of mold edges. This algorithm innovatively integrates normal vector difference analysis and neighborhood curvature mutation analysis to construct a dual-condition joint discrimination model. First, the angle difference between the normal vector of a single point cloud on the mold plane and the normal vector of the overall reference plane of the mold is calculated, setting the normal vector angle threshold to 5°-10° to screen out candidate boundary points with normal vector mutations. Then, a secondary verification is performed by combining the curvature change rate within the k-neighborhood of the point cloud, setting the curvature threshold to 0.05-0.1. Accurate locking of boundary points is achieved through dual-threshold linkage discrimination. To address the challenges of weak boundaries and low contrast scenarios, a weighted smoothing process for neighborhood point clouds and boundary continuity constraints are introduced to correct boundary breakage caused by single-point noise, overcoming the limitation of traditional algorithms failing to extract boundaries in low-quality point clouds. After extracting the mold boundary points, an improved RANSAC algorithm is used for straight line fitting, innovatively incorporating geometric constraints of adjacent edges being perpendicular and opposite edges being parallel. This accurately removes boundary burrs and stray interference points, effectively avoiding deviation and distortion of the fitted straight line, and ultimately outputting a stable and accurate mold reference edge straight line, providing a reliable benchmark for subsequent positioning deviation calculations.

[0025] Spatial distance calculation: The Euclidean distance calculation method from a spatial point to a straight line is adopted to calculate the shortest distance from the center of the reserved hole tube to the straight line of the mold reference edge, with a calculation accuracy of ≤ ±0.1mm; and the calculation results are projected onto the mold plane to eliminate the influence of spatial elevation differences on positioning accuracy, and obtain the in-plane positioning deviation to ensure that the deviation calculation is consistent with the engineering acceptance specifications.

[0026] Deviation Judgment and Report Output: The measured positioning deviation is compared with the design specification value and the design value. The allowable deviation is set according to GB 50204 and the acceptance specification for prefabricated components. The system automatically judges whether the deviation is qualified, slightly exceeding the standard, or unqualified. A digital acceptance report containing point cloud visualization annotations, deviation data, and acceptance conclusions is generated. The report supports PDF format export and realizes full-process digital traceability.

[0027] The advantages of this method for detecting the location of occupiers in prefabricated buildings based on 3D point clouds are as follows:

[0028] (1) The model optimization is highly targeted: the traditional RANSAC cylindrical fitting model is optimized by elevation prior, and the vertical protrusion feature of the reserved hole pipe is introduced as a prior constraint to reduce the random sampling interval, greatly reduce the interference of noisy points such as steel bars and debris, improve the success rate and accuracy of cylindrical fitting, and adapt to the scene of dense steel bar occlusion.

[0029] (2) High robustness of boundary extraction: An improved mold boundary extraction algorithm is proposed, which integrates the dual-condition judgment of normal vector difference and neighborhood curvature to solve the problems of inaccurate and offset mold edge extraction under weak boundary, low contrast and noise interference, and ensure the edge fitting accuracy.

[0030] (3) Integrated testing process: Construct an integrated testing process of "elevation segmentation - optimization fitting - boundary improvement - spatial solution" to achieve a fully automatic closed loop from point cloud input to acceptance report output, without manual intervention, and adaptable to complex construction site scenarios.

[0031] (4) Full-area non-contact inspection: One scan covers all hidden areas, with no blind spots for manual inspection. It does not require contact with holes, pipes, steel bars and mold surfaces, does not affect the construction process, and avoids disturbance to components and steel bars caused by contact measurement.

[0032] (5) Digital closed-loop acceptance: Point cloud model + 3D annotation + test report + data archive form a complete chain of acceptance evidence, which can be linked with BIM model to meet the digital management requirements of smart construction site and acceptance archiving.

[0033] The improvement effect of this method for detecting the location of occupiers in prefabricated buildings based on 3D point clouds is as follows:

[0034] Significant model innovation: The RANSAC cylindrical fitting model is optimized by introducing prior elevation constraints, which improves the fitting accuracy by more than 20% compared with traditional methods.

[0035] Enhanced algorithm robustness: The improved boundary extraction algorithm can stably extract mold edges in scenarios with weak boundaries, low contrast, and dense reinforcement.

[0036] Outstanding anti-interference capability: It makes full use of the vertical protrusion geometric characteristics and is not affected by the obstruction of steel bars, stirrups, or embedded parts.

[0037] Fully automated process: From point cloud input to acceptance output, it can be completed with one click without manual intervention, greatly improving acceptance efficiency.

[0038] High precision and traceability: The positioning deviation detection accuracy reaches ±0.5mm, and the results can be linked with BIM for archiving, meeting the requirements of digital quality management.

[0039] Non-contact testing: avoids disturbance to components and reinforcing bars caused by traditional contact measurement, ensuring the quality of finished construction products. Attached Figure Description

[0040] Figure 1 This is a flowchart of a method for detecting the location of occupant pipes in prefabricated buildings based on 3D point clouds.

[0041] Figure 2This is a schematic diagram of a prefabricated building hole occupancy pipe location detection system based on 3D point cloud.

[0042] Figure 3 This is a field inspection diagram of a method for detecting the location of occupant pipes in prefabricated buildings based on 3D point clouds. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention. The core objective of this embodiment is to achieve fully automatic, high-precision, non-contact detection of the location of reserved holes and pipes during the concealed acceptance stage, forming a complete digital acceptance closed loop. The specific operation is as follows:

[0044] Step (1) Global point cloud acquisition

[0045] An industrial-grade 3D laser scanner was used for point cloud acquisition. The scanner has advantages such as high measurement accuracy (±0.1mm), fast scanning speed (≥1 million points / second), and strong resistance to environmental interference. It can be adapted to the complex environment of prefabricated building construction sites and precast component factories. The following parameters and operating requirements were strictly controlled during the acquisition process:

[0046] S11 Scanning distance: Set to 1500mm. This distance range balances scanning accuracy and scanning range. When the distance is less than 600mm, the scanner lens is prone to interference with the mold and steel frame, and the point cloud data is prone to oversaturation. When the distance is greater than 1800mm, the point cloud density will be significantly reduced, which cannot meet the accuracy requirements for extracting detailed features of the hole tube.

[0047] S12 Point Cloud Density: Set to ≥80 points / cm² 2 This density ensures sufficient point cloud data for the top contour, edge, and mold details of the pre-reserved hole tube, avoiding feature loss due to sparse point cloud, and providing reliable support for subsequent cylinder fitting and edge extraction.

[0048] S13 Data Acquisition Method: Multi-station scanning mode is adopted. The number of stations is set according to the size of the detection area. Small detection areas (area ≤ 3m²) 2 Set up 3 stations for large detection areas (area > 3m²). 2 Six stations are set up, with an overlap of no less than 30% between each station to ensure the continuity of the scanned data; during the scanning process, the scanner’s automatic stitching function is activated, and the stitching accuracy is controlled within ±0.05mm to ensure full coverage of the reserved hole pipe, steel reinforcement cage, and mold, with no scanning blind spots;

[0049] S14 Data Output: After scanning, the original 3D point cloud file with 3D coordinate information is output, supporting three mainstream formats: PLY, PCD, and LAS. The PLY format is used for subsequent visualization processing, the PCD format is used for algorithm calculation, and the LAS format is used for data archiving, ensuring the universality and scalability of the data.

[0050] Step (2) Point cloud preprocessing

[0051] The original point cloud data contains a large number of discrete noise points (such as dust, light reflection interference points, surrounding debris points, and reflective points on the surface of steel bars), and the data volume is large, which will affect the accuracy and efficiency of subsequent feature extraction, cylinder fitting, and distance calculation. This step obtains clean, concise, and feature-clear point cloud data through denoising and downsampling processing, which is specifically divided into the following two sub-steps:

[0052] S21 Point Cloud Denoising: A statistical filtering algorithm is used to traverse each point and its k-neighborhood (k=32, adaptively adjusted according to the point cloud density), calculate the average distance between the point and its neighboring points, set the distance threshold to 1.5 times the average distance, remove discrete noise points whose distance exceeds the threshold, and retain the effective point cloud of reserved hole pipes, steel bars and molds.

[0053] S22 Downsampling: A voxel downsampling algorithm is adopted, with the voxel grid size set to 0.8mm. While preserving key geometric features such as the outline of the reserved hole tube and the edge of the mold to the maximum extent, the amount of point cloud data is reduced, the computational redundancy of subsequent algorithms is reduced, and the processing efficiency is improved.

[0054] Step (3) Elevation Domain Feature Segmentation

[0055] S31 Elevation Statistics: Perform elevation distribution statistics on the pre-processed point cloud. Using the mold plane as the reference plane (elevation set to 0), count the elevation values ​​of all point clouds to determine the elevation range of the reserved hole tube protruding vertically from the mold plane (usually 5mm to 50mm).

[0056] S32 Threshold Segmentation: Set the elevation segmentation threshold to 35mm, filter out point clouds with elevation values ​​within this range as candidate point clouds, and remove background point clouds such as steel skeletons, ground, and debris with elevation values ​​below 5mm to obtain a set of candidate point clouds containing only reserved hole pipes.

[0057] S33 Candidate Point Cloud Optimization: Perform connected component analysis on the selected candidate point clouds, and remove those with an area smaller than a set threshold (set according to the size of the hole tube, usually 10cm). 2 Small connected regions are further eliminated to eliminate noise interference and ensure that the candidate point cloud only contains reserved hole tubes.

[0058] Step (4) Optimize the RANSAC cylindrical fitting model and solve for the center of the circle.

[0059] S41 Model Initialization: Load candidate point cloud, start optimizing RANSAC cylindrical fitting model, set the number of iterations to 1800, the in-point threshold to 0.2mm, and the cylinder radius range to be set according to the design drawings (usually 50mm to 200mm).

[0060] S42 Elevation Prior Sampling: Introducing elevation prior constraints, only 3 points are randomly selected within the candidate point cloud range (elevation 5mm~50mm) to construct the initial cylindrical model, avoiding sampling in the background area and reducing the probability of misfitting;

[0061] S43 Topology Consistency Check: Perform topology consistency check on the fitted initial cylindrical model, calculate the curvature and normal direction of the points inside the model, remove the points that do not conform to the topological characteristics of the cylinder, and ensure that the fitting result is a tubular structure;

[0062] S44 Center Optimization: After fitting, extract the point cloud of the cylinder top contour, use the least squares method to fit the contour point cloud into a circle, with a fitting error ≤ ±0.1mm, calculate the three-dimensional coordinates (X, Y, Z) of the center of the reserved hole tube, and record and archive them.

[0063] Step (5) Improved boundary extraction algorithm fits the mold edge

[0064] S51 Mold Point Cloud Segmentation: Segment the mold plane point cloud (point cloud with elevation value close to 0) from the preprocessed point cloud as the target area for edge extraction;

[0065] S52 Dual-condition boundary discrimination: Perform normal vector difference analysis and neighborhood curvature change analysis on each point of the mold plane point cloud. The threshold for the angle between the normal vectors is set to 5° to 10°, and the curvature threshold is set to 0.05 to 0.1. When the angle between the normal vector of a point and the normal vector of a neighboring point exceeds the threshold, and the curvature change reaches the set threshold, the point is determined to be a mold boundary point.

[0066] S53 Edge Line Fitting: Aggregate all mold boundary points, use the improved RANSAC algorithm for line fitting, add geometric constraints of adjacent edges being perpendicular and opposite edges being parallel, with parallelism error ≤0.3° and perpendicularity error ≤0.5°, remove burrs and interference points, output 4 stable mold reference edge lines, and record the spatial equation of each line.

[0067] Step (6) Spatial distance calculation

[0068] S61 Spatial distance calculation: Using the Euclidean distance formula from a spatial point to a straight line, calculate the shortest distance from the center (X, Y, Z) of the reserved hole tube to the straight line of each mold reference edge, with a calculation accuracy of ≤ ±0.1mm;

[0069] S62 Plane Projection Deviation: Project the spatial distance onto the mold plane (Z=0) to eliminate the influence of spatial elevation differences on positioning accuracy and obtain the positioning deviation value in the plane. This deviation value is the actual positioning deviation from the center of the reserved hole tube to the edge of the mold.

[0070] S63 Data Recording: Record the center coordinates and measured positioning deviation of each reserved hole tube, and associate them with the corresponding mold edge straight line information to form a complete set of measured data.

[0071] Step (7) Deviation Judgment and Report Output

[0072] S71 Deviation Judgment: Based on GB 50204 "Code for Acceptance of Construction Quality of Concrete Structures" and the acceptance code for prefabricated components, the allowable deviation for the positioning of reserved hole pipes is set (usually ±5mm). The measured positioning deviation is compared with the allowable deviation and the design value: if the measured deviation is ≤ the allowable deviation and the deviation from the design value is ≤ ±0.5mm, it is judged as qualified; if the measured deviation is > the allowable deviation but ≤ 1.2 times the allowable deviation, it is judged as slightly exceeding the standard; if the measured deviation is > 1.2 times the allowable deviation, it is judged as unqualified.

[0073] S72 Visual Annotation: Generates 3D text labels in the point cloud visualization interface, with annotation content including: hole tube number, design positioning value, measured positioning value, deviation value, and acceptance conclusion; at the same time, it annotates the mold reference edge line, the center position of the hole tube, and the fitted cylinder, intuitively displaying the positioning deviation.

[0074] S73 Report Generation and Archiving: Automatically generates the "Inspection Report on the Location of Reserved Holes and Pipes for Concealed Acceptance", which includes: project name, component number, inspection area, inspection time, hole and pipe number, design positioning value, measured positioning value, deviation value, acceptance conclusion, and point cloud thumbnail; the report supports export in PDF format, enabling digital traceability.

[0075] like Figure 2 As shown, this embodiment also relates to a prefabricated building hole occupant pipe location detection system based on three-dimensional point clouds, including:

[0076] (1) Point cloud acquisition module, used to acquire full-area three-dimensional point cloud data of the concealed acceptance area;

[0077] (2) Point cloud preprocessing module, used for noise filtering, outlier removal and data simplification;

[0078] (3) Optimized cylinder fitting module: equipped with a RANSAC model with elevation prior constraints, used for pre-reserved hole tube identification and center calculation;

[0079] (4) Improved boundary extraction module: used for high-precision extraction and straight line fitting of mold edges;

[0080] (5) Distance calculation module, used to calculate the spatial distance from the center of the hole tube to the edge of the mold;

[0081] (6) Result output module, used for deviation comparison, qualification judgment and acceptance report generation.

[0082] The optimized cylinder fitting module improves the anti-interference ability of cylinder recognition through the elevation prior sampling strategy, adapts to the scene of dense steel bar occlusion, and can effectively identify reserved hole pipes partially obscured by steel bars and stirrups, with an accuracy rate of ≥93%.

[0083] The improved boundary extraction module is based on dual conditions of normal vector difference and neighborhood curvature to achieve stable extraction of mold edges under weak boundary and low contrast conditions. The edge extraction error is ≤ ±0.3mm, ensuring the accuracy of subsequent distance calculation.

Claims

1. A method for detecting the location of occupiers in prefabricated buildings based on 3D point clouds, comprising the following steps: (1) Obtain three-dimensional point cloud data of the concealed acceptance area using a three-dimensional laser scanning device; (2) Preprocess the original point cloud to achieve noise filtering, outlier removal and dataset optimization; (3) Based on the vertical protrusion geometric representation of the pre-embedded sleeve, the target feature point cloud is extracted by elevation domain segmentation; (4) The optimized RANSAC cylindrical fitting model is used to fit the candidate point cloud to the surface, and the three-dimensional coordinates of the center of the reserved hole tube are obtained by combining the circular contour analysis. (5) The mold region is segmented from the point cloud by an improved boundary feature extraction algorithm, and the reference edge line of the mold is extracted and fitted; (6) Calculate the shortest distance from the center of the hole tube to the edge of the mold through spatial geometric operations to obtain the measured positioning value; (7) Compare the measured values ​​with the design values, and output the deviation results and acceptance judgment conclusions.

2. The method for detecting the location of occupant pipes in prefabricated buildings based on three-dimensional point clouds according to claim 1, characterized in that: The preprocessing method in step (2) is as follows: statistical filtering is used to remove point cloud noise points, the k-neighborhood value is 16-32, data simplification is achieved by voxel downsampling, the voxel grid size is set to 0.8mm, and computational redundancy is reduced while preserving the geometric features of the target component, thereby improving the efficiency of subsequent processing.

3. The method for detecting the location of occupant pipes in prefabricated buildings based on three-dimensional point clouds according to claim 1, characterized in that: In step (4), the optimization of the RANSAC cylindrical fitting model is as follows: On the traditional RANSAC random sampling fitting framework, a prior constraint mechanism for mold elevation is introduced. Based on the vertical protruding structural characteristics of the pre-reserved hole pipe in prefabricated buildings, the vertically protruding area where the point cloud elevation is higher than the mold reference plane is defined as the core priority sampling interval, and the elevation range of the sampling point cloud is strictly limited to 35mm above the mold plane. Within the specified interval, non-target point cloud interference such as the mold reference plane and exposed rebar are directly eliminated through pre-filtering with an elevation threshold, avoiding invalid iterations from the sampling source. At the same time, the model iteration count is adaptively set to 1800 times, balancing fitting efficiency and global optimal solution search capability. The interior point distance threshold is finely set to 0.2mm to meet the high-precision fitting requirements of mold hole tubes, significantly improving the stability of cylinder fitting and the accuracy of size recognition under complex point cloud noise. On this basis, a cylinder topology consistency verification rule is added. By verifying three types of topological features of the point cloud—axial continuity, radial cross-section roundness, and tubular radial wall thickness consistency—non-tubular stray interference points and local protrusion noise points are automatically eliminated. The fitting model is strictly constrained to only match the standard cylindrical topology structure of the reserved hole tube, eliminating misfitting of non-target structures.

4. The method for detecting the location of occupant pipes in prefabricated buildings based on three-dimensional point clouds according to claim 1, characterized in that: The improved boundary feature extraction algorithm in step (5) is as follows: Combining the difference of normal vectors and the curvature analysis of the neighborhood, the angle difference between the normal vector of a single point cloud on the mold plane and the normal vector of the overall reference plane of the mold is first calculated, and candidate boundary points with abrupt changes in normal vectors are selected. Then, the curvature change rate in the k neighborhood of the point cloud is combined for secondary verification, and the boundary points are accurately locked through the dual threshold linkage discrimination. For the actual engineering scenario of weak reflection, low contrast and local sparseness of the edge point cloud of the mold, the neighborhood point cloud weighted smoothing processing and boundary continuity constraint are introduced to correct the boundary breakage, false detection and missed detection caused by single point noise. This breaks through the limitation of the traditional algorithm in the failure of boundary extraction under low quality point cloud, and greatly improves the robustness and completeness of the extraction of mold edge features under complex working conditions.

5. The method for detecting the location of occupant pipes in prefabricated buildings based on three-dimensional point clouds according to claim 1, characterized in that: Step (6) uses the Euclidean distance from a spatial point to a straight line for calculation with an accuracy of ≤ ±0.1 mm. The result is then projected onto the mold plane to form an in-plane positioning deviation, eliminating the influence of spatial elevation differences on positioning accuracy and ensuring that the deviation calculation is consistent with the engineering acceptance specifications.

6. The method for detecting the location of occupant pipes in prefabricated buildings based on three-dimensional point clouds according to claim 1, characterized in that: In step (4), the least squares method is used to optimize the fitting of the circular contour analysis. The fitting optimization of the point cloud of the top contour of the cylinder is performed, and the fitting error is ≤ ±0.1mm, which improves the accuracy of the center calculation and ensures the accuracy of the hole tube center positioning.

Images