A shield segment geometric precision detection method and system based on structured light scanning and a medium

Abstract

The application relates to a shield segment geometric precision detection method and system based on structured light scanning and a medium, and relates to the technical field of component maintenance. The shield segment geometric precision detection method comprises the following steps: according to the segment type and the mold number of a to-be-detected shield segment, a standard segment three-dimensional model is matched and extracted; the to-be-detected shield segment is analyzed, a plurality of segment functional parts are divided and scanned in combination with the segment geometric structure, and corresponding segment part point clouds are obtained; the segment part point clouds are cleaned and sampled, part simplified point clouds are obtained, and a shield segment point cloud model is spliced and constructed in combination with the segment functional form; the shield segment point cloud model and the standard segment three-dimensional model are fitted and aligned, multi-dimensional deviation values are calculated, and a segment part deviation array is generated; according to a preset allowed deviation interval, the segment position deviation array is compared, and a segment precision detection table is generated; and the automatic and high-precision detection of the geometric size of the shield segment is realized, and the quality inspection efficiency and data traceability are significantly improved.

Description

Technical Field

[0001] This application relates to the field of component maintenance technology, and in particular to a method, system and medium for detecting the geometric accuracy of tunnel lining segments based on structured light scanning. Background Technology

[0002] Early methods for detecting the geometric accuracy of tunnel boring machine (TBM) segments relied primarily on manual point-to-point measurements using conventional measuring tools such as micrometers, templates, and feeler gauges. This approach was highly complex, had limited coverage, resulted in significant errors in measuring complex angles, and was susceptible to environmental temperature and humidity fluctuations, as well as subjective judgment by operators. Furthermore, data calculations were cumbersome and prone to errors, making it difficult to comprehensively reflect the true shape and position deviations of the segments. With increasing engineering precision requirements, equipment such as API laser trackers, high-precision total stations, and 3D laser scanners have become increasingly common. These devices enable 3D data acquisition with micron-level precision, automatic compensation for temperature and humidity errors, and the generation of visual color difference analysis reports. Detection accuracy can reach 0.25mm, significantly improving the objectivity and comprehensiveness of the inspection.

[0003] Existing patents disclose a detection and fitting method and device for horizontal assembly of tunnel boring machine (TBM) segments. The method includes the following steps: optimizing the design of the vertical lifting fixture and positioning the rollers; adjusting the base height based on the wedge distribution of the segments, adjusting the height of the base for horizontal assembly using a total station, and adjusting the height of each point calculated based on the wedge distribution; performing three-dimensional detection on the horizontally assembled segments, creating a three-dimensional model, and acquiring data in a scanning, omnidirectional manner; processing the data using an optimal fitting algorithm to reflect the dimensional deviations of the segments or molds at various locations, and generating a color difference vector map. This invention can intuitively and effectively reflect the dimensions and flatness of the tunnel mold. After adjusting the mold precision based on the detection results, production can begin. Horizontal assembly fitting not only verifies the rationality of the segment structure design but also provides a direct indication of precision control.

[0004] The existing technical solutions mentioned above have the following drawbacks: 1. Existing monitoring methods are difficult to fully reflect the overall deviation of the tunnel segment surface, especially the local concavity or distortion of the curved contour; 2. The detection of a single tunnel segment is time-consuming and greatly affected by human reading errors. Paper records of detection data are not easy to perform statistical analysis and cannot be directly compared with the design model. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this application is to provide a method, system, and medium for detecting the geometric accuracy of tunnel boring machine (TBM) segments based on structured light scanning. By using multi-angle splicing to achieve full surface coverage of the segments, the detection time is shortened. Automated deviation calculation and color cloud map display provide an intuitive presentation of the overall geometric quality of the segments, avoiding blind spots in manual sampling inspection. This provides data support for improving the segment production process and enhances detection efficiency and reliability.

[0006] This was achieved using the following technical solutions: In a first aspect, this application provides a method for detecting the geometric accuracy of tunnel boring machine segments based on structured light scanning, including: Based on the tunnel segment model and mold number of the shield tunnel segment to be inspected, a standard three-dimensional model of the tunnel segment is matched and extracted; The tunnel segment to be inspected is analyzed, and combined with the segment geometry, several functional parts of the segment are divided and scanned to obtain the corresponding segment point cloud. Clean and sample the point cloud of the tunnel segment to obtain a simplified point cloud of the segment, and combine it with the functional shape of the segment to construct a point cloud model of the shield tunnel segment; Fit and align the point cloud model of the tunnel segment with the standard 3D model of the segment, calculate the multidimensional deviation value, and generate the segment location deviation array. A segment accuracy inspection table is generated by comparing the segment position deviation array with the preset allowable deviation range.

[0007] By adopting the above technical solution, based on the standard three-dimensional model of the tunnel segment model, the point cloud of the part is obtained by partition scanning and fitted and aligned using the ICP registration algorithm. After calculating the multidimensional deviation value, it is compared with the preset allowable deviation to generate a precision detection table. This realizes the automated and high-precision detection of the geometric dimensions of the tunnel segment, which significantly improves the quality inspection efficiency and data traceability.

[0008] This application is further configured to: match and extract a standard segment 3D model based on the segment type and mold number of the tunnel boring machine segment to be inspected, including: The identification codes of the tunnel segments to be inspected are parsed to extract the segment model and mold number; The model of the tunnel segment is matched and searched according to the shield tunnel segment model library, and the storage path of the model file is extracted. The shield tunnel segment model is extracted and parsed according to the model file storage path to obtain the model vertices, triangular facets and normal vectors, and three-dimensional mesh data is constructed. The 3D mesh data is correlated based on the mold number to obtain the standard segment 3D model.

[0009] By adopting the above technical solution, the segment model is matched based on the identifier code parsing and hash mapping algorithm, the model storage path is obtained by querying the file system, vertices, triangles and normals are extracted to construct three-dimensional mesh data, and standard segment three-dimensional models are generated based on the mold number, which significantly improves the model retrieval efficiency and data association accuracy.

[0010] This application further specifies: analyzing the tunnel segment to be inspected, combining the segment geometry, dividing and scanning several functional parts of the segment, and obtaining the corresponding segment point cloud, including: Images of the tunnel segments to be inspected are acquired, resulting in several images of the tunnel segments. Boundary segmentation and optical flow registration are performed on the shield tunnel segment images to form complete segment images; The complete segment image is divided into surface regions based on the segment geometry to obtain the segment functional parts; the segment functional parts include the inner arc surface, outer arc surface, left end face, right end face, upper edge, and lower edge; The functional parts of the tunnel segment are sorted and planned to generate a scanning guide path for each part. Based on the part scanning guide path and the structured light scanning distance, point cloud scanning is performed on the functional parts of the pipe segment to generate an initial part point cloud and calculate the regional point cloud coverage. If the point cloud coverage rate of the area is less than the area coverage threshold, it is determined that the current functional part of the pipe segment is incomplete in the scan, and a prompt to rescan is generated. If the area point cloud coverage is greater than or equal to the area coverage threshold, it is determined that the current segment functional part has been scanned completely, and the initial part point cloud is used as the corresponding segment part point cloud.

[0011] By adopting the above technical solution, a complete tunnel segment image is constructed through image acquisition and optical flow registration. Functional parts are divided according to geometric structure and structured light scanning paths are generated. After obtaining the point cloud, the scan integrity is automatically judged using the coverage threshold, which improves the automation level and data reliability of point cloud acquisition for tunnel segment parts.

[0012] This application further specifies: cleaning and sampling the point cloud of the tunnel segment to obtain a simplified point cloud of the segment, and combining it with the functional morphology of the segment to construct a shield tunnel segment point cloud model, including: The distance between each sampling point and its neighboring points in the point cloud of the tunnel segment is calculated to obtain the average neighborhood distance. If the average distance of the neighborhood is greater than the preset aggregation limit distance threshold, the current sampling point is determined to be an outlier noise point and is removed to obtain the point cloud of the aggregation part. Voxel downsampling is performed on the point cloud of the aggregation region to obtain a simplified point cloud of the region, and the point cloud curvature of each sampling point is calculated. If the curvature of the point cloud is greater than the curvature threshold, the current sampling point is determined to be a positioning marker point, and the coordinates of the marker point are extracted; Based on the coordinates of the marked points, the point clouds of different parts are simplified and homogeneous matching is performed to calculate the transformation parameters between the point clouds. Based on the registration constraint parameters and the transformation parameters between point clouds, iterative registration is performed between the simplified point clouds of the parts to obtain the rotation transformation matrix. Based on the rotation transformation matrix and the functional morphology of the tunnel segments, the simplified point cloud of each part is transformed and spliced ​​to construct a point cloud model of the tunnel segment.

[0013] By adopting the above technical solutions, outlier noise points are eliminated through neighborhood distance filtering, and voxel downsampling and curvature thresholding are used to extract positioning marker points. The coordinates of the marker points are used to achieve source matching and iterative registration (such as the ICP algorithm). Combined with the functional shape of the tunnel segments, a high-precision shield tunnel segment point cloud model is constructed, which significantly improves the automation level of data processing and the integrity of the model.

[0014] This application further specifies: fitting and aligning the point cloud model of the tunnel segment with the standard three-dimensional model of the segment, calculating multidimensional deviation values, and generating a segment location deviation array, including: The shield tunnel segment point cloud model is uniformly and randomly sampled according to the preset sampling ratio to form a control point set; Statistical calculations were performed on the control point set and the standard segment 3D model to obtain the squared signed distance, which was used as the objective function. The point cloud model of the tunnel segment and the 3D model of the standard segment are iteratively aligned according to the objective function, and the parameters of the rotation transformation matrix are updated to obtain the optimal transformation matrix. The shield tunnel segment point cloud model is registered and transformed according to the optimal transformation matrix to obtain the transformed segment point cloud model. The point cloud model of the converted tunnel segment and the standard tunnel segment 3D model are interpolated point by point to obtain the point cloud normal deviation. The point cloud model of the converted tunnel segment is analyzed based on its functional parts. The corresponding measured dimensions are calculated and compared with the theoretical dimensions to obtain multidimensional deviation values, including: The end faces of the point cloud model of the converted tunnel segment are extracted, the chord length between the end faces is calculated, and the chord length deviation is calculated by comparing it with the theoretical chord length. Curve interpolation is performed on the outer arc surface of the point cloud model of the converted tunnel segment to reconstruct the measured arc length value, which is then compared with the theoretical arc length value to calculate the arc length deviation. The distance between the outer and inner arc surfaces of the point cloud model of the converted tunnel segment is calculated, the measured thickness value is calculated, and it is compared with the theoretical thickness value to calculate the thickness deviation. The edge points of the holes in the point cloud model of the converted pipe segment are identified and fitted, the center coordinates of the measured bolt holes are calculated, and compared with the theoretical center coordinates to calculate the hole position deviation. Based on the coordinates of the sampling points, the point cloud normal deviation and multidimensional deviation values ​​are combined to construct a segment deviation array.

[0015] By adopting the above technical solution, high-precision alignment of the shield tunnel segment point cloud model with the standard three-dimensional model is achieved based on control point sampling and ICP iterative registration algorithm. By performing point-by-point difference calculation and multi-dimensional size comparison of end face chord length, arc length, thickness, and hole position, a deviation array is constructed, realizing automated and quantitative evaluation of the segment geometric quality, and improving detection accuracy and quality inspection efficiency.

[0016] This application is further configured to: generate a segment accuracy detection table by comparing the segment position deviation array with a preset allowable deviation range, including: Based on the segment type, the preset allowable deviation ranges are extracted and arranged to form a tolerance threshold array; The preset allowable deviation ranges include flatness deviation range of [-a,+a]mm, chord length deviation range of [-b,+b]mm, thickness deviation range of [-c,+c]mm, and hole position deviation range of [-d,+d]mm. The segment position deviation array is compared item by item according to the tolerance threshold array, and the number of abnormal segments is marked. If the number of abnormal parts of the tunnel segment is within the first quantity threshold range, the accuracy abnormality level of the current shield tunnel segment point cloud model is determined to be low, and a tunnel segment repair suggestion is generated. If the number of abnormal parts of the tunnel segment is within the second threshold range, the accuracy abnormality level of the current shield tunnel segment point cloud model is determined to be medium, and a segment repair prompt is generated. If the number of abnormal parts of the tunnel segment is within the third threshold range, the accuracy level of the current shield tunnel segment point cloud model is determined to be high, and a segment remodeling warning is generated. Based on the shield tunnel segment identification, the accuracy anomaly level and segment position deviation array are mapped to generate a pseudo-color cloud map of deviation distribution. Combined with the functional parts of the segments, a segment accuracy detection table is constructed.

[0017] By adopting the above technical solution, a tolerance threshold array for flatness, chord length, thickness, and hole deviation is extracted based on the segment type. The number of abnormal parts is generated by comparing each item, and low-level, medium-level, or high-level anomalies are classified according to the number. Repair suggestions, repair prompts, or reconstruction warnings are given respectively. At the same time, a pseudo-color cloud map of deviation distribution is generated by combining the segment identification mapping. This realizes the automation, visualization, and hierarchical early warning of segment accuracy detection, which significantly improves the detection efficiency and anomaly location accuracy.

[0018] Secondly, this application also provides a shield tunnel segment geometric accuracy detection system based on structured light scanning, which adopts the following technical solution: A shield tunnel segment geometric accuracy detection system based on structured light scanning, used to implement a shield tunnel segment geometric accuracy detection method, comprising: The model extraction module is used to match and extract the standard three-dimensional model of the tunnel segment based on the segment model and mold number of the shield tunnel segment to be inspected. The local acquisition module is used to analyze the shield tunnel segment to be inspected, and, in combination with the segment geometry, divide and scan several functional parts of the segment to obtain the corresponding segment point cloud. The global processing module is used to clean and sample the point cloud of the tunnel segment, obtain a simplified point cloud of the segment, and combine it with the functional shape of the segment to splice and construct the point cloud model of the shield tunnel segment. The deviation alignment module is used to fit and align the point cloud model of the tunnel segment with the standard 3D model of the segment, calculate the multidimensional deviation value, and generate a segment location deviation array. The accuracy detection module is used to compare the segment position deviation array with the preset allowable deviation range and generate a segment accuracy detection table.

[0019] By adopting the above technical solution, based on the three-dimensional model of the tunnel segment model matching benchmark, the point cloud of the part is obtained by partition scanning and a global point cloud model is constructed by sampling and splicing. After fitting and aligning with the ICP iterative nearest point algorithm, the multidimensional deviation value is calculated, and the accuracy detection table is automatically generated according to the preset tolerance range. This realizes the automated and high-precision detection of the geometric dimensions of the shield tunnel segment, and improves the quality inspection efficiency and data traceability.

[0020] Thirdly, this application also provides an electronic device, comprising: One or more processors; Memory, used to store one or more programs; When one or more programs are executed by one or more processors, the one or more processors implement any of the methods in the above scheme.

[0021] Fourthly, this application also provides a storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the shield tunnel segment geometric accuracy detection method based on structured light scanning as described above.

[0022] In summary, the beneficial technical effects of this application are as follows: Multi-angle splicing achieves full surface coverage of the tube segment, shortening the inspection time. Automated deviation calculation and color cloud map display present the overall geometric quality of the tube segment, avoiding blind spots in manual sampling. Through the quality traceability and statistical analysis of digital inspection reports, data support is provided for the improvement of segment production processes, significantly enhancing inspection efficiency and reliability. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the method for detecting the geometric accuracy of tunnel segments in this application; Figure 2 This is a flowchart illustrating step S3 in this application; Figure 3 This is a flowchart illustrating step S4 in this application; Figure 4This is a schematic diagram of the shield tunnel segment geometric accuracy detection system in this application. Detailed Implementation

[0024] The present application will be further described in detail below with reference to the accompanying drawings.

[0025] Reference Figure 1 This application discloses a method for detecting the geometric accuracy of tunnel lining segments based on structured light scanning, comprising: S1: Match and extract the standard segment 3D model based on the segment model and mold number of the shield tunnel segment to be inspected; S2: Analyze the tunnel segment to be detected, combine the segment geometry, divide and scan several functional parts of the segment, and obtain the corresponding segment point cloud; S3: Clean and sample the point cloud of the tunnel segment to obtain a simplified point cloud of the segment, and combine it with the functional shape of the segment to construct a point cloud model of the shield tunnel segment; S4: Fit and align the point cloud model of the tunnel segment with the standard 3D model of the segment, calculate the multidimensional deviation value, and generate the segment location deviation array; S5: Generate a segment accuracy inspection table by comparing the segment position deviation array with the preset allowable deviation range.

[0026] In this embodiment, the corresponding standard 3D design model file (in STL or OBJ format) is loaded from the "Design Model Library" folder of the database based on the model number (drop-down menu) and mold number (manual input or barcode scanning) of the segment to be inspected. The design model file contains the theoretical geometry of the segment, with the origin of the coordinate system defined at the center of the segment. An inspection task ID (format "JC_Year / Month / Day_Serial Number") is automatically generated, and the task status is set to "Pending Acquisition". Simultaneously, a prompt message is displayed on the interface: "Please place the segment on the calibration platform of the inspection station, ensuring the platform is clean and aligning the four corners of the segment with the four positioning marks (four circular indentations) on the platform." After the operator confirms, click "Next".

[0027] Based on the 3D schematic diagram of the tunnel segment, the area to be scanned is divided into six parts: inner arc surface, outer arc surface, left end face, right end face, upper edge, and lower edge. The operator uses a structured light scanning device (connected via USB or network) to scan each part sequentially as prompted. During scanning, the operator holds the device approximately 30-50cm away from the tunnel segment surface and moves it slowly, ensuring the structured light projected by the device covers the entire area. The operator receives the point cloud data stream transmitted from the scanning device in real time. After each scan, the point cloud data for the current segment is temporarily stored in memory in binary format (each point includes X, Y, Z coordinates and reflection intensity I), and the area is marked in green (acquired) on the schematic diagram. The acquisition time, scanning angle (approximate direction), and number of points for that segment's point cloud are also recorded. If an area is not scanned completely (point cloud coverage is below a preset threshold, such as area coverage <80%), a message "Please rescan this area" is displayed. After all six areas have been scanned, a "Acquisition Complete" confirmation box pops up.

[0028] Preprocessing is performed sequentially for each point cloud. First, outlier removal is performed: for each point cloud, the average distance from each point to its k nearest neighbors (k=50) is calculated. Assuming these average distances follow a Gaussian distribution, the global mean μ and standard deviation σ are calculated. Points with an average distance greater than μ+3σ are identified as outliers and deleted. Next, voxel downsampling is performed: a cubic grid with a side length of 1 mm is created. The coordinates of all points within each grid are averaged, retaining a representative point, thus reducing the point cloud density to approximately one point per square millimeter. The amount of preprocessed point cloud data is typically reduced to about 30% of the original. Next, point cloud stitching is performed. Four positioning markers on the calibration platform (which appear as four distinct circular depressions or protrusions in the point cloud, identifiable by reflection intensity or geometric features) are used. The 3D coordinates of these four marker points are automatically searched within each point cloud (by calculating the curvature or intensity changes of the point cloud). Using the first scanned point cloud (e.g., the inner curved surface) as a reference, its origin is translated to the first detected marker point within that point cloud. Then, for each subsequent point cloud, the coordinates of the same marker point within that point cloud are found, and the rotation matrix and translation vector are calculated (using the ICP iterative nearest-point algorithm, with 50 iterations and a convergence threshold of 0.01 mm), transforming all subsequent point clouds to the coordinate system of the reference point cloud. After stitching, all transformed point clouds are merged into a complete point cloud dataset, and duplicate points in overlapping regions are removed (keeping points that are farther apart). Finally, the origin of the stitched complete point cloud is transformed to the central coordinate system of the design model: by calculating the center coordinates of all marker points and aligning them with the theoretical center coordinates of the marker points in the design model, a rigid body transformation is applied.

[0029] The stitched measured point cloud is best-fitted and aligned with the design model. First, approximately 10% of the points in the measured point cloud are randomly sampled as control points. Using the least squares method, the rotation and translation parameters of the measured point cloud are iteratively adjusted to minimize the sum of squared distances from all control points to the surface of the design model. The iteration stops when the parameter change is less than 0.001 mm or the number of iterations reaches 100. After alignment, the normal distance from each measured point to the surface of the design model is calculated. Method: For each measured point P, the nearest triangular facet is searched in the design model, and the perpendicular distance from P to that facet is calculated (signed, positive on the outside of the model, negative on the inside). A deviation array is generated, with each element containing (x, y, z, deviation value). Based on the deviation values, a deviation distribution map is generated: the deviation interval is divided into 7 segments (<-3mm, -3~-2, -2~-1, -1~+1, +1~+2, +2~+3, >+3mm), assigned colors from blue to red respectively. The measured point cloud is projected onto a 2D unfolded diagram (along the arc length and width directions of the segment), with the color of each pixel representing the average deviation at that location. Simultaneously, key dimensions are extracted: 1) Chord length: From the measured point cloud, point sets from both ends of the segment are extracted, two planes are fitted, and the distance between the two planes is calculated; 2) Arc length: The centerline of the outer arc surface (a series of points along the longitudinal direction of the arc surface) is extracted, and the curve length is calculated; 3) Thickness: For each sampling location, the normal distance from the outer arc surface to the inner arc surface is calculated; 4) Bolt hole location: Using the point set of hole edge points in the point cloud, the center coordinates of the circle are calculated using a circle fitting algorithm (least squares method), and then compared with the theoretical center in the design model to obtain the positional deviation. All dimensional calculation results are compared with the design values, and the deviation values ​​are output.

[0030] The judgment is based on the preset allowable deviation range. The allowable deviation range is stored in a parameter table (e.g., segment width deviation ±1mm, thickness deviation ±2mm, surface flatness deviation (i.e., normal distance) ±3mm, bolt hole position deviation ±2mm). Judgment logic: Iterate through the normal distance of all measured points. If the absolute value of the deviation at any point exceeds ±3mm, the surface flatness is unqualified; if the chord length deviation exceeds ±1mm, the width is unqualified; if the thickness deviation exceeds ±2mm, the thickness is unqualified; if the bolt hole position deviation exceeds ±2mm, the hole position is unqualified. Any unqualified result results in an overall "unqualified" judgment; otherwise, it is "qualified". An inspection report is generated, including: segment number, inspection task ID, inspection time, overall judgment result, deviation distribution cloud map (embedded as a PNG image), a list of key dimension deviations (in tabular form), and a summary of out-of-tolerance points (listing the 3D coordinates and deviation values ​​of all points whose absolute deviation exceeds the allowable value). The report is saved as a PDF in the "Inspection Report" folder, with the filename format "Segment Number_Inspection Time.pdf". Simultaneously, the file path of the report is written back to the segment quality record table in the production database. For defective segments, the out-of-tolerance areas are highlighted in red on the interface, with a prompt "Please manually review or repair". Operators can click to export the coordinates of the out-of-tolerance points for subsequent repairs.

[0031] Preferably, step S1 includes: The identification codes of the tunnel segments to be inspected are parsed to extract the segment model and mold number; The model of the tunnel segment is matched and searched according to the shield tunnel segment model library, and the storage path of the model file is extracted. The shield tunnel segment model is extracted and parsed according to the model file storage path to obtain the model vertices, triangular facets and normal vectors, and three-dimensional mesh data is constructed. The 3D mesh data is correlated based on the mold number to obtain the standard segment 3D model.

[0032] In this embodiment, the model of the segment to be tested is selected from the drop-down menu in the interface. The options in this drop-down menu are derived from a list of all segment models stored in the "Design Model Library" database, which is automatically queried and populated when the interface loads.

[0033] The mold number can be entered manually or by scanning a QR code / barcode. The mold number is used to trace the mold used in the production of this segment and to assist in subsequent quality analysis.

[0034] After receiving the segment model selected by the operator, the system uses it as the primary key to perform an exact match search in the "Design Model Library" database. If the model does not exist in the design model library, a message will pop up saying "No standard model corresponding to this model was found. Please check if the model is correct," and the system will prevent the user from proceeding to the next step.

[0035] After the model verification is passed, the complete storage path of the standard 3D design model file corresponding to the model is constructed according to the preset folder structure rules (such as " / design model library / [segment model] / [segment model]_standard.STL").

[0036] Read the 3D model file (in STL or OBJ format) from the storage medium and load it into the model data structure in memory.

[0037] The loaded model file is parsed to extract vertex, triangle, and normal information, and a 3D mesh data structure is constructed. Simultaneously, the integrity of the model is verified (e.g., whether the file is corrupted, and whether the vertex / face data is complete); if verification fails, an error message is triggered.

[0038] The entered mold number is used as a query condition to retrieve detailed information about the mold (such as mold production date, cumulative number of uses, last calibration time, etc.) from the mold management database, which can be used to trace the mold status during subsequent quality analysis.

[0039] Segment-Mold Association Record: Automatically associates and binds the current segment model with the mold number and records it in the metadata of this inspection task, which facilitates subsequent statistical analysis of quality data by mold dimension.

[0040] The origin of the standard model coordinate system was confirmed to be defined at the center of the tunnel segment, which meets the design requirements. This coordinate system will serve as the benchmark for subsequent alignment of inspection data and calculation of deviations.

[0041] The loaded standard 3D model is bound to the current detection task, and a model pointer is stored in memory for subsequent point cloud comparison, deviation analysis and other modules to call.

[0042] Preferably, step S2 includes: Images of the tunnel segments to be inspected are acquired, resulting in several images of the tunnel segments. Boundary segmentation and optical flow registration are performed on the shield tunnel segment images to form complete segment images; The complete segment image is divided into surface regions based on the segment geometry to obtain the segment functional parts; the segment functional parts include the inner arc surface, outer arc surface, left end face, right end face, upper edge, and lower edge; The functional parts of the tunnel segment are sorted and planned to generate a scanning guide path for each part. Based on the part scanning guide path and the structured light scanning distance, point cloud scanning is performed on the functional parts of the pipe segment to generate an initial part point cloud and calculate the regional point cloud coverage. If the point cloud coverage rate of the area is less than the area coverage threshold, it is determined that the current functional part of the pipe segment is incomplete in the scan, and a prompt to rescan is generated. If the area point cloud coverage is greater than or equal to the area coverage threshold, it is determined that the current segment functional part has been scanned completely, and the initial part point cloud is used as the corresponding segment part point cloud.

[0043] In this embodiment, the standard 3D design model of the tunnel segment to be inspected is first analyzed to obtain its geometric parameters. As a basic component of tunnel lining, the tunnel segment has an arc-shaped plate-like geometry, with the origin of the design coordinate system defined at the segment's center. Based on the segment's geometric characteristics and considering the requirements of subsequent point cloud stitching and deviation analysis, the complete surface of the tunnel segment to be scanned is divided into six functional regions: inner arc surface, outer arc surface, left end face, right end face, upper edge, and lower edge. This division scheme covers all key working surfaces of the tunnel segment, while also taking into account the completeness requirements of subsequent assembly simulation and quality assessment.

[0044] Each area to be scanned is highlighted on the 3D schematic. The inner and outer arc surfaces are the upper and lower surfaces of the tube segment, respectively, and are the core areas for detecting the bending accuracy and fitting quality of the tube segment. The left and right end faces are the mating surfaces when the tube segments are joined circumferentially; their flatness and perpendicularity directly affect the assembly and sealing performance between adjacent tube segments. The upper and lower edges are the mating surfaces when the tube segments are connected longitudinally, used to evaluate the straightness and edge integrity of the tube segment's side profile.

[0045] Using a structured light scanning device, the operator moves slowly along the indicated path at a working distance of approximately 30 to 50 centimeters from the surface of the tube segment, ensuring that the structured light projected by the device completely covers the current area. During operation, the device encodes and decodes the projected grating stripes and the deformation stripes captured by the camera, using the principle of triangulation to calculate the high-precision three-dimensional point cloud coordinates of the object's surface, while simultaneously acquiring the reflection intensity information of each point.

[0046] The system receives binary point cloud data streams from the scanning device in real time via USB or network interface. Each scan point contains X, Y, Z three-dimensional spatial coordinates and reflection intensity I information, forming a complete geometric-optical record. The point cloud data of the current area is temporarily stored in memory, and metadata such as the acquisition time, scanning angle, and number of points in the point cloud are recorded simultaneously.

[0047] The completeness of the scan in a given area is determined based on the coverage rate of the point cloud data. A coverage threshold is preset for each area, and the effective area percentage of the received point cloud is compared with the expected area. When the point cloud coverage of an area is lower than the preset threshold (e.g., area coverage < 80%), it is considered an incomplete scan, and a message "Please rescan this area" is displayed, and the area's status in the diagram remains incomplete. If the coverage meets the requirements, the area is marked in green, indicating "collected".

[0048] After all six regions have been effectively scanned, a "Collection Complete" confirmation box pops up, prompting the operator to confirm the end of this collection task. Simultaneously, the current collection task status is updated from "Pending Collection" to "Collection Complete." While maintaining independent storage for the point cloud data of each region, a coordinate association index is established between the regions.

[0049] Reference Figure 2 Preferably, step S3 includes: S31: Calculate the distance between each sampling point and its neighboring points in the point cloud of the tunnel segment to obtain the average neighborhood distance; If the average distance of the neighborhood is greater than the preset aggregation limit distance threshold, the current sampling point is determined to be an outlier noise point and is removed to obtain the point cloud of the aggregation part. S32: Perform voxel downsampling on the point cloud of the aggregation region to obtain a simplified point cloud of the region, and calculate the point cloud curvature of each sampling point; If the curvature of the point cloud is greater than the curvature threshold, the current sampling point is determined to be a positioning marker point, and the coordinates of the marker point are extracted; S33: Based on the coordinates of the marked points, simplify the point clouds of different parts and perform source matching to calculate the transformation parameters between the point clouds; S34: Based on the registration constraint parameters and the transformation parameters between point clouds, iterative registration is performed between the simplified point clouds of the parts to obtain the rotation transformation matrix. S35: Based on the rotation transformation matrix and the functional shape of the tunnel segment, the simplified point cloud of the part is transformed and spliced ​​to construct the point cloud model of the tunnel segment.

[0050] In this embodiment, statistical filtering is performed sequentially on each independent point cloud segment to remove noise points and outliers generated during the measurement process. For the currently processed point cloud segment, the average distance from each sampling point to its k nearest neighbors is calculated. This step is based on the assumption that the average neighborhood distance of all points follows a Gaussian distribution. Based on this distribution, the global mean μ and standard deviation σ are calculated, and points with an average distance greater than μ + 3σ are identified as outliers and removed.

[0051] Point clouds after outlier removal typically have high density, which is detrimental to subsequent stitching and processing. A voxel downsampling method is used to sparsify the point cloud. First, a cubic voxel grid with a side length of 1 mm is created, dividing the entire point cloud space into regular grid cells. For all points contained in each grid cell, the average of their 3D coordinates is calculated, and this average point is used as the representative point of that grid; all other points are discarded. After voxel downsampling, the data volume of each point cloud is typically reduced to about 30% of the original data volume, and the density is reduced to about one point per square millimeter, significantly improving the efficiency of subsequent computations while preserving geometric features.

[0052] Four pre-set positioning markers on the calibration platform are used as stitching references. These markers appear as four distinct circular depressions or convex features in the point cloud data, which can be automatically identified by calculating the curvature changes or reflection intensity differences of the point cloud. The three-dimensional spatial coordinates of these four marker points are sequentially searched and extracted in the point cloud of each region, serving as the spatial positioning reference for that region of the point cloud.

[0053] For each subsequent point cloud patch, the 3D coordinates of the same marker point within that patch are located using a marker point identification algorithm. Based on the correspondence between source marker points in the reference patch and the current patch, a preliminary coarse transformation parameter between the two point clouds is calculated.

[0054] Fine registration is performed using the Iterative Closest Point (ICP) algorithm. The ICP iterative process uses coarse transformation parameters as initial values, with 50 iterations and a convergence threshold of 0.01 mm. In each iteration, the correspondence between the nearest points of the reference piece and the current piece is established. By minimizing the sum of squared distances between corresponding points, the rotation matrix and translation vector are solved and updated until the convergence condition or the maximum number of iterations is reached.

[0055] Based on the final obtained rotation matrix and translation vector, all subsequent point cloud patches are transformed to the coordinate system of the reference point cloud. After all point cloud patches have their coordinates unified, they are merged into a complete point cloud dataset, and duplicate points are removed in overlapping areas, while points that are farther apart are retained to ensure the uniformity of the point cloud distribution.

[0056] The calculated actual center coordinates of the marker points are compared with the theoretical center coordinates of the marker points in the design model to solve for the rigid body transformation parameters from the spliced ​​coordinate system to the design model coordinate system. After applying this transformation, the origins of the spliced ​​segment point cloud model coordinate system and the design model coordinate system are unified, both defined at the center of the segment, providing a unified reference for subsequent deviation analysis, dimensional inspection, and other processes.

[0057] Reference Figure 3 Preferably, step S4 includes: S41: Perform uniform random sampling on the shield tunnel segment point cloud model according to the preset sampling ratio to obtain the control point set; S42: Perform statistical calculations on the control point set and the standard segment 3D model to obtain the squared signed distance, which is used as the objective function; S43: Iteratively align the shield tunnel segment point cloud model with the standard segment 3D model according to the objective function, and update the parameters of the rotation transformation matrix to obtain the optimal transformation matrix; S44: Register and transform the shield tunnel segment point cloud model according to the optimal transformation matrix to obtain the transformed segment point cloud model; S45: Perform point-by-point difference calculation between the converted segment point cloud model and the standard segment 3D model to obtain the point cloud normal deviation; S46: Analyze the point cloud model of the converted tunnel segment based on its functional parts, calculate the corresponding measured dimensions, and compare them with the theoretical dimensions to obtain multidimensional deviation values, including: The end faces of the point cloud model of the converted tunnel segment are extracted, the chord length between the end faces is calculated, and the chord length deviation is calculated by comparing it with the theoretical chord length. Curve interpolation is performed on the outer arc surface of the point cloud model of the converted tunnel segment to reconstruct the measured arc length value, which is then compared with the theoretical arc length value to calculate the arc length deviation. The distance between the outer and inner arc surfaces of the point cloud model of the converted tunnel segment is calculated, the measured thickness value is calculated, and it is compared with the theoretical thickness value to calculate the thickness deviation. The edge points of the holes in the point cloud model of the converted pipe segment are identified and fitted, the center coordinates of the measured bolt holes are calculated, and compared with the theoretical center coordinates to calculate the hole position deviation. S47: Combine the point cloud normal deviation and multidimensional deviation values ​​based on the sampling point coordinates to construct a segment deviation array.

[0058] In this embodiment, approximately 10% of the point cloud data is extracted as a control point set from the stitched complete measured point cloud dataset using a uniform random sampling strategy. The sampling process ensures that the control points spatially cover the entire geometric feature area of ​​the tunnel segment, avoiding local over-density or sparseness.

[0059] Using the central coordinate system of the design model as a reference, the spatial position and orientation of the measured point cloud are taken as variables to be optimized. A least-squares optimization algorithm is employed, with the objective function being the minimization of the sum of squared signed distances from all control points to the surface of the design model. The rotation matrix and translation vector of the measured point cloud are iteratively adjusted. In each iteration, the normal distance from each point in the control point set to the nearest triangular facet on the design model surface is calculated, the transformation parameters are updated, and the updated transformation parameters are applied to all measured point clouds. The iteration stopping condition is set as follows: the change in transformation parameters between two consecutive iterations is less than 0.001 mm, or the number of iterations reaches the upper limit of 100. When either stopping condition is met, the alignment process is completed, and the final transformation parameters and iteration convergence status are recorded.

[0060] After registration and alignment are completed, the normal deviation of each point in the measured point cloud relative to the surface of the design model is calculated independently. For the current measured point P, the triangular facet index of the design model is traversed, and the nearest triangle facet to point P is found using a spatial nearest neighbor search strategy. The perpendicular distance from point P to that facet is then calculated. This distance is defined as a signed deviation value: positive if the point is located outside the normal vector pointing to the model surface (i.e., material is redundant), and negative if the point is located inside the normal vector pointing to the model surface (i.e., material is missing).

[0061] The three-dimensional spatial coordinates of each measured point and its corresponding signed deviation value are combined into a complete record. Each record contains four fields: X coordinate, Y coordinate, Z coordinate, and deviation value. The collection of all records constitutes the deviation array of the pipe segment, which serves as the basic data for subsequent deviation visualization and statistical analysis.

[0062] Based on the distribution range of deviation values ​​in the deviation array, the deviation intervals are divided into seven consecutive levels: less than -3mm, -3mm to -2mm, -2mm to -1mm, -1mm to +1mm, +1mm to +2mm, +2mm to +3mm, and greater than +3mm. Each interval is color-coded from blue to red according to the deviation value from negative to positive, generating a pseudo-color mapping scheme for the deviation distribution.

[0063] The measured point cloud is parametrically unfolded along the arc length and width directions of the pipe segment, establishing a mapping relationship from three-dimensional spatial coordinates to two-dimensional unfolded plane coordinates. A regular pixel grid is then divided on the two-dimensional unfolded plane, and the color value of each pixel is determined by the average deviation value of all measured points falling within that pixel region. Finally, a two-dimensional unfolded deviation map is generated that intuitively reflects the deviation distribution across the entire surface of the pipe segment.

[0064] Extract the point cloud subsets corresponding to the left and right end faces of the pipe segment from the measured point cloud, respectively. Use the least squares plane fitting algorithm to fit the spatial plane equations of the two end faces respectively. Calculate the spatial distance between the fitted planes of the two end faces as the measured chord length value. Compare it with the theoretical chord length value of the design model and output the chord length deviation.

[0065] The point sequence along the longitudinal centerline of the arc surface is extracted from the point cloud of the outer arc surface. The spatial curve of the centerline of the arc surface is reconstructed by the curve interpolation algorithm. The length of the curve is calculated by numerical integration as the measured arc length value. The measured arc length value is compared with the theoretical arc length value of the design model to output the arc length deviation.

[0066] Several sampling positions are uniformly selected on the arc surface of the segment. At each sampling position, corresponding point pairs of the outer arc surface and the inner arc surface are extracted. The normal distance from the outer arc surface to the inner arc surface is calculated as the measured thickness value at that position. The measured thickness value at that sampling position is compared with the theoretical thickness value of the design model, and the thickness deviation sequence of each sampling position is output.

[0067] The boundary contour of bolt holes is identified by using the point cloud's hole edge point set. A least-squares circle fitting algorithm is then used to fit a spatial circle to the boundary point set, and the coordinates of the fitted circle's center are calculated as the measured center position of the bolt hole. The measured center coordinates are compared with the theoretical center coordinates of the corresponding bolt hole in the design model to calculate the position deviation vector and deviation distance in three-dimensional space.

[0068] The calculation results of all multidimensional deviations, such as chord length deviation, arc length deviation, thickness deviation sequence and bolt hole position deviation, are summarized and combined with the aforementioned point cloud normal deviation array to form a segment detection deviation dataset, which is then stored in the database.

[0069] Preferably, step S5 includes: Based on the segment type, the preset allowable deviation ranges are extracted and arranged to form a tolerance threshold array; The preset allowable deviation ranges include flatness deviation range of [-a,+a]mm, chord length deviation range of [-b,+b]mm, thickness deviation range of [-c,+c]mm, and hole position deviation range of [-d,+d]mm. The segment position deviation array is compared item by item according to the tolerance threshold array, and the number of abnormal segments is marked. If the number of abnormal parts of the tunnel segment is within the first quantity threshold range, the accuracy abnormality level of the current shield tunnel segment point cloud model is determined to be low, and a tunnel segment repair suggestion is generated. If the number of abnormal parts of the tunnel segment is within the second threshold range, the accuracy abnormality level of the current shield tunnel segment point cloud model is determined to be medium, and a segment repair prompt is generated. If the number of abnormal parts of the tunnel segment is within the third threshold range, the accuracy level of the current shield tunnel segment point cloud model is determined to be high, and a segment remodeling warning is generated. Based on the shield tunnel segment identification, the accuracy anomaly level and segment position deviation array are mapped to generate a pseudo-color cloud map of deviation distribution. Combined with the functional parts of the segments, a segment accuracy detection table is constructed.

[0070] In this embodiment, preset allowable deviation ranges for each detection dimension are read from a parameter table in the database. This parameter table is stored categorized by detection item: the allowable range for flatness deviation is set to [-3mm, +3mm], the allowable range for chord length deviation is set to [-1mm, +1mm], the allowable range for thickness deviation is set to [-2mm, +2mm], and the allowable range for bolt hole position deviation is set to [-2mm, +2mm]. The threshold parameters for the four dimensions are loaded into memory as the comparison benchmark for subsequent judgment logic.

[0071] Iterate through the normal deviation value of each measured point in the deviation array and compare the absolute value of the deviation point with the upper limit of the allowable flatness range. If the absolute value of the deviation at any measured point is greater than 3mm, the surface flatness of the segment is determined to be unqualified.

[0072] The difference between the calculated measured chord length and the theoretical design chord length is calculated, and the absolute value is compared with the allowable deviation range of chord length [-1mm, +1mm]. If the absolute value of the difference is greater than 1mm, the segment width (chord length) is deemed unqualified.

[0073] The measured thickness at each sampling location is compared with the theoretical thickness. The absolute value of the difference is then compared with the allowable thickness deviation range [-2mm, +2mm]. If the absolute value of the thickness difference at any sampling location is greater than 2mm, the thickness of the segment is deemed unqualified.

[0074] The spatial distance between the measured center coordinates and the theoretical center coordinates of each bolt hole is compared one by one, and the distance value is compared with the allowable deviation range of the hole position [-2mm, +2mm]. If the positional deviation of any bolt hole is greater than 2mm, the bolt hole position of that segment is deemed unqualified.

[0075] Three-dimensional point cloud data of the demolded tunnel segments is collected using structured light scanning (accuracy 0.1mm), and the positional deviation array of each measuring point is calculated by comparing it with the design model (measuring point spacing 50mm, total 80 measuring points). The deviation array is then compared item by item with the tolerance threshold array (±2mm) to count the number of abnormal parts exceeding the tolerance (total 80 measuring points). The accuracy abnormality level is determined based on the number of abnormal parts: the first threshold range is defined as ≤3 abnormal parts, which is considered a low-level abnormality, and a "segment repair suggestion" is generated to suggest local grinding or filling of the protruding points; the second threshold range is defined as 4 ≤ ≤8 abnormal parts, which is considered a medium-level abnormality, and a "segment repair suggestion" is generated to suggest overall grouting treatment of the abnormal area; the third threshold range is defined as ≥9 abnormal parts, which is considered a high-level abnormality, and a "segment remodeling warning" is generated, requiring the segment to be scrapped and recast. After the inspection is completed, the abnormal parts are marked with red and yellow highlights, generating an inspection report, which effectively improves the efficiency of quality control before the tunnel segments leave the factory.

[0076] A logical AND operation is performed based on the individual judgment results of the four dimensions: the overall judgment result is "qualified" if and only if all four dimensions, flatness, chord length, thickness, and bolt hole position, are judged as qualified; if any one dimension is unqualified, the overall judgment is "unqualified".

[0077] Reference Figure 4 A shield tunnel segment geometric accuracy detection system based on structured light scanning, applied to a shield tunnel segment geometric accuracy detection method, includes: The model extraction module is used to match and extract the standard three-dimensional model of the tunnel segment based on the segment model and mold number of the shield tunnel segment to be inspected. The local acquisition module is used to analyze the shield tunnel segment to be inspected, and, in combination with the segment geometry, divide and scan several functional parts of the segment to obtain the corresponding segment point cloud. The global processing module is used to clean and sample the point cloud of the tunnel segment, obtain a simplified point cloud of the segment, and combine it with the functional shape of the segment to splice and construct the point cloud model of the shield tunnel segment. The deviation alignment module is used to fit and align the point cloud model of the tunnel segment with the standard 3D model of the segment, calculate the multidimensional deviation value, and generate a segment location deviation array. The accuracy detection module is used to compare the segment position deviation array with the preset allowable deviation range and generate a segment accuracy detection table.

[0078] In this embodiment, the model extraction module matches and extracts the corresponding standard segment 3D model from the 3D design database based on the model of the segment to be tested (e.g., a standard ring with an outer diameter of 6.2m) and the mold number (e.g., M-2025-03). Subsequently, the local acquisition module uses a handheld structured light scanner (0.05mm accuracy) to scan the five functional parts of the segment in sequence, namely the insertion area, hand hole area, sealing groove area, arc surface area and end face area, to obtain the original point cloud data of each part, wherein the acquisition point density of each part is not less than 200 points / cm².

[0079] The global processing module performs voxel filtering downsampling (filter radius set to 0.5mm) and outlier removal on the point cloud of each location, resulting in a simplified point cloud for that location (point cloud size controlled within 50,000 points). Based on the circumferential and longitudinal geometric features of the tunnel segment, the point clouds of each location are stitched together to form a complete shield tunnel segment point cloud model. The deviation alignment module uses the ICP algorithm to perform precise registration of this point cloud model with the standard model (registration accuracy ≤0.1mm), generating multidimensional deviation values ​​(including X, Y, and X-rays) in the detection areas such as the inner arc surface, outer arc surface, end face, and sealing groove of the tunnel segment. The module generates a segment deviation array based on the Y, Z three-dimensional deviations and normal deviations. Finally, the precision inspection module compares each item against the deviation array according to the preset allowable deviation ranges—where the allowable deviation in the segment width direction is ±1.0mm, the allowable deviation in the thickness direction is ±1.5mm, the allowable deviation in the manhole position is ±2.0mm, and the allowable deviation in the sealing groove depth is ±0.5mm. It marks the out-of-tolerance areas and generates a segment precision inspection table that includes a deviation distribution heatmap and a pass / fail judgment, assisting quality inspectors in quickly determining whether the segments meet the factory standards.

[0080] Thirdly, an electronic device includes: One or more processors; Memory, used to store one or more programs; When one or more programs are executed by one or more processors, the one or more processors implement any of the methods in the above scheme.

[0081] Fourthly, a storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the shield tunnel segment geometric accuracy detection method as described above.

[0082] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for detecting the geometric accuracy of tunnel lining segments based on structured light scanning, characterized in that, include: Based on the tunnel segment model and mold number of the shield tunnel segment to be inspected, a standard three-dimensional model of the tunnel segment is matched and extracted; The tunnel segment to be inspected is analyzed, and combined with the segment geometry, several functional parts of the segment are divided and scanned to obtain the corresponding segment point cloud. Clean and sample the point cloud of the tunnel segment to obtain a simplified point cloud of the segment, and combine it with the functional shape of the segment to construct a point cloud model of the shield tunnel segment; Fit and align the point cloud model of the tunnel segment with the standard 3D model of the segment, calculate the multidimensional deviation value, and generate the segment location deviation array. A segment accuracy inspection table is generated by comparing the segment position deviation array with the preset allowable deviation range.

2. The method for detecting the geometric accuracy of tunnel lining segments based on structured light scanning according to claim 1, characterized in that, The process of matching and extracting a standard three-dimensional model of a tunnel segment based on its segment type and mold number includes: The identification codes of the tunnel segments to be inspected are parsed to extract the segment model and mold number; The model of the tunnel segment is matched and searched according to the shield tunnel segment model library, and the storage path of the model file is extracted. The shield tunnel segment model is extracted and parsed according to the model file storage path to obtain the model vertices, triangular facets and normal vectors, and three-dimensional mesh data is constructed. The 3D mesh data is correlated based on the mold number to obtain the standard segment 3D model.

3. The method for detecting the geometric accuracy of tunnel lining segments based on structured light scanning according to claim 1, characterized in that, The process involves analyzing the tunnel lining segments to be inspected, combining the segment geometry, dividing and scanning several functional areas of the segments, and obtaining the corresponding point clouds of these areas, including: Images of the tunnel segments to be inspected are acquired, resulting in several images of the tunnel segments. Boundary segmentation and optical flow registration are performed on the shield tunnel segment images to form complete segment images; The complete segment image is divided into surface regions based on the segment geometry to obtain the segment functional parts; the segment functional parts include the inner arc surface, outer arc surface, left end face, right end face, upper edge, and lower edge; The functional parts of the tunnel segment are sorted and planned to generate a scanning guide path for each part. Based on the part scanning guide path and the structured light scanning distance, point cloud scanning is performed on the functional parts of the pipe segment to generate an initial part point cloud and calculate the regional point cloud coverage. If the point cloud coverage rate of the area is less than the area coverage threshold, it is determined that the current functional part of the pipe segment is incomplete in the scan, and a prompt to rescan is generated. If the area point cloud coverage is greater than or equal to the area coverage threshold, it is determined that the current segment functional part has been scanned completely, and the initial part point cloud is used as the corresponding segment part point cloud.

4. The method for detecting the geometric accuracy of tunnel lining segments based on structured light scanning according to claim 1, characterized in that, The point cloud data of the cleaned and sampled tunnel segments is used to obtain simplified point clouds of the segments. These simplified point clouds are then combined with the functional morphology of the tunnel segments to construct a point cloud model of the shield tunnel segments, including: The distance between each sampling point and its neighboring points in the point cloud of the tunnel segment is calculated to obtain the average neighborhood distance. If the average distance of the neighborhood is greater than the preset aggregation limit distance threshold, the current sampling point is determined to be an outlier noise point and is removed to obtain the point cloud of the aggregation part. Voxel downsampling is performed on the point cloud of the aggregation region to obtain a simplified point cloud of the region, and the point cloud curvature of each sampling point is calculated. If the curvature of the point cloud is greater than the curvature threshold, the current sampling point is determined to be a positioning marker point, and the coordinates of the marker point are extracted; Based on the coordinates of the marked points, the point clouds of different parts are simplified and homogeneous matching is performed to calculate the transformation parameters between the point clouds. Based on the registration constraint parameters and the transformation parameters between point clouds, iterative registration is performed between the simplified point clouds of the parts to obtain the rotation transformation matrix. Based on the rotation transformation matrix and the functional morphology of the tunnel segments, the simplified point cloud of each part is transformed and spliced ​​to construct a point cloud model of the tunnel segment.

5. The method for detecting the geometric accuracy of tunnel lining segments based on structured light scanning according to claim 1, characterized in that, The fitting and alignment of the shield tunnel segment point cloud model with the standard segment 3D model is performed, multidimensional deviation values ​​are calculated, and a segment location deviation array is generated, including: The shield tunnel segment point cloud model is uniformly and randomly sampled according to the preset sampling ratio to form a control point set; Statistical calculations were performed on the control point set and the standard segment 3D model to obtain the squared signed distance, which was used as the objective function. The point cloud model of the tunnel segment and the 3D model of the standard segment are iteratively aligned according to the objective function, and the parameters of the rotation transformation matrix are updated to obtain the optimal transformation matrix. The shield tunnel segment point cloud model is registered and transformed according to the optimal transformation matrix to obtain the transformed segment point cloud model. The point cloud model of the converted tunnel segment and the standard tunnel segment 3D model are interpolated point by point to obtain the point cloud normal deviation. The point cloud model of the converted segment is analyzed based on the functional parts of the segment, the corresponding measured size values ​​are calculated, and compared with the theoretical size values ​​to obtain multidimensional deviation values. Based on the coordinates of the sampling points, the point cloud normal deviation and multidimensional deviation values ​​are combined to construct a segment deviation array.

6. The method for detecting the geometric accuracy of tunnel lining segments based on structured light scanning according to claim 5, characterized in that, The process involves analyzing the point cloud model of the converted tunnel segment based on its functional parts, calculating the corresponding measured dimensions, and comparing these dimensions with theoretical dimensions to obtain multidimensional deviation values, including: The end faces of the point cloud model of the converted tunnel segment are extracted, the chord length between the end faces is calculated, and the chord length deviation is calculated by comparing it with the theoretical chord length. Curve interpolation is performed on the outer arc surface of the point cloud model of the converted tunnel segment to reconstruct the measured arc length value, which is then compared with the theoretical arc length value to calculate the arc length deviation. The distance between the outer and inner arc surfaces of the point cloud model of the converted tunnel segment is calculated, the measured thickness value is calculated, and it is compared with the theoretical thickness value to calculate the thickness deviation. The edge points of the holes in the point cloud model of the converted pipe segment are identified and fitted. The coordinates of the measured bolt hole center are calculated and compared with the theoretical center coordinates to calculate the hole position deviation.

7. The shield tunnel segment geometric accuracy detection system based on structured light scanning according to claim 1, characterized in that: The preset allowable deviation ranges include a flatness deviation allowable range of [-a, +a] mm, a chord length deviation allowable range of [-b, +b] mm, a thickness deviation allowable range of [-c, +c] mm, and a hole position deviation allowable range of [-d, +d] mm.

8. The shield tunnel segment geometric accuracy detection system based on structured light scanning according to claim 7, characterized in that: The step of generating a segment accuracy detection table by comparing the segment position deviation array with a preset allowable deviation range includes: Based on the segment type, the preset allowable deviation ranges are extracted and arranged to form a tolerance threshold array; The segment position deviation array is compared item by item according to the tolerance threshold array, and the number of abnormal segments is marked. If the number of abnormal parts of the tunnel segment is within the first quantity threshold range, the accuracy abnormality level of the current shield tunnel segment point cloud model is determined to be low, and a tunnel segment repair suggestion is generated. If the number of abnormal parts of the tunnel segment is within the second threshold range, the accuracy abnormality level of the current shield tunnel segment point cloud model is determined to be medium, and a segment repair prompt is generated. If the number of abnormal parts of the tunnel segment is within the third threshold range, the accuracy level of the current shield tunnel segment point cloud model is determined to be high, and a segment remodeling warning is generated. Based on the shield tunnel segment identification, the accuracy anomaly level and segment position deviation array are mapped to generate a pseudo-color cloud map of deviation distribution. Combined with the functional parts of the segments, a segment accuracy detection table is constructed.

9. A shield tunnel segment geometric accuracy detection system based on structured light scanning, used to implement the shield tunnel segment geometric accuracy detection method as described in any one of claims 1-8, characterized in that, include: The model extraction module is used to match and extract the standard three-dimensional model of the tunnel segment based on the segment model and mold number of the shield tunnel segment to be inspected. The local acquisition module is used to analyze the shield tunnel segment to be inspected, and, in combination with the segment geometry, divide and scan several functional parts of the segment to obtain the corresponding segment point cloud. The global processing module is used to clean and sample the point cloud of the tunnel segment, obtain a simplified point cloud of the segment, and combine it with the functional shape of the segment to splice and construct the point cloud model of the shield tunnel segment. The deviation alignment module is used to fit and align the point cloud model of the tunnel segment with the standard 3D model of the segment, calculate the multidimensional deviation value, and generate a segment location deviation array. The accuracy detection module is used to compare the segment position deviation array with the preset allowable deviation range and generate a segment accuracy detection table.

10. A storage medium storing at least one instruction, at least one program, a code set, or an instruction set, wherein the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the shield tunnel segment geometric accuracy detection method as described in any one of claims 1 to 8.

Images