A turnout pre-assembly measurement design method and system based on a turnout global point cloud

Abstract

The application discloses a kind of based on switch area global point cloud's switch preassembly measurement design method and system, method includes: using three-dimensional scanning equipment to collect switch global original point cloud, effective point cloud is obtained after filtering denoising processing, according to elevation feature segmentation extraction sleeper and fastener region of interest point cloud.Based on fastener geometric feature segmentation bolt end surface point cloud, determine bolt center point by surface fitting, obtain initial coordinates under sensor coordinate system.Select reference bolt to establish engineering reference coordinate system, build coordinate conversion model conversion point coordinate, according to bolt coordinate solving sleeper actual spacing and cumulative spacing.Based on calibrated point cloud extraction track feature, measure and calculate gauge, high, triangular pit and other geometric parameters, measured data contrast design standard, generate switch preassembly adjustment instruction and laying scheme.Provide reliable technical support for railway line safe operation.

Description

Technical Field

[0001] This invention belongs to the field of turnout pre-assembly technology, and more specifically, relates to a turnout pre-assembly measurement and design method and system based on the global point cloud of the turnout area. Background Technology

[0002] As a core component of railway lines, turnouts directly determine the efficiency and safety of train passage. Currently, railway construction generally adopts a factory pre-assembly and on-site laying operation mode. Dimensional inspection during the pre-assembly stage is a crucial link in controlling the quality of turnouts before they leave the factory. The domestic turnout manufacturing industry has long relied on traditional manual measurement methods, with workers using traditional measuring tools such as steel tape measures, levels, and optical levels to collect various parameters. Turnout sections have complex structures, with frogs, rails, sleepers, and various fasteners interwoven, resulting in numerous measuring points. Manually measuring each point is time-consuming and labor-intensive. Furthermore, prolonged fieldwork is susceptible to reading deviations due to human error and environmental factors, leading to insufficient stability of measured data and making it difficult to meet the precision control standards of high-speed railways.

[0003] With the continuous development of 3D laser scanning technology, point cloud measurement is gradually being applied to the rail transit field. However, most existing track point cloud detection technologies are geared towards the inspection of defects on existing operating lines, and there is a lack of dedicated measurement technologies adapted to the pre-assembly scenarios of turnout factories. Existing general point cloud processing algorithms lack targeted identification logic for turnout bolts and fasteners, and cannot automatically extract key positioning points based on point cloud geometric features. Feature point selection still relies on manual interactive annotation, limiting automated detection capabilities. Conventional point cloud coordinate registration relies solely on point cloud morphology to complete matching calculations, without considering the inherent engineering constraints such as track alignment and gauge in the track design. When collecting data over long distances along the turnout area, it is prone to cumulative deformation errors, leading to systematic deviations in the calculation of basic parameters such as sleeper spacing.

[0004] Insufficient dimensional accuracy in turnout inspection directly leads to the inability to locate out-of-limit defects during the pre-assembly stage. Various assembly deviations are then carried over to the on-site laying process, forcing construction units to invest significant manpower in repeated disassembly and adjustments, substantially increasing construction time and production costs. Against the backdrop of my country's continuously expanding railway network and the ongoing upgrading of high-speed rail construction standards, turnout prefabrication production is developing towards high precision and intelligentization. Traditional manual inspection and generalized point cloud inspection solutions are no longer adequate for the industry's development needs. Based on the aforementioned pain points in the industry, there is an urgent need to innovate measurement methods, relying on the full-area point cloud of the turnout area to achieve fully automated measurement throughout the entire process. This will allow for the analysis of assembly deviations and the intelligent generation of adjustment plans, optimizing the quality of turnout prefabrication from the source, reducing subsequent on-site construction and rectification costs, and solidifying the foundation for safe railway line operation. Summary of the Invention

[0005] This invention aims to overcome the industry problems of low efficiency and large error in traditional manual measurement of turnouts, insufficient automation of conventional point cloud detection, and prominent cumulative error in long-distance measurement. It relies on full-domain three-dimensional point cloud to realize automatic identification and coordinate transformation of turnout components, automatically calculate sleeper spacing and track geometric parameters, and generate assembly adjustment schemes by comparing with design standards, thereby improving the intelligence and measurement accuracy of turnout pre-assembly inspection.

[0006] To address the aforementioned deficiencies or improvement needs of existing technologies, as a first aspect of this invention, the present invention provides a turnout pre-assembly measurement and design method based on a global point cloud of the turnout area, comprising: S1. Control the 3D scanning equipment to move along the turnout area to be measured and continuously collect the original 3D point cloud data of the entire turnout area; perform filtering and noise reduction preprocessing on the original 3D point cloud data to obtain effective point cloud data; based on the elevation undulation characteristics of the effective point cloud data in the vertical direction, segment and extract the region of interest point cloud where the sleeper and fastener are located. S2. In the point cloud of the region of interest, based on the local geometric features of the fastener region shifting to both sides, identify and segment the bolt end face point cloud set; perform three-dimensional geometric surface fitting on the bolt end face point cloud set to calculate the geometric center point of each bolt end face; use the geometric center point as the bolt center to obtain the initial three-dimensional coordinates of each bolt center in the sensor coordinate system; S3. Select a preset reference bolt from the three-dimensional point cloud data of the entire turnout area, and establish a track engineering reference coordinate system in combination with the track design line parameters; construct a spatial transformation model between the sensor coordinate system and the track engineering reference coordinate system, and convert the initial three-dimensional coordinates of each bolt center into actual physical coordinates; based on the difference in actual physical coordinates of the corresponding bolt centers on adjacent sleepers, calculate the actual spacing and cumulative spacing of each sleeper. S4. Using the coordinate-transformed three-dimensional point cloud data of the turnout area, extract the spatial position features of the rails and sleepers, and calculate the track gauge, level, elevation, track direction, and triangular pit geometric parameters of the turnout area; compare and analyze the actual spacing and cumulative spacing of the sleepers and the turnout geometric parameters with the turnout design standards to generate turnout pre-assembly adjustment instructions and laying guidance schemes.

[0007] Further, in S1, controlling the three-dimensional scanning device to move along the area of ​​the turnout to be measured, and continuously acquiring the original three-dimensional point cloud data of the entire turnout area, including: An coded odometer is integrated into the three-dimensional scanning device to simultaneously collect mileage data and linear structure light cross-section data as the device moves along the rail. Based on the mileage data, pose compensation and multi-frame stitching are performed on the continuously acquired line structured light cross-section data to obtain the original three-dimensional point cloud data. The preprocessing of filtering and denoising the original three-dimensional point cloud data includes: using a statistical filtering algorithm to remove outlier noise caused by the switch area environment, and using a radius filtering algorithm to remove local dense noise caused by dust and water vapor, to obtain the effective point cloud data.

[0008] Further, in step S1, based on the vertical elevation undulation features of the effective point cloud data, the region of interest (ROI) point cloud containing the sleepers and fasteners is segmented and extracted, including: The effective point cloud data is projected onto a horizontal plane to construct an elevation histogram. Based on the peak distribution characteristics in the elevation histogram, determine the elevation threshold ranges for the top surface of the rail, the top surface of the sleeper / fastener, and the surface of the ballast. The effective point cloud data is initially segmented based on the elevation threshold range corresponding to the top surface of the sleeper / fastener to obtain candidate region point clouds; Perform three-dimensional connected component analysis on the candidate region point cloud, and remove isolated point cloud clusters with a volume smaller than a preset threshold to obtain the region of interest point cloud.

[0009] Further, in step S2, based on the local geometric features of the fastener region shifting to both sides in the point cloud of the region of interest, a set of bolt end face point clouds is identified and segmented, including: Local elevation normalization is performed on the point cloud of the region of interest to eliminate the elevation benchmark deviation caused by the tilt of the sleeper top surface and the laying slope, and a relative elevation point cloud is obtained. Based on the density distribution and connected component characteristics of the relative elevation point cloud, the largest connected component with a volume greater than a preset volume threshold is extracted and used as the main body area of ​​the fastener pad. Obtain the three-dimensional bounding box of the main area of ​​the fastener pad in the longitudinal and transverse directions of the sleeper. Based on the prior assembly ratio of the turnout fastener structure, expand the three-dimensional bounding box outward by a preset ratio distance along the transverse direction of the sleeper to both sides to construct an adaptive bolt search window. Within the adaptive bolt search window, point clouds with relative elevation values ​​greater than a preset elevation threshold are extracted as candidate bolt top point clouds; The point cloud at the top of the candidate bolt is subjected to region growing clustering based on normal vector consistency to obtain several independent point cloud clusters; the three-dimensional spatial compactness of each independent point cloud cluster is calculated, and the independent point cloud clusters with a three-dimensional spatial compactness greater than a preset confidence threshold are determined as the bolt end face point cloud set.

[0010] Further, in step S2, a three-dimensional geometric surface fitting is performed on the bolt end face point cloud set to calculate the geometric center point of each bolt end face, including: Determine the contour features of the point cloud set on the bolt end face; If the contour feature is circular or cylindrical, the Random Sample Consensus (RANSAC) algorithm is used to fit the cylindrical surface, and the intersection of the central axis of the fitted cylinder and the bolt end face is taken as the geometric center point. If the contour feature is polygonal, then polygon vertex extraction and least squares plane fitting are performed, and the geometric centroid of the fitted plane is taken as the geometric center point. By fitting the cylindrical surface or the polygonal plane, the interference of oil stains, rust, or local defects on the bolt surface on the center positioning is eliminated.

[0011] Furthermore, in step S3, a spatial transformation model is constructed between the sensor coordinate system and the track engineering reference coordinate system to convert the initial three-dimensional coordinates of each bolt center into actual physical coordinates, including: Multiple reference bolts are selected at equal intervals along the longitudinal direction of the turnout to obtain their initial three-dimensional coordinates in the sensor coordinate system and their theoretical physical coordinates in the track design drawings. Based on the initial three-dimensional coordinates and theoretical physical coordinates, the initial global rigid body transformation matrix is ​​solved using the singular value decomposition (SVD) algorithm, and the three-dimensional point cloud data of the bifurcation region is initially aligned. By introducing turnout design alignment parameters, a registration objective function integrating track geometric constraints is constructed. The track geometric constraints include design gauge constraints, longitudinal alignment smoothing constraints, and adjacent sleeper spacing constraints. Using the pre-aligned point cloud data as input, the registration objective function is solved by the constrained iterative nearest point (ICP) algorithm to obtain the nonlinear deformation compensation matrix; By using the cascaded matrix of the initial global rigid body transformation matrix and the nonlinear deformation compensation matrix, the initial three-dimensional coordinates of all bolt centers are uniformly transformed to the track engineering reference coordinate system to obtain the actual physical coordinates that eliminate long-distance cumulative errors.

[0012] Further, in step S3, based on the actual physical coordinate difference between the centers of corresponding bolts on adjacent sleepers, the actual spacing and cumulative spacing of each sleeper are calculated, including: Extract the actual physical coordinates of the centers of the corresponding bolts on the same side of two adjacent sleepers in the longitudinal direction of the track; The difference between two adjacent actual physical coordinate values ​​is calculated to obtain the actual spacing of a single sleeper; Starting from the sleeper where the preset reference bolt is located, the actual spacing of each sleeper is accumulated sequentially along the longitudinal direction of the track to obtain the cumulative spacing of each sleeper relative to the starting point, so as to eliminate the cumulative error caused by measuring each sleeper individually.

[0013] Furthermore, in step S4, the actual spacing and cumulative spacing of the sleepers, as well as the geometric parameters of the turnout, are compared and analyzed with the turnout design standards to generate turnout pre-assembly adjustment instructions and laying guidance schemes, including: The actual spacing, cumulative spacing, and turnout geometric parameters are compared with the preset turnout design standard tolerance range to identify the out-of-limit parameters and their corresponding out-of-limit locations. Based on the aforementioned out-of-limit parameters and out-of-limit locations, calculate the required adjustment amounts for the sleepers or rails in the lateral, longitudinal, and vertical directions. Based on the adjustment amount, generate the turnout pre-assembly adjustment instruction containing the adjustment position, adjustment direction and adjustment value, and generate the laying guidance scheme containing the overall turnout alignment deviation distribution diagram.

[0014] As a second aspect of the present invention, a turnout pre-assembly measurement and design system based on the global point cloud of the turnout area is also provided, comprising: The point cloud acquisition and region extraction unit is used to control the 3D scanning equipment to move along the turnout area to be measured and continuously acquire the original 3D point cloud data of the entire turnout area; perform filtering and noise reduction preprocessing on the original 3D point cloud data to obtain effective point cloud data; and segment and extract the region of interest where the sleepers and fasteners are located based on the vertical elevation undulation characteristics of the effective point cloud data. The bolt segmentation and center calculation unit is used to identify and segment the bolt end face point cloud set in the point cloud of the region of interest based on the local geometric morphology features of the fastener area shifting to both sides; perform three-dimensional geometric surface fitting on the bolt end face point cloud set to calculate the geometric center point of each bolt end face; and use the geometric center point as the bolt center to obtain the initial three-dimensional coordinates of each bolt center in the sensor coordinate system. The coordinate transformation and spacing calculation unit is used to select a preset reference bolt in the three-dimensional point cloud data of the entire turnout area, and establish a track engineering reference coordinate system in combination with the track design line parameters; construct a spatial transformation model between the sensor coordinate system and the track engineering reference coordinate system, and convert the initial three-dimensional coordinates of each bolt center into actual physical coordinates; and calculate the actual spacing and cumulative spacing of each sleeper based on the difference in actual physical coordinates of the corresponding bolt centers on adjacent sleepers. The parameter calculation and scheme generation unit is used to extract the spatial position features of the rails and sleepers using the three-dimensional point cloud data of the turnout area after coordinate transformation, and calculate the track gauge, level, elevation, track direction and triangular pit geometric parameters of the turnout area; compare and analyze the actual spacing and cumulative spacing of the sleepers and the turnout geometric parameters with the turnout design standards, and generate turnout pre-assembly adjustment instructions and laying guidance schemes.

[0015] As a third aspect of the invention, a computer-readable storage medium is also provided, on which a computer program is stored, which is executed by a processor as described in any one of the inventions: a turnout pre-assembly measurement and design method based on a global point cloud of the turnout area.

[0016] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: 1. The present invention relates to a turnout pre-assembly measurement and design method based on a full-area point cloud of the turnout area. This method continuously collects raw point clouds along the turnout line using a 3D scanning device equipped with an coded odometer. Statistical filtering and radius filtering are used to remove noise of different causes in layers to generate effective point clouds. Then, elevation threshold intervals are divided according to the vertical elevation distribution of the point clouds, and the corresponding regions of interest (ROIs) for sleepers and fasteners are extracted using 3D connected domain filtering. This entire data preprocessing process can completely collect spatial data across the entire turnout area, avoiding the problem of missing local data that occurs in traditional manual point selection measurements. Layered filtering can specifically eliminate invalid noise points caused by environmental flypoints, dust, and moisture. Defining the target area based on elevation features can filter out irrelevant point data such as ballast, simplifying the data volume for subsequent calculations and reducing interference from irrelevant features on subsequent bolt identification. This ensures the validity of input data from the source, reduces the computational burden caused by redundant data, and relies on the continuous point cloud to completely replicate the actual spatial morphology of the turnout, ensuring the reliability of the data foundation for subsequent geometric parameter calculations and steadily improving the completeness of the original measurement data.

[0017] 2. The present invention provides a turnout pre-assembly measurement and design method based on the global point cloud of the turnout area. This method eliminates elevation deviations caused by sleeper slopes by performing local elevation normalization on the point cloud of the region of interest. An adaptive bolt retrieval window is constructed based on the fastener boundary dimensions and assembly ratio. Bolt end-face point clouds are screened using elevation thresholds, normal vector clustering, and 3D tightness verification. The bolt center is determined by selecting either cylindrical or planar fitting methods based on the end-face contour, resulting in the 3D coordinates of the bolt in the sensor coordinate system. This process automatically screens bolt locations based on the turnout's structural features, eliminating the need for manual point selection. Elevation normalization offsets judgment errors caused by sleeper laying slopes. The adaptive search window constrains the retrieval range, and the tightness index filters out interference points formed by gravel and stains. Differential fitting algorithms mitigate positioning deviations caused by bolt corrosion and surface defects, extract key control point coordinates, and avoid subjective measurement errors caused by manual point marking.

[0018] 3. The present invention provides a turnout pre-assembly measurement and design method based on the full-domain point cloud of the turnout area. This method establishes an engineering reference coordinate system by selecting reference bolts and combining them with track design parameters. Coordinate registration is then completed using SVD rigid body transformation and ICP algorithm with engineering constraints. Bolt coordinates are converted to actual engineering coordinates, and sleeper spacing is calculated. The calibrated full-domain point cloud is used to calculate track geometric parameters and compare them with design tolerances to generate an assembly adjustment scheme. The two-level coordinate transformation integrates three types of inherent track constraints: gauge, alignment, and sleeper spacing. This improves upon conventional registration methods by mitigating long-distance data accumulation deformation errors. The converted coordinates enable automated calculation of sleeper spacing. Batch extraction of track features from the full point cloud allows for simultaneous calculation of multiple geometric parameters. Automatic comparison of measured data with standard parameters identifies out-of-limit locations and adjustment amounts. Quantitative data guides turnout pre-assembly rectification, reducing the workload of manual item-by-item verification and improving the standardization of turnout pre-assembly acceptance and adjustment. Attached Figure Description

[0019] Figure 1 This is a flowchart of a turnout pre-assembly measurement and design method based on the global point cloud of the turnout area, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of a three-dimensional scanning device according to an embodiment of the present invention; Figure 3 This is a schematic diagram of standard hexagonal fitting according to an embodiment of the present invention; Figure 4 This is a schematic diagram of non-standard hexagonal fitting according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the system units in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0021] Example 1 Please refer to Figure 1 This embodiment 1 provides a turnout pre-assembly measurement and design method based on the global point cloud of the turnout area, including: S1. Control the 3D scanning equipment to move along the turnout area to be measured and continuously collect the original 3D point cloud data of the entire turnout area; perform filtering and noise reduction preprocessing on the original 3D point cloud data to obtain effective point cloud data; based on the elevation undulation characteristics of the effective point cloud data in the vertical direction, segment and extract the region of interest point cloud where the sleeper and fastener are located. S2. In the point cloud of the region of interest, based on the local geometric features of the fastener region shifting to both sides, identify and segment the bolt end face point cloud set; perform three-dimensional geometric surface fitting on the bolt end face point cloud set to calculate the geometric center point of each bolt end face; use the geometric center point as the bolt center to obtain the initial three-dimensional coordinates of each bolt center in the sensor coordinate system; S3. Select a preset reference bolt from the three-dimensional point cloud data of the entire turnout area, and establish a track engineering reference coordinate system in combination with the track design line parameters; construct a spatial transformation model between the sensor coordinate system and the track engineering reference coordinate system, and convert the initial three-dimensional coordinates of each bolt center into actual physical coordinates; based on the difference in actual physical coordinates of the corresponding bolt centers on adjacent sleepers, calculate the actual spacing and cumulative spacing of each sleeper. S4. Using the coordinate-transformed three-dimensional point cloud data of the turnout area, extract the spatial position features of the rails and sleepers, and calculate the track gauge, level, elevation, track direction, and triangular pit geometric parameters of the turnout area; compare and analyze the actual spacing and cumulative spacing of the sleepers and the turnout geometric parameters with the turnout design standards to generate turnout pre-assembly adjustment instructions and laying guidance schemes.

[0022] This embodiment 1 further elaborates on the above steps.

[0023] (1) Point cloud acquisition and region extraction The structure of a turnout is complex, with various components arranged in an interwoven manner. Traditional measurement methods can only collect scattered data from local points, making it difficult to fully reconstruct the overall spatial morphology of the turnout area. Please refer to... Figure 2 This is a schematic diagram of the mobile 3D scanning device used for full-area point cloud acquisition in the turnout area according to an embodiment of the present invention. The device is integrated into a dedicated inspection trolley platform that can travel along the rails. The top of the device is equipped with a GPS antenna and a GNSS board to obtain initial global positioning values, and the timing of each sensor is unified through a synchronization board. The inertial navigation (IMU) compensates for device attitude disturbances in real time to improve the geometric continuity of the point cloud. The core measurement module uses a line structured light sensor, which, together with an industrial camera, achieves millimeter-level resolution 3D shape capture. The encoder, as an odometer unit, provides a high-precision displacement reference for point cloud stitching. Three embedded hosts are deployed in a distributed manner, respectively undertaking edge computing tasks such as point cloud preprocessing, automatic bolt identification, and coordinate transformation calculation, forming a "acquisition-processing-feedback" closed loop. All modules are firmly installed on the trolley crossbeam with rigid brackets to ensure system calibration consistency and operational stability. It is suitable for full-area high-precision, automated measurement of turnout components in factory pre-assembly lines or on-site tooling environments.

[0024] Based on this, the 3D scanning equipment is first controlled to move along the area of ​​the turnout to be measured, and the original 3D point cloud data of the entire turnout area is continuously collected. The specific process of collecting the original 3D point cloud data is as follows: First, a line structured light sensor is mounted on a mobile carrier, and an coded odometer is integrated onto the carrier. As the carrier moves along the rails in the area of ​​the turnout to be measured, the line structured light sensor continuously collects line structured light cross-sectional data of the track at a preset frequency, while the coded odometer simultaneously records the carrier's real-time mileage and pose information. Then, a global three-dimensional coordinate system is established with the starting point of data acquisition as the origin. Using the mileage and pose data provided by the coded odometer, the spatial transformation matrix of each frame of line structured light cross-sectional data in the global coordinate system is calculated. This transformation matrix is ​​used to compensate for the pose of the continuous cross-sectional data, eliminating bumps and yaw errors during the carrier's movement. The compensated multi-frame cross-sectional data are then spatially stitched together to generate original three-dimensional point cloud data containing complete turnout geometry information.

[0025] After acquiring the original 3D point cloud data, the data is preprocessed by filtering and denoising to obtain valid point cloud data. First, a statistical filtering algorithm is used to handle outlier noise caused by the switch area environment. Specifically, the process involves calculating the distance from each data point in the point cloud to its neighboring points. The average distance between the nearest neighbors is calculated, and the mean and standard deviation of the global point cloud average distance are also calculated. By setting a standard deviation multiple threshold, points whose average distance exceeds the range of the mean plus the standard deviation multiple threshold are identified as outliers and removed. Subsequently, a radius filtering algorithm is used to process local dense noise caused by dust and water vapor. The specific process is as follows: a search radius and a minimum neighbor number threshold are set, each point in the point cloud is traversed, and the number of neighbor points within the sphere of the search radius is counted. If the number of neighbor points is less than the minimum neighbor number threshold, the point is identified as a local isolated noise point and removed. The final set of points constitutes the valid point cloud data.

[0026] Subsequently, based on the vertical elevation undulation characteristics of the effective point cloud data, the region of interest (ROI) point cloud containing the sleepers and fasteners is segmented and extracted. The specific process for segmenting and extracting the ROI point cloud containing the sleepers and fasteners is as follows: First, all points in the valid point cloud data The axis coordinate values ​​are extracted, projected onto the horizontal plane, and adjusted according to the preset elevation resolution. Axial coordinates are used to perform interval statistics to construct a one-dimensional elevation histogram. Due to the significant physical height differences in the vertical direction between the rails, sleepers / fasteners, and ballast in the turnout structure, the elevation histogram will exhibit three distinct peaks. Based on this, a peak-finding algorithm is used to identify the peak positions of these three peaks. Then, using each peak as a center and combining the variance of the point cloud distribution, elevation threshold intervals are defined for the top surfaces of the rails, sleepers / fasteners, and ballast.

[0027] Then, extract the elevation threshold range corresponding to the top surface of the sleeper / fastener, traverse the effective point cloud data, and retain only the relevant data. Points whose axis coordinates fall within this interval are used to complete the initial segmentation and obtain the candidate region point cloud.

[0028] Subsequently, three-dimensional connected component analysis was performed on the candidate region point clouds obtained from the initial segmentation to remove interference. A region growing clustering algorithm based on Euclidean distance was used to group the candidate region point clouds. A distance threshold was set, and adjacent points with a spatial distance less than the threshold were grouped into the same point cloud cluster, thus dividing the candidate region point clouds into multiple independent three-dimensional connected components. For each independent point cloud cluster, the volume enclosed in three-dimensional space or the total number of points it contains was calculated. A preset volume threshold was set based on the actual physical dimensions of the sleepers and fasteners. All point cloud clusters were traversed, and those with a volume or number of points less than the preset threshold were identified as isolated point cloud clusters formed by ballast splashes, oil stains, or residual noise, and were removed. Point cloud clusters with a volume greater than or equal to the preset threshold were retained, thus obtaining the final region of interest point cloud containing the sleepers and fasteners.

[0029] (2) Bolt division and center calculation After obtaining the final region of interest (ROI) point cloud containing the sleeper and fastener, the bolt end face point cloud set is identified and segmented from the ROI point cloud based on the local geometric features of the fastener area shifting to both sides. In this embodiment, the specific process of identifying and segmenting the bolt end face point cloud set is as follows: First, the point cloud of the region of interest is subjected to local elevation normalization to eliminate elevation benchmark deviations caused by the inclination of the sleeper top surface and the laying slope. Let the point cloud of the region of interest be... ,for any of the first Points Its three-dimensional coordinates are represented as The least squares method is used to... Plane fitting is performed on the bottom point cloud to obtain the plane equation of the local reference plane for the sleeper. ,in , , These are the three components of the plane normal vector. Let be the plane intercept, and satisfy . Based on this plane equation, calculate the point. The vertical distance to the local reference plane of the sleeper is used as the relative elevation value. The calculation formula is: Traversal All points in the original data, the original data of each point. Coordinates replaced with relative elevation values This yields a relative elevation point cloud that eliminates elevation datum deviation. .

[0030] Secondly, based on the density distribution and connected component features of the relative elevation point cloud, the main body region of the fastener pad is extracted. A point cloud neighborhood search radius is set. and a minimum neighborhood threshold .for Any point in the radius, if it lies within the radius The number of neighboring points within is greater than or equal to If any of these points are connected, they are marked as core points. Euclidean clustering is then used to aggregate all interconnected core points and their neighbors into several independent connected domains. The volume of each independent connected domain in three-dimensional space is calculated and denoted as the connected domain volume. Set a preset volume threshold. In all Select from independent connected components The largest connected region is used as the main area of ​​the fastener pad. .

[0031] Subsequently, the three-dimensional bounding boxes of the main area of ​​the fastener pad in the longitudinal and transverse directions of the sleeper are obtained, and an adaptive bolt search window is constructed. Calculation The smallest axis-aligned bounding box in three-dimensional space is denoted as the three-dimensional bounding box. .extract The width of the boundary along the transverse direction of the sleeper is denoted as the transverse width of the sleeper pad. Simultaneously extract The length of the boundary along the longitudinal direction of the sleeper is denoted as the longitudinal length of the pad. Introducing a priori assembly ratio coefficient for turnout fastener structure. ,Will The track extends laterally outwards to both sides, with the distance of the outward expansion on one side being... Thus, the width is updated to Length remains as Adaptive bolt search window .

[0032] Next, the point cloud of the candidate bolt tops is extracted within the adaptive bolt search window. A preset elevation threshold is set to distinguish the bolt tops from the surface of the pad. Traversing the locations Select relative elevation values ​​from all points within the spatial range. satisfy The points are defined as the top point cloud of candidate bolts. .

[0033] Finally, clustering and compactness verification are performed on the point clouds at the top of the candidate bolts to determine the final set of bolt end face point clouds. (Calculation) The local surface normal vector of each point in the vector is denoted as . Set a threshold for the angle between the normal vectors. The region growing algorithm is used, with the angle between the normal vectors being less than 1. For growth conditions, Clustered into several independent point cloud clusters, denoted as the first... Each independent point cloud cluster is To remove irregularly shaped interfering point clouds (such as ballast or oil stains), each... Three-dimensional spatial compactness Specifically, calculation 3D convex hull volume as well as Minimum enclosing sphere volume The formula for calculating the compactness of three-dimensional space is: Set a preset reliability threshold. To satisfy all Independent point cloud clusters Retain and merge them to determine the final bolt end face point cloud set. .

[0034] Please refer to Figure 3 as well as Figure 4 After identifying and segmenting the bolt end face point cloud set, a three-dimensional geometric surface is fitted to the final bolt end face point cloud set to calculate the geometric center point of each bolt end face. Subsequently, the geometric center point is used as the bolt center to obtain the initial three-dimensional coordinates of each bolt center in the sensor coordinate system.

[0035] The specific process of 3D geometric surface fitting is as follows: First, the contour features of the bolt end face point cloud set are determined. The bolt end face point cloud set is projected onto its local fitting plane to obtain a 2D projection point set. The 2D convex hull of the 2D projection point set is calculated, and the area and perimeter of the convex hull are obtained. Based on the area and perimeter, a roundness index is calculated. The calculation formula is: roundness equals four times pi multiplied by the area divided by the square of the perimeter. A roundness judgment threshold is set. If the calculated roundness index is greater than or equal to the threshold, the contour features of the bolt end face point cloud set are determined to be circular or cylindrical; if the roundness index is less than the threshold, a polygon approximation algorithm is used to fit the projection point set. If the polygon fitting residual is less than a preset residual threshold, the contour features are determined to be polygonal.

[0036] When the contour feature is determined to be circular or cylindrical, the Random Sample Consensus (RANSAC) algorithm is used for cylindrical surface fitting. During the iteration process, a minimum set of points is randomly selected from the bolt end face point cloud set each time to calculate the cylindrical surface model parameters, including the direction vector of the cylinder's central axis, the coordinates of a point on the axis, and the cylinder radius. The distance from all points in the point cloud set to the cylindrical surface is calculated, and points with a distance less than a preset distance threshold are marked as inliers, and the total number of inliers is counted. After multiple iterations, the cylindrical surface model with the largest number of inliers is selected as the optimally fitted cylindrical surface. Subsequently, the equation of the central axis of the optimally fitted cylindrical surface is extracted, and the average elevation plane of the bolt end face point cloud set along the axis direction is calculated. The three-dimensional intersection of the central axis and this average elevation plane is taken as the geometric center point of the bolt end face.

[0037] When the contour features are determined to be polygonal, polygon vertex extraction and least-squares plane fitting are performed. First, the bolt end face point cloud set is projected onto a two-dimensional plane, and the polygon vertex sequence of the two-dimensional projection point set is extracted using a polygon approximation algorithm. Next, least-squares plane fitting is performed on the original bolt end face point cloud set in three-dimensional space to obtain the three-dimensional plane equation. The extracted two-dimensional polygon vertices are back-projected onto this three-dimensional fitting plane to obtain the coordinates of the three-dimensional polygon vertices. Based on these three-dimensional vertex coordinates, the geometric centroid of the polygon on the three-dimensional fitting plane is calculated, and this geometric centroid is used as the geometric center point of the bolt end face.

[0038] By fitting the cylindrical surface or the polygonal plane, the interference of oil stains, rust, or local defects on the bolt surface on the center positioning is eliminated.

[0039] (3) Coordinate transformation and spacing calculation After completing the acquisition, noise reduction, and initial coordinate extraction of the bolt center in the entire three-dimensional point cloud data of the turnout area, a track engineering reference coordinate system was first established in conjunction with the track design alignment parameters. Then, the measurement data in the sensor coordinate system was converted to this reference system and the sleeper spacing was calculated. The specific implementation process is as follows: First, a baseline coordinate system for the track engineering is established based on the alignment parameters in the track design drawings. Key geometric elements in the turnout design alignment are selected as the basis for the baseline definition, including parameters such as the turnout starting mileage, the theoretical centerline equations of the straight and curved tracks, the design gauge, and the standard sleeper spacing. A right-hand rectangular coordinate system conforming to engineering conventions is constructed, with the theoretical center point of the turnout starting as the origin, the tangent direction of the theoretical centerline of the straight track as the longitudinal axis, the direction perpendicular to the longitudinal axis and pointing outwards from the curved track as the transverse axis, and the vertically upward axis as the elevation axis. This coordinate system serves as a unified reference framework for the alignment and deviation calculation of all subsequent measured data; its spatial orientation and position are entirely determined by the design alignment parameters and do not depend on the actual on-site measurement conditions.

[0040] Secondly, reference bolts for spatial transformation are selected within this reference coordinate system. Multiple reference bolts are selected at equal intervals along the longitudinal direction of the turnout, and their corresponding coordinates are obtained. Let the total number of selected reference bolts be... , of which The initial three-dimensional coordinates of the reference bolts in the sensor coordinate system are as follows: Its theoretical physical coordinates in the orbital engineering reference coordinate system are as follows: , where index .

[0041] Secondly, the initial global rigid body transformation matrix is ​​solved based on the Singular Value Decomposition (SVD) algorithm to perform preliminary alignment of the point cloud data. The centroid coordinates of all reference bolts in the sensor coordinate system are then calculated. And the centroid coordinates of all reference bolts in the track engineering reference coordinate system. Calculate the centroid coordinates separately. and Construct the covariance matrix For the covariance matrix Perform singular value decomposition to obtain ,in It is a left singular vector matrix. It is a singular value diagonal matrix. Let be a right singular vector matrix. Calculate the initial global rotation matrix. To ensure it is a pure rotation matrix, if the determinant Then the matrix The last column element multiplied by Recalculate Calculate the initial global translation vector. Therefore, the initial global rigid body transformation matrix is ​​constructed. ,in It is a three-dimensional zero column vector. Using... Perform a preliminary alignment transformation on the coordinates of all point clouds in the sensor coordinate system.

[0042] Next, turnout design alignment parameters are introduced to construct a registration objective function that integrates track geometric constraints. Let the first point be the point on the left rail extracted from the point cloud after initial alignment. The coordinates of the discrete points are The coordinates of the corresponding point on the right rail are , where index , This represents the total number of discrete points on the rail. The design gauge constraint term is then constructed. ,in Let be the standard track gauge value for track design. Let the be the [value] arranged longitudinally along the track. The coordinates of the center point of each sleeper are: , where index , Given the total number of sleepers. Construct a longitudinal alignment smoothing constraint term. Construct the spacing constraint term between adjacent sleepers. ,in This represents the standard sleeper spacing value for track design. A track gauge constraint weighting coefficient is introduced. Smoothing constraint weight coefficient and spacing constraint weight coefficient Construct the final registration objective function. ,in This is the standard point-to-point or point-to-surface distance error term in the Iterative Closest Point (ICP) algorithm.

[0043] Subsequently, using the initially aligned point cloud data as input, the constrained Iterative Closest Point (ICP) algorithm is employed to solve the registration objective function. Let the first... The current transformation matrix at the next iteration is: Calculate the corresponding point set between the current point cloud and the target model, and calculate the registration objective function. Minimize the incremental transformation matrix Update the transformation matrix as follows: Repeat the above iterative process. When the difference between the registration objective function of two adjacent iterations is less than the preset convergence threshold, stop the iteration, and define the final transformation matrix as the nonlinear deformation compensation matrix. .

[0044] Finally, the final coordinate transformation is performed using a concatenated matrix. The initial global rigid body transformation matrix is ​​then used. With nonlinear deformation compensation matrix Performing cascaded multiplication yields the final space transformation matrix. For the initial three-dimensional coordinates of any bolt center in the sensor coordinate system. Convert it to homogeneous coordinate form Calculating actual physical homogeneous coordinates using matrix multiplication. ,extract The first three dimensions are the actual physical coordinates located in the track engineering reference coordinate system after eliminating long-distance cumulative errors. .

[0045] Finally, the sleeper spacing is calculated based on the transformed actual physical coordinates. First, the center points of bolts on the same side of the rail need to be paired and verified to confirm that the bolts selected on adjacent sleepers are located on the same side of the rail and in the same installation position, avoiding mismatches that could lead to distorted spacing calculations. Then, the projected coordinate values ​​of the center points of the successfully paired adjacent bolts on the longitudinal axis of the track engineering reference coordinate system are extracted. This projection operation eliminates the interference of lateral offset and elevation fluctuations on the longitudinal spacing calculation, ensuring that the obtained spacing purely represents the relative positional relationship of the sleepers along the track direction. The difference between the two projected coordinates is calculated to obtain the actual spacing of a single sleeper within that section. Next, using the sleeper containing the preset reference bolt as the starting point for the cumulative spacing, the actual spacing of each single sleeper is sequentially accumulated along the longitudinal direction of the track according to the sleeper arrangement order, obtaining the cumulative spacing of each sleeper relative to the starting point. Since all the individual spacings are derived from the difference of projected coordinates under a unified reference coordinate system, rather than the absolute distances measured independently, the spacings of each segment share the same set of coordinate systems optimized by geometric constraints. This fundamentally prevents the problem of error propagation and superposition caused by local errors introduced by each measurement in the traditional method of measuring each individual segment independently, ensuring that the cumulative spacing is consistent with the design line type across the entire domain.

[0046] This method first establishes an engineering reference coordinate system independent of the actual measurement state, and then incorporates the measured data into the system through two-level registration. This effectively avoids the error propagation and accumulation problems caused by the traditional unit-by-unit measurement method, and ensures the accuracy and consistency of the pre-assembly measurement data.

[0047] (4) Parameter calculation and scheme generation Based on the completion of sleeper spacing calculation and unified spatial coordinate transformation, geometric parameter calculation and comparative analysis were further carried out using the three-dimensional point cloud data of the turnout area converted to the track engineering reference coordinate system. By fitting the contour lines of the rail top surface and working edge, and identifying the center position of the sleeper bearing groove, the spatial position characteristics of the rail and sleeper were extracted, and then the geometric parameters such as track gauge, level, elevation, track orientation, and triangular pit of the turnout area were calculated.

[0048] Subsequently, the calculated actual sleeper spacing, cumulative spacing, and the aforementioned geometric parameters are compared item by item with the preset turnout design standard tolerance range to identify parameters exceeding the tolerance range and their corresponding specific mileage locations and component numbers. For each exceeding location, considering the direction and magnitude of the deviation between the theoretical design value and the measured value at that location, the required adjustment amounts for the sleepers or rails in the lateral, longitudinal, and vertical directions are calculated. The lateral adjustment amount is used to correct gauge and alignment deviations, the vertical adjustment amount is used to correct level, elevation, and triangular groove deviations, and the longitudinal adjustment amount is used to correct sleeper spacing deviations.

[0049] Based on the calculated adjustment amounts, a turnout pre-assembly adjustment instruction is generated, clearly indicating the location of the components to be adjusted, the adjustment direction, and the specific adjustment values, for direct execution during on-site pre-assembly operations. Simultaneously, a laying guidance plan is generated, including a turnout overall alignment deviation distribution diagram, visually displaying the deviation trends of various geometric parameters along the turnout's longitudinal direction and the distribution of over-limit areas. This provides a global reference for subsequent fine-tuning operations, ensuring consistency between the pre-assembly adjustment and the final laying state.

[0050] Example 2 Please refer to Figure 5 This embodiment 2 provides a turnout pre-assembly measurement and design system based on the global point cloud of the turnout area, including: The point cloud acquisition and region extraction unit is used to control the 3D scanning equipment to move along the turnout area to be measured and continuously acquire the original 3D point cloud data of the entire turnout area; perform filtering and noise reduction preprocessing on the original 3D point cloud data to obtain effective point cloud data; and segment and extract the region of interest where the sleepers and fasteners are located based on the vertical elevation undulation characteristics of the effective point cloud data. The bolt segmentation and center calculation unit is used to identify and segment the bolt end face point cloud set in the point cloud of the region of interest based on the local geometric morphology features of the fastener area shifting to both sides; perform three-dimensional geometric surface fitting on the bolt end face point cloud set to calculate the geometric center point of each bolt end face; and use the geometric center point as the bolt center to obtain the initial three-dimensional coordinates of each bolt center in the sensor coordinate system. The coordinate transformation and spacing calculation unit is used to select a preset reference bolt in the three-dimensional point cloud data of the entire turnout area, and establish a track engineering reference coordinate system in combination with the track design line parameters; construct a spatial transformation model between the sensor coordinate system and the track engineering reference coordinate system, and convert the initial three-dimensional coordinates of each bolt center into actual physical coordinates; and calculate the actual spacing and cumulative spacing of each sleeper based on the difference in actual physical coordinates of the corresponding bolt centers on adjacent sleepers. The parameter calculation and scheme generation unit is used to extract the spatial position features of the rails and sleepers using the three-dimensional point cloud data of the turnout area after coordinate transformation, and calculate the track gauge, level, elevation, track direction and triangular pit geometric parameters of the turnout area; compare and analyze the actual spacing and cumulative spacing of the sleepers and the turnout geometric parameters with the turnout design standards, and generate turnout pre-assembly adjustment instructions and laying guidance schemes.

[0051] Example 3 This embodiment 3 also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement any step of a turnout pre-assembly measurement and design method based on the global point cloud of the turnout area.

[0052] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0053] For a description of the computer-readable storage medium provided in this application, please refer to the above method embodiments; further details will not be repeated here.

[0054] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for pre-assembly measurement and design of turnouts based on the global point cloud of the turnout area, characterized in that, include: S1. Control the 3D scanning equipment to move along the turnout area to be measured and continuously collect the original 3D point cloud data of the entire turnout area; The original 3D point cloud data is preprocessed by filtering and denoising to obtain effective point cloud data; based on the elevation undulation features of the effective point cloud data in the vertical direction, the region of interest where the sleeper and fastener are located is segmented and extracted. S2. In the point cloud of the region of interest, based on the local geometric features of the fastener region shifting to both sides, identify and segment the bolt end face point cloud set; perform three-dimensional geometric surface fitting on the bolt end face point cloud set to calculate the geometric center point of each bolt end face; use the geometric center point as the bolt center to obtain the initial three-dimensional coordinates of each bolt center in the sensor coordinate system; S3. Select a preset reference bolt from the three-dimensional point cloud data of the entire turnout area, and establish a track engineering reference coordinate system in combination with the track design line parameters; construct a spatial transformation model between the sensor coordinate system and the track engineering reference coordinate system, and convert the initial three-dimensional coordinates of each bolt center into actual physical coordinates; based on the difference in actual physical coordinates of the corresponding bolt centers on adjacent sleepers, calculate the actual spacing and cumulative spacing of each sleeper. S4. Using the coordinate-transformed three-dimensional point cloud data of the turnout area, extract the spatial position features of the rails and sleepers, and calculate the track gauge, level, elevation, track direction, and triangular pit geometric parameters of the turnout area; compare and analyze the actual spacing and cumulative spacing of the sleepers and the turnout geometric parameters with the turnout design standards to generate turnout pre-assembly adjustment instructions and laying guidance schemes.

2. The turnout pre-assembly measurement and design method based on the global point cloud of the turnout area according to claim 1, characterized in that, In step S1, the three-dimensional scanning device is controlled to move along the area of ​​the turnout to be measured, continuously acquiring raw three-dimensional point cloud data of the entire turnout area, including: An coded odometer is integrated into the three-dimensional scanning device to simultaneously collect mileage data and linear structure light cross-section data as the device moves along the rail. Based on the mileage data, pose compensation and multi-frame stitching are performed on the continuously acquired line structured light cross-section data to obtain the original three-dimensional point cloud data. The preprocessing of filtering and denoising the original three-dimensional point cloud data includes: using a statistical filtering algorithm to remove outlier noise caused by the switch area environment, and using a radius filtering algorithm to remove local dense noise caused by dust and water vapor, to obtain the effective point cloud data.

3. The turnout pre-assembly measurement and design method based on the global point cloud of the turnout area according to claim 1, characterized in that, In step S1, based on the vertical elevation undulation features of the effective point cloud data, the region of interest (ROI) point cloud containing the sleepers and fasteners is segmented and extracted, including: The effective point cloud data is projected onto a horizontal plane to construct an elevation histogram. Based on the peak distribution characteristics in the elevation histogram, determine the elevation threshold ranges for the top surface of the rail, the top surface of the sleeper / fastener, and the surface of the ballast. The effective point cloud data is initially segmented based on the elevation threshold range corresponding to the top surface of the sleeper / fastener to obtain candidate region point clouds; Perform three-dimensional connected component analysis on the candidate region point cloud, and remove isolated point cloud clusters with a volume smaller than a preset threshold to obtain the region of interest point cloud.

4. The turnout pre-assembly measurement and design method based on the global point cloud of the turnout area according to claim 1, characterized in that, In step S2, based on the local geometric features of the fastener region shifting to both sides in the point cloud of the region of interest, a set of bolt end face point clouds is identified and segmented, including: Local elevation normalization is performed on the point cloud of the region of interest to eliminate the elevation benchmark deviation caused by the tilt of the sleeper top surface and the laying slope, and a relative elevation point cloud is obtained. Based on the density distribution and connected component characteristics of the relative elevation point cloud, the largest connected component with a volume greater than a preset volume threshold is extracted and used as the main body area of ​​the fastener pad. Obtain the three-dimensional bounding box of the main area of ​​the fastener pad in the longitudinal and transverse directions of the sleeper. Based on the prior assembly ratio of the turnout fastener structure, expand the three-dimensional bounding box outward by a preset ratio distance along the transverse direction of the sleeper to both sides to construct an adaptive bolt search window. Within the adaptive bolt search window, point clouds with relative elevation values ​​greater than a preset elevation threshold are extracted as candidate bolt top point clouds; The point cloud at the top of the candidate bolt is subjected to region growing clustering based on normal vector consistency to obtain several independent point cloud clusters; the three-dimensional spatial compactness of each independent point cloud cluster is calculated, and the independent point cloud clusters with a three-dimensional spatial compactness greater than a preset confidence threshold are determined as the bolt end face point cloud set.

5. The turnout pre-assembly measurement and design method based on the global point cloud of the turnout area according to claim 1, characterized in that, In step S2, a three-dimensional geometric surface is fitted to the point cloud set of the bolt end faces to calculate the geometric center points of each bolt end face, including: Determine the contour features of the point cloud set on the bolt end face; If the contour feature is circular or cylindrical, the Random Sample Consensus (RANSAC) algorithm is used to fit the cylindrical surface, and the intersection of the central axis of the fitted cylinder and the bolt end face is taken as the geometric center point. If the contour feature is polygonal, then polygon vertex extraction and least squares plane fitting are performed, and the geometric centroid of the fitted plane is taken as the geometric center point. By fitting the cylindrical surface or the polygonal plane, the interference of oil stains, rust, or local defects on the bolt surface on the center positioning is eliminated.

6. The turnout pre-assembly measurement and design method based on the global point cloud of the turnout area according to claim 1, characterized in that, In step S3, a spatial transformation model is constructed between the sensor coordinate system and the track engineering reference coordinate system, converting the initial three-dimensional coordinates of each bolt center into actual physical coordinates, including: Multiple reference bolts are selected at equal intervals along the longitudinal direction of the turnout to obtain their initial three-dimensional coordinates in the sensor coordinate system and their theoretical physical coordinates in the track design drawings. Based on the initial three-dimensional coordinates and theoretical physical coordinates, the initial global rigid body transformation matrix is ​​solved using the singular value decomposition (SVD) algorithm, and the three-dimensional point cloud data of the bifurcation region is initially aligned. By introducing turnout design alignment parameters, a registration objective function integrating track geometric constraints is constructed. The track geometric constraints include design gauge constraints, longitudinal alignment smoothing constraints, and adjacent sleeper spacing constraints. Using the pre-aligned point cloud data as input, the registration objective function is solved by the constrained iterative nearest point (ICP) algorithm to obtain the nonlinear deformation compensation matrix; By using the cascaded matrix of the initial global rigid body transformation matrix and the nonlinear deformation compensation matrix, the initial three-dimensional coordinates of all bolt centers are uniformly transformed to the track engineering reference coordinate system to obtain the actual physical coordinates that eliminate long-distance cumulative errors.

7. The turnout pre-assembly measurement and design method based on the global point cloud of the turnout area according to claim 1, characterized in that, In step S3, based on the actual physical coordinate difference between the centers of corresponding bolts on adjacent sleepers, the actual spacing and cumulative spacing of each sleeper are calculated, including: Extract the actual physical coordinates of the centers of the corresponding bolts on the same side of two adjacent sleepers in the longitudinal direction of the track; The difference between two adjacent actual physical coordinate values ​​is calculated to obtain the actual spacing of a single sleeper; Starting from the sleeper where the preset reference bolt is located, the actual spacing of each sleeper is accumulated sequentially along the longitudinal direction of the track to obtain the cumulative spacing of each sleeper relative to the starting point, so as to eliminate the cumulative error caused by measuring each sleeper individually.

8. The turnout pre-assembly measurement and design method based on the global point cloud of the turnout area according to claim 1, characterized in that, In step S4, the actual spacing and cumulative spacing of the sleepers, as well as the geometric parameters of the turnout, are compared and analyzed with the turnout design standards to generate turnout pre-assembly adjustment instructions and laying guidance schemes, including: The actual spacing, cumulative spacing, and turnout geometric parameters are compared with the preset turnout design standard tolerance range to identify the out-of-limit parameters and their corresponding out-of-limit locations. Based on the aforementioned out-of-limit parameters and out-of-limit locations, calculate the required adjustment amounts for the sleepers or rails in the lateral, longitudinal, and vertical directions. Based on the adjustment amount, generate the turnout pre-assembly adjustment instruction containing the adjustment position, adjustment direction and adjustment value, and generate the laying guidance scheme containing the overall turnout alignment deviation distribution diagram.

9. A turnout pre-assembly measurement and design system based on the global point cloud of the turnout area, characterized in that, include: The point cloud acquisition and region extraction unit is used to control the 3D scanning equipment to move along the turnout area to be measured and continuously acquire the original 3D point cloud data of the entire turnout area. The original 3D point cloud data is preprocessed by filtering and denoising to obtain effective point cloud data; based on the elevation undulation features of the effective point cloud data in the vertical direction, the region of interest where the sleeper and fastener are located is segmented and extracted. The bolt segmentation and center calculation unit is used to identify and segment the bolt end face point cloud set in the point cloud of the region of interest based on the local geometric morphology features of the fastener area shifting to both sides; perform three-dimensional geometric surface fitting on the bolt end face point cloud set to calculate the geometric center point of each bolt end face; and use the geometric center point as the bolt center to obtain the initial three-dimensional coordinates of each bolt center in the sensor coordinate system. The coordinate transformation and spacing calculation unit is used to select a preset reference bolt in the three-dimensional point cloud data of the entire turnout area, and establish a track engineering reference coordinate system in combination with the track design line parameters; construct a spatial transformation model between the sensor coordinate system and the track engineering reference coordinate system, and convert the initial three-dimensional coordinates of each bolt center into actual physical coordinates; and calculate the actual spacing and cumulative spacing of each sleeper based on the difference in actual physical coordinates of the corresponding bolt centers on adjacent sleepers. The parameter calculation and scheme generation unit is used to extract the spatial position features of the rails and sleepers using the three-dimensional point cloud data of the turnout area after coordinate transformation, and calculate the track gauge, level, elevation, track direction and triangular pit geometric parameters of the turnout area; compare and analyze the actual spacing and cumulative spacing of the sleepers and the turnout geometric parameters with the turnout design standards, and generate turnout pre-assembly adjustment instructions and laying guidance schemes.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is executed by a processor as described in any one of claims 1-8: a turnout pre-assembly measurement and design method based on the global point cloud of the turnout area.

Images