A method and system for point cloud registration of a wading infrastructure adjunct descriptor

Abstract

The present application belongs to the field of data processing, and relates to a point cloud registration method and system for a water-related infrastructure accessory structure descriptor, the method comprising: acquiring point cloud data; preprocessing the point cloud data; extracting a tunnel accessory structure descriptor based on the preprocessed point cloud data; screening feature points through the tunnel accessory structure descriptor; correcting the feature point positions through a feature region according to the feature points; and performing point cloud registration based on the corrected feature point positions. The method can effectively remove noise and redundant information, ensuring the accuracy and reliability of the data, accurately capturing the uniqueness and key information of the tunnel structure, improving the recognition accuracy of the feature points, and significantly enhancing the stability and robustness of the registration. The method can further reduce errors and improve registration accuracy, fully utilizing the geometric features and spatial distribution of the tunnel accessory structure to achieve accurate adjustment and optimization of the feature points.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a point cloud registration method and system of a water-related infrastructure accessory structure descriptor. BACKGROUND

[0002] With the development of laser radar technology, it has gradually become more widely used in tunnel engineering such as water-related infrastructure accessory structures, such as reconstructing a three-dimensional tunnel and creating a completed building information model using laser point cloud data, constructing an accurate model of the tunnel using the correspondence between images and point clouds, and monitoring the deformation and structural damage of the tunnel using multiple point cloud data. Tunnels can be divided into rail tunnels such as subway tunnels and high-speed rail tunnels, and non-rail tunnels such as air defense tunnels and highway tunnels. The commonly used laser scanners for these two types of tunnels have some differences. Rail tunnels generally use mobile laser systems (MLS) to collect three-dimensional information of the tunnel, and non-rail tunnels generally use terrestrial laser scanners (TLS). MLS generally has an inertial navigation system built-in and scans along the tunnel, which can directly obtain the overall point cloud. Terrestrial laser scanners use multiple scans to obtain point cloud data, and in order to obtain the overall tunnel, the point clouds of two scans must be registered.

[0003] Point cloud accurate registration in a tunnel scene is a very challenging task. The main reason is that the features in the tunnel scene have a high degree of repetition, which greatly affects the correctness of the corresponding point estimation, and weak geometric features, which make it difficult for descriptors to accurately describe the feature information contained in the key points. Existing point cloud registration schemes mainly place multiple targets during data collection and use target positions to achieve point cloud registration, but few studies use the features present in the tunnel to solve the problem of accurate point cloud registration. The difficulty of registration due to the small number of key points and high similarity of features in the tunnel will inevitably result in a loss of data value. SUMMARY

[0004] To solve the above technical problems, the present application provides a point cloud registration method for a water-related infrastructure accessory structure descriptor, which adopts the following technical solution, comprising:

[0005] Obtain point cloud data;

[0006] Preprocess the point cloud data;

[0007] Extract a tunnel accessory structure descriptor based on the preprocessed point cloud data;

[0008] Screen feature points through the tunnel accessory structure descriptor;

[0009] Correct the positions of the feature points according to the feature points through a feature region;

[0010] Perform point cloud registration based on the corrected feature point positions.

[0011] Preferably, the step of obtaining point cloud data specifically includes:

[0012] According to the length of the tunnel, set the number, spacing and scanning parameters of the measuring stations;

[0013] Perform scanning operations according to the set number of measuring stations, spacing, and scanning parameters to collect point cloud data.

[0014] Preferably, the step of preprocessing the point cloud data specifically includes:

[0015] removing noise points in the point cloud data;

[0016] filtering the point cloud data;

[0017] Segmenting the point cloud data into a plurality of subsets, each subset representing a different portion of the tunnel or ancillary structures;

[0018] The point cloud data is simplified.

[0019] Preferably, the step of extracting tunnel-attached structure descriptors based on the pre-processed point cloud data specifically includes:

[0020] Identify the main structure of the tunnel;

[0021] Extracting key points based on the main structure of the tunnel, wherein the key points refer to feature points representing the structural characteristics of the tunnel;

[0022] For each extracted key point, calculate the feature area around it;

[0023] According to the point cloud distribution and geometric features in the feature area, the local information of the key points is extracted;

[0024] Calculate the global information of the key points based on the positional relationship between the key points and the tunnel axis;

[0025] Filter key points based on their local and global information;

[0026] The local information, global information and filtered key points of the extracted key points are encoded and quantized to generate a tunnel accessory structure descriptor.

[0027] Preferably, the step of screening feature points using the tunnel-attached structure descriptor specifically includes:

[0028] Use threshold segmentation and cluster analysis to identify the initial feature point set;

[0029] Optimize and filter feature points from the initial feature point set.

[0030] Preferably, the step of correcting the position of the feature point through the feature area based on the feature point specifically includes:

[0031] Extracting feature areas according to the feature points;

[0032] Based on the feature area extraction, the correspondence between the feature points in the two point cloud data sets is established to perform feature point matching;

[0033] Based on the feature area and the matched feature point pairs, the initial feature point positions are corrected.

[0034] Preferably, the step of performing point cloud registration based on the corrected feature point positions specifically includes:

[0035] Based on the corrected feature point positions, perform point cloud coarse registration;

[0036] Based on the position of the feature points after coarse registration, perform fine registration of the point cloud;

[0037] Evaluate and optimize the point cloud registration results after fine registration.

[0038] In order to solve the above technical problems, the present invention also provides a point cloud registration system for water infrastructure accessory structure descriptors, which adopts the following technical solution, including:

[0039] Acquisition module, used to obtain point cloud data;

[0040] A preprocessing module, used for preprocessing the point cloud data;

[0041] an extraction module, configured to extract tunnel-attached structure descriptors based on the preprocessed point cloud data;

[0042] A screening module, configured to screen feature points using the tunnel-attached structure descriptor;

[0043] A correction module, configured to correct the position of the feature point through the feature area according to the feature point;

[0044] The registration module is used to perform point cloud registration based on the corrected feature point positions.

[0045] In order to solve the above technical problems, the present invention also provides a computer device, which adopts the technical solution described below, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, it implements the steps of the above-mentioned point cloud registration method of the auxiliary structure descriptor of the water-related infrastructure.

[0046] In order to solve the above technical problems, the present invention also provides a computer-readable storage medium, which adopts the technical solution described below, and the computer-readable storage medium stores computer-readable instructions, which, when executed by the processor, implement the steps of the above-mentioned point cloud registration method of the water-related infrastructure accessory structure descriptor.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] By acquiring point cloud data and performing preprocessing, noise and redundant information can be effectively removed, providing a high-quality data foundation for subsequent steps, ensuring the accuracy and reliability of the data, and laying a solid foundation for subsequent descriptor extraction and feature point screening;

[0049] By extracting feature points using tunnel accessory structure descriptors, the uniqueness and key information of the tunnel structure can be accurately captured, which not only improves the recognition accuracy of feature points but also significantly enhances the stability and robustness of registration.

[0050] Correcting the position of feature points through feature regions can further reduce errors and improve registration accuracy. This fully utilizes the geometric characteristics and spatial distribution of tunnel ancillary structures and achieves precise adjustment and optimization of feature points.

[0051] By performing point cloud registration based on the corrected feature point positions, the registration speed and efficiency can be significantly improved, effectively addressing the complexity and diversity of tunnel structures and providing strong technical support for the design, construction, and maintenance of tunnel projects.

[0052] It has multiple advantages such as high precision, high efficiency and high robustness, providing a new solution for 3D modeling and analysis of tunnel engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the solutions in the present invention, a brief introduction is given below to the drawings required for use in describing the embodiments of the present invention. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0054] Figure 1 is a flow chart of an embodiment of a method for point cloud registration of water infrastructure accessory structure descriptors of the present invention;

[0055] Figure 2 It is a structural diagram of an embodiment of a point cloud registration system for water infrastructure accessory structure descriptors of the present invention;

[0056] Figure 3is a structural schematic diagram of one embodiment of the computer device of the present application. DETAILED DESCRIPTION

[0057] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used in the description herein is for describing particular embodiments only and is not intended to be limiting of the application; the description herein and the claims of the application and the above description of the drawings herein use the term "comprising" and "having" and their any variations thereof to mean "including but not limited to". The description herein and the claims of the application and the above description of the drawings herein use the term "first", "second", and the like to mean "different" and are not intended to imply that a particular order.

[0058] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another.

[0059] In order to make the technical personnel in the art better understand the application scheme, the technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings.

[0060] It should be noted that the point cloud registration method of the water-related infrastructure accessory structure descriptor provided by the embodiments of the application is generally executed by a high-precision laser scanner, a high-definition camera, a positioning system, and a data processing device, and accordingly, the point cloud registration system of the water-related infrastructure accessory structure descriptor is generally arranged in the positioning system or the data processing device.

[0061] Embodiment one

[0062] Please refer to Figure 1 , Figure 1 is a flowchart of one embodiment of the point cloud registration method of the water-related infrastructure accessory structure descriptor of the present application. The point cloud registration method of the water-related infrastructure accessory structure descriptor comprises the following steps:

[0063] Step S1, acquiring point cloud data.

[0064] Point cloud data refers to a large number of discrete three-dimensional coordinate point sets collected in the tunnel interior and surrounding environment through three-dimensional laser scanning, photogrammetry, etc. These point cloud data can accurately reflect the geometric shape, structural characteristics of the tunnel and the position and state of the accessory facilities. In tunnel engineering, point cloud data is the basis for constructing three-dimensional models, conducting structural analysis, monitoring deformation, planning maintenance, etc.

[0065] In this embodiment, step S1, obtaining point cloud data may specifically include the following steps:

[0066] S11, set the number, spacing and scanning parameters of the measuring stations according to the length of the tunnel.

[0067] Setting the number and spacing of survey stations: The number and spacing of survey stations directly impact the coverage and accuracy of point cloud data. Generally speaking, longer tunnels require more survey stations, and spacing should be appropriately reduced to ensure data continuity and integrity. Furthermore, the spacing of survey stations must consider the effective scanning range of the scanner and the size of the overlap area to ensure seamless data flow between adjacent stations.

[0068] Scan parameter settings: Scanning parameters include scanning speed, resolution, and scanning angle. These parameters should be adjusted based on the actual tunnel conditions and scanning requirements. For example, for tunnels with complex structures, the scanning resolution should be increased to capture more details; whereas for longer tunnels with relatively simple structures, the resolution can be appropriately reduced to improve scanning efficiency.

[0069] When setting the number and spacing of measurement stations, factors such as the tunnel geometry, the location of ancillary structures, and the performance of the scanning instrument must also be considered. Proper layout and parameter settings ensure that the collected point cloud data is both comprehensive and accurate.

[0070] In step S11, by setting reasonable measurement station layout and scanning parameters, it can be ensured that the collected point cloud data not only comprehensively covers all parts of the tunnel but also accurately reflects its geometric shape and structural characteristics.

[0071] S12, performing scanning operations according to the set number of measuring stations, spacing and scanning parameters to collect point cloud data.

[0072] Scanning instrument selection and calibration: Select an appropriate 3D laser scanner or photogrammetry system based on the tunnel's actual conditions and scanning requirements. Before use, the scanner must be calibrated to ensure data accuracy.

[0073] Scanning: At each station, activate the scanning instrument and collect data according to the preset scanning parameters. During the scanning process, ensure that the instrument is stable and the line of sight is clear, avoiding obstructions and interference. Also, be sure to record the location information of each station for subsequent data processing.

[0074] Data quality inspection and control: After scanning is complete, the collected point cloud data must undergo a quality inspection. This includes checking the data's integrity, continuity, and accuracy. Data that does not meet quality requirements requires additional scanning or rescanning to ensure data reliability.

[0075] Laser scanner according to the preset number of measuring stations ,spacing and scanning parameters (such as resolution Field of view ) to collect point cloud data. The coordinate systems of each survey station need to be initially aligned using targets or feature points.

[0076] The station spacing constraint is set as: ,in is the effective range of the scanner (unit: meter), is the horizontal field of view of the scanner (unit: radians), is the overlapping area redundancy coefficient (preferably 0.2~0.5).

[0077] Step S12 can improve scanning efficiency by optimizing scanning strategies and parameter settings to reduce unnecessary scanning time and data processing. At the same time, strict data quality inspection and control measures can ensure that the collected point cloud data has high accuracy and reliability to meet the needs of subsequent work.

[0078] In practice, datasets collected from the real world can also be used, such as the tunnel data in the open-source WH-TLS dataset from Wuhan University, or measurements taken in a constructed tunnel using the BLK360. The WU-TLS dataset has an accuracy of 5mm and contains approximately 20 million points per station. The BLK360 dataset has an accuracy of 4mm and contains approximately 3 million points per station.

[0079] Step S2: pre-process the point cloud data.

[0080] In this embodiment, step S2, pre-processing the point cloud data may specifically include the following steps:

[0081] S21, remove noise points in point cloud data.

[0082] Noise points are usually caused by scanning equipment errors, environmental interference or improper human operation. The denoising method varies depending on the structure type of point cloud data (ordered or scattered).

[0083] In specific implementation, ordered point cloud denoising and scattered point cloud denoising can be performed.

[0084] Ordered point cloud data includes grid-type and array-type point clouds, which are organized regularly and have a clear structure. De-noising can be performed using curve inspection, isolated point rejection, and filtering.

[0085] The curve checking method refers to: using the first and last points of the point cloud data, obtaining a curve based on the least squares principle, solving the distance from each point to this line, and setting a threshold. If the distance value is less than the set threshold, the point is considered a normal point; otherwise, it is considered a noise point.

[0086] The isolated point rejection method refers to: eliminating points or isolated points that deviate greatly from the scan line based on the displayed point cloud scan line.

[0087] Filtering refers to applying signal processing principles to select an appropriate filter function to process noise. Gaussian filtering is commonly used for denoising, followed by median filtering and average filtering.

[0088] In specific implementation, neighborhood mean filtering can be used to filter the points of -Point set within the neighborhood To smooth: ,in , is the coordinate of the original point (unit: meter), for point of The nearest neighbor set, are the coordinates after denoising.

[0089] Statistical outlier removal and calculation of neighborhood distance mean and standard deviation : , .

[0090] Elimination conditions (if met, it is considered noise): ,in is the threshold coefficient (preferably ).

[0091] Since there is no topological relationship between points in a scattered point cloud, the aforementioned ordered point cloud denoising methods are not applicable to scattered point clouds. The key to denoising scattered point clouds lies in establishing topological relationships between points and then applying appropriate algorithms based on these topological relationships. These algorithms can be used, for example, using the K-neighborhood Hardy function denoising algorithm or adaptive point cloud denoising.

[0092] The principle of denoising scattered point clouds based on density clustering is to use density clustering algorithms (such as DBSCAN) to separate noise and valid points. of -The density of points in the neighborhood is defined as: , core point determination: if ,but is the core point. is the neighborhood radius (needs to be set according to the point cloud density, unit: meter), is the minimum neighborhood point threshold (preferably ).

[0093] S22, filtering the point cloud data.

[0094] Filtering smoothes point cloud data and removes minor noise. Unlike denoising, filtering focuses on smoothing the entire point cloud rather than removing individual noise points. The choice of filtering method depends on the type of point cloud data and the characteristics of the noise.

[0095] Gaussian filtering can be used to smooth the point cloud data based on a Gaussian function, preserving the overall shape while suppressing high-frequency noise. Gaussian filtering can effectively remove high-frequency noise while preserving the overall shape characteristics of the point cloud data.

[0096] Point The filtered coordinates of The calculation is as follows:

[0097] ,in for point The neighborhood point set (radius or recently point), is the standard deviation of the Gaussian function (controls the smoothing strength, unit: meter), is the normalization factor, For each point Determine the neighborhood , calculate the Gaussian weight of each point in the neighborhood, and then take the weighted average to obtain the filtered coordinates.

[0098] Alternatively, median filtering can be used. This nonlinear filtering method smoothes the data by taking the median value of each point in the point cloud data within its neighborhood. This method then replaces the current point with the median of the coordinate components within the neighborhood, effectively removing impulse noise. Median filtering is particularly effective for removing impulse noise, such as outliers caused by scanning equipment failures.

[0099] Point , filtered coordinates for:

[0100] ,in is the median operator, is a set of neighborhood points (which can be nearest neighbor, is an odd number such as 5, 7). For each point , collect all points in the neighborhood Components, calculate the median of the three components respectively, and combine them into the filtered coordinates.

[0101] You can also use average filtering, which is a linear filtering method that smoothes the data by taking the average value of each point in its neighborhood. Although average filtering can remove some noise, it may also cause blurred edges in the point cloud data.

[0102] Filtered coordinates The calculation is as follows: ,in Neighborhood Number of inliers, neighborhood definition: fixed radius or fixed points For each point Determine the neighborhood range and calculate the average coordinates of all points in the neighborhood.

[0103] S23, segmenting the point cloud data into a plurality of subsets, each subset representing a different part or auxiliary structure of the tunnel.

[0104] The purpose of segmentation is to simplify subsequent processing steps and improve the accuracy of registration and modeling.

[0105] In specific implementation, segmentation based on geometric features, segmentation based on regions, and segmentation based on models can be performed.

[0106] Geometric feature-based segmentation refers to a segmentation method that uses the geometric characteristics of point cloud data (such as curvature and normal vectors) for segmentation. It uses the local geometric properties of the point cloud (such as differences in curvature and normal vectors) to cluster regions with similar characteristics into the same subset. This method is suitable for tunnel ancillary structures with significant geometric differences.

[0107] Normal vector difference segmentation, calculation point With neighboring points The normal vector angle of : . .

[0108] Split condition (if into different subsets):

[0109] ,in for point The normal vector of (obtained by the corresponding vector of the minimum eigenvalue of PCA), is the normal vector angle threshold (which can be ), For the A split subset.

[0110] Curvature clustering, calculation points Curvature For each point First, the covariance matrix of the point set in its neighborhood is calculated. This covariance matrix reflects the statistical properties of the point distribution in the neighborhood. Then, the covariance matrix is decomposed into three eigenvalues: , and , which correspond to the variances of the point set in the three principal component directions, respectively. Among them, represents the minimum value of the three eigenvalues, which reflects the tightest distribution of the point set in the corresponding principal component direction, i.e., the smallest degree of change in that direction.

[0111] The curvature value actually reflects the flatness or curvature of the local region where the point is located. In the region growing clustering process, points with similar curvature values are merged into the same region, and these points are likely to belong to the same geometric structure or surface. In this way, complex point cloud data can be effectively divided into different regions with similar geometric properties.

[0112] Region-based segmentation divides point cloud data into multiple regions, each with similar attributes (such as color, texture, etc.). Similarity clustering is performed based on local attributes (such as color, intensity, density) of the point cloud. This method is suitable for cases where the tunnel auxiliary structure has obvious texture or color differences.

[0113] Region growing algorithm, seed point selection condition (e.g., minimum curvature point): .

[0114] Growth condition (neighborhood points satisfy attribute similarity):

[0115] and , where is the attribute vector (such as RGB color or intensity) of point , , are the similarity thresholds of attributes and distances, respectively.

[0116] Voxel segmentation (VCCS), which divides point cloud into uniform voxels: , where is the center of voxel , and is the voxel radius (which can be 2-3 times the average distance of the point cloud).

[0117] Predefined models are used to match and segment point cloud data. Predefined model-based segmentation utilizes point cloud matching techniques (e.g., cylinders, planes) to extract structures that conform to the model. It should be noted that this method requires the shape and size of tunnel ancillary structures to be known in advance and is therefore suitable for ancillary structures with standard shapes.

[0118] RANSAC plane fitting, model parameters (normal vector) and (Distance to the origin) satisfies:

[0119] and ,in is the inlier distance threshold (can be 0.01~0.1 meters).

[0120] Cylinder segmentation (for pipes) minimizes the distance from a point to the cylinder surface:

[0121] ,in is the starting point of the cylinder axis, is the axis direction vector, is the radius of the cylinder.

[0122] S24, simplifying the point cloud data.

[0123] Simplification is the last step in preprocessing, which aims to reduce the amount of point cloud data and improve the efficiency of subsequent processing. The choice of simplification method depends on the type of point cloud data and the requirements of subsequent processing.

[0124] According to actual needs, scan line point cloud simplification, array point cloud simplification, triangulated point cloud simplification and scattered point cloud simplification can be performed.

[0125] Scanline point cloud simplification refers to the process of resampling scanline point cloud data using the chord length of the data to reduce the number of data points. This method preserves the geometric characteristics of the scanline while reducing redundant data.

[0126] Array point cloud simplification refers to the process of reducing the amount of data or streamlining the array point cloud data by reducing the spacing or magnification. This method is suitable for situations where the point cloud data is evenly distributed and has a high density.

[0127] Triangulated point cloud simplification refers to the ability to simplify triangulated point cloud data based on point cloud density or by using bounding boxes. This method preserves the topological structure of the point cloud data while reducing the number of data points.

[0128] Scattered point cloud simplification refers to the simplification of scattered point cloud data based on data curvature, meshing, and other methods. It should be noted that this method requires establishing topological relationships between points and then using corresponding algorithms for simplification calculations.

[0129] Step S3: extracting tunnel-attached structure descriptors based on the pre-processed point cloud data.

[0130] Tunnel ancillary structure descriptors refer to the point cloud data features of auxiliary facilities in addition to the main tunnel structure, such as lighting facilities, drainage systems, support structures, etc. These features can serve as important reference information in point cloud registration, improving the accuracy and robustness of registration.

[0131] In this embodiment, step S3, extracting the tunnel-attached structure descriptor based on the pre-processed point cloud data, may specifically include the following steps:

[0132] S31, identify the main structure of the tunnel.

[0133] The identification of the main structure of the tunnel is the basis for extracting the descriptors of the tunnel's auxiliary structures. The purpose is to identify the main structural parts of the tunnel, including the tunnel body, portal, etc.

[0134] The main structure of a tunnel can be identified using either a geometric feature-based recognition method or a machine learning algorithm. The geometric feature-based recognition method can identify the main structure based on the tunnel's geometric features and structural characteristics; the machine learning algorithm can use existing point cloud data to train the model and automatically recognize new data.

[0135] The principle of geometric feature-based recognition is to use the geometric characteristics of the tunnel main structure (such as continuous curvature and axial consistency) for segmentation and extraction.

[0136] For example, for tunnels, a cylindrical / arch model fitting method can be used. The RANSAC algorithm is used to fit the cylindrical surface and minimize the distance from the point to the model:

[0137] ,in is the coordinate of the starting point of the cylinder axis, is the axis unit direction vector ( ), is the radius of the cylinder (unit: meter), The first The coordinates of a point.

[0138] Applicable to tunnel portals, plane detection can be used, and the covariance matrix can be analyzed through PCA The minimum eigenvalue of Identify the plane: , .like , determined to be a flat area ( Usually 0.01 is used).

[0139] In terms of axial consistency constraints, the axial direction of the tunnel body and the fitting direction The deviation angle should satisfy:

[0140] .in is the design axis direction (known a priori), is the angle threshold (e.g. )

[0141] The recognition principle based on machine learning is to use supervised learning models (such as PointNet++ and RandLA-Net) to perform semantic segmentation on point clouds.

[0142] Perform feature extraction (local geometric descriptors) and calculate points FPFH eigenvector : , including statistics such as normal vector angle and distance.

[0143] For network training (taking PointNet++ as an example), the hierarchical feature aggregation formula is:

[0144] .in, is the vector concatenation operation, For the Layers of multi-layer perceptrons, It is the maximum pooling operation.

[0145] The loss function (cross entropy loss) is: ,in for point The true category (such as tunnel main body and auxiliary structures), is the predicted probability.

[0146] S32, extracting key points based on the main structure of the tunnel, where the key points refer to feature points representing the structural characteristics of the tunnel.

[0147] In point cloud data, key points are those important points that represent the structural characteristics of a tunnel. These points are typically located at the edges, corners, or locations with significant geometric changes. By extracting these key points, we can more effectively describe the overall structure and local characteristics of the tunnel.

[0148] Based on the main structure of the tunnel, key points are extracted through edge detection, corner detection, or geometric features. Edge detection can identify the edge features of the tunnel; corner detection can identify the corner features of the tunnel; and geometric feature extraction can extract key points based on the geometric features of the tunnel.

[0149] The principle of edge detection is to identify the edge points of the tunnel structure (such as portal contours and segment joints) through local curvature or normal vector mutations.

[0150] Curvature extreme value detection, calculation point Curvature (Proportion of the minimum eigenvalue of the PCA covariance matrix): , .

[0151] Edge point determination conditions: and .in, is the PCA eigenvalue, , is the curvature threshold (such as 0.1), is the curvature gradient (the rate of change of curvature within the neighborhood).

[0152] Normal vector mutation detection, calculate the average angle of neighborhood normal vectors: .

[0153] Edge point determination conditions: ,in for point The normal vector of is the angle threshold (such as 45°).

[0154] The principle of corner detection is to identify corners in the tunnel structure (such as segment connections and inspection opening corners) based on local curvature or eigenvalue analysis.

[0155] Perform Harris3D corner detection. Construct covariance matrix And calculate the corner response :

[0156] ,

[0157] .

[0158] Corner point determination conditions: .in, is the Gaussian weight ,

[0159] is the empirical coefficient (can be 0.05), and is the response threshold.

[0160] Curvature extremum combination detection, if a point satisfies the maximum curvature radius and the minimum curvature radius at the same time, it is determined as a corner point. And: .

[0161] The principle of geometric feature extraction is to extract key points based on the geometric invariance of tunnel structure, such as the intersection of cylindrical surface and plane.

[0162] Cylinder-plane intersection extraction, if the tunnel body (cylindrical surface) intersects with the tunnel portal (plane), the intersection point set satisfies: where, is the cylindrical parameter (as above), is the plane equation parameter.

[0163] ISS (Intrinsic Shape Signature) based key point detection, calculate the ISS feature value ratio , , , , and are all eigenvalues of the weighted covariance matrix.

[0164] Key point determination condition: and . Wherein is a threshold value (such as 0.8), is a proportion threshold (such as 0.5).

[0165] S33, for each extracted key point, calculate its surrounding feature region.

[0166] The feature region refers to the point cloud region within a certain range around the key point. This region contains the geometric and positional information closely related to the key point, and is an important basis for extracting local information. Through the feature region, the local features of the key point can be analyzed more effectively, and the accuracy of registration can be improved.

[0167] For each extracted key point, its surrounding feature region can be calculated by neighborhood search, point cloud segmentation or feature extraction, etc. Neighborhood search method can search the surrounding point cloud data according to the position of key point; point cloud segmentation method can segment point cloud data into different regions; feature extraction method can extract feature information according to the point cloud distribution and geometric features in the feature region.

[0168] The role of step S33 is to provide accurate feature region for subsequent local information extraction.

[0169] ​​S34, extracting local information of key points based on the point cloud distribution and geometric features within the feature area.

[0170] The local information of a key point refers to the geometric and positional information within the feature region adjacent to the key point. This information includes the point cloud distribution, curvature changes, and normal direction within the feature region. Using this local information, the position of the key point can be further corrected, improving the upper limit of registration accuracy.

[0171] Based on the point cloud distribution and geometric features within the feature region, local information about key points can be extracted through methods such as curvature calculation, normal estimation, or local feature description. Curvature calculation methods can calculate the curvature change within the feature region; normal estimation methods can estimate the normal direction within the feature region; and local feature description methods can describe the local characteristics of key points based on the point cloud distribution and geometric features within the feature region. These techniques provide accurate local information support for the subsequent registration process.

[0172] S35, calculating the global information of the key point based on the positional relationship between the key point and the tunnel centerline.

[0173] The global information of a key point refers to its positional relationship with the tunnel's central axis. This information includes the distance and direction from the key point to the central axis. This global information helps determine the key point's position within the overall tunnel structure, thereby more accurately describing the tunnel's geometry.

[0174] Based on the positional relationship between key points and the tunnel's central axis, global information about the key points is calculated through methods such as tunnel central axis extraction and distance and direction calculation. The tunnel central axis extraction method extracts the central axis based on the tunnel's geometric characteristics, while the distance and direction calculation method calculates the distance and direction from the key point to the central axis.

[0175] The function of step S35 is to provide accurate global information support for the subsequent registration process.

[0176] S36, filtering key points according to local and global information of the key points.

[0177] Key points are screened based on their local and global information using feature-based screening or statistical filtering. Feature-based screening can select high-quality key points based on their local and global information, while statistical filtering can remove noise points and redundant information using statistical principles.

[0178] The function of step S36 is to provide stable and representative key point support for the subsequent registration process.

[0179] S37 , encoding and quantizing the local information and global information of the extracted key points and the filtered key points to generate a tunnel-attached structure descriptor.

[0180] Through methods such as feature encoding or vector quantization, the local and global information of the extracted key points, as well as the filtered key points, are encoded and quantized to generate a tunnel accessory structure descriptor. Feature encoding can encode the extracted feature information into a vector form; vector quantization can quantize the vector data to generate the final descriptor.

[0181] The function of step S37 is to provide effective data support for the subsequent registration and recognition processes.

[0182] Step S4: filtering feature points using the tunnel-attached structure descriptor.

[0183] In this embodiment, step S4, screening feature points using the tunnel attached structure descriptor, may specifically include the following steps:

[0184] S41, using threshold segmentation and cluster analysis to identify the initial feature point set.

[0185] Based on the extracted feature descriptors, methods such as threshold segmentation and cluster analysis are used to identify an initial set of feature points. Threshold segmentation sets a threshold based on the numerical range of the feature descriptors, and uses points exceeding the threshold as initial feature points. Cluster analysis can group points with similar feature descriptors into the same category, with the center point or representative point of each category serving as the initial feature point.

[0186] The purpose of step S41 is to quickly identify points with potential feature significance from a large amount of point cloud data by setting a reasonable threshold or using cluster analysis, thereby providing a basis for subsequent feature point screening and optimization.

[0187] S42, optimizing and screening feature points from the initial feature point set.

[0188] After identifying the initial feature points, further optimization and screening are required. The optimization process can include removing spurious feature points (such as outliers caused by scanning errors or improper data processing) and merging adjacent feature points with similar characteristics. The screening process can comprehensively evaluate the feature points based on factors such as their spatial distribution, geometry, and physical properties, selecting the most representative and stable set of feature points.

[0189] An improved Freeman chain code can be used to convert crack information into a chain code sequence. The absolute value of the chain code mean difference is calculated to determine the location of suspicious inflection points, and false and redundant inflection points are eliminated. For structures such as tunnel linings and steel arches, methods such as geometric shape analysis and distance ratio calculation can be used to assess the stability and representativeness of characteristic points.

[0190] Step S42 reduces redundancy and interference factors in the feature point set by eliminating pseudo feature points and merging similar feature points; by comprehensively evaluating the stability and representativeness of the feature points, the most valuable feature point set is selected, providing a reliable reference benchmark for the subsequent alignment process.

[0191] Step S5: Correct the position of the feature point through the feature area according to the feature point.

[0192] In point cloud registration, feature points are key points in 3D space that possess significant geometric properties or can stably represent the structure of a point cloud. These points are typically automatically detected or manually annotated within the point cloud data and are characterized by uniqueness, stability, and repeatability. In point cloud data, feature points may include protruding structures on tunnel walls and geometric features of tunnel ancillary facilities.

[0193] The accuracy of feature point positions has a direct impact on point cloud registration results. If there are deviations in feature point positions, the registered point cloud data will be misaligned or overlapped, which will affect subsequent structural analysis and safety assessments.

[0194] In this embodiment, step S5, correcting the position of the feature point through the feature area based on the feature point may specifically include the following steps:

[0195] S51: Extract feature areas based on feature points.

[0196] Feature region extraction is the first step in correcting feature point positions. It aims to identify regions with significant geometric characteristics from point cloud data. These regions often contain rich structural information, which helps improve the accuracy and stability of feature point matching.

[0197] Geometric feature analysis can be used to extract feature regions: Geometric properties of point clouds (such as curvature and normal vectors) can be used to identify feature regions. For example, protruding structures or recessed areas on a tunnel wall can be identified by calculating the local curvature of the point cloud.

[0198] Machine learning can also be used to extract feature regions: supervised or unsupervised learning methods are used to train models to identify feature regions. These methods can automatically extract feature information from point clouds and have high generalization capabilities.

[0199] Through step S51, the extracted feature area provides a stable reference benchmark for subsequent feature point matching, which helps to improve the accuracy and robustness of matching.

[0200] S52: Based on the feature region extraction, a correspondence relationship between feature points in the two point cloud data sets is established to perform feature point matching.

[0201] A feature descriptor is calculated for each feature point to characterize the geometric structure and texture information around it. Feature descriptors include FPFH (Fast Point Feature Histogram) and SHOT (Signature of Histograms of Orientations).

[0202] Alternatively, methods such as nearest neighbor search and RANSAC (Random Sample Consensus) can be used to find matching feature point pairs in the feature descriptor space. These methods can efficiently process large amounts of feature point data and have a certain degree of noise resistance and robustness.

[0203] In step S52, a correspondence between feature points in two point cloud data sets can be established through feature point matching, providing a basis for subsequent feature point position correction.

[0204] S53: Correct the initial feature point positions based on the feature region and the matched feature point pairs.

[0205] Feature point position correction is based on the feature area and the matched feature point pairs, and the initial feature point positions are adjusted to eliminate errors and improve accuracy.

[0206] Rigid body transformation estimation can be used: using matched feature point pairs, the rigid body transformation matrix (including rotation and translation) is estimated through least squares or other optimization algorithms. This matrix can align one point cloud dataset to another point cloud dataset.

[0207] Alternatively, the Iterative Closest Point (ICP) algorithm can be used to iteratively optimize the locations of feature points based on the rigid body transformation estimate. The ICP algorithm iteratively adjusts the correspondence between point clouds to minimize the distance error between them, thereby accurately correcting the locations of feature points.

[0208] Feature region constraints can be used: During the feature point position correction process, feature region constraints can be introduced. For example, the geometric shape and size information of the feature region can be used to constrain the corrected feature point position to ensure that it conforms to the actual structural characteristics.

[0209] Step S53 can eliminate errors in the initial feature point positions by feature point position correction, improving the accuracy and robustness of the registration result. Meanwhile, the introduction of the feature region constraint helps to maintain the consistency of the registered point cloud data with the actual structure.

[0210] Step S6, based on the corrected feature point positions, performs point cloud registration.

[0211] Point cloud registration refers to the process of aligning two or more point cloud datasets in the same coordinate system through geometric transformation. Specifically, the goal of point cloud registration is to find a transformation T (including rotation R and translation t) that makes the transformed source point cloud Ps and the target point cloud Pt as coincident as possible. This technique is particularly important in tunnel accessory structure measurement, modeling and maintenance, as it can accurately align point cloud data acquired from different perspectives or different time points, generating a complete and consistent three-dimensional model.

[0212] In this embodiment, step S6, based on the corrected feature point positions, performs point cloud registration, which can specifically include the following steps:

[0213] S61, based on the corrected feature point positions, performs point cloud coarse registration.

[0214] Coarse registration is a relatively rough registration performed when the transformation between two point clouds is completely unknown, with the goal of providing a better initial value for fine registration. Coarse registration can be implemented through feature point-based matching, geometric feature-based matching, and statistical method-based matching, etc.

[0215] Use feature point-based matching methods such as RANSAC algorithm or 4PC algorithm. RANSAC algorithm estimates the transformation matrix through random sampling consensus, which can handle data containing a large amount of noise and outliers; 4PC algorithm estimates the transformation matrix by finding four coplanar corresponding points, which has high robustness and accuracy.

[0216] Through step S61, the two point clouds can be initially aligned to provide a better initial position for subsequent fine registration. Through coarse registration, the computational load and iteration number of fine registration can be reduced, improving registration efficiency.

[0217] S62, based on the feature point positions after coarse registration, performs point cloud fine registration.

[0218] Fine registration is a further optimization based on coarse registration to obtain a more accurate transformation. Fine registration can use the Iterative Closest Point algorithm (ICP) or its variants to perform point cloud fine registration based on the feature point positions after coarse registration.

[0219] First, determine the initial set of corresponding points. Before fine registration begins, it is necessary to determine the initial set of corresponding points between the source point cloud and the target point cloud. These corresponding points can be obtained from the results of coarse registration or manually selected.

[0220] Use the result of coarse registration as the initial corresponding point set; or manually select corresponding points in the source point cloud and the target point cloud, ensuring that these points have significant similarity in geometry or texture.

[0221] Determining the initial corresponding point set is the premise and basis for precise registration, and provides reliable input data for the subsequent optimization process.

[0222] Secondly, remove incorrect point pairs. Since the initial corresponding point set may contain incorrect point pairs (such as those caused by noise, occlusion, or mismatching), these point pairs will negatively impact the results of fine registration. Therefore, these incorrect point pairs need to be removed before fine registration.

[0223] You can use conditions such as distance threshold and normal vector angle threshold to remove incorrect point pairs. Specifically, you can calculate the distance and normal vector angle between each point pair and set a reasonable threshold to filter out valid point pairs.

[0224] Removing erroneous point pairs can improve the accuracy and robustness of fine registration and avoid registration errors caused by erroneous point pairs.

[0225] Then solve the transformation matrix. After determining the valid corresponding point set, it is necessary to solve the transformation matrix (including the rotation matrix R and the translation matrix T) so that the transformed source point cloud and the target point cloud have the highest possible overlap.

[0226] The transformation matrix is ​​solved using methods such as least squares or singular value decomposition (SVD). The least squares method solves the transformation matrix by minimizing the error between corresponding point pairs; the SVD method solves the optimal rotation and translation parameters by decomposing the covariance matrix.

[0227] Solving the transformation matrix is ​​one of the core steps of precise registration. By solving the transformation matrix, the optimal transformation relationship from the source point cloud to the target point cloud can be obtained, thereby achieving point cloud alignment.

[0228] Finally, iterative optimization is performed. Since the initial corresponding point set may have errors and the process of solving the transformation matrix may be affected by noise and errors, iterative optimization is required to further improve the accuracy of the registration.

[0229] The ICP algorithm or its variants can be used for iterative optimization. The ICP algorithm minimizes the error function by continuously iteratively updating the corresponding point set and the transformation matrix until the iterative termination condition is met (such as reaching the maximum number of iterations or the error change is less than a threshold).

[0230] During the iteration process, data structures such as KD trees can be used to accelerate the search for the nearest point and improve the efficiency of iterative optimization. In addition, error metrics such as point-to-line and point-to-surface can be introduced to improve the accuracy and robustness of the registration.

[0231] Iterative optimization can further reduce the impact of errors and noise on the registration results, and improve the accuracy and stability of the registration. Through iterative optimization, more accurate and consistent point cloud alignment results can be obtained.

[0232] S63, evaluate and optimize the point cloud registration results after fine registration.

[0233] After completing fine registration, the registration results need to be evaluated and optimized. The purpose of evaluation is to check the accuracy and consistency of the registration results to ensure the effectiveness of the registration process; the purpose of optimization is to further improve the accuracy and robustness of the registration.

[0234] The accuracy of the registration results can be evaluated using indicators such as overlap and root mean square error (RMSE); the registration results can be further optimized using filtering algorithms, optimization algorithms, and other methods.

[0235] Overlap refers to the proportion of overlapping points in the transformed source and target point clouds. This can be calculated by calculating the ratio of the number of overlapping points to the total number of points. The root mean square error (RMSE) is a commonly used metric for evaluating registration accuracy. It calculates the average error between pairs of corresponding points.

[0236] Registration result evaluation and optimization can ensure the effectiveness and accuracy of the registration process and improve the precision and consistency of the registration results. Through evaluation and optimization, the impact of errors and noise can be further reduced, resulting in more reliable and accurate point cloud alignment results.

[0237] The implementation of this embodiment has the following beneficial effects:

[0238] By acquiring point cloud data and performing preprocessing, noise and redundant information can be effectively removed, providing a high-quality data foundation for subsequent steps, ensuring the accuracy and reliability of the data, and laying a solid foundation for subsequent descriptor extraction and feature point screening;

[0239] By extracting feature points using tunnel accessory structure descriptors, the uniqueness and key information of the tunnel structure can be accurately captured, which not only improves the recognition accuracy of feature points but also significantly enhances the stability and robustness of registration.

[0240] Correcting the position of feature points through feature regions can further reduce errors and improve registration accuracy. This fully utilizes the geometric characteristics and spatial distribution of tunnel ancillary structures and achieves precise adjustment and optimization of feature points.

[0241] By performing point cloud registration based on the corrected feature point positions, the registration speed and efficiency can be significantly improved, effectively addressing the complexity and diversity of tunnel structures and providing strong technical support for the design, construction, and maintenance of tunnel projects.

[0242] It has multiple advantages such as high precision, high efficiency and high robustness, providing a new solution for 3D modeling and analysis of tunnel engineering.

[0243] The present invention can be used in a wide variety of general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. The present invention can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present invention can also be practiced in distributed computing environments in which tasks are performed by remote processing devices connected via a communications network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0244] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware using computer-readable instructions. The computer-readable instructions can be stored in a computer-readable storage medium. When executed, the program can include the processes in the above-described method embodiments. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0245] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0246] Example 2

[0247] Further referring to Figure 2 , Figure 2 As a structural diagram of an embodiment of the point cloud registration system of the water-related infrastructure accessory structure descriptor, as described above Figure 1 As an implementation of the method shown in Figure 1 The present application provides an embodiment of a point cloud registration system of a water-related infrastructure accessory structure descriptor, which corresponds to the method embodiment shown in

[0248] As shown in Figure 2 The embodiment of the point cloud registration system of the water-related infrastructure accessory structure descriptor 70 includes an acquisition module 71, a preprocessing module 72, an extraction module 73, a screening module 74, a correction module 75, and a registration module 76. Among them:

[0249] The acquisition module 71 is used to acquire point cloud data;

[0250] The preprocessing module 72 is used to preprocess the point cloud data;

[0251] The extraction module 73 is used to extract tunnel accessory structure descriptors based on the preprocessed point cloud data;

[0252] The screening module 74 is used to screen feature points through the tunnel accessory structure descriptors;

[0253] The correction module 75 is used to correct the feature point positions through the feature regions according to the feature points;

[0254] The registration module 76 is used to perform point cloud registration based on the corrected feature point positions.

[0255] Implementing the embodiment has the beneficial effects that:

[0256] By acquiring point cloud data and preprocessing, noise and redundant information can be effectively removed, providing a high-quality data basis for subsequent steps, ensuring the accuracy and reliability of the data, and laying a solid foundation for subsequent descriptor extraction and feature point screening;

[0257] By extracting feature points using tunnel accessory structure descriptors, the uniqueness and key information of the tunnel structure can be accurately captured, improving the recognition accuracy of feature points and significantly enhancing the stability and robustness of registration;

[0258] By correcting the feature point positions through the feature regions, errors can be further reduced and registration accuracy can be improved, fully utilizing the geometric features and spatial distribution of the tunnel accessory structure to achieve accurate adjustment and optimization of the feature points;

[0259] By performing point cloud registration based on the corrected feature point positions, the registration speed and efficiency can be significantly improved, effectively dealing with the complexity and diversity of tunnel structures, and providing strong technical support for the design, construction and maintenance of tunnel engineering.

[0260] With multiple advantages such as high precision, high efficiency and high robustness, it provides a new solution for three-dimensional modeling and analysis of tunnel engineering.

[0261] Embodiment three

[0262] To solve the above technical problems, the embodiment of the present application also provides a computer device. For details, please refer to Figure 3 , Figure 3 The structural diagram of an embodiment of the computer device of the present application.

[0263] The above computer device 8 includes a memory 81, a processor 82, and a network interface 83, which are connected to each other through a system bus. It should be noted that only the computer device 8 with components memory 81, processor 82 and network interface 83 is shown in the figure, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be alternatively implemented. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculation and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessor, application specific integrated circuit (ASIC), field programmable gate array (FPGA), digital signal processor (DSP), embedded device, etc.

[0264] The above computer device can be a desktop computer, a notebook computer, a palm computer and a cloud server, etc. The above computer device can interact with the user through a keyboard, a mouse, a remote controller, a touchpad or a voice control device, etc.

[0265] The memory 81 includes at least one type of readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 81 can be an internal storage unit of the computer device 8, such as a hard disk or a memory of the computer device 8. In other embodiments, the memory 81 can also be an external storage device of the computer device 8, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device 8. Of course, the memory 81 can also include both the internal storage unit and the external storage device of the computer device 8. In this embodiment, the memory 81 is generally used to store an operating system and various application software installed on the computer device 8, such as computer readable instructions of the point cloud registration method of the water-related infrastructure accessory structure descriptor, etc. In addition, the memory 81 can also be used to temporarily store various data that have been output or will be output.

[0266] The processor 82 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 82 is generally used to control the overall operation of the computer device 8. In this embodiment, the processor 82 is used to run computer readable instructions or process data stored in the memory 81, such as computer readable instructions of the point cloud registration method of the water-related infrastructure accessory structure descriptor.

[0267] The network interface 83 can include a wireless network interface or a wired network interface, and is generally used to establish a communication connection between the computer device 8 and other electronic devices.

[0268] By implementing this embodiment, the following beneficial effects can be achieved:

[0269] By obtaining point cloud data and preprocessing, noise and redundant information can be effectively removed, providing a high-quality data basis for subsequent steps, ensuring the accuracy and reliability of the data, and laying a solid foundation for subsequent descriptor extraction and feature point selection;

[0270] By extracting feature points using tunnel accessory structure descriptors, the uniqueness and key information of the tunnel structure can be accurately captured, which not only improves the recognition accuracy of feature points but also significantly enhances the stability and robustness of registration.

[0271] Correcting the position of feature points through feature regions can further reduce errors and improve registration accuracy. This fully utilizes the geometric characteristics and spatial distribution of tunnel ancillary structures and achieves precise adjustment and optimization of feature points.

[0272] By performing point cloud registration based on the corrected feature point positions, the registration speed and efficiency can be significantly improved, effectively addressing the complexity and diversity of tunnel structures and providing strong technical support for the design, construction, and maintenance of tunnel projects.

[0273] It has multiple advantages such as high precision, high efficiency and high robustness, providing a new solution for 3D modeling and analysis of tunnel engineering.

[0274] Example 4

[0275] The present invention also provides another embodiment, namely, providing a computer-readable storage medium, which stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable at least one processor to perform the steps of the point cloud registration method of the water infrastructure accessory structure descriptor as described above.

[0276] The implementation of this embodiment has the following beneficial effects:

[0277] By acquiring point cloud data and performing preprocessing, noise and redundant information can be effectively removed, providing a high-quality data foundation for subsequent steps, ensuring the accuracy and reliability of the data, and laying a solid foundation for subsequent descriptor extraction and feature point screening;

[0278] By extracting feature points using tunnel accessory structure descriptors, the uniqueness and key information of the tunnel structure can be accurately captured, which not only improves the recognition accuracy of feature points but also significantly enhances the stability and robustness of registration.

[0279] Correcting the position of feature points through feature regions can further reduce errors and improve registration accuracy. This fully utilizes the geometric characteristics and spatial distribution of tunnel ancillary structures and achieves precise adjustment and optimization of feature points.

[0280] By performing point cloud registration based on the corrected feature point positions, the registration speed and efficiency can be significantly improved, effectively addressing the complexity and diversity of tunnel structures and providing strong technical support for the design, construction, and maintenance of tunnel projects.

[0281] It has multiple advantages such as high precision, high efficiency and high robustness, providing a new solution for 3D modeling and analysis of tunnel engineering.

[0282] Through the above description of the embodiments, those skilled in the art will clearly understand that the above-mentioned embodiments and methods can be implemented using software and a necessary general-purpose hardware platform. Of course, hardware can also be used, but in many cases, the former is a more preferred embodiment. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk) and includes a number of instructions for enabling a terminal device (such as a mobile phone, computer, server, air conditioner, or network device) to execute the various embodiments and methods of the present invention.

[0283] Obviously, the embodiments described above are only some embodiments of the present invention, rather than all embodiments. The preferred embodiments of the present invention are given in the accompanying drawings, but they do not limit the patent scope of the present invention. The present invention can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive. Although the present invention has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions described in the aforementioned specific embodiments, or to make equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the present invention specification and drawings, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present invention.

Claims

1. A point cloud registration method for water infrastructure accessory structure descriptors, characterized in that: The steps include: Get point cloud data; Preprocessing the point cloud data; extracting tunnel-attached structure descriptors based on the preprocessed point cloud data; Filtering feature points using the tunnel accessory structure descriptor; According to the feature points, correcting the positions of the feature points through the feature areas; Perform point cloud registration based on the corrected feature point positions; The step of extracting the tunnel-attached structure descriptor based on the pre-processed point cloud data specifically includes: Identify the main structure of the tunnel; Extracting key points based on the main structure of the tunnel, wherein the key points refer to feature points representing the structural characteristics of the tunnel; For each extracted key point, calculate the feature area around it; According to the point cloud distribution and geometric features in the feature area, the local information of the key points is extracted; Calculate the global information of the key points based on the positional relationship between the key points and the tunnel axis; Filter key points based on their local and global information; The local information, global information and filtered key points of the extracted key points are encoded and quantized to generate a tunnel accessory structure descriptor.

2. The point cloud registration method for water infrastructure accessory structure descriptors according to claim 1 is characterized in that: The step of obtaining point cloud data specifically includes: According to the length of the tunnel, set the number, spacing and scanning parameters of the measuring stations; Perform scanning operations according to the set number of measuring stations, spacing, and scanning parameters to collect point cloud data.

3. The point cloud registration method for water infrastructure accessory structure descriptors according to claim 1 is characterized in that: The step of preprocessing the point cloud data specifically includes: removing noise points in the point cloud data; filtering the point cloud data; Segmenting the point cloud data into a plurality of subsets, each subset representing a different portion of the tunnel or ancillary structures; Simplifying the point cloud data; Use the bottom surface point cloud to correct the tilt error generated during measurement, and estimate the rotation angle around the X-axis and Y-axis and the height difference between the two point clouds through the bottom surface; The side point clouds at the same position of the source tunnel and the target tunnel are extracted and used to estimate the transfer amount.

4. The point cloud registration method for water infrastructure accessory structure descriptors according to claim 1 is characterized in that: The step of screening feature points using the tunnel-attached structure descriptor specifically includes: Use threshold segmentation and cluster analysis to identify the initial feature point set; Optimize and filter feature points from the initial feature point set.

5. The point cloud registration method for water infrastructure accessory structure descriptors according to claim 1 is characterized in that: The step of correcting the position of the feature point through the feature area according to the feature point specifically includes: Extracting feature areas according to the feature points; Based on the feature area extraction, the correspondence between the feature points in the two point cloud data sets is established to perform feature point matching; Based on the feature area and the matched feature point pairs, the initial feature point positions are corrected.

6. The point cloud registration method for water infrastructure accessory structure descriptors according to any one of claims 1 to 5, characterized in that: The step of performing point cloud registration based on the corrected feature point positions specifically includes: Based on the corrected feature point positions, perform point cloud coarse registration; Based on the position of the feature points after coarse registration, perform fine registration of the point cloud; Evaluate and optimize the point cloud registration results after fine registration.

7. A point cloud registration system for water infrastructure accessory structure descriptors, characterized by: include: Acquisition module, used to obtain point cloud data; A preprocessing module, used for preprocessing the point cloud data; an extraction module, configured to extract tunnel-attached structure descriptors based on the preprocessed point cloud data; A screening module, configured to screen feature points using the tunnel-attached structure descriptor; A correction module, configured to correct the position of the feature point through the feature area according to the feature point; Registration module, used to perform point cloud registration based on the corrected feature point positions; The extraction module is specifically used for: Identify the main structure of the tunnel; Extracting key points based on the main structure of the tunnel, wherein the key points refer to feature points representing the structural characteristics of the tunnel; For each extracted key point, calculate the feature area around it; According to the point cloud distribution and geometric features in the feature area, the local information of the key points is extracted; Calculate the global information of the key points based on the positional relationship between the key points and the tunnel axis; Filter key points based on their local and global information; The local information, global information and filtered key points of the extracted key points are encoded and quantized to generate a tunnel accessory structure descriptor.

Images