Coating thickness detection method in poor control point scene

Abstract

The invention discloses a coating thickness detection method in a poor control point scene. The method comprises the following steps: respectively collecting point cloud data of the inner wall of a closed space before and after construction by adopting a standing type three-dimensional laser scanner, and respectively recording the point cloud data as a point cloud before construction and a point cloud after construction; manufacturing cross-shaped marks with different reflectivity on the surface of the coating after construction, and measuring the thickness of the coating at each mark; identifying a mark center coordinate from the constructed point cloud and calculating a unit normal vector of the mark center coordinate; reversely deducing an original surface coordinate point set according to a formula; carrying out coarse registration on the point set and the point cloud before construction, and then carrying out fine registration by adopting a horizontal constraint iterative nearest point algorithm to obtain transformation parameters; and performing coordinate transformation on the point cloud after construction, and calculating the point-by-point thickness based on the triangular mesh of the point cloud before construction to realize three-dimensional visual detection of the coating thickness. According to the method, the problem of high-precision registration of two-stage point clouds in a poor control point scene is solved, and millimeter-level precision coating thickness distribution can be efficiently and comprehensively obtained.

Description

Technical Field

[0001] This invention relates to the field of engineering construction quality inspection technology, specifically to a high-precision full-coverage detection method for coating thickness in enclosed spaces lacking control points. This method comprehensively utilizes point cloud data acquired by a high-precision 3D laser scanner and discrete thickness data collected by a thickness gauge, and achieves 3D visualization detection of coating thickness through an innovative point cloud registration algorithm. It is particularly suitable for engineering scenarios lacking measurement control points, such as long straight tunnels, underground pipe corridors, cylindrical chambers, and elliptical chambers. Background Technology

[0002] In the construction and operation of water conservancy and hydropower projects, transportation tunnels, and underground utility tunnels, the thickness of coatings or linings is one of the key indicators for evaluating construction quality. For the thickness detection of covering layers such as tunnel secondary linings, anti-corrosion coatings, and insulation materials, it directly affects the safety performance and service life of the engineering structure. However, the core technical challenge of this type of detection work lies in the fact that the areas being inspected are often enclosed or semi-enclosed underground spaces, characterized by narrow spaces, poor lighting conditions, and the inability to establish traditional measurement control points. Due to the limitations of the engineering characteristics, it is impossible to establish mapping benchmarks as corresponding points within the limited space. Furthermore, the environment before and after spraying or applying the material or coating is completely different; the interior is completely encased in material, lacking obvious structural objects as references, resulting in the absence of mapping benchmarks for measurement work. This environment is called a control-point-deficient scenario, and achieving millimeter-level accuracy in coating thickness measurement under such conditions is an extremely significant challenge.

[0003] Currently, two main technical approaches are used for coating thickness inspection in scenarios lacking control points. The first is the traditional manual point-by-point measurement method using thickness gauges. Operators need to erect scaffolding to approach the surface being measured, using electromagnetic or ultrasonic thickness gauges to perform point measurements at predetermined locations, and then using a total station to determine the spatial coordinates of the measurement points. This method has significant limitations. First, it is inefficient; large-area inspections require substantial manpower and resources, and the time cost is very high. Second, scaffolding is required to reach a higher position to measure close to the surface, resulting in a high operational risk. Furthermore, erecting scaffolding covering the entire surface of large engineering projects is often impractical. In addition, the measurement results are discrete point data, which can only be presented in tabular or report form, showing only point-like quality inspection results and unable to provide surface or three-dimensional visualizations. The understanding and application of the inspection results are difficult, and blind spots exist, making it difficult to achieve full-coverage inspection.

[0004] The second method is a point cloud comparison method based on 3D laser scanning. This involves scanning the target surface before and after coating application to obtain point cloud data for both phases. Point cloud registration is then performed using geometric features such as center points and bounding box coordinates, and the distance difference is calculated as the coating thickness. This method has relatively high measurement efficiency and can obtain spatially continuous thickness data. However, existing point cloud comparison methods still face significant technical challenges in practical applications. First, in scenarios lacking control points, stable common feature points are missing as registration benchmarks, making it difficult to accurately unify the coordinate systems of the two point clouds. Second, when using geometric features such as center points and bounding boxes for registration, these points can shift due to the non-uniformity of the coating during application. This is fatal for millimeter-level precision quality inspection, introducing unacceptable systematic errors and leading to unstable registration results and inaccurate accuracy estimation. Furthermore, traditional iterative nearest-point algorithms are prone to mismatches when dealing with special geometric shapes such as cylindrical or elliptical caverns due to uneven point cloud distribution, affecting the final thickness calculation accuracy. Summary of the Invention

[0005] To address the problems of insufficient registration accuracy, low detection efficiency, and inability to achieve full coverage in existing coating thickness detection technologies for scenarios lacking control points, this invention provides a coating thickness detection method based on a high-precision 3D laser scanner and thickness gauge. The core innovation of this method lies in: setting marker points on the coating surface after construction and measuring their thickness; combining this with point cloud normal vector estimation technology to reverse-calculate the 3D coordinates of the original surface beneath the coating. This achieves accurate registration of the two-stage point clouds without relying on geometric feature points such as center points or bounding boxes that are prone to shift due to uneven construction, ultimately obtaining a point-by-point full-coverage distribution result of the coating thickness.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for detecting coating thickness in a scenario lacking control points, characterized by comprising the following steps:

[0007] A stand-alone 3D laser scanner was used to perform a high-precision 3D laser scan of the inner wall of the enclosed space before construction. The 3D laser scanner has an automatic leveling function, and performs registration, stitching, and noise reduction processing on the collected multi-point cloud to ensure that the point cloud surface is smooth and free of error points. The processed point cloud is recorded as the point cloud before construction. ;

[0008] After the coating is applied, several marks are made on the surface. These marks are drawn in a cross shape using a special paint pen with a different reflectivity than the coating surface. The marks are evenly distributed across the area to be inspected, according to the spatial size. At the center of each mark, a thickness gauge is used to vertically measure the coating thickness, and the coating thickness value at each mark location is recorded. ,in To mark the total number;

[0009] The surface of the coating after construction was scanned using a 3D laser scanner. The collected point clouds were registered, stitched, and color-coded. The complete point cloud was then denoised and noise removed. The difference in reflectivity between the marked paint and the coating surface was used to identify the marked locations during point cloud intensity rendering. The processed point cloud was recorded as the post-construction point cloud. ;

[0010] Cloud point analysis after construction The three-dimensional coordinates of each marker center are determined through manual interaction or image recognition. Query the neighborhood point set of each marked point The unit normal vector at this point is calculated using principal component analysis. The normal vector points to the outside of the coating;

[0011] According to the marked coordinates Normal vector and coating thickness The coordinates of the original surface under the coating are calculated along the opposite direction of the normal vector. The calculation formula is as follows: ,in These are the original surface coordinates obtained by reverse engineering;

[0012] The original surface coordinates obtained by reverse engineering are used as registration reference feature points and compared with the point cloud before construction. For registration, coarse registration is first performed through manual interaction, involving 3D translation and in-plane rotation. Then, fine registration is performed using an iterative nearest-point algorithm with horizontal constraints. This algorithm estimates only the rotation angle around the vertical axis, assuming both point clouds have been horizontally leveled. With three-dimensional translation vector , to obtain the transformation parameters;

[0013] Using transformed parameters to analyze post-construction point clouds Perform coordinate transformation to obtain the transformed point cloud The Delaunay triangulation algorithm was used to analyze the point cloud before construction. Constructing triangular mesh surfaces, for Calculate the distance to each point in the equation. The nearest distance of the triangular mesh is used as the coating thickness value at that point. The thickness value is stored in the attribute field of the point cloud and the thickness feature is displayed by color rendering through point cloud visualization software.

[0014] In the above technical solution, the steps The specific process of calculating the surface normal vector using principal component analysis is as follows: query the marked points. In point cloud Neighborhood point set Principal component analysis is performed on the neighborhood point set to calculate the eigenvalues ​​and eigenvectors of the covariance matrix. The eigenvector corresponding to the smallest eigenvalue is taken as the unit normal vector of the marked point. .

[0015] In the above technical solution, the steps The transformation model of the iterative nearest point algorithm with medium-level constraints is expressed as follows: ,in To bypass Axis rotation Rotation matrix of angle, , The rotation matrix is ​​a three-dimensional translation vector, and the first row is a three-row, three-column matrix. The second line The third line .

[0016] In the above technical solution, the optimization objective of the iterative nearest point algorithm for horizontal constraints is to minimize the transformed inverse original surface coordinates and the point cloud before construction. The sum of squared distances between corresponding points in the algorithm reduces the degree of freedom from 6 to 4 by horizontal constraints compared to the traditional six-degree-of-freedom iterative nearest-point algorithm, thus improving the registration stability for special geometric shapes such as cylindrical and elliptical caverns.

[0017] In the above technical solution, the steps The markings are made by using a special paint pen to draw cross-shaped marks on the inner wall surface. The significant difference in reflectivity between the paint and the inner wall material ensures that the marks can be clearly identified when rendering the point cloud intensity. The number and distribution of the marks are determined comprehensively based on the size of the space, the expected complexity of the coating thickness variation, and the detection accuracy requirements.

[0018] In the above technical solution, the steps The specific method for calculating the distance from a point to the triangular network is as follows: For point clouds after construction... any point in Before construction, cloud observation was conducted. Find the nearest triangle ABC in the constructed triangular mesh, and calculate the vertical distance from point P to the plane containing triangle ABC as the coating thickness at that location. The distance calculation takes into account the vertical distance from the point to the nearest triangular facet.

[0019] In the above technical solutions, the enclosed space includes tunnels, underground utility tunnels, cylindrical chambers or elliptical chambers. Due to the engineering characteristics, it is impossible to set up traditional measurement control points in these areas within a limited space, which is a scenario with few control points.

[0020] In the above technical solution, the leveling accuracy of the stand-up 3D laser scanner used in steps S1 and S3 reaches the second level, and the vertical axis of the scanner is precisely aligned with the direction of gravity, ensuring that the vertical direction of each station cloud is consistent with the direction of gravity, and providing an accurate horizontal reference for the horizontal constraint iterative nearest point algorithm.

[0021] In the above technical solution, the coarse registration stage in step S6 uses a manual interactive method to perform preliminary alignment of the two sets of data through three-dimensional translation and horizontal rotation, so that the original surface coordinate points obtained by back-reasoning are consistent with the point cloud before construction. The visual alignment is basically consistent, providing a good initial value for fine registration. Since both scans have been automatically leveled, only translation and rotation in the horizontal plane need to be adjusted during coarse registration, without adjusting the pitch and roll angles.

[0022] In the above technical solution, the coating thickness distribution results are stored in the form of point cloud attribute fields, which can be used for three-dimensional visualization rendering of the thickness distribution in any point cloud processing software. The spatial distribution characteristics of the thickness can be intuitively displayed through pseudo-color rendering and other methods, which facilitates the identification of thickness abnormal areas and provides comprehensive and intuitive data support for engineering quality assessment.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] First, this invention innovatively proposes a technical solution to infer the original surface coordinates from thickness gauge data, breaking through the technical bottleneck of two-stage point cloud registration in scenarios lacking control points. It does not rely on geometric feature points such as the center of the circle and the bounding box, which are prone to shift due to uneven construction, thus fundamentally improving the registration accuracy and stability and ensuring the reliability of the detection results.

[0025] Secondly, this invention employs a point-to-surface detection strategy. Only a few marks need to be made on the coating surface and a limited number of thickness gauge measurements performed to obtain a comprehensive, point-by-point thickness distribution result through point cloud comparison. Compared to traditional point-by-point measurement methods, this significantly reduces on-site work intensity and labor costs, substantially improves detection efficiency, avoids the blind spot problem inherent in traditional point-by-point measurement methods, and eliminates the need for large-scale scaffolding and other high-risk operations, thus improving construction safety.

[0026] Third, the horizontal constraint iterative nearest point algorithm makes full use of the automatic leveling function of the high-precision laser scanner, reducing the transformation degrees of freedom from 6 to 4. It is particularly suitable for enclosed spaces with axisymmetric features such as tunnels, underground pipe corridors, and circular tanks, significantly improving the registration convergence and accuracy in such scenarios and solving the problem that traditional iterative nearest point algorithms are prone to mismatches under special geometric shapes.

[0027] Fourth, this invention can acquire point-by-point thickness data of the inner wall after construction, forming a spatially continuous thickness distribution result, overcoming the limitation of traditional methods that can only obtain discrete point thickness data. The detection results are stored in the form of point cloud attribute fields, and the thickness distribution can be visualized and rendered in three dimensions in any point cloud processing software, which facilitates intuitive identification of thickness anomaly areas and provides clear guidance for quality control and rectification.

[0028] Fifth, the method of the present invention has wide applicability and can carry out high-precision thickness detection on the inner walls of curved surfaces with different geometric shapes such as cylindrical caverns and elliptical caverns. In actual engineering applications, the detection accuracy can reach the millimeter level, meeting the quality detection requirements of high-value coatings. It has achieved good application results in several tunnel projects. Attached Figure Description

[0029] Figure 1 This is a technical flowchart of the method of the present invention, which shows the complete process from point cloud acquisition before construction to final coating thickness calculation.

[0030] Figure 2 This diagram illustrates how to calculate the distance from a point to a triangle, showing how to calculate the distance from a point in the post-construction point cloud to the triangular mesh constructed from the pre-construction point cloud, thereby determining the coating thickness. In the diagram, P represents a point in the post-construction inner wall point cloud to be calculated, and triangle ABC is a triangle constructed from the pre-construction inner wall point cloud. By calculating the Euclidean distance d from point P to the plane containing triangle ABC, the coating thickness at that location can be obtained.

[0031] Figure 3 This is a schematic diagram illustrating the principle of reverse calculation of the original surface coordinates.

[0032] Figure 4 A schematic diagram of horizontal constraint ICP registration (top view). Detailed Implementation

[0033] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0034] See Figure 1 , Figure 2 It can be seen that the coating thickness detection method in the scenario of lack of control points of the present invention includes the following steps:

[0035] S1. Use a stand-up 3D laser scanner to perform a high-precision 3D laser scan of the inner wall of the enclosed space before construction. The 3D laser scanner has an automatic leveling function, and performs registration, stitching, and noise reduction processing on the collected multi-point cloud to ensure that the point cloud surface is smooth and free of error points. The processed point cloud is recorded as the point cloud before construction. ;

[0036] S2. After the coating is applied, several marks are made on the surface. The marks are drawn in a cross shape using a special paint pen with a different reflectivity than the coating surface. The marks are evenly distributed in the area to be inspected according to the spatial size. At the center of each mark, a thickness gauge is used to vertically measure the coating thickness, and the coating thickness value at each mark location is recorded. ,in To mark the total number;

[0037] S3. Use a 3D laser scanner to scan the surface of the coating after construction. Register, stitch, and color-grade the collected point clouds. Denoise and remove noise from the complete point cloud. Use the difference in reflectivity between the marked paint and the coating surface to identify the marked positions during point cloud intensity rendering. Record the processed point cloud as the post-construction point cloud. ;

[0038] S4, Cloud Pointing after Construction The three-dimensional coordinates of each marker center are determined through manual interaction or image recognition. Query the neighborhood point set of each marked point The unit normal vector at this point is calculated using principal component analysis. The normal vector points to the outside of the coating;

[0039] S5. Based on the marked coordinates Normal vector and coating thickness The coordinates of the original surface under the coating are calculated along the opposite direction of the normal vector. The calculation formula is as follows: ,in These are the original surface coordinates obtained by reverse engineering;

[0040] like Figure 3 As shown: The core innovation of this invention is the geometric principle of the original surface coordinate inverse calculation method. The blue curve in the figure represents the original surface before construction. The red curve represents the surface after coating application. The three-dimensional coordinates of the marked points on the coating surface are known. The unit normal vector of that point And the coating thickness measured by the thickness gauge Offset thickness in the opposite direction of the normal vector This allows us to calculate the corresponding coordinates of the original surface under the coating.

[0041] S6. Use the original surface coordinate points obtained by reverse engineering as registration reference feature points and the point cloud before construction. For registration, coarse registration of 3D translation and horizontal plane rotation is first performed through manual interaction. Then, fine registration is performed using the horizontally constrained iterative nearest point algorithm. The horizontally constrained iterative nearest point algorithm estimates only the rotation angle θ around the vertical axis and the 3D translation vector t to obtain the transformation parameters, provided that both point clouds have been horizontally leveled.

[0042] S7. Using transformation parameters to analyze post-construction point clouds Perform coordinate transformation to obtain the transformed point cloud The Delaunay triangulation algorithm was used to analyze the point cloud before construction. Constructing triangular mesh surfaces, for Calculate the distance to each point in the equation. The nearest distance of the triangular mesh is used as the coating thickness value at that point. The thickness value is stored in the attribute field of the point cloud and the thickness feature is displayed by color rendering through point cloud visualization software.

[0043] This invention creatively solves the technical challenge of coating thickness detection in scenarios lacking control points. The method cleverly combines 3D laser scanning technology with point measurement data from a thickness gauge, overcoming the bottleneck of traditional methods that rely on geometric feature point registration. Specifically, this method utilizes markers created on the coating surface by a thickness gauge to perform a limited number of thickness measurements. Combined with point cloud normal vector estimation technology, it reverse-calculates the 3D coordinates of the original surface beneath the coating, thus achieving accurate registration of point clouds before and after construction without the need for traditional measurement control points. This approach ensures registration accuracy while significantly reducing on-site workload, realizing a highly efficient detection strategy that uses points instead of surfaces. Compared to traditional point-by-point measurement methods, this method reduces the detection workload by over 95% while obtaining spatially continuous, full-coverage thickness data, providing millimeter-level precision 3D visualization results for engineering quality assessment. Furthermore, the horizontally constrained iterative nearest-point registration algorithm employed in this method fully utilizes the automatic leveling function of the high-precision scanner, significantly improving the stability and convergence of the registration, making it particularly suitable for enclosed spaces with axisymmetric geometries such as cylindrical and elliptical shapes.

[0044] The specific process of calculating the surface normal vector using principal component analysis in step S4 is as follows: querying the marker points. In point cloud Neighborhood point set Principal component analysis is performed on the neighborhood point set to calculate the eigenvalues ​​and eigenvectors of the covariance matrix. The eigenvector corresponding to the smallest eigenvalue is taken as the unit normal vector of the marked point. .

[0045] This invention employs the mathematically rigorous and robust principal component analysis (PCA) method to estimate surface normal vectors. As a mature data dimensionality reduction technique, PCA can accurately extract key geometric features from the spatial distribution of a neighborhood point set. By calculating the eigenvalue decomposition of the covariance matrix, the eigenvector corresponding to the smallest eigenvalue represents the direction of least change in the neighborhood point set, i.e., the surface normal direction. This method exhibits good noise resistance; even with a small number of noise points or measurement errors in the neighborhood point set, stable and reliable normal vector estimation results can be obtained through statistical analysis. Compared to other normal vector calculation methods such as least squares plane fitting, PCA is mathematically more elegant, computationally more efficient, and does not require prior assumptions about the local geometry of the surface. For minor unevenness or spraying inconsistencies that may exist on the coating surface, PCA can effectively smooth these local disturbances through neighborhood statistical properties, ensuring the accuracy and stability of the normal vector estimation, thus providing reliable basic data for subsequent back-calculation of the original surface coordinates.

[0046] The transformation model of the iterative nearest point algorithm for the horizontal constraint in step S6 is expressed as follows: ,in Let be the rotation matrix for rotating about the Z-axis by an angle θ. , The rotation matrix is ​​a three-dimensional translation vector, and the first row is a three-row, three-column matrix. The second line The third line .

[0047] like Figure 4 The diagram illustrates the working principle of the horizontally constrained ICP registration algorithm using a top-down view of a tunnel cross-section. The left side shows the state before registration, represented by the point cloud before construction. There is a positional deviation between the original surface points obtained by reverse engineering and the right side; the right side shows the registered state, and the two are accurately aligned after optimization by the horizontal constraint ICP algorithm. Compared with the traditional 6-DOF ICP algorithm, the horizontal constraint ICP only needs to optimize 4 parameters (1 rotation angle + 3 translation components), which significantly improves the registration stability of special geometric shapes such as cylindrical and elliptical caverns.

[0048] The rotation matrix around the Z-axis is in the following form: ; ;

[0049] The objective constraint is:

[0050] in, This represents the i-th point in the source point cloud. Indicates the target point cloud and The corresponding nearest neighbor, This represents the point-to-point mapping established by the nearest neighbor search. This represents the set of point-to-point pairs established during the ICP process. This represents the rotation matrix about the vertical axis. Let be the rotation angle about the vertical axis. Let be the rotation vector. It is the Euclidean L2 norm, used to measure the distance between pairs of points.

[0051] This invention precisely defines the mathematical form of the horizontal constraint transformation model, giving the registration algorithm a clear mathematical foundation and repeatability. By restricting the transformation degrees of freedom to a single rotation angle θ around the vertical axis and a three-dimensional translation vector t, the model fully utilizes the prior knowledge provided by the automatic leveling function of a high-precision 3D laser scanner, namely, that the point clouds before and after construction are precisely aligned in the vertical direction. The introduction of this constraint condition significantly simplifies the complexity of the registration problem, transforming the traditional six-degree-of-freedom registration problem into a four-degree-of-freedom problem, significantly improving the convergence speed and stability of the algorithm. The explicit form of the rotation matrix ensures the orthogonality of the transformation and the clarity of its geometric meaning, avoiding the matrix degradation problem that may occur in numerical calculations. For slender enclosed spaces such as tunnels and utility tunnels, as well as cylindrical and elliptical caverns with axisymmetric features, the horizontal constraint model can effectively prevent the problems of axial slippage or incorrect rotation around non-vertical axes that are common in traditional iterative nearest-point algorithms in these scenarios, thereby ensuring the geometric correctness and physical rationality of the registration results.

[0052] The optimization objective of the iterative nearest point algorithm for the horizontal constraints is to minimize the difference between the transformed, back-calculated original surface coordinates and the point cloud before construction. The sum of squared distances between corresponding points in the algorithm reduces the degree of freedom from 6 to 4 by horizontal constraints compared to the traditional six-degree-of-freedom iterative nearest-point algorithm, thus improving the registration stability for special geometric shapes such as cylindrical and elliptical caverns.

[0053] This invention clarifies the form of the optimization objective function, providing a clear mathematical guideline for the implementation of the registration algorithm. By minimizing the sum of squared distances, the algorithm can find transformation parameters that best match the original surface coordinate point set obtained through back-reaming with the point cloud before construction, thus ensuring registration accuracy. The square-form optimization objective has good mathematical properties; its gradient and Hessian matrix are easy to calculate, facilitating the use of efficient numerical optimization methods such as gradient descent or Newton's method. More importantly, by reducing the degrees of freedom from six to four, the parameter space dimension of the algorithm is significantly reduced. This not only reduces computational complexity but, more importantly, significantly improves the condition number and convergence of the optimization problem. When dealing with geometric shapes with rotational or approximate symmetry, such as cylindrical or elliptical caverns, traditional six-degree-of-freedom algorithms often face the problem of multiple local optima, easily falling into incorrect registration states. Horizontal constraints, by introducing prior geometric knowledge, effectively eliminate these ambiguities, making the optimization problem have a more explicit global optimum, significantly improving the robustness and reliability of registration, and ensuring correct registration results in various complex geometric scenarios.

[0054] The markings in step S2 are made by using a special paint pen to draw cross-shaped markings on the inner wall surface. The significant difference in reflectivity between the paint and the inner wall material ensures that the markings can be clearly identified during point cloud intensity rendering. The number and distribution of the markings are determined comprehensively based on the spatial size, the expected complexity of coating thickness variation, and the detection accuracy requirements.

[0055] This invention provides a simple, practical, and low-cost marking method while ensuring the recognizability of the markings in subsequent point cloud processing. The cross-shaped pattern design makes the center position of the marking easy to identify manually or automatically located through image processing algorithms, improving the accuracy of marking point coordinate extraction. The design cleverly solves the marking recognition problem by utilizing the difference in reflectivity between a special paint pen and the inner wall material, resulting in a clear contrast in the point cloud intensity image acquired by a 3D laser scanner. This facilitates rapid positioning of the marking through simple threshold segmentation or intensity rendering, eliminating the need for complex image recognition algorithms. Compared to traditional marking methods such as physical targets or reflective stickers, paint pen marking offers advantages such as ease of production, extremely low cost, resistance to peeling, and no damage to the coating surface, making it particularly suitable for large-area inspection operations. The flexible configuration strategy for the number and distribution of markings fully considers the actual needs of different engineering scenarios. It allows for increased marking density in areas with complex thickness variations to improve registration accuracy, while appropriately reducing the number of markings in areas with uniform thickness distribution to lower operating costs, achieving an optimal balance between detection accuracy and operational efficiency. This adaptive marking layout scheme provides excellent flexibility and operability for the practical application of the method.

[0056] The specific method for calculating the distance from the point to the triangular network in step S7 is as follows: For the point cloud after construction... For any point P in the cloud, before construction... Find the nearest triangle ABC in the constructed triangular mesh, and calculate the vertical distance from point P to the plane containing triangle ABC as the coating thickness at that location. The distance calculation takes into account the vertical distance from the point to the nearest triangular facet.

[0057] This invention employs a distance calculation method based on triangular meshes. This method fully utilizes the spatial distribution information of the point cloud before construction to achieve accurate reconstruction and representation of the original surface geometry. The triangular meshes generated by the Delaunay triangulation algorithm have excellent geometric properties, providing a continuous surface representation while preserving the geometric features of the original point cloud. This avoids the discontinuities and instabilities that may arise from directly using discrete point clouds for distance calculations. By finding the nearest triangle and calculating the perpendicular distance from a point to the triangle plane, this method ensures that the geometric meaning of the thickness value is clear, representing the true thickness of the coating surface point along its normal direction to the original surface. This definition of perpendicular distance is completely consistent with the physical meaning of coating thickness, avoiding measurement errors caused by surface tilt compared to simple nearest-point distance calculations. Furthermore, the triangular mesh-based method has good computational efficiency. The nearest triangle can be quickly located using spatial indexing structures such as octrees or kd-trees, making it possible to perform point-by-point thickness calculations on large-scale point clouds containing hundreds of millions of points. This method also has good interpolation characteristics. Even in areas with low density in the original point cloud, reasonable thickness estimates can be obtained through the continuous surface representation of the triangular mesh, ensuring the spatial integrity and continuity of the detection results.

[0058] The enclosed space includes tunnels, underground utility tunnels, cylindrical chambers, or elliptical chambers. Due to their engineering characteristics, traditional measurement control points cannot be set up in these areas within a limited space, which constitutes a scenario lacking control points.

[0059] This invention clearly defines its application scope and technical focus, highlighting the professionalism and practical value of the method in solving specific engineering problems. By listing typical control point-deficient scenarios such as tunnels, underground utility tunnels, cylindrical caverns, and elliptical caverns, the claims clearly indicate the target application areas of the method. These scenarios are widely representative and have significant engineering value in industries such as water conservancy and hydropower, transportation, and urban infrastructure. These enclosed spaces share the common characteristics of being long and narrow or having unique geometric shapes, making it difficult for traditional surveying methods to establish stable mapping benchmark networks. Significant differences exist between the pre- and post-construction scenarios, and there is a lack of obvious corresponding feature points. This is precisely the core technical challenge that the method of this invention aims to solve. The clear definition of the application scenarios provides clear constraints and optimization objectives for the technical solution design of the method, enabling technical details such as the horizontal constraint registration algorithm to fully utilize the geometric characteristics of these scenarios and achieve targeted performance optimization. Furthermore, the claims provide clear directional guidance for the technical promotion and market application of this invention, helping potential users quickly determine the applicability of the method, reducing the communication costs of technology transfer, and demonstrating good engineering practical value and market application prospects.

[0060] The leveling accuracy of the gantry-type 3D laser scanner used in steps S1 and S3 reaches the second level. The vertical axis of the scanner is precisely aligned with the direction of gravity, ensuring that the vertical direction of each station cloud is consistent with the direction of gravity, thus providing an accurate horizontal reference for the horizontal constraint iterative nearest point algorithm.

[0061] This invention clarifies the technical requirements for the scanning equipment, ensuring the hardware foundation and accuracy guarantee for the method's implementation. Second-level leveling accuracy means that the deviation between the scanner's vertical axis and the actual gravity direction is controlled within arcseconds. This extremely high leveling accuracy provides a crucial prior condition for the successful implementation of the horizontal constraint registration algorithm. By ensuring that the point cloud data from the two phases of scanning before and after construction maintain a consistent gravity reference in the vertical direction, this requirement significantly simplifies the complexity of the point cloud registration problem, allowing the registration process to reasonably ignore changes in pitch and roll angles, reducing the six-degree-of-freedom problem to a four-degree-of-freedom problem. This dimensionality reduction not only improves the computational efficiency of the registration algorithm but, more importantly, significantly enhances its robustness and convergence, effectively avoiding the registration instability problems commonly encountered by traditional registration methods when dealing with slender or axisymmetric geometries. The second-level leveling accuracy requirement is consistent with the current technical level of mainstream high-precision 3D laser scanners, ensuring the feasibility and practicality of the method. Meanwhile, this requirement also sets clear equipment selection standards for the promotion and application of the method, making it easier for engineering implementation units to select appropriate scanning equipment and ensuring the consistency and reliability of the method in different projects. Through the organic combination of hardware accuracy assurance and algorithm innovation, this invention achieves system-level technical optimization.

[0062] In step S6, the coarse registration stage uses a manual interactive method to perform preliminary alignment of the two sets of data through three-dimensional translation and horizontal rotation, so that the original surface coordinate points obtained by back-reasoning are consistent with the point cloud before construction. The visual alignment is basically consistent, providing a good initial value for fine registration. Since both scans have been automatically leveled, only translation and rotation in the horizontal plane need to be adjusted during coarse registration, without adjusting the pitch and roll angles.

[0063] This invention employs a human-machine collaborative registration strategy, fully leveraging the advantages of human interaction in initial alignment while laying a solid foundation for subsequent automated fine registration. The manual interaction in the coarse registration stage is intuitive and simple; operators can quickly achieve initial point cloud alignment through a visual software interface, without complex mathematical calculations or parameter settings. This interactive operation method lowers the technical requirements for operators, reducing the barrier to entry for the method. More importantly, manual coarse registration can utilize the operator's intuitive understanding of the scene's geometric features to quickly eliminate obviously erroneous registration directions, preventing subsequent fine registration from getting stuck in local optima. By providing visually consistent initial alignment results, coarse registration significantly reduces the search space of the fine registration algorithm, improving convergence speed and success rate. Since both scans have been leveled, coarse registration only requires translation and rotation adjustments within the horizontal plane, greatly reducing the degree of operational freedom and making the alignment operation simpler and more intuitive. This staged registration strategy ensures robustness and improves overall efficiency, achieving a complementary advantage between automated processing and human experience. The coarse registration stage also provides operators with an opportunity to check data quality, which can promptly detect anomalies or errors in the scanned data and ensure the reliability of subsequent processing.

[0064] The coating thickness distribution results are stored in the form of point cloud attribute fields, which can be used for three-dimensional visualization rendering of the thickness distribution in any point cloud processing software. The spatial distribution characteristics of the thickness can be intuitively displayed through pseudo-color rendering and other methods, making it easy to identify areas with abnormal thickness and providing comprehensive and intuitive data support for engineering quality assessment.

[0065] This invention provides a standardized, highly compatible, intuitive, and efficient way to express test results. Storing thickness values ​​as attribute fields of point clouds is a universal data organization method that conforms to international standard formats for point cloud data, such as LAS and E57, ensuring the exchangeability and long-term availability of the results data. This storage method allows test results to be directly opened and processed in various mainstream point cloud processing software such as CloudCompare, RiScan, and ReCap, without the need for dedicated software or format conversion, greatly improving the ease of application of the results data. Pseudo-color rendering technology maps continuous thickness values ​​to intuitive color gradients, enabling complex three-dimensional spatial data to be quickly conveyed visually. Operators can easily identify the overall pattern and local anomalies in the thickness distribution. Compared to traditional tabular or two-dimensional planar diagram representations, the thickness rendering of three-dimensional point clouds retains complete spatial location information and geometric relationships, making the location of quality problems more accurate and the formulation of rectification measures more targeted. This visual representation method is particularly suitable for reporting test results to managers or owners without technical backgrounds, lowering the barrier to communication of professional technical information. Furthermore, the storage method of point cloud attribute fields facilitates subsequent data analysis and statistics, such as calculating quantitative indicators like average thickness, standard deviation, area and distribution of areas with insufficient thickness, providing a reliable data foundation for a comprehensive assessment of engineering quality.

[0066] This invention systematically solves the technical challenge of coating thickness detection in enclosed spaces, achieving an optimal balance between accuracy, efficiency, and cost. The solution innovatively integrates 3D laser scanning technology, thickness gauge point measurement technology, principal component analysis, horizontal constraint iterative nearest-point registration algorithm, and 3D visualization technology to construct a complete technical system. This overcomes multiple technical bottlenecks faced by traditional methods in scenarios lacking control points, such as insufficient registration accuracy, low detection efficiency, and limited coverage.

[0067] This invention represents the forefront of technological development in this field, and has significant theoretical and practical value for improving the technical level of engineering construction quality inspection, reducing inspection costs, and ensuring engineering safety. The successful development and application of this technical solution not only provides related industries with efficient and reliable quality inspection methods, but also opens up new directions for the in-depth application of 3D laser scanning technology in the engineering field, yielding substantial social and economic benefits.

[0068] Example

[0069] This embodiment uses an underground chamber project in a certain province as an example to illustrate the specific implementation process and technical effects of the method of the present invention. To ensure high consistency and sealing of the inner wall of the underground chamber, a high-performance anti-corrosion coating material needs to be sprayed onto the surface of the existing metal inner wall. The inner wall of the chamber has a cylindrical structure with an inner diameter of approximately 8 meters and a length of approximately 50 meters. The sprayed material is a high-performance anti-corrosion coating with a designed thickness of approximately 20 millimeters. Because the chamber is a closed space and relatively long, it is a typical uncontrolled area where stable mapping benchmarks cannot be established. Traditional thickness detection methods are insufficient to meet the requirements of full-coverage, high-precision detection.

[0070] The coating thickness of the chamber was detected using the method proposed in this invention. The specific implementation steps are as follows:

[0071] Step S1: Point Cloud Acquisition Before Construction. Before coating application, the interior walls of the chamber were scanned using a FARO Focus S350 stand-alone 3D laser scanner. This scanner has a ranging accuracy of ±2 mm, an angle accuracy of 19 seconds, and a high-precision automatic leveling function with leveling accuracy down to the second level. Considering the length and geometric features of the chamber, eight scanning stations were set along the chamber's axis, with an adjacent station spacing of approximately 7 meters to ensure the integrity of the scan coverage and the uniformity of the point cloud density. The scanning resolution for each station was set to 1 / 4, and the scanning range was set to a 360-degree horizontal viewing angle and a 300-degree vertical viewing angle. After scanning, FARO Scene software was used for automatic registration of the multi-station point clouds, with the registration error controlled within 1 mm. Subsequently, statistical filtering denoising was performed on the point clouds to remove noise points generated during the scanning process and to delete irrelevant point cloud data from outside the chamber. The final pre-construction point cloud was obtained. It contains approximately 200 million points, with a point cloud density of about 10 points per square centimeter, and a smooth surface without any errors.

[0072] Step S2: Marking and Thickness Measurement. After the coating spraying is completed and cured, markings are uniformly made on the inner wall surface of the chamber. Considering the geometric dimensions of the chamber and the accuracy requirements for coating thickness measurement, a cross-section is set every 5 meters along the chamber axis, with 9 marking points evenly distributed along the circumference of each cross-section, for a total of 10 cross-sections and 90 marking points. The markings are done using a yellow paint pen, drawing a cross shape on the inner wall surface. The cross is approximately 10 cm long and the line width is approximately 1 cm. Due to the significant difference in reflectivity between yellow paint and the metal surface, the markings exhibit a distinct high-brightness feature in the point cloud intensity rendering, facilitating subsequent identification. A PosiTector 6000 ultrasonic thickness gauge is used to measure the coating thickness at the center of each marking, with a measurement accuracy of ±0.1 mm. During measurement, the thickness gauge probe is kept perpendicular to the surface, and each point is measured 3 times and the average value is taken to improve measurement accuracy. The coating thickness values ​​of all 90 marking points are recorded. The measurement results showed that the coating thickness ranged from 18.5 mm to 22.0 mm, with an average thickness of 20.2 mm.

[0073] Step S3: Post-construction point cloud acquisition. After construction, the interior walls of the tunnel were scanned using the same FARO Focus S350 scanner. The scanning station locations were kept as consistent as possible with those before construction, but slight differences occurred due to environmental changes after construction. Eight scanning stations were used, with scanning parameters set identically to those before construction. After scanning, FARO Scene software was used for automatic registration and stitching of the multi-site point clouds, with registration errors controlled within 1 mm. The point clouds were then color-coded using the RGB color information acquired by the scanner; yellow markers were clearly visible in the colored point cloud. Statistical filtering was performed to denoise the point clouds, removing noise and outliers. The final post-construction point cloud was obtained. It also contains approximately 200 million dots, covering the entire interior wall surface of the tunnel.

[0074] Step S4, marker recognition and normal vector calculation. The post-construction point cloud... The point cloud was loaded into CloudCompare point cloud processing software and displayed using intensity rendering mode. Yellow markers appeared as distinctly highlighted areas due to differences in reflectivity. The 3D coordinates of each marker were precisely extracted from its center using a manual point-prick method. To ensure the accuracy of the puncture points, the point cloud was magnified locally before extracting the coordinates to ensure that the puncture point positions were accurately located at the center of the marker. For each marker point... A spherical neighborhood query with a radius of 5 cm is used to find the neighborhood point set. The neighborhood contains approximately 500 points. Principal component analysis (PCA) is applied to this neighborhood point set to calculate the covariance matrix, and the eigenvalues ​​and eigenvectors are obtained. The eigenvector corresponding to the smallest eigenvalue is the direction of the surface normal vector. The normal vector is then normalized to obtain the unit normal vector. Since this chamber has a cylindrical structure, the normal vector should point radially towards the chamber's axis. Therefore, a uniform correction of the normal vector direction is performed to ensure that the normal vector points towards the outer side of the coating. The same treatment is applied to all 90 marked points.

[0075] Step S5, reverse calculation of original surface coordinates. Based on the coordinates of the marked points on the coating surface. Normal vector Coating thickness measured by a thickness gauge Calculate the corresponding coordinates of the original surface under the coating according to the formula: For example, if the coordinates of a marker point are (1000.000, 2000.000, 500.000) mm, the normal vector is (0.707, 0.707, 0.000), and the measured coating thickness is 20.5 mm, then the original surface coordinates obtained by reverse calculation are (1000.000-20.5×0.707, 2000.000-20.5×0.707, 500.000-20.5×0.000) = (985.494, 1985.494, 500.000) mm. Performing the same calculation on all 90 marker points yields the original coordinate set of the inner wall.

[0076] Step S6, Point Cloud Registration. First, perform coarse registration. Simultaneously load the point cloud data from before construction into the CloudCompare software. The point cloud is adjusted manually using the original surface coordinate point set obtained by reverse engineering. The position and orientation of the point cloud. Because the two scans may have used different starting sites and scan routes, the initial coordinate system of the point cloud differs significantly. The operator translates and rotates the point cloud around the vertical axis. This ensures that the point cloud visually overlaps with the coordinate point set. Since both scans were automatically leveled, coarse registration only requires adjusting translation and rotation within the horizontal plane; pitch and roll angles do not need to be adjusted. After coarse registration, the point cloud... The average distance between the coordinate point set and the coordinate point set is approximately 10 centimeters. Next, fine registration is performed. An iterative nearest-point algorithm with horizontal constraints is employed. This algorithm utilizes the leveling function of a high-precision scanner, assuming the point cloud is precisely aligned vertically, and only requires estimating the rotation angle around the vertical axis. and three-dimensional translation vector The optimization objective is to minimize the difference between the original coordinate point set of the inner wall and the point cloud before construction. The sum of squared distances between corresponding points. Specifically, for each coordinate point, after passing through the current rotation matrix... Translation vector Transformed point cloud The algorithm finds the nearest neighbor, establishes point correspondences, and calculates the sum of squared distances between all point pairs as the objective function value. Through iterative optimization, the rotation angle is gradually adjusted. Translation vector The goal is to minimize the objective function value. The iteration process is set to a maximum of 100 iterations, terminating when the root mean square error between two consecutive iterations is less than 0.1 mm. After 23 iterations, the algorithm converges, finally obtaining a rotation angle θ of approximately -3.7 degrees and a translation vector. Approximately (-82 mm, 45 mm, 12 mm). After registration, the original coordinate point set of the inner wall and the point cloud before construction are... The average distance is approximately 1.5 mm, and the registration accuracy meets the requirements.

[0077] Step S7, Thickness Calculation and Result Output. The rotation matrix obtained through precise registration is used. Translation vector For point clouds after construction Perform a 3D transformation to obtain the transformed point cloud. At this moment, the clouds... With point clouds Within the same coordinate system, precise alignment of point clouds before and after construction was achieved. (This refers to the point cloud before construction.) Delaunay triangulation was performed using the Delaunay 2.5D triangulation function in CloudCompare software. The point cloud was projected onto the best-fit plane and then triangulated to generate a 3D triangular mesh model containing approximately 400 million triangles. This process was applied to the point cloud after construction. Each point in the cloud was selected before construction. Find the nearest triangle in the triangulation network and calculate the perpendicular distance from that point to the plane containing the triangle. The specific calculation method is as follows: Let P be the point in the point cloud after construction, and let A, B, and C be the three vertices of the nearest triangle. First, calculate the normal vector of the triangle. Then calculate the distance from point P to the plane containing the triangle. The calculated distance d is used as the coating thickness value at that point and stored in the scalar field of the point cloud. This is then applied to the point cloud after construction. All points are processed in the same way to obtain the coating thickness value for each point. The thickness data is exported as a CSV file, and the thickness distribution is displayed in CloudCompare using pseudo-color rendering. The color gradually changes from blue (smaller thickness) to red (larger thickness), which visually shows the spatial distribution characteristics of the coating thickness.

[0078] To verify the detection accuracy of the method of the present invention, after the coating thickness calculation was completed, nine representative locations on a typical cross-section in the middle of the chamber were selected for actual measurement using an ultrasonic thickness gauge. These nine locations are evenly distributed in the circumferential direction and are numbered P0, P40, P80, P120, P160, P200, P240, P280, and P320, with an angular interval of 40 degrees. The thickness was repeatedly measured at each location using the thickness gauge to obtain the measured thickness value. Simultaneously, the thickness value at the corresponding location was extracted from the thickness calculation results of the method of the present invention. The accuracy comparison results are shown in the table below:

[0079]

[0080] The data in the table shows that the difference between the coating thickness calculated by the method of this invention and the measured value by the ultrasonic thickness gauge is between -0.9 mm and 2.0 mm, with an average absolute error of approximately 1.1 mm and a root mean square error of approximately 1.2 mm. The error is relatively small, within 2 mm, and close to the random error level (±2 mm) of a high-precision 3D laser scanner. The main sources of error include: the ranging accuracy of the laser scanner, the cumulative error of point cloud registration, the approximate error of triangulation interpolation, and the random error of the thickness gauge measurement. This fully verifies the effectiveness and reliability of the method of this invention. Overall, the method of this invention can meet the accuracy requirements of engineering quality inspection and provides a reliable technical means for full-coverage detection of coating thickness.

[0081] This embodiment fully demonstrates the application effect of the method of the present invention in coating thickness detection in uncontrolled areas. This method only requires the creation of 90 marker points and measurement with a thickness gauge, reducing workload by more than 95% compared to the traditional comprehensive point-by-point measurement method (which typically requires measuring thousands of points), significantly reducing labor and time costs. Simultaneously, this method obtains full-coverage thickness data containing approximately 200 million points, which can reflect the spatial distribution characteristics of coating thickness in detail, providing rich information for quality assessment. The detection accuracy reaches the millimeter level, meeting the quality control requirements of high-value coatings. The entire detection process requires no scaffolding, ensuring high operational safety and significantly improving detection efficiency. The thickness distribution results obtained by the method of the present invention can be exported in multiple formats, facilitating analysis and display in different point cloud processing software, providing reliable data support for project acceptance and quality assessment.

[0082] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., 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 detecting coating thickness in a scenario lacking control points, characterized in that, Includes the following steps: S1. Use a stand-up 3D laser scanner to perform a high-precision 3D laser scan of the inner wall of the enclosed space before construction. The 3D laser scanner has an automatic leveling function, and performs registration, stitching, and noise reduction processing on the collected multi-point cloud to ensure that the point cloud surface is smooth and free of error points. The processed point cloud is recorded as the point cloud before construction. ; S2. After the coating is applied, several marks are made on the surface. The marks are drawn in a cross shape using a special paint pen with a different reflectivity than the coating surface. The marks are evenly distributed in the area to be inspected according to the spatial size. At the center of each mark, a thickness gauge is used to vertically measure the coating thickness, and the coating thickness value at each mark location is recorded. ,in To mark the total number; S3. Use a 3D laser scanner to scan the surface of the coating after construction. Register, stitch, and color-grade the collected point clouds. Denoise and remove noise from the complete point cloud. Use the difference in reflectivity between the marked paint and the coating surface to identify the marked positions during point cloud intensity rendering. Record the processed point cloud as the post-construction point cloud. ; S4, Cloud Pointing after Construction The three-dimensional coordinates of each marker center are determined through manual interaction or image recognition. Query the neighborhood point set of each marked point The unit normal vector at this point is calculated using principal component analysis. The normal vector points to the outside of the coating; S5. Based on the marked coordinates Normal vector and coating thickness The coordinates of the original surface under the coating are calculated along the opposite direction of the normal vector. The calculation formula is as follows: ,in These are the original surface coordinates obtained by reverse engineering; S6. Use the original surface coordinate points obtained by reverse engineering as registration reference feature points and the point cloud before construction. For registration, coarse registration is first performed through manual interaction, involving 3D translation and in-plane rotation. Then, fine registration is performed using an iterative nearest-point algorithm with horizontal constraints. This algorithm estimates only the rotation angle around the vertical axis, assuming both point clouds have been horizontally leveled. With three-dimensional translation vector , to obtain the transformation parameters; S7. Using transformation parameters to analyze post-construction point clouds Perform coordinate transformation to obtain the transformed point cloud The Delaunay triangulation algorithm was used to analyze the point cloud before construction. Constructing triangular mesh surfaces, for Calculate the distance to each point in the equation. The nearest distance of the triangular mesh is used as the coating thickness value at that point. The thickness value is stored in the attribute field of the point cloud and the thickness feature is displayed by color rendering through point cloud visualization software.

2. The method according to claim 1, characterized in that, The specific process of calculating the surface normal vector using principal component analysis in step S4 is as follows: querying the marker points. In point cloud Neighborhood point set Principal component analysis is performed on the neighborhood point set to calculate the eigenvalues ​​and eigenvectors of the covariance matrix. The eigenvector corresponding to the smallest eigenvalue is taken as the unit normal vector of the marked point. .

3. The method according to claim 1, characterized in that, The transformation model of the iterative nearest point algorithm for the horizontal constraint in step S6 is expressed as follows: ,in Let be the rotation matrix for rotating about the Z-axis by an angle θ. , The rotation matrix is ​​a three-dimensional translation vector, and the first row is a three-row, three-column matrix. The second line The third line .

4. The method according to claim 3, characterized in that, The optimization objective of the iterative nearest point algorithm for the horizontal constraints is to minimize the difference between the transformed, back-calculated original surface coordinates and the point cloud before construction. The sum of squared distances between corresponding points.

5. The method according to claim 1, characterized in that, The markings in step S2 are made by using a paint pen to draw cross-shaped markings on the inner wall surface to ensure that the markings can be clearly identified when rendering point cloud intensity. The number and distribution of the markings are determined comprehensively based on the spatial size, the expected complexity of coating thickness changes, and the detection accuracy requirements.

6. The method according to claim 1, characterized in that, The specific method for calculating the distance from the point to the triangular network in step S7 is as follows: For the point cloud after construction... any point in Before construction, cloud observation was conducted. Find the nearest triangle ABC in the constructed triangular mesh, and calculate the vertical distance from point P to the plane containing triangle ABC as the coating thickness at that location. The distance calculation takes into account the vertical distance from the point to the nearest triangular facet.

7. The method according to claim 1, characterized in that, The enclosed space includes tunnels, underground utility tunnels, cylindrical chambers, or elliptical chambers. Due to their engineering characteristics, traditional measurement control points cannot be set up in these areas within a limited space, which constitutes a scenario lacking control points.

8. The method according to claim 1, characterized in that, The leveling accuracy of the gantry-type 3D laser scanner used in steps S1 and S3 reaches the second level. The vertical axis of the scanner is precisely aligned with the direction of gravity, ensuring that the vertical direction of each station cloud is consistent with the direction of gravity, thus providing an accurate horizontal reference for the horizontal constraint iterative nearest point algorithm.

9. The method according to claim 1, characterized in that, In step S6, the coarse registration stage uses a manual interactive method to perform preliminary alignment of the two sets of data through three-dimensional translation and horizontal rotation, so that the original surface coordinate points obtained by back-reasoning are consistent with the point cloud before construction. The visual alignment is basically consistent, providing a good initial value for fine registration. Since both scans have been automatically leveled, only translation and rotation in the horizontal plane need to be adjusted during coarse registration, without adjusting the pitch and roll angles.

10. The method according to claim 1, characterized in that, The coating thickness distribution results are stored in the form of point cloud attribute fields, and the thickness distribution can be visualized and rendered in three dimensions in any point cloud processing software. The spatial distribution characteristics of the thickness are intuitively displayed through pseudo-color rendering, which makes it easy to identify areas with abnormal thickness.

Images