Intelligent detection method for thickness of protective layer of power cement pole based on laser measurement

Abstract

The application discloses a laser measurement-based intelligent detection method for the thickness of a protective layer of a power cement pole, relates to the field of nondestructive testing technology in civil engineering, and comprises the following steps: projecting structured light onto the surface of the power cement pole and receiving a reflected signal, and obtaining surface point cloud data by processing; separating two types of sub-point clouds corresponding to the main body of the pole and the reinforcing steel features from the point cloud; fitting a theoretical base cylindrical surface model by using the main body point cloud; projecting the reinforcing steel feature point cloud to the model to obtain the two-dimensional position index of the reinforcing steel; intercepting the local high-precision point cloud around each reinforcing steel according to the index, extracting the physical boundary of the concrete and the reinforcing steel through geometric analysis, and calculating the vertical distance from the physical boundary to the base surface as the thickness of the protective layer; and finally generating a thickness spatial distribution report. The method realizes non-contact, full-field and high-precision automatic detection of the thickness of the protective layer.

Description

Technical Field

[0001] This invention belongs to the field of non-destructive testing technology in civil engineering, specifically a method for intelligent detection of the protective layer thickness of power cement poles based on laser measurement. Background Technology

[0002] Currently, the thickness of the concrete protective layer on power poles is mainly measured using manual sampling or electromagnetic and ultrasonic methods. Manual measurement is inefficient and cannot reflect the overall condition. Among non-contact testing methods, ultrasonic methods are affected by the internal inhomogeneity of the material, resulting in insufficient measurement stability. Electromagnetic induction methods are easily affected by environmental interference and have poor adaptability to unknown rebar arrangements.

[0003] While 3D laser scanning can acquire high-precision surface data, the resulting point cloud is a mixture of the main curved surface of the rod and the surface deformation caused by the internal reinforcing steel. Conventional processing methods treat the entire point cloud as a single object. When directly fitting the cylindrical surface, distortion points in the reinforcing steel region contaminate the model, causing the reference plane to deviate from the true geometric axis. Furthermore, on homogeneous cement surfaces with single reflective properties, traditional point cloud segmentation algorithms based on optical properties fail, making it difficult to directly and robustly separate the weak geometric features representing the reinforcing steel from the 3D point cloud. In addition, directly calculating the position of the reinforcing steel and the thickness of the protective layer in 3D space is complex and sensitive to noise.

[0004] A method is needed to automatically distinguish between rod and reinforcing bar features based on geometric characteristics from a single mixed point cloud to obtain clean main structural data. Simultaneously, a mechanism needs to be established to map reinforcing bar features to a stable and accurate geometric reference surface, thereby simplifying the positioning process and providing a reliable benchmark for thickness calculation. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes an intelligent detection method for the protective layer thickness of power cement poles based on laser measurement, including: A structured laser pattern is projected onto the surface of a cement pole and the reflected laser signal is received. The resulting purified surface point cloud data is then processed. From the purified surface point cloud data, the support point cloud representing the main structure of the cement pole and the masking point cloud representing the distribution characteristics of the reinforcing bars are separated. Cylindrical surface fitting is performed on the point cloud of the support to construct a theoretical reference cylindrical surface model of the cement pole; Project the point cloud of the mask onto the theoretical reference cylindrical surface model, and calculate the set of projection positions of the reinforcing bars on the theoretical reference cylindrical surface model; Based on the set of projection positions, local point cloud data corresponding to the protective layer area around each steel bar is extracted from the cleaned surface point cloud data. Perform geometric analysis on each of the local point cloud data to extract the physical boundary features of the protective layer region; Based on the spatial relationship between the physical boundary characteristics and the theoretical reference cylindrical surface model, the actual thickness of the concrete cover at each point of reinforcement is calculated. The actual thickness values ​​calculated for all locations are summarized to generate a spatial distribution report of the protective layer thickness of the power cement pole.

[0006] Furthermore, the obtained purified surface point cloud data includes: A structured laser pattern is projected onto the surface of a power cement pole, and laser signals reflected from the surface of the power cement pole are received; The received laser signal is subjected to photoelectric conversion and signal conditioning to generate raw depth point cloud data; Spatial filtering and outlier removal are performed on the original depth point cloud data to obtain purified surface point cloud data. The step of performing photoelectric conversion and signal conditioning on the received laser signal to generate raw depth point cloud data includes: The reflected laser signal is converted into a corresponding analog electrical signal using an array of photoelectric sensors; The analog electrical signal is amplified, filtered, and noise suppressed to obtain a conditioned electrical signal; The conditioned electrical signal is subjected to high-speed analog-to-digital conversion to obtain digitized laser intensity and phase information; By combining the laser projector's emission angle, scanning timing parameters, and the digitized laser intensity and phase information, the three-dimensional spatial coordinates are calculated point by point using the triangulation principle. All the calculated three-dimensional spatial coordinates are bound to the corresponding laser reflection intensity values ​​and organized according to the scanning order to form the original depth point cloud data.

[0007] Further, the step of performing spatial filtering and outlier removal on the original depth point cloud data to obtain purified surface point cloud data includes: Establish a spatial index structure for the original depth point cloud data; Traverse each point in the original depth point cloud data and query the distribution density of other points within its neighborhood; Based on a preset distance threshold, isolated points with a distribution density significantly lower than the normal level are identified and removed. For points that are not removed, statistical analysis is performed on the distance distribution of each point in its neighborhood along the normal direction to identify and remove outliers that deviate from the distribution center by more than a preset range. After removing outliers, a smoothing filter is applied to the remaining point set to smooth the data, and the purified surface point cloud data is output.

[0008] Further, separating the support point cloud representing the main structure of the cement pole and the masking point cloud representing the distribution characteristics of the reinforcing bars from the purified surface point cloud data includes: Normal estimation is performed on the purified surface point cloud data, and the normal vector of each point is calculated; The purified surface point cloud data is initially clustered based on the point cloud curvature characteristics. In the initial clustering results, large clusters of point clouds with low curvature and spatial continuity are selected and classified as candidate main structure point clouds. A secondary analysis is performed on the point cloud of the candidate main structure. A cylindrical model is fitted by a random sampling consistency algorithm, and points that conform to the cylindrical model and whose residuals are within a set range are classified as the point cloud of the support body. The support point cloud is removed from the purified surface point cloud data. The remaining point cloud is then re-clustered based on spatial proximity and normal consistency. Point cloud clusters that appear as elongated strips or have periodic distribution characteristics are identified as the mask point cloud.

[0009] Furthermore, the step of performing cylindrical surface fitting on the point cloud of the support to construct a theoretical reference cylindrical surface model of the cement pole includes: Randomly select several minimum subsets from the point cloud of the support, each of the minimum subsets containing the minimum number of points required to define the cylindrical geometric model; An initial cylinder model parameter is calculated using each of the minimum subsets, the cylinder model parameter including the central axis direction vector, the coordinates of a point on the central axis, and the cylinder radius; Calculate the distance from all points in the entire support point cloud to each initial cylindrical model, and mark the points whose distance is less than a set threshold as local points of the initial cylindrical model; The initial cylindrical model with the most inliers is selected as the optimal model candidate. The random sampling and optimization process is performed iteratively until the number of local points of the candidate optimal model reaches a preset stability condition. Using all the finally determined interior points, the precise parameters of the theoretical reference cylindrical model are calculated by least squares optimization.

[0010] Further, the step of projecting the point cloud of the mask onto the theoretical reference cylindrical surface model and calculating the set of projected positions of the reinforcing bars on the theoretical reference cylindrical surface model includes: Obtain the geometric parameters of the theoretical reference cylindrical surface model; For each point in the point cloud of the mask, calculate the shortest distance and the foot of the perpendicular from the point to the theoretical reference cylindrical surface model along the normal direction of the theoretical reference cylindrical surface model. All calculated perpendicular points are used as projection points of the point cloud of the mask onto the theoretical reference cylindrical surface model; The projection points are clustered by density based on their unfolded coordinates on the cylindrical surface to form several dense clusters; Calculate the geometric center of each of the dense point clusters, and record the three-dimensional coordinates of the geometric center on the theoretical reference cylindrical surface model as the projected position of a reinforcing bar. The projected positions of all the reinforcing bars constitute the set of projected positions.

[0011] Further, the step of extracting local point cloud data corresponding to the protective layer area around each rebar from the cleaned surface point cloud data based on the projection position set includes: For each projection position in the set of projection positions, a rectangular or sector-shaped neighborhood search window is defined on the theoretical reference cylindrical surface model with the projection position as the center. The neighborhood search window is projected backwards into three-dimensional space to determine a three-dimensional neighborhood search volume; In the purified surface point cloud data, all points whose spatial coordinates are located within the neighborhood search volume of the three dimensions are extracted to form an initial candidate point set; Based on the theoretical reference cylindrical surface model, calculate the distance from each point in the initial candidate point set to the theoretical reference cylindrical surface model; Points whose distance values ​​are within a reasonable range of the preset protective layer thickness are selected to form the final local point cloud data, which represents the protective layer area from the surface of the cement pole to the outer edge of the reinforcing bar.

[0012] Further, the step of performing geometric analysis on each of the local point cloud data to extract the physical boundary features of the protective layer region includes: Principal component analysis is performed on the local point cloud data to determine the main direction and extent of the point distribution; An analysis profile is constructed along the direction perpendicular to the normal of the theoretical reference cylindrical surface model and passing through the corresponding reinforcement projection position; The points in the local point cloud data are projected onto the analysis profile to form a set of two-dimensional scattered points. Statistical histogram analysis was performed on the two-dimensional scatter points to identify abrupt changes in data distribution density along the depth direction; The first abrupt change in data distribution from dense to sparse is identified as the location feature of the outer surface of the cement pole, and the second abrupt change in data distribution from sparse to dense is identified as the location feature of the surface of the reinforcing steel bar. These two location features together constitute the physical boundary feature.

[0013] Furthermore, the calculation of the actual thickness of the concrete cover for each reinforcement layer based on the spatial relationship between the physical boundary features and the theoretical reference cylindrical surface model includes: Obtain the position features of the outer surface of the cement pole and the surface of the reinforcing bar from the physical boundary features; Calculate the shortest distance from the feature point on the outer surface of the cement pole to the theoretical reference cylindrical surface model, and denot it as the outer surface distance; Calculate the shortest distance from the feature point on the surface of the reinforcing bar to the theoretical reference cylindrical surface model, and denot it as the reinforcing bar surface distance; Calculate the difference between the distance to the outer surface and the distance to the surface of the reinforcing bar; the difference is the actual thickness value. The calculated actual thickness values ​​are standardized in terms of unit and format.

[0014] Furthermore, the process of summarizing the actual thickness values ​​calculated from all locations to generate a spatial distribution report of the protective layer thickness of the power cement pole includes: Establish a mapping table between spatial coordinates and the actual thickness value, wherein the spatial coordinates are based on the set of projection positions; The data in the mapping table is visualized and rendered on the two-dimensional unfolded diagram of the theoretical reference cylindrical surface model, with different colors or contour lines representing the thickness distribution. Statistical characteristics of all the actual thickness values ​​are calculated, including minimum, maximum, average and standard deviation. The visualization rendering results are integrated with statistical feature data, and a spatial distribution report containing spatial distribution maps and statistical data is generated according to a preset report template.

[0015] Compared with the prior art, the beneficial effects of the present invention are: Automatic segmentation based on point cloud geometric features separates the support point cloud representing the main body of the pole from the original mixed data, separating it from the masking point cloud representing the reinforcing steel. This process relies on the analysis of the local curvature and normal vectors of the point cloud, clustering points with different geometric properties. The result is a pure support point cloud containing only information about the smooth main body, thus eliminating the interference of distortion points in the reinforcing steel region during subsequent cylindrical surface fitting. The fitted theoretical benchmark cylindrical surface model more accurately represents the true ideal geometry of the concrete pole, establishing a precise spatial benchmark for all subsequent calculations.

[0016] Using this high-precision theoretical reference cylindrical model, the separated point cloud of the masking body is projected perpendicularly along the model's normal direction to generate a set of projected positions of the reinforcing bars on the two-dimensional cylindrical surface. This operation transforms the originally scattered and complex three-dimensional spatial positioning problem into a coordinate mapping problem on a regular two-dimensional curved surface. The set of projected positions provides a standardized representation of the reinforcing bar distribution, directly and clearly indicating the axial and circumferential positions of each reinforcing bar on the rod, and using this as an index to back-locate to precise regions in the original three-dimensional point cloud.

[0017] Based on the index of the projection location set, precise local point cloud data around each rebar is extracted from the original full-width point cloud. This data range is clearly focused on the target area, excluding a large number of irrelevant point clouds, thus reducing the data size for subsequent calculations. Geometric analysis is performed on the local point cloud to extract the physical boundary features between the concrete outer surface and the rebar surface. By calculating the vertical distance from these boundary feature points to the theoretical reference cylindrical surface model, the protective layer thickness is directly obtained. Since the reference benchmark for distance calculation is an accurately fitted theoretical cylindrical surface, and the data is a high-fidelity local original point cloud, this calculation method has clear geometric meaning, avoids the cumulative error of indirect estimation, achieves direct and high-precision acquisition of thickness values, and can ultimately generate a systematic spatial distribution report. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the steps of the intelligent detection method for the protective layer thickness of power cement poles based on laser measurement as described in this invention. Figure 2 A flowchart for separating the point cloud of the support body from the point cloud of the mask body; Figure 3 Curves showing the performance of the cylindrical surface fitting algorithm in laser detection of cement poles for power transmission. Figure 4 A 3D point cloud visualization of a power cement pole laser inspection scenario; Figure 5 This is a professional analysis diagram showing the distribution of the protective layer thickness along the height of a power cement pole. Detailed Implementation

[0019] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] See Figure 1A structured laser pattern is projected onto the surface of a cement pole, and the reflected laser signals are received. The laser signals are then processed to obtain purified surface point cloud data. From this purified point cloud data, the support point cloud, representing the main structure of the cement pole, and the masking point cloud, representing the distribution characteristics of the reinforcing bars, are separated. A cylindrical surface fitting is performed on the support point cloud to construct a theoretical reference cylindrical model of the cement pole. The masking point cloud is projected onto this theoretical reference cylindrical model, and the set of projection positions of the reinforcing bars on the theoretical reference cylindrical model is calculated. Based on this set of projection positions, local point cloud data corresponding to the protective layer area around each reinforcing bar is extracted from the purified surface point cloud data. Geometric analysis is performed on each local point cloud data to extract the physical boundary features of the protective layer area. Based on the spatial relationship between the extracted physical boundary features and the theoretical reference cylindrical model, the actual thickness of the protective layer at each reinforcing bar is calculated. The actual thickness values ​​calculated for all locations are summarized to generate a spatial distribution report of the protective layer thickness of the cement pole.

[0021] In one embodiment of the present invention, a structured laser pattern is projected onto the surface of a power cement pole and the laser signal reflected from the surface of the power cement pole is received. The received laser signal is photoelectrically converted and conditioned to generate raw depth point cloud data. The photoelectric conversion and signal conditioning convert the reflected laser signal into a corresponding analog electrical signal through a photoelectric sensor array. The analog electrical signal is amplified, filtered, and noise suppressed to obtain a conditioned electrical signal. The conditioned electrical signal is then subjected to high-speed analog-to-digital conversion to obtain digitized laser intensity and phase information. Combining the laser projector's emission angle, scanning timing parameters, and the digitized laser intensity and phase information, the three-dimensional spatial coordinates are calculated point by point using the triangulation principle. All the calculated three-dimensional spatial coordinates are bound to the corresponding laser reflection intensity values ​​and organized according to the scanning order to form raw depth point cloud data.

[0022] In specific implementation, spatial filtering and outlier removal are performed on the original depth point cloud data to obtain purified surface point cloud data. A spatial index structure for the original depth point cloud data is established. Each point in the original depth point cloud data is traversed, and the distribution density of other points within its neighborhood is queried. Based on a preset distance threshold, isolated points with distribution densities significantly lower than normal levels are identified and removed. For the points that are not removed, the distance distribution along the normal direction of each point in its neighborhood is statistically analyzed to identify and remove outliers that deviate from the distribution center beyond a preset range. After outlier removal, a smoothing filter is applied to the remaining point set to smooth the data, and the purified surface point cloud data is output. In some embodiments, the preset distance threshold is dynamically adjusted according to the global statistical characteristics of the point cloud data. For example, in specific implementations, the distribution density is evaluated by calculating the average distance from each point to its k nearest neighbors, expressed by the formula: in: Indicates the average distance. This represents the coordinate vector of the current point. This represents the coordinate vector of the j-th nearest neighbor. This represents the number of nearest neighbors. If the value exceeds a preset threshold, the current point is determined to be an isolated point and is removed. It is understandable that this distance-based density assessment can effectively identify noisy points. Optionally, the smoothing filter uses Gaussian filtering or median filtering to preserve the main geometric features of the point cloud. Smoothing helps reduce data fluctuations.

[0023] In the example scenario, for laser scanning of power cement poles, the original depth point cloud data may contain outliers caused by surface stains, corrosion, or environmental interference. Through spatial filtering and outlier removal, the purified surface point cloud data more accurately reflects the geometry of the pole surface. Data comparison shows that isolated points and outliers in the filtered point cloud are removed, and the point cloud distribution is more uniform. It can be understood that point cloud purification improves data quality. In specific implementation, the outlier removal process is based on point cloud density and distance distribution to ensure that only reliable data points are retained. The spatial index structure uses kd-trees or octrees to accelerate neighborhood queries. The distance distribution along the normal direction is estimated by calculating the principal direction of the neighborhood point cloud to identify outliers along the surface normal direction.

[0024] See Figure 2 In one embodiment of the present invention, normal estimation is performed on the purified surface point cloud data to calculate the normal vector of each point. Normal estimation is accomplished by searching the set of neighboring points of each point and fitting a local tangent plane. In some embodiments, principal component analysis is used to calculate the eigenvector corresponding to the minimum eigenvalue of the local neighborhood covariance matrix of the point as the normal vector. Optionally, the normal direction is unified through viewpoint consistency constraints. In specific implementation, the purified surface point cloud data is initially clustered based on the point cloud curvature features. The curvature features of a point are calculated from the eigenvalues ​​of the covariance matrix of its local neighborhood. It can be understood that the curvature value reflects the degree of curvature of the local surface, and the calculation formula is: in: Represents the curvature of a point. , , Describe the eigenvalues ​​of the local neighborhood covariance matrix of a point and satisfy the following conditions: Regions with curvature values ​​close to zero typically correspond to flat surfaces. In practice, in the initial clustering results, large clusters of point clouds with low curvature and spatial continuity are selected and classified as candidate main structure point clouds. The clustering algorithm can employ a region growing method based on Euclidean distance and normal vectors. It can be understood that the main surface of the power cement pole is relatively flat and continuous, therefore, low-curvature and connected point cloud clusters are highly likely to correspond to the main structure.

[0025] In practice, a secondary analysis is performed on the point cloud of the candidate main structure. A cylindrical model is fitted using a random sampling consensus algorithm, and points that conform to the cylindrical model and whose residuals are within a set range are classified as support point clouds. The random sampling consensus algorithm randomly selects the minimum set of points from the candidate main structure point cloud to fit the initial cylindrical model and evaluates the number of points in the entire point cloud that conform to the model. This process is iterated to find the cylindrical model parameter with the largest number of interior points. In the example scenario, for a standard circular cement pole, the point cloud of its main body can be well fitted by a cylindrical model. The fitting residual is set to a range of, for example, 3 mm. Points that meet this condition are classified as support point clouds. Data comparison shows that in the candidate main structure point cloud obtained from the initial clustering, most points can well conform to the finally fitted cylindrical model. In specific implementation, the support point cloud is removed from the purified surface point cloud data, and the remaining point cloud is re-clustered based on spatial proximity and normal consistency. Point cloud clusters that appear as elongated shapes or have periodic distribution characteristics are identified as masking point clouds. In some embodiments, clustering is performed based on the three-dimensional coordinates and normal vector direction of the points in the remaining point cloud. Optionally, a density-based clustering algorithm is used. Elongated point cloud clusters may correspond to exposed or near-surface steel bars, and the periodic distribution characteristics are consistent with the regular arrangement of steel bars in the cement pole.

[0026] In one embodiment of the present invention, several minimum subsets are randomly selected from the support point cloud. Each minimum subset contains the minimum number of points required to define the cylindrical geometric model. For a spatial cylindrical model, the minimum number of points required is three non-collinear points. In a specific implementation, an initial cylindrical model parameter is calculated using each minimum subset. The cylindrical model parameter includes the central axis direction vector, the coordinates of a point on the central axis, and the cylinder radius. The central axis direction vector is estimated by the cross product of the vectors formed by the points in the minimum subset. The coordinates of a point on the central axis can be solved by geometric constraints. The cylinder radius is obtained by calculating the average distance from each point in the minimum subset to the estimated axis. In some embodiments, the number of random samplings is preset to a fixed value based on the size of the support point cloud. Optionally, the number of samplings can also be dynamically adjusted according to the convergence of the algorithm.

[0027] In practice, the distance from all points in the entire support point cloud to each initial cylindrical model is calculated. Points with a distance less than a set threshold are marked as local points of the initial cylindrical model. The formula for calculating the distance d from a point to the cylindrical model is as follows: in: This represents the distance from a point to the surface of the cylinder. Represents the coordinate vector of a point. This represents the coordinate vector of a point on the central axis of the cylinder. The unit direction vector representing the central axis of the cylinder. Represents the radius of a cylinder, symbol Represents the cross product of vectors. The magnitude of the vector is represented by a threshold value determined based on the point cloud accuracy and the flatness of the cement pole surface. This distance can be understood as measuring the radial deviation between the point and the cylindrical model.

[0028] In practical implementation, the random sampling and optimization process is iteratively executed until the number of inliers of the optimal model candidate reaches a preset stability condition. The stability condition is defined as the number of inliers of the optimal model candidate no longer increasing in N consecutive iterations; in some embodiments, N is set to 50. In the example scenario, the point cloud of the support structure of a power concrete pole is fitted, with a distance threshold of 5 mm. After multiple random samplings and inlier statistics, a cylindrical model with high consistency across a large number of point clouds is finally found. Data comparison shows that compared with directly using all points for least squares fitting, this method can effectively eliminate the interference of a few outliers or non-main structure points through the random sampling consistency algorithm. In practical implementation, using all the finally determined inliers, the precise parameters of the theoretical benchmark cylindrical surface model are calculated using the least squares method. The least squares method optimizes the central axis direction vector, the coordinates of a point on the central axis, and the cylinder radius parameters by minimizing the sum of the squares of the distances from all inliers to the cylinder surface.

[0029] See Figure 3This is a graph used to analyze the performance of the cylindrical surface fitting algorithm in laser inspection of power cement poles. When the distance threshold increases from 1mm to 5mm, the proportion of in-situ points rises rapidly from 0.4 to approximately 0.95, indicating that more points are included in the cylindrical model. Once the distance threshold exceeds 5mm, the increase in the proportion of in-situ points slows down, and the curve enters a plateau, indicating that further increasing the threshold has little effect on increasing the number of in-situ points. This graph visually verifies the effectiveness of the RANSAC algorithm in fitting point clouds of power cement poles. By setting a reasonable distance threshold, the main body of the support point cloud can be efficiently selected, providing a reliable theoretical reference cylindrical surface for subsequent calculations of the protective layer thickness. Based on this curve, engineers can avoid computational redundancy caused by excessively increasing the threshold and prevent the loss of effective points due to an excessively small threshold. This helps to balance computational efficiency and model accuracy in actual inspection, improving the overall efficiency of the inspection process.

[0030] In one embodiment of the present invention, the geometric parameters of the theoretical reference cylindrical surface model are obtained. These parameters include the direction vector of the cylinder's central axis, the coordinates of a point on the central axis, and the cylinder radius. It can be understood that the geometric parameters are derived from the fitting results of the previous embodiment. In a specific implementation, for each point in the mask point cloud, the shortest distance from the point to the theoretical reference cylindrical surface model along the normal direction of the theoretical reference cylindrical surface model and the foot of the perpendicular are calculated. The shortest distance calculation involves the radial deviation of the point from the cylindrical surface, and the foot of the perpendicular is the orthogonal projection of the point onto the cylindrical surface. In some embodiments, the coordinates of the foot of the perpendicular... Calculated using the following formula: in: Represents the perpendicular foot coordinate vector. This represents the coordinate vector of a point in the point cloud of the mask. This represents the coordinate vector of a point on the central axis of the theoretical reference cylindrical surface model. The unit direction vector representing the central axis of the theoretical reference cylindrical surface model. The radius of the theoretical reference cylindrical surface model is represented by the symbol. Represents the vector dot product. The vector's magnitude is represented. In a specific implementation, all calculated perpendicular points are used as projection points of the mask point cloud onto the theoretical reference cylindrical model. Optionally, the projection points are stored in three-dimensional coordinates. In a specific implementation, the projection points are density-clustered based on their unfolded coordinates on the cylindrical surface, forming several dense point clusters. The unfolded coordinates are obtained by unfolding the cylindrical surface into a plane. Density clustering algorithms such as DBSCAN can identify spatially clustered point groups. In a specific implementation, the geometric center of each dense point cluster is calculated, and the three-dimensional coordinates of the geometric center on the theoretical reference cylindrical model are recorded as the projection position of a reinforcing bar. The projection positions of all reinforcing bars constitute a projection position set. Refer to Table 1, which shows an exemplary projection position set: Table 1: Set of Reinforcing Bar Projection Locations In practical implementation, based on the projection position set, local point cloud data corresponding to the protective layer area around each rebar is extracted from the cleaned surface point cloud data. For each projection position in the projection position set, a rectangular or fan-shaped neighborhood search window is defined on the theoretical reference cylindrical surface model, centered on the projection position. In some embodiments, the rectangular search window extends by a fixed dimension along the axial and circumferential directions of the cylindrical surface, for example, 0.2 meters axially and 0.15 meters circumferentially. In practical implementation, the neighborhood search window is back-projected into three-dimensional space to determine a three-dimensional neighborhood search volume. Back-projection is achieved by extending the window boundary points outward by a certain distance along the normal direction of the cylindrical surface. It can be understood that the three-dimensional neighborhood search volume covers the area from the surface of the cement pole to the interior that may contain rebar. In practical implementation, all points whose spatial coordinates lie within the three-dimensional neighborhood search volume are extracted from the cleaned surface point cloud data to form an initial candidate point set. Optionally, a spatial index structure can be used to accelerate point querying. In practice, based on the theoretical reference cylindrical surface model, the distance from each point in the initial candidate point set to the theoretical reference cylindrical surface model is calculated, and the distance calculation formula is the same as described above.

[0031] In practice, points with distance values ​​within a preset reasonable range for the protective layer thickness are selected to form the final local point cloud data. The reasonable range for the protective layer thickness is set according to design specifications, for example, 0.02 meters to 0.08 meters. The local point cloud data represents the protective layer area from the surface of the cement pole to the outer edge of the reinforcing bars. In the example scenario, for a single power cement pole, the set of projected reinforcing bar positions indicates the approximate distribution of the reinforcing bars on the cylindrical surface. Through a cropping operation, local point clouds around each projected position are obtained. Data comparison shows that the local point cloud data concentrates the points in the protective layer area and reduces interference from irrelevant background points.

[0032] See Figure 4This is a 3D point cloud visualization of a power concrete pole laser inspection scenario, clearly showing the layered structure of the inspection data. The red mask point cloud is embedded within the blue support point cloud, visually representing the actual distribution of rebar within the concrete pole. The green theoretical outline envelops the entire point cloud, clearly demonstrating the algorithm's fitting effect on the main structure of the concrete pole. By visualizing the spatial relationship between the support and mask, the accuracy of point cloud segmentation can be intuitively evaluated, ensuring the reliability of subsequent protective layer thickness calculations. By observing the completeness of the extracted red mask point cloud, the clustering parameters of the point cloud segmentation algorithm can be optimized accordingly; by observing the fit between the green outline and the point cloud, the threshold and number of iterations for cylindrical surface fitting can be iteratively adjusted, thereby continuously improving algorithm performance.

[0033] In one embodiment of the present invention, principal component analysis (PCA) is performed on local point cloud data to determine the main direction and extent of its point distribution. PCA is accomplished by calculating the covariance matrix of the local point cloud data and its eigenvalues ​​and eigenvectors. The eigenvector corresponding to the largest eigenvalue indicates the main direction of the point cloud along the depth direction, which helps to determine the reasonable orientation of the analysis profile. In a specific implementation, an analysis profile is constructed along a direction perpendicular to the normal of the theoretical reference cylindrical surface model and passing through the corresponding reinforcement projection position. This profile is a plane, and its normal vector is defined by the cross product of the tangential direction of the theoretical reference cylindrical surface model at the reinforcement projection position and the normal of the theoretical reference cylindrical surface model. In a specific implementation, points in the local point cloud data are projected onto the analysis profile to form a set of two-dimensional scattered points. The projection is achieved by calculating the perpendicular projection point of each point to the analysis profile. The two-dimensional coordinates of the projected points can be represented in a local coordinate system with a point within the profile as the origin. In some embodiments, one axis of the local coordinate system is defined as the radial direction of the theoretical reference cylindrical model at the projection position, pointing outwards, and the other axis is defined as the axial direction of the theoretical reference cylindrical model at the projection position. Optionally, the axial direction may be along the height direction of the cement pole.

[0034] In specific implementation, statistical histogram analysis is performed on the two-dimensional scattered points, and abrupt changes in data distribution density are identified along the depth direction. The depth direction is defined as the radial direction from the theoretical reference cylindrical model to the outside of the cement pole. The histogram is divided into intervals along the depth direction, and the number of projection points in each interval is counted. In some embodiments, the abrupt change location is determined by finding the interval boundary where the absolute value of the histogram gradient exceeds a preset threshold, expressed by the formula: in: This represents the count of the i-th histogram interval. This represents the count of the (i+1)th histogram interval. This represents a preset gradient threshold. When the condition is met, the interval boundary is considered a potential abrupt change location. In specific implementation, the first abrupt change location where the data distribution changes from dense to sparse is identified as the location feature of the outer surface of the cement pole, and the second abrupt change location where the data distribution changes from sparse to dense is identified as the location feature of the steel reinforcement surface. These two location features together constitute the physical boundary features. It can be understood that the outer surface of the cement pole usually has a large number of point clouds forming a high-density distribution, and the surface of the steel reinforcement may also form a relatively concentrated point cloud due to its reflective properties. In the example scenario, for the local protective layer point cloud of a power cement pole, the histogram along the depth direction shows two obvious density change boundaries. The first boundary corresponds to the transition of the point cloud from the cement pole surface to the air, and the second boundary corresponds to the transition of the point cloud from the cement matrix to the steel reinforcement. Data comparison shows that compared with the traditional method based on a single threshold, this method can locate the boundary more stably.

[0035] In practical implementation, the actual thickness of the concrete cover at each rebar location is calculated based on the spatial relationship between the physical boundary features and the theoretical reference cylindrical surface model. The location feature points on the outer surface of the cement pole and the surface of the rebar are obtained from the physical boundary features. In practical implementation, the shortest distance from the location feature points on the outer surface of the cement pole to the theoretical reference cylindrical surface model is calculated and recorded as the outer surface distance. Similarly, the shortest distance from the location feature points on the rebar surface to the theoretical reference cylindrical surface model is calculated and recorded as the rebar surface distance. The calculation method for the shortest distance is consistent with the calculation principle for the distance from a point to a cylindrical surface. In practical implementation, the difference between the outer surface distance and the rebar surface distance is calculated; this difference is the actual thickness value. Optionally, the calculated actual thickness values ​​are standardized in unit and format, for example, by converting all thickness values ​​to millimeters and retaining one decimal place. In practical implementation, the actual thickness values ​​calculated for all locations are summarized to generate a spatial distribution report of the concrete cover thickness of the power cement pole. A mapping table between spatial coordinates and actual thickness values ​​is established, with the spatial coordinates based on a set of projected locations. In practical implementation, the data in the mapping table is visualized on a two-dimensional unfolded diagram of the theoretical benchmark cylindrical surface model, with different colors or contour lines representing the thickness distribution. The two-dimensional unfolded diagram is obtained by cutting and flattening the cylindrical surface along a generatrix. Optionally, the color mapping uses a gradient from green to red to represent thicknesses from large to small. In practical implementation, the statistical characteristics of all actual thickness values ​​are collected, including the minimum, maximum, average, and standard deviation. These statistics provide an overall overview of the protective layer thickness.

[0036] See Figure 5This is a professional analysis chart showing the distribution of the protective layer thickness along the height of a power concrete pole, possessing multi-dimensional core value in engineering inspection and decision-making. The actual thickness fluctuates between 20mm and 42mm, with most locations meeting design requirements, but significant deviations exist in some areas. At approximately 2.5 meters, the thickness drops to about 20mm, far below the design value, representing a typical weak point. Between 0.5 and 1.5 meters, the thickness is generally higher than the design value, indicating good quality pouring of the protective layer in this area. The chart, presented as a continuous curve, comprehensively shows the protective layer thickness distribution throughout the entire pole, providing objective and quantitative quality assessment data for project acceptance. The uneven thickness distribution reflects the stability of the pouring process during concrete pole production, providing direct feedback for optimizing the production process and improving product quality.

[0037] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A smart detection method for the protective layer thickness of power cement poles based on laser measurement, characterized in that, The method includes: A structured laser pattern is projected onto the surface of a cement pole and the reflected laser signal is received. The resulting purified surface point cloud data is then processed. From the purified surface point cloud data, the support point cloud representing the main structure of the cement pole and the masking point cloud representing the distribution characteristics of the reinforcing bars are separated. Cylindrical surface fitting is performed on the point cloud of the support to construct a theoretical reference cylindrical surface model of the cement pole; Project the point cloud of the mask onto the theoretical reference cylindrical surface model, and calculate the set of projection positions of the reinforcing bars on the theoretical reference cylindrical surface model; Based on the set of projection positions, local point cloud data corresponding to the protective layer area around each steel bar is extracted from the cleaned surface point cloud data. Perform geometric analysis on each of the local point cloud data to extract the physical boundary features of the protective layer region; Based on the spatial relationship between the physical boundary characteristics and the theoretical reference cylindrical surface model, the actual thickness of the concrete cover at each point of reinforcement is calculated. The actual thickness values ​​calculated for all locations are summarized to generate a spatial distribution report of the protective layer thickness of the power cement pole.

2. The intelligent detection method for the protective layer thickness of power cement poles based on laser measurement according to claim 1, characterized in that, The obtained purified surface point cloud data includes: A structured laser pattern is projected onto the surface of a power cement pole, and laser signals reflected from the surface of the power cement pole are received; The received laser signal is subjected to photoelectric conversion and signal conditioning to generate raw depth point cloud data; Spatial filtering and outlier removal are performed on the original depth point cloud data to obtain purified surface point cloud data. The step of performing photoelectric conversion and signal conditioning on the received laser signal to generate raw depth point cloud data includes: The reflected laser signal is converted into a corresponding analog electrical signal using an array of photoelectric sensors; The analog electrical signal is amplified, filtered, and noise suppressed to obtain a conditioned electrical signal; The conditioned electrical signal is subjected to high-speed analog-to-digital conversion to obtain digitized laser intensity and phase information; By combining the laser projector's emission angle, scanning timing parameters, and the digitized laser intensity and phase information, the three-dimensional spatial coordinates are calculated point by point using the triangulation principle. All the calculated three-dimensional spatial coordinates are bound to the corresponding laser reflection intensity values ​​and organized according to the scanning order to form the original depth point cloud data.

3. The intelligent detection method for the protective layer thickness of power cement poles based on laser measurement according to claim 2, characterized in that, The process of performing spatial filtering and outlier removal on the original depth point cloud data to obtain purified surface point cloud data includes: Establish a spatial index structure for the original depth point cloud data; Traverse each point in the original depth point cloud data and query the distribution density of other points within its neighborhood; Based on a preset distance threshold, isolated points with a distribution density significantly lower than the normal level are identified and removed. For points that are not removed, statistical analysis is performed on the distance distribution of each point in its neighborhood along the normal direction to identify and remove outliers that deviate from the distribution center by more than a preset range. After removing outliers, a smoothing filter is applied to the remaining point set to smooth the data, and the purified surface point cloud data is output.

4. The intelligent detection method for the protective layer thickness of power cement poles based on laser measurement according to claim 3, characterized in that, The step of separating the support point cloud, which characterizes the main structure of the cement pole, and the masking point cloud, which characterizes the distribution characteristics of the reinforcing bars, from the purified surface point cloud data includes: Normal estimation is performed on the purified surface point cloud data, and the normal vector of each point is calculated; The purified surface point cloud data is initially clustered based on the point cloud curvature characteristics. In the initial clustering results, large clusters of point clouds with low curvature and spatial continuity are selected and classified as candidate main structure point clouds. A secondary analysis is performed on the point cloud of the candidate main structure. A cylindrical model is fitted by a random sampling consistency algorithm, and points that conform to the cylindrical model and whose residuals are within a set range are classified as the point cloud of the support body. The support point cloud is removed from the purified surface point cloud data. The remaining point cloud is then re-clustered based on spatial proximity and normal consistency. Point cloud clusters that appear as elongated strips or have periodic distribution characteristics are identified as the mask point cloud.

5. The intelligent detection method for the protective layer thickness of power cement poles based on laser measurement according to claim 4, characterized in that, The step of fitting a cylindrical surface to the point cloud of the support body to construct a theoretical reference cylindrical surface model of the cement pole includes: Randomly select several minimum subsets from the point cloud of the support, each of the minimum subsets containing the minimum number of points required to define the cylindrical geometric model; An initial cylinder model parameter is calculated using each of the minimum subsets, the cylinder model parameter including the central axis direction vector, the coordinates of a point on the central axis, and the cylinder radius; Calculate the distance from all points in the entire support point cloud to each initial cylindrical model, and mark the points whose distance is less than a set threshold as local points of the initial cylindrical model; The initial cylindrical model with the most inliers is selected as the optimal model candidate. The random sampling and optimization process is performed iteratively until the number of local points of the candidate optimal model reaches a preset stability condition. Using all the finally determined interior points, the precise parameters of the theoretical reference cylindrical model are calculated by least squares optimization.

6. The intelligent detection method for the protective layer thickness of power cement poles based on laser measurement according to claim 5, characterized in that, The step of projecting the point cloud of the mask onto the theoretical reference cylindrical surface model and calculating the set of projected positions of the reinforcing bars on the theoretical reference cylindrical surface model includes: Obtain the geometric parameters of the theoretical reference cylindrical surface model; For each point in the point cloud of the mask, calculate the shortest distance and the foot of the perpendicular from the point to the theoretical reference cylindrical surface model along the normal direction of the theoretical reference cylindrical surface model. All calculated perpendicular points are used as projection points of the point cloud of the mask onto the theoretical reference cylindrical surface model; The projection points are clustered by density based on their unfolded coordinates on the cylindrical surface to form several dense clusters; Calculate the geometric center of each of the dense point clusters, and record the three-dimensional coordinates of the geometric center on the theoretical reference cylindrical surface model as the projected position of a reinforcing bar. The projected positions of all the reinforcing bars constitute the set of projected positions.

7. The intelligent detection method for the protective layer thickness of power cement poles based on laser measurement according to claim 6, characterized in that, The step of extracting local point cloud data corresponding to the protective layer area around each rebar from the cleaned surface point cloud data based on the projection position set includes: For each projection position in the set of projection positions, a rectangular or sector-shaped neighborhood search window is defined on the theoretical reference cylindrical surface model with the projection position as the center. The neighborhood search window is projected backwards into three-dimensional space to determine a three-dimensional neighborhood search volume; In the purified surface point cloud data, all points whose spatial coordinates are located within the neighborhood search volume of the three dimensions are extracted to form an initial candidate point set; Based on the theoretical reference cylindrical surface model, calculate the distance from each point in the initial candidate point set to the theoretical reference cylindrical surface model; Points whose distance values ​​are within a reasonable range of the preset protective layer thickness are selected to form the final local point cloud data, which represents the protective layer area from the surface of the cement pole to the outer edge of the reinforcing bar.

8. The intelligent detection method for the protective layer thickness of power cement poles based on laser measurement according to claim 7, characterized in that, The step of performing geometric analysis on each of the local point cloud data to extract the physical boundary features of the protective layer region includes: Principal component analysis is performed on the local point cloud data to determine the main direction and extent of the point distribution; An analysis profile is constructed along the direction perpendicular to the normal of the theoretical reference cylindrical surface model and passing through the corresponding reinforcement projection position; The points in the local point cloud data are projected onto the analysis profile to form a set of two-dimensional scattered points. Statistical histogram analysis was performed on the two-dimensional scatter points to identify abrupt changes in data distribution density along the depth direction; The first abrupt change in data distribution from dense to sparse is identified as the location feature of the outer surface of the cement pole, and the second abrupt change in data distribution from sparse to dense is identified as the location feature of the surface of the reinforcing steel bar. These two location features together constitute the physical boundary feature.

9. The intelligent detection method for the protective layer thickness of power cement poles based on laser measurement according to claim 8, characterized in that, Based on the spatial relationship between the physical boundary features and the theoretical reference cylindrical surface model, the actual thickness of the concrete cover at each location is calculated, including: Obtain the position features of the outer surface of the cement pole and the surface of the reinforcing bar from the physical boundary features; Calculate the shortest distance from the feature point on the outer surface of the cement pole to the theoretical reference cylindrical surface model, and denot it as the outer surface distance; Calculate the shortest distance from the feature point on the surface of the reinforcing bar to the theoretical reference cylindrical surface model, and denot it as the reinforcing bar surface distance; Calculate the difference between the distance to the outer surface and the distance to the surface of the reinforcing bar; the difference is the actual thickness value. The calculated actual thickness values ​​are standardized in terms of unit and format.

10. The intelligent detection method for the protective layer thickness of power cement poles based on laser measurement according to claim 9, characterized in that, The actual thickness values ​​calculated from all locations are aggregated to generate a spatial distribution report of the protective layer thickness of the power cement pole, including: Establish a mapping table between spatial coordinates and the actual thickness value, wherein the spatial coordinates are based on the set of projection positions; The data in the mapping table is visualized and rendered on the two-dimensional unfolded diagram of the theoretical reference cylindrical surface model, with different colors or contour lines representing the thickness distribution. Statistical characteristics of all the actual thickness values ​​are calculated, including minimum, maximum, average and standard deviation. The visualization rendering results are integrated with statistical feature data, and a spatial distribution report containing spatial distribution maps and statistical data is generated according to a preset report template.

Images