Holographic Data-Based Method for Detecting Vertical Shaft Defects

Abstract

This invention relates to a method for detecting defects in vertical shafts based on holographic data. Existing methods cannot detect surface defects such as water ingress, flaking, and efflorescence on the inner wall of shafts. Therefore, this method for detecting defects in vertical shafts based on holographic data includes the following steps: (1) detecting the shaft using an inertial measurement unit and a laser scanner; (2) point cloud filtering; (3) segmented fitting of the central axis; (4) cutting sections based on the known central axis mileage and cross-sectional width to complete the shaft cross-section fitting and obtain a three-dimensional shaft model; (5) converting the point cloud data of the shaft wall in the three-dimensional shaft model into a grayscale image; (6) identifying blocky defects in the shaft based on the grayscale image of the three-dimensional shaft model. This method for detecting defects in vertical shafts based on holographic data is suitable for detecting defects in coal mine vertical shafts.

Description

Technical Field

[0001] This invention relates to a method for detecting defects in vertical shafts based on holographic data. Background Technology

[0002] Currently, well inspection is mainly conducted manually. On the one hand, manual inspection requires a large amount of manpower. Working in well environments with poor lighting, dampness, and frequent water intrusion can easily increase labor intensity and pose personal safety hazards. On the other hand, manual inspection relies mainly on human eyes to detect disease characteristics, which cannot promptly and accurately locate all well defects. Sometimes, due to subjective reasons, the existence of some defects may be overlooked, resulting in low defect identification capabilities and difficulty in quantitative statistical analysis.

[0003] Chinese invention patent CN 114739311 B, published on September 20, 2022, discloses a method for rapid deformation monitoring of mine shafts. The device employs multiple sensors to measure the initial coordinates of a 2D lidar, which is then gradually lowered into the shaft. Data at different times is recorded. Data recorded by a wheeled odometer and an inertial measurement unit (IMU) is preprocessed to calculate the lowering distance of the steel wire and the pose data of the IMU at different times. The lowering distance data calculated by the wheeled odometer and the pose data calculated by the IMU are interpolated by timestamps to obtain the pose of the 2D lidar at each timestamp. The precise coordinates of the shaft wall at different depths are calculated, a three-dimensional mesh model of the shaft is established, and the deformation data of the shaft is obtained. This invention uses a 2D lidar and can only detect shaft deformation; it cannot detect surface defects such as water ingress, spalling, and efflorescence on the shaft wall. Summary of the Invention

[0004] The technical problem to be solved by the present invention is that the existing technology cannot detect surface defects such as water, peeling, and efflorescence on the inner wall of the well. The present invention provides a method for detecting defects in vertical wells based on holographic data.

[0005] To address the aforementioned technical problems, this method for detecting defects in vertical shafts based on holographic data includes the following steps:

[0006] (1) Use an inertial measurement unit and a laser scanner to detect the well shaft and obtain high-precision POS data and point cloud data. POS data provides the real-time pose of the well shaft, and point cloud data can reflect the three-dimensional model of the well shaft. The POS data and point cloud data are fused to obtain the spatial coordinate information of the original laser scatter points of the well shaft in various directions. The three-dimensional coordinate information based on the local horizontal coordinate system is obtained through coordinate transformation.

[0007] (2) Point cloud filtering is performed using the principle of the cylinder model in the RANSAC algorithm. The mathematical model formula for the well shaft can be expressed as:

[0008]

[0009] In the formula, (x0, y0, z0) is a point on the central axis L of the target object, r is the radius of the target object, and (l, m, n) is a non-zero vector parallel to the central axis L.

[0010] The accuracy of this extraction method depends on the number of edge points. Before projection, a coordinate transformation is performed to align the wellbore's extension direction with the coordinate axes, thereby increasing the number of edge points. The coordinate transformation formula is as follows:

[0011]

[0012] Where, λ θ for:

[0013]

[0014] In the formula, θ is the transformation angle between coordinate systems;

[0015] (3) Piecewise fitting of the central axis: Each segmented curve is treated as a straight line. Each curve segment has a corresponding normal plane. The intersection of these two perpendicular normal planes is the central axis. Then, the discrete points on the intersection line are superimposed to obtain the central axis of the wellbore point cloud. The specific process of the fitting algorithm is as follows:

[0016] ①. Coordinate transformation: Before projecting the point cloud, adjust the direction of the well shaft extension so that the extension direction of the well shaft is along the direction of the coordinate axis;

[0017] ②. Point cloud projection: Project the point cloud of the well shaft onto the XOZ plane and YOZ plane respectively;

[0018] ③. Extract the edge point sets on both sides of the projection plane and fit them using a quadratic continuous function:

[0019] ④. Find the edge point set A by searching for its maximum value on the projection plane. Calculate the normal space vector of any point cloud in point set A. The intersection of this vector with the lower boundary line yields the edge point set B. Calculate the mean of point set A and point set B to obtain the midpoint set C1. Similarly, calculate the midpoint set C2 from the lower edge point set B using the same process. Finally, extract the points on the central axis using the mean of midpoint sets C1 and C2.

[0020] ⑤. Find the normal plane P orthogonal to the XOZ projection plane of the wellbore and the normal plane P' orthogonal to the YOZ projection plane of the wellbore. The intersection of plane P and plane P' is the central axis of the wellbore. The curve formed by summing multiple discrete points is the fitted central axis.

[0021] (4) Based on the known mileage of the central axis and the cross-sectional width, the cross-section is cut and the well section is fitted to obtain a three-dimensional well model;

[0022] The point cloud data of the well wall is truncated along the central axis at a certain thickness and projected onto a plane perpendicular to the central axis. Then, the least squares method is used to fit the cross-section and perform roundness analysis on the wellbore.

[0023] (5) Convert the point cloud data of the well wall in the 3D wellbore model into a grayscale image.

[0024] The point cloud data of the well wall is converted into a grayscale image through four steps: cross-section extraction, point cloud projection, image grayscale conversion, and mileage correction. During the conversion process, different weights are assigned to RGB values, and the results are calculated. Then, the weighted mean is converted to grayscale. This invention uses the following weighting formula to grayscale the input image signal. After the 3D point cloud is projected onto a 2D plane, a corresponding grayscale value is assigned based on the reflection intensity of the point cloud. Then, it is rasterized, with one grid cell representing one pixel in the image. The grayscale value of each pixel is determined by the point cloud intensity.

[0025] f(x,y)=0.3R(x,y)+0.59G(x,y)+0.11B(x,y)

[0026] (6) Identify blocky defects in the wellbore based on the grayscale image of the three-dimensional wellbore model.

[0027] Then, the grayscale image of the wellbore is processed using a connected component algorithm to obtain the number of connected components in the image, i.e., the number of leakage areas. The connected component algorithm is used to count a series of attributes of the marked regions. Its basic attribute information includes the number of pixels in different numbered regions, the centroid coordinates of each numbered region, and the calculation of the minimum bounding circle for each region. Finally, the minimum bounding circle is imported into the grayscale image, enabling the visualization of automatic detection of well wall leakage and intuitively obtaining the spatial distribution location of well wall leakage.

[0028] This invention also proposes an automatic identification method for seepage, detachment, and efflorescence diseases based on disease characteristics. This method combines average grayscale value, connected region area, and the number of connected sub-blocks within a moving window. First, minimum thresholds for average grayscale value and connected region area are set. Then, a structural circle is constructed as a moving window. This structural circle moves horizontally from left to right and from top to bottom across the entire grayscale image region. At each position of the moving structural circle, the number of connected regions enclosed by the circle is calculated. Based on statistical experience data from wellbore disease images, a minimum average grayscale threshold of 30 and a minimum connected region threshold of 50 are selected. The geometric semantic descriptions of seepage, detachment, and efflorescence are as follows:

[0029] a. If the gray value is less than 40 and the area of ​​the connected region is greater than 0.15 square meters, then the disease is considered to be a seepage area;

[0030] b. If the gray value is greater than 130 and the area of ​​the connected region is greater than 0.05 square meters, then the disease is considered to be an efflorescence area;

[0031] c. If the gray value is greater than 130, the area of ​​the connected region is less than 0.05 square meters, and the number of connected sub-blocks in the moving window is less than 10, then the disease is considered to be an efflorescence area.

[0032] d. If the gray value is greater than 130, the area of ​​the connected region is less than 0.05 square meters, and the number of connected sub-blocks in the moving window is greater than 10, then the disease is considered to be a detached area.

[0033] As an optimization, the point cloud filtering process includes the following steps:

[0034] ①. Randomly select a sample set from the target point cloud, use the least squares method to estimate the mathematical model parameters of the random sample set, and fit the model.

[0035] ②Then, the deviation between the current point cloud set and the remaining point cloud set is calculated iteratively, and the deviation is then compared with the algorithm threshold.

[0036] If the deviation value is greater than the threshold, the sample set is classified as an inlier; otherwise, the above process is repeated iteratively.

[0037] ③. Finally, the data model is fitted using a sample set of all local points. When the performance of the mathematical model of the target object reaches a certain threshold, the iteration ends and the optimal model parameters are determined by calculation.

[0038] As an optimization, the inertial measurement unit is equipped with a high-precision positioning attitude system, and the laser scanner is equipped with a CCD panoramic camera.

[0039] This invention utilizes an inertial measurement unit combined with a laser scanner to inspect the wellbore, acquiring high-precision point cloud data and spatial coordinate information of the original laser scatter points in various directions. Through coordinate transformation, three-dimensional coordinate information based on the local horizontal coordinate system is derived. By incorporating multiple sensors, including a CCD panoramic camera and a high-precision positioning attitude system, along with an assembly control system, a high-performance computer, and control devices, a method for detecting defects in vertical shafts based on holographic data has been developed.

[0040] The present invention provides a method for detecting defects in vertical shafts based on holographic data, which is suitable for detecting defects in vertical shafts of coal mines. Attached Figure Description

[0041] The present invention, based on holographic data, for detecting defects in vertical shafts will be further described below with reference to the accompanying drawings:

[0042] Figure 1 This is a schematic diagram of a cylindrical wellbore model;

[0043] Figure 2 This is the RANSAC algorithm flow;

[0044] Figure 3 This is a schematic diagram of the centerline of the well shaft. Detailed Implementation

[0045] Implementation method one: such as Figure 1-3 As shown, this method for detecting defects in vertical shafts based on holographic data includes the following steps:

[0046] (1) Use an inertial measurement unit and a laser scanner to detect the well shaft and obtain high-precision POS data and point cloud data. POS data provides the real-time pose of the well shaft, and point cloud data can reflect the three-dimensional model of the well shaft. The POS data and point cloud data are fused to obtain the spatial coordinate information of the original laser scatter points of the well shaft in various directions. The three-dimensional coordinate information based on the local horizontal coordinate system is obtained through coordinate transformation.

[0047] (2) Point cloud filtering is performed using the principle of the cylinder model in the RANSAC algorithm. The mathematical model formula for the well shaft can be expressed as:

[0048]

[0049] In the formula, (x0, y0, z0) is a point on the central axis L of the target object, r is the radius of the target object, and (l, m, n) is a non-zero vector parallel to the central axis L.

[0050] The accuracy of this extraction method depends on the number of edge points. Before projection, a coordinate transformation is performed to align the wellbore's extension direction with the coordinate axes, thereby increasing the number of edge points. The coordinate transformation formula is as follows:

[0051]

[0052] Where, λ θ for:

[0053]

[0054] In the formula, θ is the transformation angle between coordinate systems;

[0055] (3) Piecewise fitting of the central axis: Each segmented curve is treated as a straight line. Each curve segment has a corresponding normal plane. The intersection of these two perpendicular normal planes is the central axis. Then, the discrete points on the intersection line are superimposed to obtain the central axis of the wellbore point cloud. The specific process of the fitting algorithm is as follows:

[0056] ①. Coordinate transformation: Before projecting the point cloud, adjust the direction of the well shaft extension so that the extension direction of the well shaft is along the direction of the coordinate axis;

[0057] ②. Point cloud projection: Project the point cloud of the well shaft onto the XOZ plane and YOZ plane respectively;

[0058] ③. Extract the edge point sets on both sides of the projection plane and fit them using a quadratic continuous function:

[0059] ④. Find the edge point set A by searching for its maximum value on the projection plane. Calculate the normal space vector of any point cloud in point set A. The intersection of this vector with the lower boundary line yields the edge point set B. Calculate the mean of point set A and point set B to obtain the midpoint set C1. Similarly, calculate the midpoint set C2 from the lower edge point set B using the same process. Finally, extract the points on the central axis using the mean of midpoint sets C1 and C2.

[0060] ⑤. Find the normal plane P orthogonal to the XOZ projection plane of the wellbore and the normal plane P' orthogonal to the YOZ projection plane of the wellbore. The intersection of plane P and plane P' is the central axis of the wellbore. The curve formed by summing multiple discrete points is the fitted central axis.

[0061] (4) Based on the known mileage of the central axis and the cross-sectional width, the cross-section is cut and the well section is fitted to obtain a three-dimensional well model;

[0062] The point cloud data of the well wall is truncated along the central axis at a certain thickness and projected onto a plane perpendicular to the central axis. Then, the least squares method is used to fit the cross-section and perform roundness analysis on the wellbore.

[0063] (5) Convert the point cloud data of the well wall in the 3D wellbore model into a grayscale image.

[0064] The point cloud data of the well wall is converted into a grayscale image through four steps: cross-section extraction, point cloud projection, image grayscale conversion, and mileage correction. During the conversion process, different weights are assigned to RGB values, and the results are calculated. Then, the weighted mean is converted to grayscale. This invention uses the following weighting formula to grayscale the input image signal. After the 3D point cloud is projected onto a 2D plane, a corresponding grayscale value is assigned based on the reflection intensity of the point cloud. Then, it is rasterized, with one grid cell representing one pixel in the image. The grayscale value of each pixel is determined by the point cloud intensity.

[0065] f(x,y)=0.3R(x,y)+0.59G(x,y)+0.11B(x,y)

[0066] (6) Identify blocky defects in the wellbore based on the grayscale image of the three-dimensional wellbore model.

[0067] Then, the grayscale image of the wellbore is processed using a connected component algorithm to obtain the number of connected components in the image, i.e., the number of leakage areas. The connected component algorithm is used to count a series of attributes of the marked regions. Its basic attribute information includes the number of pixels in different numbered regions, the centroid coordinates of each numbered region, and the calculation of the minimum bounding circle for each region. Finally, the minimum bounding circle is imported into the grayscale image, enabling the visualization of automatic detection of well wall leakage and intuitively obtaining the spatial distribution location of well wall leakage.

[0068] This invention also proposes an automatic identification method for seepage, detachment, and efflorescence diseases based on disease characteristics. This method combines average grayscale value, connected region area, and the number of connected sub-blocks within a moving window. First, minimum thresholds for average grayscale value and connected region area are set. Then, a structural circle is constructed as a moving window. This structural circle moves horizontally from left to right and from top to bottom across the entire grayscale image region. At each position of the moving structural circle, the number of connected regions enclosed by the circle is calculated. Based on statistical experience data from wellbore disease images, a minimum average grayscale threshold of 30 and a minimum connected region threshold of 50 are selected. The geometric semantic descriptions of seepage, detachment, and efflorescence are as follows:

[0069] a. If the gray value is less than 40 and the area of ​​the connected region is greater than 0.15 square meters, then the disease is considered to be a seepage area;

[0070] b. If the gray value is greater than 130 and the area of ​​the connected region is greater than 0.05 square meters, then the disease is considered to be an efflorescence area;

[0071] c. If the gray value is greater than 130, the area of ​​the connected region is less than 0.05 square meters, and the number of connected sub-blocks in the moving window is less than 10, then the disease is considered to be an efflorescence area.

[0072] d. If the gray value is greater than 130, the area of ​​the connected region is less than 0.05 square meters, and the number of connected sub-blocks in the moving window is greater than 10, then the disease is considered to be a detached area.

[0073] As an optimization, the point cloud filtering process includes the following steps:

[0074] ①. Randomly select a sample set from the target point cloud, use the least squares method to estimate the mathematical model parameters of the random sample set, and fit the model.

[0075] ②. Then, iteratively calculate the deviation between the current point cloud set and the remaining point cloud sets, and then compare the deviation with the algorithm threshold.

[0076] If the deviation value is greater than the threshold, the sample set is classified as an inlier; otherwise, the above process is repeated iteratively.

[0077] ③. Finally, the data model is fitted using a sample set of all local points. When the performance of the mathematical model of the target object reaches a certain threshold, the iteration ends and the optimal model parameters are determined by calculation.

[0078] As an optimization, the inertial measurement unit is equipped with a high-precision positioning attitude system, and the laser scanner is equipped with a CCD panoramic camera.

[0079] This invention utilizes an inertial measurement unit combined with a laser scanner to inspect the wellbore, acquiring high-precision point cloud data and spatial coordinate information of the original laser scatter points in various directions. Through coordinate transformation, three-dimensional coordinate information based on the local horizontal coordinate system is derived. By incorporating multiple sensors, including a CCD panoramic camera and a high-precision positioning attitude system, along with an assembly control system, a high-performance computer, and control devices, a method for detecting defects in vertical shafts based on holographic data has been developed.

[0080] (1) Establish a mathematical model and solve for wellbore deformation parameters.

[0081] This invention iteratively calculates the parameter values ​​of a mathematical model to ultimately determine the mathematical model fitted to the target object, and uses this mathematical model as the basis for point cloud filtering. The RANSAC algorithm simulates the mathematical model according to standard geometric shapes, including curved surface models, three-dimensional parabolic surface models, and cylindrical models. By improving the above method, the point cloud set on the inner wall of the well shaft can be regarded as a cylinder. Therefore, the principle of the cylindrical model in the RANSAC algorithm is adopted for point cloud filtering. The mathematical model formula for the well shaft can be expressed as:

[0082]

[0083] In the formula, (x0,y0,z0) is a point on the central axis L of the target object, r is the radius of the target object, and (l,m,n) is a non-zero vector parallel to the central axis L.

[0084] The accuracy of this extraction method depends on the number of edge points. Before projection, a coordinate transformation is performed to align the wellbore's extension direction with the coordinate axes, thereby increasing the number of edge points. The coordinate transformation formula is as follows:

[0085]

[0086] Where, λ θ for:

[0087]

[0088] In the formula, θ is the transformation angle between coordinate systems.

[0089] The target object of this invention is the vertical shaft of a coal mine. The inner wall of the shaft, with its clustered points, can be considered a cylinder. Figure 1 As shown.

[0090] The present invention should meet the following conditions when estimating the mathematical model:

[0091] (1) Existence of noise: There are some invalid noises mixed in near the target point cloud.

[0092] (2) Point clouds need to have a standard geometric shape.

[0093] (3) When performing iterative calculations of the mathematical model, only one cross-sectional model can be estimated.

[0094] (4) Set the number of iterations and algorithm constraints reasonably.

[0095] Once the above conditions are met, point cloud filtering is performed based on the improved RANSAC algorithm. The algorithm flow is as follows: Figure 2 As shown.

[0096] A sample set is randomly selected from the target point cloud. The mathematical model parameters of the random sample set are estimated using the least squares method, and the model is fitted. Then, the deviation between the current point cloud set and the remaining point clouds is iteratively calculated, and the deviation is compared with the algorithm threshold. If the deviation value is greater than the threshold, the sample set is classified as an inlier; otherwise, the above process is repeated iteratively. Finally, the data model is fitted using the sample set of all inliers. When the performance of the target mathematical model reaches a set specific threshold, the iteration ends, and the optimal model parameters are determined through calculation.

[0097] The centerline shows the overall trend of the wellbore and indicates its direction of extension, providing crucial wellbore parameter information and mileage location. Therefore, fitting the centerline is essential. Consequently, this invention will subsequently rely on the fitted wellbore centerline, extracting cross-sections based on the known centerline mileage location and cross-sectional width. After completing the wellbore cross-section fitting, the elliptical parameters of the cross-section and the coordinate origin will be used to perform wellbore deformation analysis. For example... Figure 3 As shown.

[0098] The proposed centerline fitting method employs a piecewise approach. After dividing the curve into infinitesimally small, equally spaced segments, each segment can be considered a straight line. Each curve segment has a corresponding normal plane, and the intersection of these two perpendicular normal planes is the centerline. By superimposing the discrete points along the intersection line, the centerline of the wellbore point cloud can be obtained. The specific process of the fitting algorithm is as follows:

[0099] (1) Coordinate transformation: Before projecting the point cloud, adjust the direction of the well extension so that the extension direction of the well is along the direction of the coordinate axis;

[0100] (2) Point cloud projection: Project the point cloud of the well shaft onto the XOZ plane and the YOZ plane respectively;

[0101] (3) Extract the edge point sets on both sides of the projection plane and fit them using a quadratic continuous function:

[0102] (4) The edge point set A is obtained by searching the maximum value of the edge points on the projection plane. The normal space vector of any point cloud in the point set A is calculated. The intersection of this vector with the lower boundary line can obtain the edge point set B. The midpoint set C1 is obtained by taking the mean of the point set A and the point set B. Similarly, the midpoint set C2 can be obtained by taking the lower edge point set B according to the above process. Finally, the points on the central axis are extracted by taking the mean of the midpoint set C1 and the midpoint set C2.

[0103] (5) Find the normal plane P orthogonal to the XOZ projection plane of the wellbore and the normal plane P' orthogonal to the YOZ projection plane of the wellbore. The intersection of plane P and plane P' is the central axis of the wellbore. The central axis is essentially a curve formed by the sum of multiple discrete points. The central axis of the wellbore is as follows: Figure 3 As shown.

[0104] Suppose the equations of plane P and plane P' are:

[0105] A1x+B1y+C1z+D1=0

[0106] A²x + B²y + C²z + D² = 0

[0107] If the spatial vectors of the intersection line of two planes are known, a = {A1, B1, C1}, b = {A2, B2, C2}, then:

[0108]

[0109] The intersecting lines intersect the XOZ plane. Taking z = 0, we solve the following equations:

[0110] A1x + B1y + D1 = 0

[0111] A²x + B²y + D² = 0

[0112] A point on the intersecting lines can be found as:

[0113]

[0114] Let this point be denoted as (M,N,0). Therefore, the direction vector (E,F,G) passing through the point (M,N,0) can be represented as:

[0115]

[0116] After searching for the upper and lower edge points, the RANSAC algorithm is used to fit the centerline of the edge point set. The fitting process is as follows: The upper and lower edge point sets are denoted as (x... k y k z k ) m ,(x l y l z l ) n Each of them is fitted with a quadratic curve.

[0117]

[0118] Take the average of the two curves:

[0119]

[0120] (xyz) after mean s That is, the central axis.

[0121] Given the mileage of a certain point, the mileage of any point on the axis can be determined. If the mileage of point p on the central axis is known... n (x n y n z n If the mileage position is S, then the distance to any point p can be calculated using discrete point accumulation. m (x m y m z m Mileage position S m for:

[0122]

[0123] over p m ,p n The normal plane is:

[0124]

[0125] If S and S m If the spacing is d, then the cross-sectional data within this mileage interval can be represented as:

[0126]

[0127] Similarly, the point cloud can be segmented based on the distance between two points and the known mileage position of the points, where d is the width of the cross section.

[0128] Without considering factors such as construction and mining operations, the cross-section of the well shaft can be considered a standard circle. Analysis of the stress characteristics of the well shaft reveals that its deformation is a complex process involving multiple interacting factors; its perimeter remains essentially constant, but its shape tends to become elliptical. Therefore, the generalized equation of an ellipse is used for calculation when fitting the cross-section.

[0129] The generalized equation of an ellipse is:

[0130] Ax 2 +By 2 +Cxy+Dx+Ey=0

[0131] When solving for the five parameters ABCDE, the RANSAC algorithm process described above can be followed. Five measured data points can be randomly selected and substituted into the calculation. The parameter values ​​of the mathematical model can be estimated through iteration, and finally, the mathematical model fitted to the target object can be determined. At the same time, in order to overcome the problem of excessive algorithm iterations, we impose a constraint that the major and minor axes of the ellipse are not much different from the initial diameter value, thereby improving the efficiency of the algorithm.

[0132] Let the ellipse parameters obtained from the measured points be (A) n B n C n D n E n ), the geometric center of the ellipse (x n ,y n )for:

[0133]

[0134] Converting to the standard form, the major and minor axes ab are:

[0135]

[0136]

[0137] After determining the design diameter R, a threshold is set by the difference between ab and R. If the threshold condition is met, the model is temporarily retained until the optimal solution is obtained, thereby obtaining the five parameters of the ellipse and the coordinates of the center. Deformation detection is completed by comparing the major and minor axes with the design diameter of the wellbore.

[0138] (2) Identify blocky defects in the wellbore based on the grayscale image of the well wall.

[0139] Before identifying blocky defects such as water leakage, this invention proposes the concept of a minimum enclosing circle. The basic parameters of the minimum enclosing circle are stored in an N×3 matrix, where N rows represent the number of minimum enclosing circles, and the three columns represent the three basic parameters of the minimum enclosing circle: the x-coordinate of the centroid of the blocky defect, the y-coordinate of the centroid of the blocky defect, and the radius of the minimum enclosing circle.

[0140] Grayscale images contain sufficient information to reflect both overall and local features of an image, and are computationally simple with limited information content. To facilitate subsequent defect extraction, the point cloud data of the well wall is converted into a grayscale image through four steps: cross-section extraction, point cloud projection, image grayscale conversion, and mileage correction. During the conversion process, different weights are assigned to RGB values, and the results are calculated. The weighted mean is then converted to grayscale. This invention uses the following weighting formula to grayscale the input image signal. Furthermore, after the three-dimensional point cloud is projected onto a two-dimensional plane, a corresponding grayscale value is assigned based on the reflection intensity of the point cloud, and then it is rasterized. One grid represents one pixel in the image, and the grayscale value of each pixel is determined by the point cloud intensity.

[0141] f(x,y)=0.3R(x,y)+0.59G(x,y)+0.11B(x,y)

[0142] Then, the grayscale image of the wellbore is processed using a connected component algorithm to obtain the number of connected components in the image, i.e., the number of leakage areas. The connected component algorithm is used to statistically analyze a series of attributes of the marked regions. Its basic attribute information includes the number of pixels in different numbered regions, the centroid coordinates of each numbered region, and the calculation of the minimum bounding circle for each region. Finally, the minimum bounding circle is imported into the grayscale image, enabling visualization of automatic wellbore leakage detection and intuitively obtaining the spatial distribution location of wellbore leakage.

[0143] This invention also proposes a method for automatically identifying seepage, detachment, and efflorescence damage types based on damage characteristics. This method combines average grayscale value, connected region area, and the number of connected sub-blocks within a moving window. First, minimum thresholds for average grayscale value and connected region area are set. Then, a structural circle is constructed as a moving window. The structural circle moves horizontally from left to right and from top to bottom across the entire grayscale image region. At each position of the structural circle, the number of connected regions enclosed by the structural circle is calculated. Based on statistical experience data from wellbore damage images, a minimum average grayscale threshold of 30 and a minimum connected region threshold of 50 are selected. The geometric semantic descriptions of seepage, detachment, and efflorescence are as follows:

[0144] a. If the gray value is less than 40 and the area of ​​the connected region is greater than 0.15 square meters, then the disease is considered to be a seepage area;

[0145] b. If the gray value is greater than 130 and the area of ​​the connected region is greater than 0.05 square meters, then the disease is considered to be an efflorescence area;

[0146] c. If the gray value is greater than 130, the area of ​​the connected region is less than 0.05 square meters, and the number of connected sub-blocks in the moving window is less than 10, then the disease is considered to be an efflorescence area.

[0147] d. If the gray value is greater than 130, the area of ​​the connected region is less than 0.05 square meters, and the number of connected sub-blocks in the moving window is greater than 10, then the disease is considered to be a detached area.

[0148] The present invention provides the following description regarding the semantic discrimination of blocky defects in wellbore:

[0149] (1) Analysis of defects: The purpose of vertical shaft inspection is to quickly detect and locate well shaft defects that may cause well shaft damage. Among the common well shaft defects, in addition to well shaft cracks, there are also various other forms of well shaft defects such as water seepage, spalling, and efflorescence.

[0150] Water seepage refers to the infiltration of surface water or moisture adhering to the concrete structure into the surface of the well casing's inner wall. Through actual observation, the seepage area in the well casing is generally darker than the surrounding area due to water infiltration. When converted to grayscale, the grayscale value of the seepage area is significantly lower than the surrounding area. Seepage areas typically exhibit irregular outlines and are relatively large.

[0151] Spalling refers to the phenomenon where sand and cement paste detach from the concrete surface, exposing coarse aggregate. In severe cases, the aggregate and the mortar surrounding it may detach together, or the mortar on the concrete surface may peel off in flakes. The spalling area often presents as an irregularly shaped distribution.

[0152] Efflorescence: In ordinary concrete well walls, when exposed to air or moisture within the wall, some of the concrete components undergo hydrolysis, producing alkaline substances with low solubility. As the temperature gradually rises, water vapor evaporates, slowly releasing these alkaline substances from the wall's interior. With each evaporation, the alkaline substances accumulate on the concrete surface, causing whitening and peeling, forming a phenomenon known as efflorescence. Efflorescence reduces the wall's strength, lowering its load-bearing capacity, and also reduces the building's cross-sectional area, increasing the load per unit area and posing a potential hazard to the well shaft. Efflorescence areas typically exhibit irregular shapes.

[0153] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as herein.

[0154] The above embodiments are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of this invention. The embodiments of the present invention have been described in detail above, but the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for detecting defects in vertical shafts based on holographic data, comprising the following steps: (1) Use an inertial measurement unit and a laser scanner to detect the well shaft and obtain high-precision POS data and point cloud data. The POS data provides the real-time pose of the well shaft, and the point cloud data reflects the three-dimensional model of the well shaft. The POS data and point cloud data are fused to obtain the spatial coordinate information of the original laser scatter points of the well shaft in various directions. The three-dimensional coordinate information based on the local horizontal coordinate system is obtained through coordinate transformation. (2) Point cloud filtering is performed using the principle of the cylinder model in the RANSAC algorithm. The mathematical model formula for the wellbore is expressed as: In the formula, Let r be a point on the central axis L of the target object, and r be the radius of the target object. A non-zero vector parallel to the central axis L. The accuracy of this extraction method depends on the number of edge points. Before projection, a coordinate transformation is performed to align the wellbore's extension direction with the coordinate axes, thereby increasing the number of edge points. The coordinate transformation formula is as follows: in, for: In the formula, The angle between coordinate systems; (3) Piecewise fitting of the central axis: Each segmented curve is regarded as a straight line. For each curve segment, there is a corresponding normal plane. The intersection of these two mutually perpendicular normal planes is the central axis. Then, the spatial discrete points on the intersection line are superimposed to obtain the central axis of the wellbore point cloud. The specific process of the fitting algorithm is as follows: ①. Coordinate transformation: Before projecting the point cloud, adjust the direction of the well shaft extension so that the extension direction of the well shaft is along the direction of the coordinate axis; ②. Point cloud projection: Project the point cloud of the well shaft onto the XOZ plane and YOZ plane respectively; ③. Extract the edge point sets on both sides of the projection plane and fit them using a quadratic continuous function: ④. Find the edge point set A by searching for its maximum value on the projection plane. Calculate the normal space vector of any point cloud in point set A. This vector intersects the lower boundary line, yielding the edge point set B. Calculate the mean of point set A and point set B to obtain the midpoint set C1. Similarly, calculate the midpoint set C2 from the lower edge point set B using the above process. Finally, extract the points on the central axis using the mean of midpoint sets C1 and C2. ⑤. Find the normal plane P orthogonal to the XOZ projection plane of the wellbore and the normal plane P' orthogonal to the YOZ projection plane of the wellbore. The intersection of plane P and plane P' is the central axis of the wellbore. The curve formed by summing multiple discrete points is the fitted central axis. (4) Based on the known mileage of the central axis and the cross-sectional width, the cross-section is cut off, the cross-section of the well is fitted, and a three-dimensional well model is obtained; The point cloud data of the well wall is truncated along the central axis at a certain thickness and projected onto a plane perpendicular to the central axis. Then, the least squares method is used to fit the cross-section and perform roundness analysis on the wellbore. (5) Convert the point cloud data of the well wall in the three-dimensional wellbore model into a grayscale image. The point cloud data of the well wall is converted into a grayscale image through four steps: cross-section extraction, point cloud projection, image grayscale conversion, and mileage correction. During the conversion process, different weights are assigned to RGB values, and the results are calculated. Then, the weighted mean is converted to grayscale, and the input image signal is grayscaled using the following weighted formula. After the 3D point cloud is projected onto a 2D plane, a corresponding grayscale value is assigned based on the reflection intensity of the point cloud. Finally, it is rasterized, with one grid cell representing one pixel in the image, and the grayscale value of each pixel is determined by the point cloud intensity. (6) Based on the grayscale image of the three-dimensional wellbore model, identify blocky defects in the wellbore. Then, the grayscale image of the wellbore is processed using a connected component algorithm to obtain the number of connected components in the image, i.e., the number of leakage areas. The connected component algorithm is used to count a series of attributes of the marked regions. Its basic attribute information includes the number of pixels in different numbered regions, the centroid coordinates of each numbered region, and the calculation of the minimum bounding circle for each region. Finally, the minimum bounding circle is imported into the grayscale image, enabling visualization of automatic wellbore leakage detection and intuitively obtaining the spatial distribution location of wellbore leakage. To automatically identify seepage, detachment, and efflorescence diseases, a method combining average grayscale value, connected region area, and the number of connected sub-blocks within a moving window is proposed based on disease characteristics. First, minimum thresholds for average grayscale value and connected region area are set. Then, a structural circle is constructed as a moving window, which moves horizontally from left to right and from top to bottom across the entire grayscale image region. At each position of the moving structural circle, the number of connected regions enclosed by the circle is calculated. Based on statistical experience data from well shaft disease images, a minimum threshold of 30 for average grayscale value and a minimum threshold of 50 for connected region area are selected. The geometric semantic descriptions of seepage, detachment, and efflorescence are as follows: a. If the gray value is less than 40 and the area of ​​the connected region is greater than 0.15 square meters, then the disease is considered to be a seepage area; b. If the gray value is greater than 130 and the area of ​​the connected region is greater than 0.05 square meters, then the disease is considered to be an efflorescence area; c. If the gray value is greater than 130, the area of ​​the connected region is less than 0.05 square meters, and the number of connected sub-blocks in the moving window is less than 10, then the disease is considered to be an efflorescence area. d. If the gray value is greater than 130, the area of ​​the connected region is less than 0.05 square meters, and the number of connected sub-blocks in the moving window is greater than 10, then the disease is considered to be a detached area.

2. The method for detecting defects in vertical shafts based on holographic data according to claim 1, characterized in that: The point cloud filtering process includes the following steps: ①. Randomly select a sample set from the target point cloud, use the least squares method to estimate the mathematical model parameters of the random sample set, and fit the model. ②. Then, iteratively calculate the deviation between the current point cloud set and the remaining point cloud sets, and then compare the deviation with the algorithm threshold. If the deviation value is greater than the threshold, the sample set is classified as an inlier; otherwise, the above process is repeated iteratively. ③. Finally, the data model is fitted using a sample set of all local points. When the performance of the mathematical model of the target object reaches a certain threshold, the iteration ends and the optimal model parameters are determined by calculation.

3. The method for detecting defects in vertical shafts based on holographic data according to claim 1 or 2, characterized in that: The inertial measurement unit uses a high-precision positioning attitude system, and the laser scanner uses a CCD panoramic camera.

Images