A full-scale detection method for bolt group loosening based on three-dimensional point cloud

Abstract

The present invention provides a full-scale detection method for bolt group loosening based on three-dimensional point cloud, which relates to the technical field of bolt loosening detection, and the specific steps include: S1. Using a laser scanner to collect three-dimensional point cloud data of a bolt group, the three-dimensional point cloud data includes point cloud data of bolts, connecting plates and surrounding environment; S2. Performing plane processing on the point cloud data, identifying the largest plane in the point cloud, and extracting the inner point set of the largest plane from the point cloud data. The present invention can effectively identify and separate the connecting plate and the plane part by performing plane processing and data segmentation on the point cloud data, thereby improving the accuracy and reliability of detection; in addition, by locating the position of the bolt on the connecting plate, slicing to obtain the three-dimensional space contour line of the bolt node, and projecting it to a two-dimensional plane for automatic clustering processing, the total length of the integer slices of the protruding part of the bolt and the remaining length after slicing can be accurately extracted; it can also avoid the influence of light and provide more stable and high-precision detection results.

Description

Technical Field

[0001] The present invention relates to the technical field of bolt loosening detection, and in particular to a full-scale detection method for bolt group loosening based on three-dimensional point cloud. Background Art

[0002] In the prior art, a method, system and storage medium for detecting loose bolts based on deep learning, with application publication number CN114742820B, are provided, which obtain a historical bolt image set and construct a target detection data set and a key point detection data set; construct a target detection network model, train the target detection network model through the target detection data set, and obtain an optimal target detection model; construct a key point detection model, train the key point detection model through the key point detection data set, and obtain an optimal key point detection model; obtain a real-time bolt image, and input the real-time bolt image into the optimal target detection model and the optimal key point detection model in turn for detection, and perform geometric correction on the real-time bolt image according to the detection result, and then calculate the bolt edge line angle value by the directed edge line value method, and judge whether the bolt is loose according to the angle value; the beneficial effect of the present invention is to improve the efficiency of bolt detection; accurately detect the coordinates of the six corner points of the bolt, and then calculate the bolt angle based on the corner point coordinates, thereby improving the accuracy and reliability of high-strength bolt loosening detection.

[0003] However, there are still some shortcomings. From the above statements, it can be seen that the deep learning model is very sensitive to image quality and lighting conditions. In practical applications, bolt detection is often limited by changes in lighting conditions on site, such as insufficient light, reflections or shadows. When the corners are blurred or blocked, the detection accuracy will decrease, which will affect the final looseness judgment result.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0005] The purpose of the present invention is to provide a full-scale detection method for bolt group loosening based on three-dimensional point cloud to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A full-scale detection method for bolt group loosening based on three-dimensional point cloud, the specific steps include:

[0008] S1. Use a laser scanner to collect three-dimensional point cloud data of the bolt group, where the three-dimensional point cloud data includes point cloud data of the bolts, connecting plates, and surrounding environment;

[0009] S2. Perform plane processing on the point cloud data, identify the largest plane in the point cloud, extract the inner point set of the largest plane from the point cloud data, and segment the point cloud data into a plane part and a non-plane part through a data segmentation algorithm, wherein the plane part is the point cloud of the connecting plate, and the non-plane part is the point cloud of the bolt and the surrounding environment;

[0010] S3. By positioning the bolt on the connecting plate, determine the slicing plane at the position where the screw extends, slice the point cloud data on the slicing plane, and obtain the contour line of the bolt node in three dimensions;

[0011] S4. Calculate the normal vector of the three-dimensional space contour line of the bolt node, select a coordinate plane perpendicular to the maximum plane normal vector of the point cloud data, project the sliced ​​three-dimensional space coordinate points onto the selected two-dimensional coordinate plane, convert them into two-dimensional plane coordinate points, and automatically cluster the point cloud data of the bolt node connection plate on the two-dimensional plane projection image of the screw extension position to obtain the two-dimensional space contour line clusters of all bolts on the bolt node connection plate;

[0012] S5. extract the contour line cluster in the two-dimensional space after projection, obtain the screw extension length and the remaining length after the screw extension part is sliced, convert the screw extension length into the bolt rotation angle, convert the remaining length after the screw extension part is sliced ​​into the bolt rotation angle, compare the bolt rotation angle converted from the total length of the integer slice of the screw extension part with its threshold, and compare the bolt rotation angle converted from the remaining length after the screw extension part is sliced ​​with its threshold;

[0013] S6. Determine whether the bolt is loose based on the comparison result.

[0014] Furthermore, the point cloud data is processed in a plane, the largest plane in the point cloud is identified, and the process of extracting the inner point set of the largest plane from the point cloud data is as follows:

[0015] The RANSAC algorithm is used to fit the point cloud data to a plane. The maximum number of iterations is set to N, the distance threshold is set to d, and three points are randomly selected from the point cloud data. 1 =(x 1 ,y 1 , z 1 ), P 2 =(x 2 ,y 2 , z 2 ), P 3 =(x 3 ,y 3 , z 3 ), then the vector V 1 =P 2 -P 1 =(x 2 -x 1, y 2 -y 1 , z 2 -z 1 ), vector V 2 = P 3 -P 1 = (x 3 -x 1 , y 3 -y 1 , z 3 -z 1 ), normal vector n = V 1 ×V 2 , set the plane equation Ax + By + Cz + D = 0, where (A, B, C) are the components of the plane normal vector, and D is the distance of the plane from the origin, A = n x , B = n y , C = n z , D = -(Ax 1 + By 1 + Cz 1 );

[0016] For each point in the point cloud data, P i = (x i , y i , z i ), calculate its distance to this plane, and the formula is as follows:

[0017]

[0018] When D(P i ) < d, then P i is considered an inlier;

[0019] Record the current number of inliers. If the current number of inliers is more than the previously recorded maximum number of inliers, then update the maximum plane and the maximum inlier set. After N iterations, select the plane with the most inliers as the finally fitted plane.

[0020] Furthermore, the process of segmenting the point cloud data into a plane part and a non - plane part through a data segmentation algorithm is as follows:

[0021] Extract the inlier set of the maximum plane to form the point cloud data of the plane part. The plane part P plane is composed of the inlier set of the maximum plane, that is, the point cloud data of the connecting plate. Delete the point cloud data of the plane part P plane from the original point cloud data set. The remaining point cloud data is the non - plane part P non-plane , that is, the point cloud data of the bolt and its surrounding environment;

[0022] The RANSAC algorithm is used to identify and fit the largest plane in the point cloud data, and the inner point set P of the plane is obtained. i ,

[0023] P i ∈P|distance(P i , plane) <d

[0024] Among them, distance(P i , plane) represents point P i The distance to the fitting plane, d is the distance threshold;

[0025] The non-planar point set is the remaining point set after deleting the planar point cloud data from the original point cloud data set. The specific steps to obtain it are as follows:

[0026] Initialize the non-planar point set and create an empty point set P non-plane ;

[0027] Traverse the original point cloud dataset, for each point P i ∈P, when Then P i Belong to the non-planar part, add the non-planar part P non-plane .

[0028] Furthermore, by locating the position of the bolt on the connecting plate, determining the slice plane of the screw extension position, slicing the point cloud data on the slice plane, and obtaining the contour line of the three-dimensional space of the bolt node, the process is as follows:

[0029] The center point of the bolt K=(x c ,y c , z c ), the bolt direction vector g=(n x , n y , n z );

[0030] When the slice plane passes through the center point of the bolt and is parallel to the bolt direction vector, the equation of the slice plane is expressed as follows:

[0031] Ax+By+Cz+D=0,

[0032] Where D = -(n x x c +n y y c +n z z c );

[0033] Slice the point cloud data to obtain the point set P on the slice plane slice, use the distance formula from point to plane to filter the points belonging to the plane:

[0034]

[0035] Where, ò is the distance threshold, when D(P i )<ò, then click P i It is considered to belong to the slice plane;

[0036] The normalization function and complex summation function are used to represent the contour extraction process;

[0037] The normalization function φ(x) is used to standardize the coordinates of a point:

[0038]

[0039] Among them, μ x and σ x The point set P slice mean and standard deviation in the x direction;

[0040] Complex summation function Ψ(P slice ) is used to represent the extraction of contour lines:

[0041]

[0042] Among them, α is the adjustment parameter, M is the number of points on the slice plane;

[0043]

[0044] Where L is the three-dimensional space contour of the bolt node, M is the number of points on the slice plane, and x i is the x coordinate of the ith point, y i is the y coordinate of the i-th point, z i is the z coordinate of the ith point, μ x The point set P slice The mean in the x direction, σ x The point set P slice The standard deviation in the x direction, μ y The point set P slice The mean in the y direction, σ y The point set P slice The standard deviation in the y direction, μ z The point set P slice The mean value in the z direction, σ z The point set P slice The standard deviation in the z direction, (A, B, C) are the three components of the normal vector of the slice plane, D is the distance of the slice plane, ò is the distance threshold, and I(·) is the indicator function, which takes the value of 1 when the condition is met and 0 otherwise.

[0045] Furthermore, the normal vector of the three-dimensional contour line of the bolt node is calculated, and the coordinate plane perpendicular to the maximum plane normal vector of the point cloud data is selected. The sliced ​​three-dimensional space coordinate points are projected onto the selected two-dimensional coordinate plane. The process of converting them into two-dimensional plane coordinate points is as follows:

[0046] The point set P on the contour line of the bolt node is P 1 , P 2 ,…,P r , where each point P i =(x i ,y i , z i ), the normal vector n of the contour line passes through P i The eigenvector of the covariance matrix is ​​obtained according to the following formula:

[0047]

[0048] in, is the centroid of the point set P, and U is the i The covariance matrix of

[0049] Among the eigenvectors of the covariance matrix U, the eigenvector corresponding to the minimum eigenvalue is the normal vector n of the contour line;

[0050] When the maximum plane normal vector of the point cloud data is m, select a coordinate plane so that its normal vector is perpendicular to m, which is achieved by selecting an orthogonal basis of m;

[0051] Project the sliced ​​3D space coordinate points onto the selected 2D coordinate plane. For each point P on the contour line i =(x i ,y i , z i ), and project it onto the selected two-dimensional coordinate plane using the following formula:

[0052] p i' =(x i ,y i , 0).

[0053] Furthermore, on the two-dimensional plane projection image of the screw extension position, the point cloud data of the bolt node connection plate is automatically clustered to obtain the two-dimensional space contour line clusters of all bolts on the bolt node connection plate as follows:

[0054] The two-dimensional plane projection image data points of the bolt node connection plate point cloud at the screw extension position are automatically clustered to obtain the contour line clusters of all bolts on the bolt node connection plate. The K-Means algorithm is used to cluster the data points based on the mean. The calculation process is as follows:

[0055] Randomly select k data points from the two-dimensional data point set as the initial cluster center points;

[0056] Calculate the distances from the remaining data points to the k cluster centers, and assign the data points to the cluster corresponding to the cluster center that is closest to them;

[0057] According to the clustering results, calculate the mean of all data points in the cluster and update the cluster center to the mean;

[0058] Re-cluster all data points in the set according to the new cluster center point until the clustering result no longer changes.

[0059] Furthermore, the process of obtaining the screw extension length and converting the screw extension length into the bolt rotation angle is as follows:

[0060] The change in the length of the screw extension is converted to the bolt loosening angle value, that is, the extension length of the screw in the bolt before and after loosening is converted into the bolt rotation angle. The bolt loosening angle is quantified by calculating the difference in the bolt rotation angle before and after loosening. The formula is as follows:

[0061] Dl=l 0 -l t

[0062] Where Dl is the change in the screw extension length, l 0 is the extension length of the screw before loosening, l t It is the extended length of the screw after loosening;

[0063] The quantification of the screw extension length is divided into two parts. The first part is the quantification of the total length of the integer slices of the screw extension part, and the second part is the quantification of the remaining length after the screw extension part is sliced. The formula is as follows:

[0064] l=l 1 +l 2

[0065] l 1 =n'd

[0066] Where, l is the extension length of the screw, l 1 is the total length of the slice extending from the screw, l 2 is the remaining length of the extended portion of the screw after slicing, and n' is the maximum integer number of slices of the extended portion of the screw when the slice thickness is d;

[0067] The single bolt point cloud is sliced, and the RANSAC algorithm is used to extract the largest plane in the point cloud. The normal vector of the largest plane is calculated, and the single bolt point cloud is sliced ​​equidistantly along the normal vector of the largest plane in the direction from outside the plane to inside the plane. The slice thickness is d, and the slice spacing is 0. The detection range of the loose angle in the bolt group looseness detection based on two-dimensional images is generally 0°-60°. In the single bolt point cloud, the moving distance on the screw when the nut rotates 60° is taken as the slice thickness of the point cloud. The calculation formula is as follows:

[0068]

[0069] Among them, h is the ratio of the side length measured in the bolt point cloud to the actual side length of the bolt, and k is the thread pitch, that is, the distance the nut moves on the screw when it rotates 360°;

[0070] The quantification of the full-scale loosening angle of the bolt also consists of two parts. The first part is to convert the total length of the integer slices of the extended part of the screw into the bolt rotation angle. The second part is to convert the remaining length of the extended part of the screw into the bolt rotation angle after slicing. The calculation formula is as follows:

[0071]

[0072] q=q 1 +q 2

[0073]

[0074] Where q is the bolt rotation angle converted from the screw extension length, q 1 The total length of the integral slice of the screw extension converted into the bolt rotation angle, q 2 is the bolt rotation angle converted from the remaining length of the extended part of the screw after slicing, Dq is the loosening angle of the bolt, The bolt rotation angle converted to the extended length of the screw after loosening;

[0075] The calculation formula for the bolt rotation angle converted from the total length of the integer slices of the screw extension is as follows:

[0076] l 1 =n'60°.

[0077] Furthermore, the remaining length of the extended portion of the screw after slicing is obtained, and the process of converting the remaining length of the extended portion of the screw after slicing into the bolt rotation angle is as follows:

[0078] Traverse the bolt contour pixel points, extract the coordinate value of each pixel point, and convert it from the coordinate value expressed in the rectangular coordinate system to the coordinate value expressed in the polar coordinate system with the centroid of the bolt connection domain as the pole. The conversion formula is as follows:

[0079]

[0080] Where ρ represents the polar diameter, that is, the distance from the bolt contour pixel point to the centroid of the connected domain, and α represents the polar angle, which ranges from 0° to 360°;

[0081] The maximum points in the polar angle-polar diameter curve generated by the pixel points of the bolt contour are extracted, and the six points on the bolt contour that are farthest from the centroid of the bolt connected domain are obtained, namely the bolt corner points;

[0082] After the bolt corner point extraction is completed, the bolt rotation angle converted from the remaining length of the extended part of the screw after slicing is calculated as follows:

[0083]

[0084] Among them, q 2 It represents the angle between the line connecting the corner point of the i-th side of the bolt and the centroid of the bolt and the positive direction of the x-axis in the local coordinate system. rem(×) is a mathematical operator for calculating the remainder after the divisor. represents the angle between the extension line of the i-th side of the bolt in the image coordinate system and the positive direction of the x-axis. According to the regular hexagonal characteristics of the bolt, q 2 The value range is 0°-60°.

[0085] Furthermore, the bolt rotation angle converted from the total length of the integral slices of the extended portion of the screw is compared with the threshold value, and the bolt rotation angle converted from the remaining length of the extended portion of the screw after slicing is compared with the threshold value. The process of judging whether the bolt is loose is as follows:

[0086] Only when q 1 ≥YZ 1 ,q 2 ≥YZ 2 , the bolts become loose;

[0087] Only when q 1 <YZ 1 ,q 2 <YZ 2 , the bolts are not loose.

[0088] Among them, YZ 1 The threshold value of the bolt rotation angle converted from the total length of the integer slice of the screw extension, YZ 2 The threshold value of the bolt rotation angle converted to the remaining length of the screw after slicing.

[0089] Compared with the prior art, the present invention has the following beneficial effects:

[0090] The present invention uses a laser scanner to collect three-dimensional point cloud data of a bolt group, and can efficiently and accurately obtain three-dimensional information of the bolts and their surroundings. By performing plane processing and data segmentation on the point cloud data, the connecting plate and the plane part can be effectively identified and separated, thereby improving the accuracy and reliability of detection. In addition, by locating the position of the bolt on the connecting plate, slicing to obtain the three-dimensional spatial contour line of the bolt node, and projecting it onto a two-dimensional plane for automatic clustering processing, the total length of the integer slices of the protruding part of the bolt and the remaining length after slicing can be accurately extracted. The influence of light can also be avoided, providing more stable and high-precision detection results. By calculating and comparing the bolt rotation angle with the threshold, it can automatically determine whether the bolt is loose, which significantly improves the efficiency and accuracy of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0092] Figure 2 It is a point cloud diagram of the bolt node connection plate of the present invention;

[0093] Figure 3 It is a point cloud diagram of a single bolt of the present invention;

[0094] Figure 4 This is a schematic diagram of the change in the extension length of the screw rod of the present invention;

[0095] Figure 5 This is a schematic diagram of the quantification of the full-scale looseness of the bolt of the present invention;

[0096] Figure 6 It is a schematic diagram of a point cloud cross section of the present invention;

[0097] Figure 7 This is a bolt corner point extraction effect diagram of the present invention;

[0098] Figure 8 It is a schematic diagram of the loosening angle of the bolt of the present invention. DETAILED DESCRIPTION

[0099] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0100] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0101] Example:

[0102] See also Figures 1 to 8 , the present invention provides a technical solution:

[0103] A full-scale detection method for bolt group loosening based on three-dimensional point cloud, such as Figure 1 As shown, the specific steps include:

[0104] S1. Use a laser scanner to collect three-dimensional point cloud data of the bolt group, where the three-dimensional point cloud data includes point cloud data of the bolts, connecting plates, and surrounding environment;

[0105] S2. Perform plane processing on the point cloud data, identify the largest plane in the point cloud, extract the inner point set of the largest plane from the point cloud data, and segment the point cloud data into a plane part and a non-plane part through a data segmentation algorithm, wherein the plane part is the point cloud of the connecting plate, and the non-plane part is the point cloud of the bolt and the surrounding environment;

[0106] S3. By positioning the bolt on the connecting plate, determine the slicing plane at the position where the screw extends, slice the point cloud data on the slicing plane, and obtain the contour line of the bolt node in three dimensions;

[0107] S4. Calculate the normal vector of the three-dimensional space contour line of the bolt node, select a coordinate plane perpendicular to the maximum plane normal vector of the point cloud data, project the sliced ​​three-dimensional space coordinate points onto the selected two-dimensional coordinate plane, convert them into two-dimensional plane coordinate points, and automatically cluster the point cloud data of the bolt node connection plate on the two-dimensional plane projection image of the screw extension position to obtain the two-dimensional space contour line clusters of all bolts on the bolt node connection plate;

[0108] S5. extract the contour line cluster in the two-dimensional space after projection, obtain the screw extension length and the remaining length after the screw extension part is sliced, convert the screw extension length into the bolt rotation angle, convert the remaining length after the screw extension part is sliced ​​into the bolt rotation angle, compare the bolt rotation angle converted from the total length of the integer slice of the screw extension part with its threshold, and compare the bolt rotation angle converted from the remaining length after the screw extension part is sliced ​​with its threshold;

[0109] S6. Determine whether the bolt is loose based on the comparison result.

[0110] Based on the above embodiment, the principle and process of the laser scanner collecting point cloud data of the bolts, connecting plates and surrounding environment of the bolt group are as follows:

[0111] Common laser scanning technology is based on the principle of triangulation. The laser scanner emits a laser beam to the target object, and then captures the reflected laser point through a camera or other sensor. Based on the angle and distance relationship between the laser emission point, receiving point and reflection point, the three-dimensional coordinates of the object surface can be calculated.

[0112] Process: The laser scanner starts and begins to emit a laser beam. The emitted laser beam irradiates the surface of the object and is reflected back at different angles according to the material and shape of the surface. The scanner's receiver captures the reflected laser signals, which are usually light signals and need to be converted into electrical signals. In order to cover the entire target area, the mirror of the laser scanner will rotate or swing to change the scanning direction of the laser beam. At the same time, the scanner will record the angle at which the laser beam is emitted. By combining the distance (calculated by the flight time or phase difference of the laser signal) and the angle information, the scanner calculates the three-dimensional coordinates (X, Y, Z) of each reflection point. All calculated three-dimensional coordinate points are combined to form a point cloud data set, which accurately represents the surface information of all objects in the scanning area, including point cloud data of bolts, connecting plates and the surrounding environment.

[0113] On the basis of the above embodiment, the process of performing plane processing on the point cloud data, identifying the largest plane in the point cloud, and extracting the inner point set of the largest plane from the point cloud data is as follows:

[0114] The RANSAC algorithm is used to fit the point cloud data to a plane. First, the maximum number of iterations is set to N, the distance threshold is set to d, and three points are randomly selected from the point cloud data. 1 =(x 1 ,y 1 , z 1 ), P 2 =(x 2 ,y 2 , z 2 ), P 3=(x 3 , y 3 , z 3 );

[0115] Then the vector V 1 = P 2 - P 1 =(x 2 - x 1 , y 2 - y 1 , z 2 - z 1 ), the vector V 2 = P 3 - P 1 =(x 3 - x 1 , y 3 - y 1 , z 3 - z 1 ), the normal vector n = V 1 × V 2 . Set the plane equation Ax + By + Cz + D = 0, where (A, B, C) are the components of the plane normal vector, and D is the distance of the plane from the origin. A = n x , B = n y , C = n z , D = -(Ax 1 + By 1 + Cz 1 ).

[0116] For each point in the point cloud data, P i =(x i , y i , z i ), calculate its distance to this plane. The formula is as follows:

[0117]

[0118] If D(P i ) < d, then P i is considered an inlier.

[0119] Record the current number of inliers. If the current number of inliers is more than the previously recorded maximum number of inliers, then update the maximum plane and the maximum inlier set; after N iterations, select the plane with the most inliers as the finally fitted plane.

[0120] Based on the above embodiments, the process of dividing the point cloud data into a plane part and a non - plane part through a data segmentation algorithm is as follows: The plane part is the point cloud of the connecting plate, and the non - plane part is the point cloud of the bolt and the surrounding environment as follows:

[0121] The inner point set of the largest plane is extracted to form the point cloud data of the plane part. plane It is composed of the inner point set of the largest plane, that is, the point cloud data of the connecting plate; the plane part P is deleted from the original point cloud data set plane The remaining point cloud data is the non-planar part P non-plane , that is, point cloud data of the bolt and its surrounding environment.

[0122] The RANSAC algorithm is used to identify and fit the largest plane in the point cloud data, and the inner point set P of the plane is obtained. i ,

[0123] P i ∈P|distance(P i , plane) <d

[0124] Among them, distance(P i , plane) represents point P i The distance to the fitted plane, d is the distance threshold.

[0125] The non-planar point set is the remaining point set after deleting the planar point cloud data from the original point cloud data set. The specific steps to obtain it are as follows:

[0126] Initialize the non-planar point set and create an empty point set P non-plane ;

[0127] Traverse the original point cloud dataset, for each point P i ∈P, if Then P i Belong to the non-planar part, add the non-planar part P non-plane .

[0128] On the basis of the above embodiment, by locating the position of the bolt on the connecting plate, determining the slice plane of the screw extension position, slicing the point cloud data on the slice plane, and obtaining the contour line of the three-dimensional space of the bolt node, the process is as follows:

[0129] The center point of the bolt K=(x c ,y c , z c ), the bolt direction vector g=(n x , n y , n z );

[0130] Assuming that the slice plane passes through the center point of the bolt and is parallel to the bolt direction vector, the equation of the slice plane can be expressed as follows:

[0131] Ax+By+Cz+D=0,

[0132] Where D = -(n x x c +n y y c +n z z c ).

[0133] Slice the point cloud data to obtain the point set P on the slice plane slice , use the distance formula from point to plane to filter the points belonging to the plane:

[0134]

[0135] Where, ò is the distance threshold, if D(P i )<ò, then click P i Considered to belong to the slice plane.

[0136] In order to obtain the three-dimensional space contour of the bolt node, it is necessary to slice the point set P on the plane slice Further processing is performed, using normalization function and complex summation function to represent the contour extraction process.

[0137] The normalization function φ(x) is used to standardize the coordinates of a point:

[0138]

[0139] Among them, μ x and σ x The point set P slice The mean and standard deviation in the x direction, the complex sum function Ψ(P slice ) is used to represent the extraction of contour lines:

[0140]

[0141] Among them, α is the adjustment parameter and M is the number of points on the slice plane.

[0142]

[0143] Where L is the three-dimensional space contour of the bolt node, M is the number of points on the slice plane, and x i is the x coordinate of the ith point, y i is the y coordinate of the i-th point, z i is the z coordinate of the ith point, μ x The point set P slice The mean in the x direction, σ x The point set P slice The standard deviation in the x direction, μ y The point set P slice The mean in the y direction, σy The point set P slice The standard deviation in the y direction, μ z The point set P slice The mean value in the z direction, σ z The point set P slice The standard deviation in the z direction, (A, B, C) are the three components of the normal vector of the slice plane, D is the distance of the slice plane, ò is the distance threshold, and I(·) is the indicator function, which takes the value of 1 when the condition is met and 0 otherwise.

[0144] On the basis of the above embodiment, the normal vector of the contour line of the bolt node in three-dimensional space is calculated, a coordinate plane perpendicular to the maximum plane normal vector of the point cloud data is selected, and the sliced ​​three-dimensional space coordinate points are projected onto the selected two-dimensional coordinate plane. The process of converting the three-dimensional space coordinate points into two-dimensional plane coordinate points is as follows:

[0145] The point set P on the contour line of the bolt node is P 1 , P 2 ,…,P r , where each point P i =(x i ,y i , z i ), the normal vector n of the contour line passes through P i The eigenvector of the covariance matrix is ​​obtained according to the following formula:

[0146]

[0147] in, is the centroid of the point set P, and U is the i The covariance matrix of

[0148] Among the eigenvectors of the covariance matrix U, the eigenvector corresponding to the minimum eigenvalue is the normal vector n of the contour line;

[0149] Assuming that the maximum plane normal vector of the point cloud data is (m), select a coordinate plane whose normal vector is perpendicular to (m). This can be achieved by selecting an orthogonal basis of (m). For example, if m = (m x ,m y ,m z ), you can select the (xy) plane, whose normal vector is (0,0,1);

[0150] Project the sliced ​​3D space coordinate points onto the selected 2D coordinate plane. For each point P on the contour line i =(x i ,y i , z i ), and project it onto the selected two-dimensional coordinate plane using the following formula:

[0151] p i' =(x i ,y i ,0)

[0152] On the basis of the above embodiment, the point cloud data of the bolt node connection plate is automatically clustered on the two-dimensional plane projection image of the screw extension position to obtain the two-dimensional space contour line clusters of all bolts on the bolt node connection plate as follows:

[0153] Automatic clustering of the two-dimensional plane projection image data points of the bolt node connection plate point cloud at the screw extension position can obtain the contour line clusters of all bolts on the bolt node connection plate. The K-Means algorithm is used to cluster the data points based on the mean. The calculation process is as follows:

[0154] (1) Randomly select k data points from the two-dimensional data point set as the initial cluster center points;

[0155] (2) Calculate the distances from the remaining data points to the k cluster centers, and assign the data points to the cluster corresponding to the cluster center that is closest to them;

[0156] (3) According to the clustering results, calculate the mean of all data points in the cluster and update the cluster center to the mean;

[0157] (4) Repeat steps (2) and (3) to re-cluster all data points in the set according to the new cluster center points until the clustering results no longer change.

[0158] like Figure 2 and Figure 3 As shown, all the bolt contour line clusters on the point cloud slice of the bolt node connection plate point cloud model at the screw extension position are finally obtained. The number of clusters is the number of bolts on the bolt node connection plate. The two-dimensional centroid point coordinates of all the bolt contour line clusters on the bolt node connection plate are extracted, and the bolt node connection plate point cloud is cropped according to the bolt centroid point coordinates to obtain a single bolt point cloud model, thereby realizing the unified extraction of group-distributed bolts.

[0159] On the basis of the above embodiment, the contour line cluster of the two-dimensional space after projection is extracted, the screw extension length and the remaining length after the screw extension part is sliced ​​are obtained, the screw extension length is converted into the bolt rotation angle, and the process of converting the remaining length after the screw extension part is sliced ​​into the bolt rotation angle is as follows:

[0160] like Figure 4 As shown, the change in the length of the screw extension is converted to obtain the bolt loosening angle value, which means that the extension length of the screw in the bolt before and after loosening is converted into the bolt rotation angle. By calculating the difference in the bolt rotation angle before and after loosening, the bolt loosening angle is quantified, based on the following formula:

[0161] Dl=l 0 -l t

[0162] Where Dl is the change in screw extension length, l 0 is the extension length of the screw before loosening, l t It is the extended length of the screw after loosening.

[0163] The quantification of the screw extension length is divided into two parts. The first part is the quantification of the total length of the integer slices of the screw extension part, and the second part is the quantification of the remaining length after the screw extension part is sliced. The formula is as follows:

[0164] l=l 1 +l 2

[0165] l 1 =n'd

[0166] Where, l is the screw extension length, l 1 is the total length of the slice extending from the screw, l 2 The remaining length of the extended part of the screw after slicing (l 2 <d), n' is the maximum integer number of slices of the extended portion of the screw when the slice thickness is d.

[0167] The single bolt point cloud is sliced, and the RANSAC algorithm is used to extract the largest plane in the point cloud. The normal vector of the largest plane is calculated, and the single bolt point cloud is sliced ​​equidistantly along the normal vector of the largest plane in the direction from outside the plane to inside the plane. The slice thickness is d, and the slice spacing is 0. Considering that the detection range of the loose angle in the current bolt group looseness detection based on two-dimensional images is generally 0°-60°, the movement distance of the nut on the screw when the nut rotates 60° in the single bolt point cloud is taken as the slice thickness of the point cloud. The calculation formula is as follows:

[0168]

[0169] Where h is the ratio of the side length measured in the bolt point cloud to the actual side length of the bolt, and k is the thread pitch, that is, the distance the nut moves on the screw when it rotates 360°.

[0170] like Figure 5 As shown in the figure, from the above analysis of the quantification of the screw extension length, it can be seen that the quantification of the full-scale loose angle of the bolt also consists of two parts. The first part is to convert the total length of the integer slices of the screw extension part into the bolt rotation angle, and the second part is to convert the remaining length of the screw extension part after slicing into the bolt rotation angle. The calculation formula is as follows:

[0171]

[0172] q=q 1 +q 2

[0173]

[0174] Where q is the bolt rotation angle converted from the screw extension length, q 1 The total length of the integral slice of the screw extension converted into the bolt rotation angle, q 2 is the bolt rotation angle converted from the remaining length of the extended part of the screw after slicing, Dq is the loosening angle of the bolt, The bolt rotation angle converted from the screw extension length after loosening.

[0175] The quantization process of the total length of the integer slices of the screw extension part is as follows:

[0176] The schematic diagram of the point cloud section obtained by equidistant slicing of a single bolt point cloud with a slice thickness of d and a slice spacing of 0 is shown in the figure below. Figure 6 As shown in the figure, the K-Means algorithm is used to cluster the point cloud slices to obtain the contour line clusters of the bolts at each section. According to the change in the area of ​​the bolt contour line clusters, the total length of the integer slices of the extended part of the screw is quantified and converted into the bolt rotation angle. The specific steps are as follows:

[0177] (1) Calculate the area of ​​all bolt contour clusters. The area of ​​the bolt contour cluster represents the cross-sectional area of ​​the bolt at that section.

[0178] (2) Sort the bolt cross-sectional areas according to the order of bolt point cloud slicing, and obtain the maximum integer slice number of the extended part of the screw when the point cloud slice thickness is d according to the change of the bolt cross-sectional area;

[0179] (3) Multiply the calculated maximum number of integer slices of the screw extension part by the point cloud slice thickness, and the product is the total length of the integer slices of the screw extension part.

[0180] (4) From the above analysis, it can be seen that the point cloud slice thickness d is the moving distance of the nut on the screw when the nut rotates 60° in the bolt point cloud. The maximum integer slice number of the screw extension part is multiplied by 60° to quantify the bolt rotation angle. Therefore, the calculation formula for the bolt rotation angle converted from the total length of the integer slices of the screw extension part is as follows:

[0181] l 1 =n'60°

[0182] The quantification process of the remaining length of the screw extension after slicing is as follows:

[0183] The point cloud slices of the remaining length of the bolt screw extension part after slicing are extracted, and the K-Means algorithm is used to cluster them to obtain the contour cluster of the bolt at the section. All data points on the bolt contour cluster are traversed, and the coordinate value of each data point is extracted. The coordinate value is converted from the coordinate value represented by the rectangular coordinate system to the coordinate value represented by the polar coordinate system with the centroid of the bolt connected domain as the pole. The polar coordinate values ​​of the data points on the bolt contour cluster are summarized, and the polar angle-polar diameter curve is fitted to extract the curve maximum point to complete the bolt corner point extraction.

[0184] Traverse the bolt contour pixel points, extract the coordinate value of each pixel point, and convert it from the coordinate value expressed in the rectangular coordinate system to the coordinate value expressed in the polar coordinate system with the centroid of the bolt connection domain as the pole. The conversion formula is as follows:

[0185]

[0186] α=arctan(y / x)

[0187] Where ρ represents the polar diameter, that is, the distance from the bolt contour pixel to the centroid of the connected domain, and α represents the polar angle, which ranges from 0° to 360°.

[0188] The polar coordinate values ​​of the bolt contour pixel points are summarized and a high-order polynomial fitting is performed on them. Since there are many pixels in the bolt contour, in order to fit the curve more accurately and avoid the curve being connected end to end, the polar coordinate values ​​are arranged from small to large according to the α value, and then divided into two parts for high-order polynomial fitting. The fitting results are then spliced ​​to obtain the polar angle-polar diameter curve of the bolt contour pixel points.

[0189] like Figure 7 As shown, the maximum points in the polar angle-polar diameter curve generated by the pixel points of the bolt contour are extracted, and the six points on the bolt contour that are farthest from the centroid of the bolt connected domain are obtained, namely the bolt corner points.

[0190] like Figure 8 As shown in the figure, after the bolt corner point extraction is completed, the bolt rotation angle converted from the remaining length of the extended part of the screw after slicing can be calculated as follows:

[0191]

[0192] In the formula, q 2 It represents the angle between the line connecting the corner point of the i-th side of the bolt and the centroid of the bolt and the positive direction of the x-axis in the local coordinate system. rem(×) is a mathematical operator for calculating the remainder after the divisor. represents the angle between the extension line of the i-th side of the bolt in the image coordinate system and the positive direction of the x-axis. According to the regular hexagonal characteristics of the bolt, q 2 The value range is 0°-60°.

[0193] On the basis of the above embodiment, the bolt rotation angle converted from the total length of the integral slices of the extended part of the screw is compared with the threshold value, and the bolt rotation angle converted from the remaining length of the extended part of the screw after slicing is compared with the threshold value, and the process of judging whether the bolt is loose is as follows:

[0194] Only when q 1 ≥YZ 1 ,q 2 ≥YZ 2 , the bolts become loose;

[0195] Only when q 1 <YZ 1 ,q 2 <YZ 2 , the bolts are not loose.

[0196] Among them, YZ 1 The threshold value of the bolt rotation angle converted from the total length of the integer slice of the screw extension, YZ 2 The threshold value of the bolt rotation angle converted to the remaining length of the screw after slicing.

[0197] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0198] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0199] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0200] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A full-scale detection method for bolt group loosening based on three-dimensional point cloud, characterized in that: The specific steps include: S1. Use a laser scanner to collect three-dimensional point cloud data of the bolt group, where the three-dimensional point cloud data includes point cloud data of the bolts, connecting plates, and surrounding environment; S2. Perform plane processing on the point cloud data, identify the largest plane in the point cloud, extract the inner point set of the largest plane from the point cloud data, and segment the point cloud data into a plane part and a non-plane part through a data segmentation algorithm, wherein the plane part is the point cloud of the connecting plate, and the non-plane part is the point cloud of the bolt and the surrounding environment; S3. By positioning the bolt on the connecting plate, determine the slicing plane at the position where the screw extends, slice the point cloud data on the slicing plane, and obtain the contour line of the bolt node in three dimensions; S4. Calculate the normal vector of the three-dimensional space contour line of the bolt node, select a coordinate plane perpendicular to the maximum plane normal vector of the point cloud data, project the sliced ​​three-dimensional space coordinate points onto the selected two-dimensional coordinate plane, convert them into two-dimensional plane coordinate points, and automatically cluster the point cloud data of the bolt node connection plate on the two-dimensional plane projection image of the screw extension position to obtain the two-dimensional space contour line clusters of all bolts on the bolt node connection plate; S5. extract the contour line cluster in the two-dimensional space after projection, obtain the screw extension length and the remaining length after the screw extension part is sliced, convert the screw extension length into the bolt rotation angle, convert the remaining length after the screw extension part is sliced ​​into the bolt rotation angle, compare the bolt rotation angle converted from the total length of the integer slice of the screw extension part with its threshold, and compare the bolt rotation angle converted from the remaining length after the screw extension part is sliced ​​with its threshold; S6. Determine whether the bolt is loose according to the comparison result; The process of obtaining the screw extension length and converting the screw extension length into the bolt rotation angle is as follows: The change in the length of the screw extension is converted to the bolt loosening angle value, that is, the extension length of the screw in the bolt before and after loosening is converted into the bolt rotation angle. The bolt loosening angle is quantified by calculating the difference in the bolt rotation angle before and after loosening. The formula is as follows: Dl＝l0-l t Where Dl is the change in the extension length of the screw, l0 is the extension length of the screw before loosening, and l t It is the extended length of the screw after loosening; The quantification of the screw extension length is divided into two parts. The first part is the quantification of the total length of the integer slices of the screw extension part, and the second part is the quantification of the remaining length after the screw extension part is sliced. The formula is as follows: l＝l1+l2 l1=n'd Wherein, l is the screw extension length, l1 is the total length of the slices of the screw extension part, l2 is the remaining length of the screw extension part after slicing, and n' is the maximum integer number of slices of the screw extension part when the slice thickness is d; The single bolt point cloud is sliced, and the RANSAC algorithm is used to extract the largest plane in the point cloud. The normal vector of the largest plane is calculated, and the single bolt point cloud is sliced ​​equidistantly along the normal vector of the largest plane in the direction from outside the plane to inside the plane. The slice thickness is d, and the slice spacing is 0. The detection range of the loose angle in the bolt group looseness detection based on two-dimensional images is 0°-60°. In the single bolt point cloud, the moving distance on the screw when the nut rotates 60° is taken as the slice thickness of the point cloud. The calculation formula is as follows: Among them, h is the ratio of the measured side length in the bolt point cloud to the actual side length of the bolt, and k is the thread pitch, that is, when the nut rotates 360°, its moving distance on the screw; The quantification of the full-scale loosening angle of the bolt also consists of two parts. The first part is to convert the total length of the integer slices of the protruding part of the screw into the rotation angle of the bolt, and the second part is to convert the remaining length after slicing the protruding part of the screw into the rotation angle of the bolt. The calculation formula is as follows: q = q1 + q2 Among them, q is the bolt rotation angle converted from the screw extension length, q1 is the bolt rotation angle converted from the total length of the integer slice of the screw extension part, q2 is the bolt rotation angle converted from the remaining length of the screw extension part after slicing, and Dq is the loosening angle of the bolt. The bolt rotation angle converted to the extended length of the screw after loosening; The calculation formula for converting the total length of the integer slices of the protruding part of the screw into the rotation angle of the bolt is as follows: l1＝n’60° Obtain the remaining length after slicing the protruding part of the screw. The process of converting the remaining length after slicing the protruding part of the screw into the rotation angle of the bolt is as follows: Traverse the pixel points of the bolt contour, extract the coordinate values of each pixel point, and convert them from the coordinate values represented in the rectangular coordinate system to the coordinate values represented in the polar coordinate system with the centroid of the bolt connected domain as the pole. The conversion formula is as follows: In the formula, ρ represents the polar radius, that is, the distance from the pixel point of the bolt contour to the centroid of the connected domain, and α represents the polar angle, and the angle range is 0° - 360°; Extract the maximum value points in the polar angle - polar radius curve graph generated by the pixel points of the bolt contour, and obtain six points on the bolt contour that are farthest from the centroid of the bolt connected domain, that is, the bolt corner points; After completing the extraction of the bolt corner points, calculate the rotation angle of the bolt converted from the remaining length after slicing the protruding part of the screw according to the following formula: Where q2 represents the angle between the line connecting the corner point corresponding to the i-th side of the bolt and the centroid of the bolt in the local coordinate system and the positive direction of the x-axis, and rem(×) is a mathematical operator for calculating the remainder after the divisor. It represents the angle between the extension line of the i-th side of the bolt in the image coordinate system and the positive direction of the x-axis. According to the regular hexagonal characteristics of the bolt, the value range of q2 is 0°-60°; Compare the rotation angle of the bolt converted from the total length of the integer slices of the protruding part of the screw with its threshold, and compare the rotation angle of the bolt converted from the remaining length after slicing the protruding part of the screw with its threshold. The process of judging whether the bolt is loose is as follows: Only when q1 ≥ YZ1 and q2 ≥ YZ2, the bolt is loose; Only when q1 < YZ1 and q2 < YZ2, the bolt is not loose; Among them, q1 is the rotation angle of the bolt converted from the total length of the integer slices of the protruding part of the screw, q2 is the rotation angle of the bolt converted from the remaining length after slicing the protruding part of the screw, YZ1 is the threshold of the rotation angle of the bolt converted from the total length of the integer slices of the protruding part of the screw, and YZ2 is the threshold of the rotation angle of the bolt converted from the remaining length after slicing the protruding part of the screw.

2. The method for full-scale detection of bolt group loosening based on three-dimensional point cloud according to claim 1 is characterized in that: Perform plane processing on the point cloud data, identify the largest plane in the point cloud, and the process of extracting the inlier set of the largest plane from the point cloud data is as follows: The point cloud data is fitted with a plane using the RANSAC algorithm. The maximum number of iterations is set to N, the distance threshold is set to d, and three points are randomly selected from the point cloud data: P1 = (x1, y1, z1), P2 = (x2, y2, z2), and P3 = (x3, y3, z3). Then, vector V1 = P2-P1 = (x2-x1, y2-y1, z2-z1), vector V2 = P3-P1 = (x3-x1, y3-y1, z3-z1), normal vector n = V1×V2, and the plane equation is set to Ax+By+Cz+D=0, where (A, B, C) are the components of the plane normal vector, D is the distance of the plane from the origin, and A = n x , B=n y , C = n z , D = -(Ax1 + By1 + Cz1); For each point in the point cloud data, P i =(x i ,y i , z i ), calculate its distance to the plane according to the following formula: When D(P i ) < d, then P i is considered an interior point; Record the current number of inliers. If the current number of inliers is more than the previously recorded maximum number of inliers, update the largest plane and the largest inlier set. After N iterations, select the plane with the most inliers as the finally fitted plane.

3. The full-scale detection method for bolt group loosening based on three-dimensional point cloud according to claim 1 is characterized in that: The process of dividing the point cloud data into a plane part and a non-plane part through a data segmentation algorithm, where the plane part is the point cloud of the connecting plate and the non-plane part is the point cloud of the bolt and the surrounding environment is as follows: The inner point set of the largest plane is extracted to form the point cloud data of the plane part. plane The inner point set of the largest plane is composed of the point cloud data of the connecting plate, and the plane part P is deleted from the original point cloud data set. plane The remaining point cloud data is the non-planar part P non-plane , i.e. point cloud data of the bolt and its surroundings; The RANSAC algorithm is used to identify and fit the largest plane in the point cloud data, and the inner point set P of the plane is obtained. i , P i ∈P|distance(P i , plane) <d Among them, distance(P i , plane) represents point P i The distance to the fitting plane, d is the distance threshold; The non-plane part point set is the remaining point set after deleting the point cloud data of the plane part from the original point cloud data set. The specific steps for obtaining it are as follows: Initialize the non-planar point set and create an empty point set P non-plane ; Traverse the original point cloud dataset, for each point P i ∈P, when Then P i Belong to the non-planar part, add the non-planar part P non-plane .

4. The full-scale detection method for bolt group loosening based on three-dimensional point cloud according to claim 1 is characterized in that: By positioning the position of the bolt on the connecting plate, determine the slicing plane of the protruding position of the screw, and slice the point cloud data on the slicing plane to obtain the contour line of the three-dimensional space of the bolt node. The process is as follows: The center point of the bolt K=(x c ,y c , z c ), the bolt direction vector g=(n x , n y , n z ); When the slice plane passes through the center point of the bolt and is parallel to the bolt direction vector, the equation of the slice plane is expressed as follows: Ax+By+Cz+D=0, where D = -(n x x c + n y y c + n z z c )); Slice the point cloud data to obtain the point set P on the slice plane slice , use the distance formula from point to plane to filter the points belonging to the plane: Where, ò is the distance threshold, when D(P i )<ò, then click P i It is considered to belong to the slice plane; The normalization function and complex summation function are used to represent the contour extraction process; The normalization function φ(x) is used to standardize the coordinates of a point: Among them, μ x and σ x The point set P slice mean and standard deviation in the x direction; Complex sum function Ψ(P slice ) is used to represent the extraction of contour lines: Among them, α is the adjustment parameter, M is the number of points on the slice plane; Where L is the three-dimensional space contour of the bolt node, M is the number of points on the slice plane, and x i is the x coordinate of the ith point, y i is the y coordinate of the i-th point, z i is the z coordinate of the ith point, μ x The point set P slice The mean in the x direction, σ x The point set P slice The standard deviation in the x direction, μ y The point set P slice The mean in the y direction, σ y The point set P slice The standard deviation in the y direction, μ z The point set P slice The mean value in the z direction, σ z The point set P slice The standard deviation in the z direction, (A, B, C) are the three components of the normal vector of the slice plane, D is the distance of the slice plane, ò is the distance threshold, and I(·) is the indicator function, which takes the value of 1 when the condition is met and 0 otherwise.

5. The method for full-scale detection of bolt group loosening based on three-dimensional point cloud according to claim 1 is characterized in that: Calculate the normal vector of the 3D contour line of the bolt node, select the coordinate plane perpendicular to the maximum plane normal vector of the point cloud data, project the sliced ​​3D space coordinate points onto the selected 2D coordinate plane, and convert them into 2D plane coordinate points as follows: The point set P on the contour line of the bolt node is P1, P2, ..., P r , where each point P i =(x i ,y i , z i ), the normal vector n of the contour line passes through P i The eigenvector of the covariance matrix is ​​obtained according to the following formula: in, is the centroid of the point set P, and U is the i The covariance matrix of Among the eigenvectors of the covariance matrix U, the eigenvector corresponding to the minimum eigenvalue is the normal vector n of the contour line; When the maximum plane normal vector of the point cloud data is m, select a coordinate plane so that its normal vector is perpendicular to m, which is achieved by selecting an orthogonal basis of m; Project the sliced ​​3D space coordinate points onto the selected 2D coordinate plane. For each point P on the contour line i =(x i ,y i , z i ), and project it onto the selected two-dimensional coordinate plane using the following formula: p i' =(x i ,y i ,0)。 6. The method for full-scale detection of bolt group loosening based on three-dimensional point cloud according to claim 1 is characterized in that: On the two-dimensional plane projection image of the screw extension position, the point cloud data of the bolt node connection plate is automatically clustered to obtain the two-dimensional space contour line clusters of all bolts on the bolt node connection plate as follows: The two-dimensional plane projection image data points of the bolt node connection plate point cloud at the screw extension position are automatically clustered to obtain the contour line clusters of all bolts on the bolt node connection plate. The K-Means algorithm is used to cluster the data points based on the mean. The calculation process is as follows: Randomly select k data points from the two-dimensional data point set as the initial cluster center points; Calculate the distances from the remaining data points to the k cluster centers, and assign the data points to the cluster corresponding to the nearest cluster center. According to the clustering results, calculate the mean of all data points in the cluster and update the cluster center to the mean; Re-cluster all data points in the set according to the new cluster center point until the clustering result no longer changes.

Images