Pole tower local deformation detection method based on different-period laser point clouds

Abstract

A tower local deformation detection method based on laser point clouds in different periods comprises the following steps: carrying out scanning detection on an overhead transmission line by adopting a three-dimensional laser scanner, and dividing tower point clouds collected in different periods into reference point cloud and comparison point cloud sequences 1, 2,... n according to a time sequence; slicing and layering the tower point cloud, and automatically identifying a first cross partition of the tower according to a tower point cloud structure judgment criterion; registering the point cloud sequences 1, 2,... n to the reference point cloud by using a Super-4PCS coarse registration algorithm and a local ICP fine registration algorithm by taking a first transverse partition surface of the reference point cloud as a reference object; and extracting the displacement of the point cloud sequences 1, 2,... n relative to the key node of the reference point cloud, and comparing the displacement with a key node structure abnormality judgment threshold to evaluate whether the tower structure is abnormal or not. According to the invention, structural abnormalities such as tower inclination, main material bending, foundation settlement and cross arm deformation caused by icing galloping, extreme strong wind and extremely severe working conditions of geological disasters can be effectively detected.

Description

Technical Field

[0001] The invention relates to the field of transmission line tower structure anomaly detection, and in particular to a tower local deformation detection method based on laser point clouds of different periods. Background Art

[0002] Since the transmission lines stretch for hundreds to thousands of kilometers, are distributed over a wide area, and have a harsh external environment, they are very susceptible to extreme natural disasters. Under the influence of harsh working conditions such as ice dancing, extreme strong winds, and geological disasters, the transmission line towers will be affected by factors such as load imbalance, over-design load, and foundation displacement and settlement, which will cause stress yielding such as deformation of the main materials of the towers and bending of auxiliary materials, as well as foundation sliding and settlement, which will cause the towers to tilt and the crossarms to tilt, threatening the safe operation of the lines. Since different parts of the towers are subjected to different forces and have different strengths, the degree of deformation of different parts is different. In particular, the crossarms and ground wire brackets are prone to local bending under heavy loads such as heavy ice, dancing, and ice shedding and jumping. If the load continues to expand, it may cause accidents such as damage to the tower head and tower collapse.

[0003] The invention patent publication number CN115810012B discloses a method, device, equipment and storage medium for detecting the inclination of a transmission tower. It includes: obtaining multiple initial point cloud data for the space where the transmission tower to be detected is located; solving the geometric features of each initial point cloud data to obtain the geometric feature information corresponding to each initial point cloud data, and based on the geometric feature information, obtaining the target point cloud data corresponding to each initial point cloud data; from the target point cloud data, obtaining the transmission tower point cloud data corresponding to the transmission tower, and the ground point cloud data corresponding to the ground where the transmission tower is located; based on the transmission tower point cloud data and the ground point cloud data, obtaining the corresponding inclination detection result of the transmission tower. This method is suitable for extracting the tower point cloud from the transmission line point cloud, and then calculating the inclination of the tower. However, for this kind of non-uniform deformation, the single state quantity of the tower inclination cannot effectively reflect the local yield deformation, and there is a possibility of missed judgment.

[0004] The invention patent publication number CN115880276A discloses a tower operation status assessment method based on multi-period point cloud comparison. A three-dimensional cone is established based on the tower point cloud to obtain the inclination angle. Based on the inclination angle and the inclination degrees of multiple towers, the tower operation status is obtained. The inclination degree of the tower is judged by extracting information from the front and side using a two-dimensional convolution kernel and from the top using a three-dimensional convolution kernel, respectively. This not only enables accurate tower shape information to be obtained from the front and side, but also enables the overall judgment of the inclination degree from the top without cutting angle problems. This method can detect and evaluate the inclination of the entire tower, but it is difficult to reflect local displacement deformations such as crossarm hanging points and ground wire brackets.

[0005] In summary, the existing methods mainly use three-dimensional laser point clouds to identify the inclination of the entire tower, and evaluate the operating status of the tower based on this. However, since the load of the tower is mainly concentrated on the ground wire hanging point and the foundation, these key nodes are also most likely to undergo local deformation under overload, and single state quantities such as the inclination of the entire tower and the skewness of the crossarm cannot effectively reflect this local deformation. In order to reflect the local deformation of the tower, the present invention analyzes multi-period point cloud data, extracts the displacement of the key nodes of the tower, and then measures the local deformation of the tower. In addition, due to the transmission and amplification of slight displacements and deformations such as foundation displacement and main material deformation through the tower body, a smaller displacement will cause obvious displacement of key nodes such as the ground wire hanging point, which is easier to detect. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention provides a method for detecting local deformation of a tower based on laser point clouds at different stages, which can be used to detect structural abnormalities of the tower, such as foundation displacement and settlement, main material bending and deformation, and stress yield of crossarms and ground wire supports, caused by external loads on the transmission tower.

[0007] A method for detecting local deformation of a tower based on laser point clouds of different periods comprises the following steps:

[0008] Step 1: Use a 3D laser scanner to scan and detect the overhead transmission line, and divide the tower point clouds collected at different times into reference point clouds and comparison point cloud sequences 1, 2, ... n according to the time series;

[0009] Step 2, slicing and layering the tower point cloud, and automatically identifying the first transverse plane of the tower according to the tower point cloud structure judgment criterion;

[0010] Step 3, taking the first transverse plane of the reference point cloud as the reference object, using the Super-4PCS coarse registration and local ICP fine registration algorithms, the point cloud sequence 1, 2, ...n is registered to the reference point cloud;

[0011] Step 4: extract the displacement of the key nodes of the point cloud sequence 1, 2, ... n relative to the reference point cloud, and compare it with the structural anomaly judgment threshold of the key nodes to evaluate whether the tower structure is abnormal.

[0012] Furthermore, step 1 includes:

[0013] The transmission line is scanned by 3D laser radar, and then solved, filtered and thinned. The point cloud classification technology is used to extract the tower point cloud, and the ground wire, ground and tree point clouds are eliminated. According to the time sequence of point cloud acquisition, the first phase of the tower point cloud is set as the reference point cloud, denoted as S 0 , set the subsequent point clouds as the comparison sequence point clouds in sequence, that is, sequence 1 point cloud, sequence 2 point cloud...sequence n point cloud, denoted as S1 , S 2 …S n .

[0014] Furthermore, step 2 includes:

[0015] Step 2.1, slice the tower point cloud: use a set of planes with a thickness of δ to intersect the point cloud, slice the point cloud from top to bottom, the thickness of the ground scanned point cloud slice δ is 50cm, the thickness of the machine scanned point cloud slice δ is 80cm, and the tower point cloud is cut from top to bottom in the vertical z-axis direction to obtain the tower horizontal point cloud slices S in sequence. i :

[0016]

[0017] Project each layer of point cloud slices onto the horizontal plane as shown below:

[0018]

[0019] Step 2.2, extract the circumscribed polygon boundary: Use the "bidirectional closest point search method" to extract the circumscribed polygon boundary of each layer of slices, and project the set of point cloud slices In S′, take any point Ps as the starting point of the boundary line, find the point Pr closest to Ps, and calculate the distance ds from Pr to Ps; i In the remaining point cloud, find the nearest point Pe of Ps and calculate the distance de from Ps to Pe; if ds≤de, insert Pr in front of Ps as a new starting point, connect PrPsPe in pairs, and replace Pr with Ps as the starting point; if ds≥de, grow to the other end Pe; then determine the set S′ i Are there any remaining points? If so, repeat the above steps; finally, traverse the set S′ i After all points are found, until PrPe overlaps, PsPe is connected to close the polygon and obtain the polygon boundary of the point cloud outer contour;

[0020] Step 2.3, calculate the vertices of the circumscribed polygon: through the linear regression equation, fit the outermost boundary point cloud into four boundary lines, which are recorded as AB, BC, CD, and DA in clockwise order; the intersection of the four straight lines is the four endpoints of the tower transverse diaphragm, and the coordinates of the endpoints are recorded as A(x A ,y A ,z A ), B(x B ,y B ,z B ), C(x C ,y C ,z C ), D(x D ,y D,z D );

[0021] Step 2.4, calculate the aspect ratio, area ratio parameters and number of points of the slice circumscribed polygon: calculate the side length l of each layer of point cloud slice respectively ix ,l iy , area i 、Number of points n i , and then calculate the area of ​​the circumscribed rectangle and aspect ratio Where l ix ,l iy Project S for the i-th slice i 'The length and width of the circumscribed rectangle;

[0022] Step 2.5, identify the first transverse plane of the tower according to the number of slice points and the aspect ratio and area ratio of the outer polygon. The identification criteria are: when slicing from top to bottom, the slice layer where the last peak of the number of slice points is located, and the number of slice points n i Should be greater than 500, and the height of the diaphragm should be less than the tower height The height from the ground is generally not more than 15m; the length-to-width ratio of the circumscribed rectangle of the section meets the following requirements: Slice area ratio

[0023] Furthermore, the Super-4PCS coarse registration algorithm in step 3 includes:

[0024] Step 3.1.1, in the reference point cloud S 0 Select 4 coplanar and non-colinear points, which are the common points of the multi-period point cloud, and the distance between them is as large as possible, denoted as B = {P a , P b , P c , P d}, calculate the distance and intersection ratio of straight lines AB and CD according to the following formula:

[0025] d1=||P a -P b ||,d2=||P c -P d || (3)

[0026]

[0027] Step 3.1.2, according to the affine invariance principle, in sequence 1 point cloud S 1 Find the point set with the same distance and intersection ratio with the straight lines AB and CD, as follows:

[0028] In sequence 1 point cloud S 1 Determine the point pair set S1, S2, and take each point qi As the center of the sphere, draw spheres with R = d1 ± ε and R = d2 ± ε as the radius respectively. The set S1 is q1 and the points distributed in the range [d1-ε, d1+ε], and q1 and the points distributed in the range [d2-ε, d2+ε] are S2. At the same time, the point cloud surface is rasterized to establish a three-dimensional grid G ​​with a unit size of ε.

[0029] Step 3.1.3, in sequence 1 point cloud S 1 Extract the reference point cloud S 0 Corresponding four point sets: traverse all candidate point pairs in point pair sets S1 and S2, calculate all intersection nodes e according to the cross ratio consistency, and store them in the grid G, based on the reference point cloud S 0 The angle between the two corresponding point pairs is θ, and the nodes extracted from the grid G ​​are approximately equal. At the same time, the angle between the two point pairs is approximately equal to θ, and the four-point basis matching is performed within the error range of ζ;

[0030] Step 3.1.4, in sequence 1 point cloud S 1 All the points that meet the conditions and the reference point cloud S 0 The corresponding four-point set U = {U1, U2, ..., U n}, find the reference point cloud S 0 With every U i The transformation matrix between them is compared to maximize the common point set LCP and select the transformation matrix with the highest registration accuracy for global transformation.

[0031] Furthermore, the local ICP fine registration algorithm in step 3 includes: for the point cloud S 0 and S 1 , from the reference point cloud S 0 Select the reference point p i , from the point cloud S 1 Find the distance point p in i The point q with the shortest Euclidean distance i , with p i and q i The corresponding point set formed calculates the transformation matrix, continuously iterates the process, uses Formula 5 as the target error function, stops iterating when the value of f is less than the threshold, and obtains the optimal transformation matrix.

[0032]

[0033] Where R represents the rotation transformation matrix, T represents the translation transformation matrix, and k represents the number of corresponding points.

[0034] Furthermore, step 4 includes:

[0035] Step 4.1, point cloud cutting: respectively cut the point cloud S 0 and S1 The principal component analysis method is used to obtain the main direction feature vector of the two-dimensional projection point cloud. On this basis, the main direction rotation angle is calculated, and the tower point cloud is rotated to obtain the redirected point cloud data. The direction perpendicular to the main direction after rotation on the horizontal plane is selected as the horizontal direction, and the point cloud is cut along the horizontal direction. The cut point cloud is divided into two parts, left and right.

[0036] Step 4.2, scattered point cloud indexing: for the reference point cloud S 0 And sequence 1 point cloud S 1 Use Octree method to build scattered point cloud index;

[0037] Step 4.3, Euclidean clustering segmentation: For the reference point cloud S with established Octree index 0 And sequence 1 point cloud S 1 The insulators in the cluster are clustered in Euclidean order to distinguish the phases of the insulators. The point cloud of each tower is set as P, which consists of n points. The Euclidean distance between the i-th sample and the j-th sample is calculated. The index number of the point with the smallest distance is classified into one class. The Euclidean distance between the new classes is calculated and iterated until the specified threshold is less than the Euclidean distance between any two classes or the number of classes is less than the specified number. The Euclidean clustering segmentation is completed.

[0038] Step 4.4, key node position extraction: extract the point cloud S of each period after clustering 0 and S 1 The three-dimensional coordinates of the hanging points on the crossarms of each phase insulator and the hanging points on the left and right phase ground wires are calculated to obtain the point cloud S 0 and S 1 The displacement of the key nodes at the same position on the x-axis, y-axis and z-axis;

[0039] Step 4.5, structural anomaly determination: compare the displacement of each key node with the key node structural anomaly determination threshold to determine whether the multi-phase point cloud key nodes are displaced, and then determine whether the tower structure is deformed.

[0040] Furthermore, the specific implementation process of step 4.3 is as follows:

[0041] Step 4.3.1, establish an Octree data structure for each tower point cloud P;

[0042] Step 4.3.2, establish an empty cluster set C and a queue;

[0043] Step 4.3.3, for each point P i ∈P, perform the following steps:

[0044] (1) Add the point to the current queue Q;

[0045] (2) For each point P i ∈Q, perform the following operations: ① Use the search radius r<d th , for P i Perform a k-nearest neighbor search and the points found are ②For each calculate With P i The Euclidean distance of each P is the same as the other two. i Check whether the point performs the above operations. If not, add the point to the queue Q;

[0046] (3) When all points in the list of queue Q have performed the above operations, add the points in Q to the list of set C and clear the list of Q;

[0047] (4) When each point P i ∈P performs the above operations, and P i are all part of set C, and the algorithm terminates.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] (1) Compared with the prior art, the present invention solves the problem that the prior art is difficult to determine the local structural abnormality of the tower. The present invention compares the three-dimensional coordinates of the key nodes of the tower at different times, extracts the relative displacement of the key nodes of the tower, and compares it with the threshold value for determining the structural abnormality of the key nodes, so as to timely detect whether the tower has local structural deformation.

[0050] (2) Compared with the prior art, the present invention can detect the local deformation of the crossarms and ground wire brackets caused by rapid ice dancing and extreme strong winds. Compared with the prior art, which mainly uses the inclination of the tower to characterize the structural abnormality of the tower, it cannot reflect the local deformation of the crossarms and ground wire brackets. The present invention uses three-dimensional laser scanning technology to perform multiple scanning and detection of the tower before and after being subjected to extremely harsh working conditions, effectively obtain the three-dimensional coordinates of the key nodes of the tower at different times, and determine whether the crossarms and ground wire brackets are locally deformed. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of the Super-4PCS coarse registration technology adopted by the present invention;

[0052] Figure 2 It is a flow chart of the local ICP precise registration technology adopted by the present invention;

[0053] Figure 3 It is a flow chart of the relative displacement extraction technology of key nodes of the tower adopted by the present invention;

[0054] Figure 4 are two phases of point clouds to be registered in an embodiment of the present invention;

[0055] Figure 5 is a schematic diagram of the first transverse plane of the tower extracted in an embodiment of the present invention;

[0056] Figure 6 is a schematic diagram of two-phase point cloud registration in an embodiment of the present invention;

[0057] Figure 7 is a schematic diagram of orienting and rotating a tower using a principal component analysis method in an embodiment of the present invention;

[0058] Figure 8 is a schematic diagram of insulator cluster analysis in an embodiment of the present invention;

[0059] Fig. 9 It is a schematic diagram of displacement extraction of key nodes of a tower in an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the present invention.

[0061] The embodiment of the present invention first uses a 3D laser scanner to scan and detect a potential danger tower twice, and divides the two phases of point clouds into a reference point cloud and a comparison point cloud; the comparison point cloud is registered to the reference point cloud through the registration algorithm proposed by the present invention, and the relative displacement of the ground wire hanging point is extracted. The specific implementation steps are as follows.

[0062] Step 1: Scan and inspect a potential danger tower twice to obtain two phases of point cloud data of the tower. Figure 4 .

[0063] Step 2: Follow Figure 1 The process shown in the figure is to slice the tower point cloud. A set of planes with a thickness of δ = 80 cm are used to intersect the point cloud, slice the point cloud, and extract the first transverse plane of the tower point cloud. For details, see Figure 5 .

[0064] Step 3: Follow Figure 2 The process shown in the figure uses Super-4PCS coarse registration and local ICP algorithm to register the sequence 1 point cloud of the tower to the reference point cloud. Figure 6 Specifically, the first transverse plane of the tower is selected as the registration reference, and the Super-4PCS coarse registration technology is used. Figure 1 , the sequence 1 point cloud S of the tower 1 Register to the reference point cloud S 0 ; Then use the local ICP algorithm for precise registration, see Figure 2 , the sequence 1 point cloud S of the tower 1 Register to the reference point cloud S 0 .

[0065] Step 4, follow Figure 3 The process shown in the figure is to extract the relative displacement of the key nodes of the tower (ground wire hanging point and insulator cross arm hanging point). First, the principal component analysis method is used to orient and rotate the tower, such as Figure 7 ; Then cluster the insulators, such as Figure 8 ; The three-dimensional coordinates of key nodes such as the side hanging point of the tower ground wire crossarm are obtained by extracting the coordinates of the point closest to the insulator and the tower. For details, see Fig. 9 , calculate the point cloud S 0 and S 1 The displacements of the key nodes at the same position on the x-axis, y-axis and z-axis are shown in Table 1.

[0066] Table 1 Displacement of key nodes of tower

[0067] Key Nodes Δx(m) Δy(m) Δz(m) Δd(m) Threshold Is it abnormal? Point #1 -41.2 12.1 -56.2 70.73 0.25 yes Point #2 2.5 3.6 0.4 4.40 0.23 no Point#3 -3.5 3.7 -5.6 7.57 0.23 no Point#4 2.4 2.6 -5.9 6.88 0.19 no Point#5 -1.3 3.2 2 3.99 0.19 no Point#6 0.2 -2.7 2.6 3.75 0.15 no Point#7 -1.1 2.9 -2.1 3.75 0.15 no Point#8 -6.4 -3.7 -6.8 10.04 0.25 no Point#9 9.8 5.4 8.9 14.30 0.23 no Point #10 -7.9 2.1 2.4 8.52 0.23 no Point#11 2.5 2.6 -6.8 7.70 0.19 no Point #12 -7.8 3.6 9.8 13.03 0.19 no Point#13 7.4 -2.9 -2.1 8.22 0.15 no Point #14 5.2 5.1 -4.7 8.67 0.15 no

[0068] Step 5: The threshold for determining the structural abnormality of the key nodes of the tower is (h×5‰), where h is the height of each node relative to the tower base. The displacement of the key nodes of the tower is compared with the threshold for determining the structural abnormality of the key nodes. It is found that the ground wire bracket on the left side of the tower (Point#1) exceeds the threshold for determining the structural abnormality of the key nodes, as shown in Table 1, and is determined to be structurally abnormal. On-site inspection found that the ground wire bracket on the left side of the tower was bent and deformed, which verified the effectiveness and feasibility of this method.

[0069] For multiple point clouds, repeat steps 2-4 respectively, and convert the sequence point cloud S 2 ,S 3 ,…S n The tower is aligned to the reference point cloud in sequence, and then the displacement of the key nodes of the tower is extracted and compared with the judgment threshold to determine whether the tower has structural abnormalities.

[0070] Compared with the prior art, the present invention has the following beneficial effects:

[0071] (1) Compared with the prior art, the present invention solves the problem that the prior art is difficult to determine the local structural abnormality of the tower. The present invention compares the three-dimensional coordinates of the key nodes of the tower at different times, extracts the displacement of the key nodes of the tower, and compares it with the threshold value for determining the structural abnormality of the key nodes, so as to timely detect whether the tower has local structural deformation.

[0072] (2) Compared with the prior art, the present invention can detect the local deformation of the crossarm and ground wire bracket caused by rapid ice dancing and extreme strong winds. Compared with the prior art, which mainly uses the inclination of the tower to characterize the structural abnormality of the tower, it cannot reflect the local deformation of the crossarm and ground wire bracket. The present invention uses three-dimensional laser scanning technology to perform multiple scanning and detection of the tower before and after being subjected to extremely harsh working conditions, effectively obtain the three-dimensional coordinates of the key nodes of the tower at different times, and determine whether the crossarm and ground wire bracket have local deformation.

[0073] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0074] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0075] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0076] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting local deformation of a tower based on laser point clouds of different periods, characterized in that: The steps include: Step 1: Use a 3D laser scanner to scan and detect the overhead transmission line, and divide the tower point clouds collected at different times into reference point clouds and comparison point cloud sequences 1, 2, ... n according to the time series; Step 2, slicing and layering the tower point cloud, and automatically identifying the first transverse plane of the tower according to the tower point cloud structure judgment criterion; Step 3, taking the first transverse plane of the reference point cloud as the reference object, using the Super-4PCS coarse registration and local ICP fine registration algorithms, the point cloud sequence 1, 2, ...n is registered to the reference point cloud; Step 4: extract the displacement of the key nodes of the point cloud sequence 1, 2, ... n relative to the reference point cloud, and compare it with the structural anomaly judgment threshold of the key nodes to evaluate whether the tower structure is abnormal.

2. The method for detecting local deformation of a tower based on laser point clouds of different periods according to claim 1, characterized in that: Step 1 includes: The transmission line is scanned by 3D laser radar, and then solved, filtered and thinned. The point cloud classification technology is used to extract the tower point cloud, and the ground wire, ground and tree point clouds are eliminated. According to the time sequence of point cloud acquisition, the first phase of the tower point cloud is set as the reference point cloud, denoted as S 0 , set the subsequent point clouds as the comparison sequence point clouds in sequence, that is, sequence 1 point cloud, sequence 2 point cloud...sequence n point cloud, denoted as S 1 , S 2 …S n .

3. The method for detecting local deformation of a tower based on laser point clouds of different periods according to claim 2, characterized in that: Step 2 includes: Step 2.1, slice the tower point cloud: use a set of planes with a thickness of δ to intersect the point cloud, slice the point cloud from top to bottom, the thickness of the ground scanned point cloud slice δ is 50cm, the thickness of the machine scanned point cloud slice δ is 80cm, and the tower point cloud is cut from top to bottom in the vertical z-axis direction to obtain the tower horizontal point cloud slices S in sequence. i : Project each layer of point cloud slices onto the horizontal plane as shown below: Step 2.2, extract the circumscribed polygon boundary: use the "bidirectional closest point search method" to extract the circumscribed polygon boundary of each layer of slices, and project the set of point cloud slices In S', take any point Ps as the starting point of the boundary line, find the point Pr closest to Ps, and calculate the distance ds from Pr to Ps; i In the remaining point cloud, find the nearest point Pe of Ps and calculate the distance de from Ps to Pe; if ds≤de, insert Pr in front of Ps as a new starting point, connect PrPsPe in pairs, and replace Pr with Ps as the starting point; if ds≥de, grow to the other end Pe; then determine the set S' i Are there any remaining points? If so, repeat the above steps; finally, traverse the set S' i After all points are found, until PrPe overlaps, PsPe is connected to close the polygon and obtain the polygon boundary of the point cloud outer contour; Step 2.3, calculate the vertices of the circumscribed polygon: through the linear regression equation, fit the outermost boundary point cloud into four boundary lines, which are recorded as AB, BC, CD, and DA in clockwise order; the intersection of the four straight lines is the four endpoints of the tower transverse diaphragm, and the coordinates of the endpoints are recorded as A(x A ,y A ,z A ), B(x B ,y B ,z B ), C(x C ,y C ,z C ), D(x D ,y D ,z D ); Step 2.4, calculate the aspect ratio, area ratio parameters and number of points of the slice circumscribed polygon: calculate the side length l of each layer of point cloud slice respectively ix ,l iy , area i 、Number of points n i , and then calculate the area of ​​the circumscribed rectangle and aspect ratio Where l ix ,l iy Project S for the i-th slice i 'The length and width of the circumscribed rectangle; Step 2.5, identify the first transverse plane of the tower according to the number of slice points and the aspect ratio and area ratio of the outer polygon. The identification criteria are: when slicing from top to bottom, the slice layer where the last peak of the number of slice points is located, and the number of slice points n i Should be greater than 500, and the height of the diaphragm should be less than the tower height The height from the ground is generally not more than 15m; the length-to-width ratio of the circumscribed rectangle of the section meets the following requirements: Slice area ratio 4. The method for detecting local deformation of a tower based on laser point clouds of different periods according to claim 3, characterized in that: The Super-4PCS coarse registration algorithm in step 3 includes: Step 3.1.1, in the reference point cloud S 0 Select 4 coplanar and non-colinear points, which are the common points of the multi-period point cloud, and the distance between them is as large as possible, denoted as B = {P a , P b , P c , P d }, calculate the distance and intersection ratio of straight lines AB and CD according to the following formula: d1=||P a -P b ||,d2=||P c -P d || (3) Step 3.1.2, according to the affine invariance principle, in sequence 1 point cloud S 1 Find the point set with the same distance and intersection ratio with the straight lines AB and CD, as follows: In sequence 1 point cloud S 1 Determine the point pair set S1, S2, and take each point q i As the center of the sphere, draw spheres with R = d1 ± ε and R = d2 ± ε as the radius respectively. The set S1 is q1 and the points distributed in the range [d1-ε, d1+ε], and q1 and the points distributed in the range [d2-ε, d2+ε] are S2. At the same time, the point cloud surface is rasterized to establish a three-dimensional grid G ​​with a unit size of ε. Step 3.1.3, in sequence 1 point cloud S 1 Extract the reference point cloud S 0 Corresponding four point sets: traverse all candidate point pairs in point pair sets S1 and S2, calculate all intersection nodes e according to the cross ratio consistency, and store them in the grid G, based on the reference point cloud S 0 The angle between the two corresponding point pairs is θ, and the nodes extracted from the grid G ​​are approximately equal. At the same time, the angle between the two point pairs is approximately equal to θ, and the four-point basis matching is performed within the error range of ξ; Step 3.1.4, in sequence 1 point cloud S 1 All the points that meet the conditions and the reference point cloud S 0 The corresponding four-point set U = {U1, U2, ..., U n }, find the reference point cloud S 0 With every U i The transformation matrix between them is compared to maximize the common point set LCP and select the transformation matrix with the highest registration accuracy for global transformation.

5. The method for detecting local deformation of a tower based on laser point clouds of different periods according to claim 3, characterized in that: The local ICP fine registration algorithm in step 3 includes: for the point cloud S 0 and S 1 , from the reference point cloud S 0 Select the reference point p i , from the point cloud S 1 Find the distance point p in i The point q with the shortest Euclidean distance i , with p i and q i The corresponding point set formed calculates the transformation matrix, continuously iterates the process, uses Formula 5 as the target error function, stops iterating when the value of f is less than the threshold, and obtains the optimal transformation matrix. Where R represents the rotation transformation matrix, T represents the translation transformation matrix, and k represents the number of corresponding points.

6. The method for detecting local deformation of a tower based on laser point clouds of different periods according to claim 1, characterized in that: Step 4 includes: Step 4.1, point cloud cutting: respectively cut the point cloud S 0 and S 1 The principal component analysis method is used to obtain the main direction feature vector of the two-dimensional projection point cloud. On this basis, the main direction rotation angle is calculated, and the tower point cloud is rotated to obtain the redirected point cloud data. The direction perpendicular to the main direction after rotation on the horizontal plane is selected as the horizontal direction, and the point cloud is cut along the horizontal direction. The cut point cloud is divided into two parts, left and right. Step 4.2, scattered point cloud indexing: for the reference point cloud S 0 And sequence 1 point cloud S 1 Use Octree method to build scattered point cloud index; Step 4.3, Euclidean clustering segmentation: For the reference point cloud S with established Octree index 0 And sequence 1 point cloud S 1 The insulators in the cluster are clustered in Euclidean order to distinguish the phases of the insulators. The point cloud of each tower is set as P, which consists of n points. The Euclidean distance between the i-th sample and the j-th sample is calculated. The index number of the point with the smallest distance is classified into one class. The Euclidean distance between the new classes is calculated and iterated until the specified threshold is less than the Euclidean distance between any two classes or the number of classes is less than the specified number. The Euclidean clustering segmentation is completed. Step 4.4, key node position extraction: extract the point cloud S of each period after clustering 0 and S 1 The three-dimensional coordinates of the hanging points on the crossarms of each phase insulator and the hanging points on the left and right phase ground wires are calculated to obtain the point cloud S 0 and S 1 The displacement of the key nodes at the same position on the x-axis, y-axis and z-axis; Step 4.5, structural anomaly determination: compare the displacement of each key node with the key node structural anomaly determination threshold to determine whether the multi-phase point cloud key nodes are displaced, and then determine whether the tower structure is deformed.

7. The method for detecting local deformation of a tower based on laser point clouds of different periods according to claim 6, characterized in that: The specific implementation process of step 4.3 is as follows: Step 4.3.1, establish an Octree data structure for each tower point cloud P; Step 4.3.2, establish an empty cluster set C and a queue; Step 4.3.3, for each point P i ∈P, perform the following steps: (1) Add the point to the current queue Q; (2) For each point P i ∈Q, perform the following operations: ① Use the search radius r<d th , for P i Perform a k-nearest neighbor search and the points found are ②For each calculate With P i The Euclidean distance of each P is the same as the other two. i Check whether the point performs the above operations. If not, add the point to the queue Q; (3) When all points in the list of queue Q have performed the above operations, add the points in Q to the list of set C and clear the list of Q; (4) When each point P i ∈P performs the above operations, and P i are all part of set C, and the algorithm terminates.

Images