A method for identifying the tilt state of a hyperbolic cooling tower

Through the combination of drone and ground laser scanner, combined with data registration, filtering, dilution and particle swarm optimization algorithms, the data quality and accuracy problems in cooling tower inclination detection are solved, and a more efficient and reliable structure is achieved Evaluate.

CN118500353BActive Publication Date: 2024-11-01THE THIRD CONSTR OF CHINA CONSTR EIGHTH ENG BUREAU
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410954653.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-01
Estimated Expiration
2044-07-17

AI Technical Summary

Technical Problem

The existing cooling tower inclination detection methods cannot comprehensively evaluate the damage of complex structures, and the data quality and accuracy are affected by equipment and environmental factors, making it difficult to obtain complete point cloud data, resulting in information loss and reduced measurement accuracy.

Method used

The tilt state of the cooling tower is calculated through data registration, filtering, dilution processing, layering processing and particle swarm optimization algorithms to ensure the accuracy and reliability of the measurement.

Benefits of technology

It significantly improves data acquisition efficiency and accuracy, can capture slight changes in cooling tower structure in detail, provide more accurate detection results, ensure the reliability of evaluation results, and reduce operational risks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118500353B_ABST
    Figure CN118500353B_ABST
Patent Text Reader

Abstract

The present invention provides a method for identifying the tilt state of a hyperbolic cooling tower based on laser point cloud data, and the identification method includes: collecting cooling tower point cloud data; performing data registration on the collected point cloud data; using a filter to perform denoising on the point cloud data; using uniform sampling technology to perform thinning and simplification processing; performing layered processing on the cooling tower point cloud data; applying a particle swarm optimization algorithm to calculate the center point of each layer of point cloud data, and fitting the center points of each layer into a straight line; calculating the angle between the fitting straight line and the Z axis, so as to determine the degree of deviation of the cooling tower from the vertical direction. The present invention utilizes a combination of unmanned aerial vehicles and ground laser scanning to improve data collection efficiency and point cloud model accuracy. The present invention can capture subtle changes in the cooling tower structure in detail, provide more accurate detection results, and ensure the reliability of the evaluation results; the non-contact detection method reduces operational risks and protects the safety of staff and equipment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of inclination measurement of industrial building cooling towers, and in particular relates to a method for identifying the inclination state of a hyperbolic cooling tower based on laser point cloud data. Background Art

[0002] Hyperbolic cooling towers are tall, thin-walled reinforced concrete buildings designed specifically for industrial facilities (including power plants) to effectively remove process heat and release it to the atmosphere. Such buildings are often exposed to extreme environmental conditions, such as significant temperature fluctuations and chemical corrosion, which can lead to material degradation and even structural damage, deformation and defects in the building, so regular inspections of the structural health of the cooling tower are particularly important. Detecting the inclination of the cooling tower is a key measure to ensure its structural safety and extend its service life. By monitoring the inclination, potential structural problems caused by foundation settlement, material aging or external forces (such as earthquakes) can be discovered in a timely manner, so that necessary corrective measures can be taken to prevent further damage.

[0003] Existing detection methods, such as visual inspection and simple non-destructive testing, often fail to fully evaluate the complex damage of cooling towers and may require expensive equipment and expertise. With the development of three-dimensional technology, the application of LiDAR in the field of building safety detection has become more efficient and versatile due to its advantages such as high precision and non-contact. Among them, structural deformation detection uses the point cloud data obtained by LiDAR scanning to evaluate the state of the object. The point cloud data obtained can be analyzed by appropriate methods to further obtain relevant information such as the safety of the building structure, but it faces many challenges in practical applications. First, the accuracy and reliability of point cloud data depend to a large extent on the quality of the data, and the data quality itself is significantly affected by the limitations of the acquisition equipment and environmental factors. For example, for some large buildings and structures, it is difficult to obtain complete point cloud data using only ground LiDAR scanning; secondly, for structures with complex shapes, the processing and analysis of point cloud data are difficult. Methods such as plane fitting and least squares often lead to a large amount of information loss and are sensitive to noise and outliers, affecting the measurement accuracy.

[0004] Therefore, for the special structure of cooling tower, how to design effective data collection methods and appropriate data processing and analysis methods is particularly critical, which is of great value for ensuring the structural safety of cooling towers and similar industrial facilities. Summary of the invention

[0005] In view of the shortcomings in the prior art, the present invention provides a method for identifying the tilt state of a hyperbolic cooling tower based on laser point cloud data, which can capture subtle changes in the cooling tower structure in detail, provide more accurate detection results, and ensure the reliability of the evaluation results.

[0006] The present invention achieves the above technical objectives through the following technical means.

[0007] A method for identifying the tilt state of a hyperbolic cooling tower based on laser point cloud data includes the following steps:

[0008] Step 1: Collect cooling tower point cloud data based on the comprehensive application of drones and ground laser scanners;

[0009] Step 2: Perform data registration on the collected point cloud data;

[0010] Step 3: Use filters to remove noise and outliers from the point cloud data;

[0011] Step 4: Sparse and simplify the point cloud data;

[0012] Step 5: Adjust the point cloud data posture and perform layered processing on the point cloud data;

[0013] Step 6: Apply the particle swarm optimization algorithm to calculate the center point of each layer of point cloud data, and fit the center points of each layer into a straight line;

[0014] Step 7: Calculate the angle between the fitting line and the Z-axis of the spatial rectangular coordinate system, and determine the degree of deviation of the cooling tower from the vertical direction based on the angle.

[0015] Furthermore, the specific process of step 1 is as follows:

[0016] Firstly, an integrated application drone equipped with multiple lenses was used to obtain image data of the upper and peripheral structures of the cooling tower. The cooling tower was divided into 24 sections for up and down flight missions. The horizontal overlap was greater than 70%, and the vertical overlap was set by the time interval of the up and down flights of the integrated application drone. Then, a ground laser scanner was used to scan the point cloud data of the cooling tower. With the cooling tower to be tested as the center, multiple stations were set up around the cooling tower. The inside of the cooling tower was also scanned using a ground laser scanner.

[0017] Furthermore, the specific process of step 2 is as follows:

[0018] First, the point cloud data collected by the ground laser scanner is imported into the 3D laser scanning software, and the point cloud data of each site is registered. After registration, it is exported and recorded as the ground point cloud ; Then, the images collected by the integrated application drone are imported into the 3D design software for photo alignment, the common feature points between the images are identified and a sparse point cloud is generated, and the 3D information is extracted from the aligned photos to finally form a high-resolution target point cloud, recorded as ; Use the iterative closest point algorithm for precise registration. The specific process is as follows:

[0019] S1: Select an initial transformation matrix , is a 3×3 rotation matrix, is a 3×1 column vector, Set to the identity matrix;

[0020] S2: For ground point cloud Every point in , in the target point cloud Find the nearest point , implemented through KD tree: , Representation and The nearest point is The index in is the number of iterations, Indicates Iteration transformation matrix;

[0021] S3: Calculate a transformation including rotation and translation to minimize arrive The registration error is: ;

[0022] S4: Use singular value decomposition to solve the optimal rotation matrix and translation variables :

[0023] S4.1: For ground point clouds and target point cloud , first calculate their respective centroids and , and then center the point set at the origin:

[0024]

[0025]

[0026] in, , Represents ground point cloud , target point cloud The original point in , Decentralized coordinates, ;

[0027] S4.2: Construct the cross-covariance matrix: , Represents a transpose operation;

[0028] S4.3: Apply singular value decomposition: , and is an orthogonal matrix, is a diagonal matrix;

[0029] S4.4: Calculate the rotation matrix and translation variables : , , and then get the corresponding transformation matrix;

[0030] S5: Based on transformation matrix Update the point cloud to achieve registration: ;

[0031] S6: Set the number of iterations, relative error, absolute error, and calculate the average error and variance after each iteration for quality control of the entire registration process.

[0032] Furthermore, the specific process of step 4 is as follows:

[0033] First, the entire point cloud space is divided into a three-dimensional cube grid, and the grid size is defined according to the amount of point cloud data; then each point is traversed, and the relationship between the point coordinates and the grid boundary is calculated to determine which grid the coordinates belong to; in the grid, the center of all points is calculated As a representative point: , is the number of points in the grid, , , are the coordinates of each point in the grid.

[0034] Furthermore, the specific process of the layered processing in step 5 is as follows:

[0035] Traverse the entire point cloud data and calculate the lowest point and the highest point , determine the vertical range of the cooling tower point cloud ;

[0036] Building a Hierarchy: Setting Layer Heights , and calculate the total number of layers : ,Will arrive The range is evenly divided into For all point cloud data, according to the Z-axis coordinate value of the point Divide into corresponding levels, for each level , select points that satisfy the following conditions: .

[0037] Furthermore, the specific process of step 6 is as follows:

[0038] Step 6.1: Initialize the particle swarm: Each particle represents a point in three-dimensional space , whose location is , its initial position and speed They are all randomly generated within the boundary of the point cloud data layer and obey a uniform distribution:

[0039]

[0040] in, Indicated in , Uniformly randomly distributed between and are the minimum and maximum bounds of the search space, respectively;

[0041] Indicated in , Uniformly randomly distributed between

[0042] Step 6.2: Evaluation function For particles The sum of the Euclidean distances to all points in the point cloud:

[0043]

[0044] in, It is the first The location of the point, is the total number of points in the point cloud, represents the Euclidean distance;

[0045] Step 6.3: Particle Update: Each Iteration Next, particles Speed Updated as follows:

[0046]

[0047] in, is the inertia weight; and is the acceleration factor; and is a random number between [0,1]; It is The best position that a particle experiences; is the global optimal position;

[0048] Each iteration Next, particles Location Updated as follows: .

[0049] Step 6.4: Individual optimal and global optimal updates:

[0050] like , then update ;

[0051] like , for any , then update ;

[0052] Step 6.5: Set the number of particles and the maximum number of iterations, repeat the update operation until the maximum number of iterations is reached, and finally return to the global optimal solution , which is used as the representative center point of the point cloud layer.

[0053] Furthermore, the specific process of step 7 is as follows:

[0054] Calculate the representative center point of the point cloud layer The mean , and transform the coordinates of the center point into a coordinate system with this mean as the origin. Apply the principal component analysis method to the transformed data to solve the eigenvector corresponding to the maximum eigenvalue of the covariance matrix , the eigenvector is the main direction of the data, that is, the first principal component, is the direction vector of the fitted straight line; the Z-axis vector is defined as , use the vector dot product formula and the vector length to calculate the cosine of the angle between the fitted line and the Z axis:

[0055]

[0056] in, , , The vector coordinates representing the main direction of the point cloud data.

[0057] And the included angle is obtained by arc cosine calculation :

[0058]

[0059] Angle Indicates the degree of deviation of the cooling tower from the vertical direction.

[0060] A computer-readable storage medium is used to store program data. When the program data is executed by a processor, the steps of the above-mentioned method for identifying the tilt state of a hyperbolic cooling tower based on laser point cloud data are implemented.

[0061] The present invention has the following beneficial effects:

[0062] (1) Improving data collection efficiency: The combination of UAV and ground laser scanning significantly improves the efficiency of data collection and the accuracy of point cloud models.

[0063] (2) Enhanced detection accuracy and reliability: The present invention can capture minute changes in the cooling tower structure in detail, providing more accurate detection results than traditional methods, thus ensuring the reliability of the evaluation results.

[0064] (3) Increased operational safety: Non-contact detection methods reduce operational risks and protect the safety of workers and equipment, especially when testing facilities with critical structural stability or severe damage. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is the visualization diagram of the cooling tower point cloud data obtained in step 1;

[0066] Figure 2 It is a layered visualization of point cloud data;

[0067] Figure 3 It is a visualization of the center point of each layer of point cloud calculated using the particle swarm algorithm;

[0068] Figure 4 It is a visualization of the center points of each layer of the cooling tower point cloud;

[0069] Figure 5 It is a visualization diagram of the fitted straight line and the Z-axis of the spatial rectangular coordinate system;

[0070] Figure 6 It is a flow chart of the method for identifying the tilt state of a cooling tower. DETAILED DESCRIPTION

[0071] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited thereto.

[0072] Embodiment 1:

[0073] The method for identifying the tilt state of a hyperbolic cooling tower based on laser point cloud data of the present invention is as follows: Figure 6 As shown, the specific process includes the following:

[0074] Step 1: Cooling tower point cloud data collection;

[0075] Firstly, an unmanned aerial vehicles (UAV) equipped with five lenses was used to obtain image data of the upper and peripheral structures of the cooling tower. Each camera had 25 million pixels, a total of 105 million pixels, and a camera focal length of 35 mm. The cooling tower was divided into 24 sections for up and down flight missions, with a horizontal overlap greater than 70%, and a vertical overlap set by the time interval of the UAV up and down flight. Then, a terrestrial laser scanner (TLS) was used to scan the point cloud data of the cooling tower. With the cooling tower to be tested as the center, multiple sites were set up around the cooling tower. The distance between the sites was set to 20 meters, so that the point position accuracy was maintained at 2.9 mm, and the point cloud data acquisition resolution was selected to be 6 mm@10 m. TLS was also used to scan the interior of the cooling tower. In order to align the internal and external point cloud data, multiple positions were measured on the tower door to achieve fine matching accuracy.

[0076] Step 2: Perform data registration on the collected point cloud data;

[0077] First, collect the TLS Figure 1 The point cloud data shown is imported into the 3D laser scanning software (Leica Cyclone software in this embodiment), and the point cloud data of each site is registered. After registration, it is exported and recorded as the ground point cloud. , , , , represents each point contained in the point cloud set; then use the 3D design software (Photoscan software in this embodiment) to create a new project and import the UAV images for photo alignment, identify the common feature points between the images and generate a sparse point cloud, and extract more detailed 3D information from the aligned photos, and finally form a high-resolution target point cloud, recorded as , , , , Represents each point contained in the target point cloud set; the UAV scanned point cloud data is used as the target point cloud, and the TLS scanned point cloud is used as the point cloud to be registered. The iterative closest point (ICP) algorithm is used for accurate registration. The specific process is as follows:

[0078] S1: Null hypothesis: Choose an initial transformation matrix ,in, is a 3×3 rotation matrix, is a 3×1 column vector, Set to the identity matrix.

[0079] S2: Closest point matching: for ground point cloud Every point in , in the target point cloud Find the nearest point , implemented through KD tree (K-dimension tree): , Representation and The nearest point is The index in is the number of iterations, Indicates Iteration transformation matrix.

[0080] S3: Minimize the error between point pairs: Calculate a transformation (including rotation and translation) to minimize the error between point pairs. arrive The registration error is: .

[0081] S4: Use Singular Value Decomposition (SVD) to solve the optimal rotation matrix and translation variables :

[0082] S4.1: For ground point clouds and target point cloud , first calculate their respective centroids, that is, the arithmetic mean of all point coordinates in the point cloud, respectively and , then, subtract the coordinates of each point in the point set from its corresponding centroid, centering the point set at the origin: , , , ( ) represent ground point clouds respectively , target point cloud The original point in , Decentralized coordinates.

[0083] S4.2: Constructing the cross-covariance matrix : , Representation vector The transpose of Represents a transpose operation.

[0084] S4.3: Apply SVD: , and is an orthogonal matrix, is a diagonal matrix.

[0085] S4.4: Calculate the rotation matrix and translation variables : , , and then get the corresponding transformation matrix.

[0086] S5: Apply transformation: Based on transformation matrix Update the point cloud to achieve registration: .

[0087] S6: Error calculation, quality control of the entire registration process: setting the number of iterations The relative error and absolute error are set to 0.001m, and the average error is calculated after each iteration. and variance :

[0088] Average Error: ;

[0089] variance: .

[0090] Step 3: When obtaining the aligned cooling tower point cloud data, these data not only include the cooling tower itself, but may also include point cloud data of other buildings, ground and vegetation; in order to focus on the analysis of the cooling tower, first use the segmentation function of the 3D point cloud processing software (CloudCompare software in this embodiment), use polygons as segmentation lines, and accurately define the segmentation area to filter out the cooling tower data; after completing the screening and extraction of the cooling tower point cloud data, further use the built-in filter (SOR filter) of the software to denoise the point cloud data to remove noise and outliers, so as to ensure the accuracy and reliability of the subsequent inclination calculation.

[0091] Step 4: The cooling tower point cloud data obtained in step 3 is further thinned and simplified using the uniform sampling function of the 3D point cloud processing software to improve data processing efficiency while ensuring the accuracy of subsequent calculations:

[0092] First, the entire point cloud space is divided into a regular three-dimensional cube grid, and the grid size is defined according to the amount of point cloud data; then each point is traversed, and the relationship between the point coordinates and the grid boundary is calculated to determine which grid the coordinates belong to; in the grid, the center of all points is calculated as the representative point: , is the number of points in the grid, , , are the coordinates of each point in the grid.

[0093] Step 5: Use the 3D point cloud processing software to adjust the point cloud data posture of the cooling tower after processing the point cloud data in step 4, ensure that the height direction of the data is aligned with the Z axis, and then perform layered processing on the cooling tower point cloud data. The specific method is as follows:

[0094] Determine the range: traverse the entire point cloud data and calculate the lowest point in the data and the highest point , determine the vertical range of the cooling tower point cloud ;

[0095] like Figure 2 As shown, establish a hierarchical structure: set the layer height , and calculate the total number of layers : ,Will arrive The range is evenly divided into levels, each level represents a horizontal segment of a fixed height; for all point cloud data, according to the z-axis coordinate value of the point The size is divided into corresponding levels, and for each level , select points that satisfy the following conditions: , ensuring that each point is correctly classified into its corresponding level.

[0096] Step 6: Reference Figure 3 , Figure 4 , Figure 5 , the particle swarm optimization (PSO) algorithm is applied to calculate the center point of each layer of point cloud data, and the center points of each layer are fitted into a straight line; the PSO algorithm is applied to calculate the center point of the point cloud data. Compared with simply finding the average value of each point, the PSO algorithm can effectively search and adjust the potential center point position. The specific process of using the PSO algorithm to calculate the center point is as follows:

[0097] Step 6.1: Initialize the particle swarm: Each particle represents a point in three-dimensional space , whose location is , its initial position and speed They are all randomly generated within the boundary of the point cloud data layer and obey a uniform distribution:

[0098]

[0099] in, Indicated in , Uniformly randomly distributed between and are the minimum and maximum bounds of the search space, respectively;

[0100] Indicated in , Uniformly randomly distributed between

[0101] Step 6.2: Evaluation Function: Evaluation Function Defined as a particle The sum of the Euclidean distances to all points in the point cloud:

[0102]

[0103] in, It is the first The location of the point, is the total number of points in the point cloud, represents the Euclidean distance.

[0104] Step 6.3: Particle Update: Each Iteration Next, particles Speed Updated as follows:

[0105]

[0106] in, is the inertia weight, set to 0.5; and are acceleration factors, which are set to 0.8 and 0.9 respectively; and is a random number between [0,1]; It is The best position that a particle experiences; is the global optimal position;

[0107] Each iteration Next, particles Location Updated as follows: .

[0108] Step 6.4: Individual optimal and global optimal updates:

[0109] like , then update ;

[0110] like , for any , then update ;

[0111] Step 6.5: Set the number of particles (points) to 50, the maximum number of iterations to 100, repeat the above update operation until the maximum number of iterations is reached, and finally return to the global optimal solution , which is used as the representative center point of the point cloud layer.

[0112] Step 7: Calculate the representative center point of the above point cloud layer The mean , and transform the coordinates of the center points into a coordinate system with this mean as the origin (the coordinates of each representative center point minus the mean), apply principal components analysis (PCA) to the transformed data, and solve the eigenvector corresponding to the maximum eigenvalue of the covariance matrix , the eigenvector is the main direction of the data, that is, the first principal component, is the direction vector of the fitted straight line; the Z-axis vector of the spatial rectangular coordinate system is defined as , use the vector dot product formula and vector length to calculate the angle between the fitted line and the Z axis of the spatial rectangular coordinate system :

[0113]

[0114] in, , , The vector coordinates representing the main direction of the point cloud data.

[0115] And the angle between the vectors is obtained by arc cosine calculation:

[0116]

[0117] Angle This indicates the degree of deviation of the cooling tower from the vertical, thus intuitively quantifying the inclination of the structure.

[0118] Embodiment 2:

[0119] This embodiment provides a computer-readable storage medium, which is used to store program data. When the program data is executed by a processor, it is used to implement the method for identifying the inclination state of a hyperbolic cooling tower based on laser point cloud data described in Example 1.

[0120] The embodiments are preferred implementations of the present invention, but the present invention is not limited to the above-mentioned implementations. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essential content of the present invention belong to the protection scope of the present invention. Figure 5 The different color markings in the figures are only for the convenience of distinguishing different illustrations and should not be understood as limiting the present invention.

Claims

1. A method for identifying the tilt state of a hyperbolic cooling tower based on laser point cloud data, characterized in that: The process includes the following: Step 1: Collect cooling tower point cloud data based on the comprehensive application of drones and ground laser scanners; Step 2: Perform data registration on the collected point cloud data; Step 3: Use filters to remove noise and outliers from the point cloud data; Step 4: Sparse and simplify the point cloud data; Step 5: Adjust the point cloud data posture and perform layered processing on the point cloud data; Step 6: Apply the particle swarm optimization algorithm to calculate the center point of each layer of point cloud data, and fit the center points of each layer into a straight line; Step 7: Calculate the angle between the fitting straight line and the Z axis of the spatial rectangular coordinate system, and determine the degree of deviation of the cooling tower from the vertical direction according to the angle; The specific process of step 2 is as follows: First, the point cloud data collected by the ground laser scanner is registered, and the registered point cloud P is exported. Then, the images collected by the comprehensive application drone are processed for photo alignment, the common feature points between the images are identified and a sparse point cloud is generated, and three-dimensional information is extracted from the aligned photos, and finally a high-resolution target point cloud is formed, which is recorded as Q. Then, the iterative closest point algorithm is used for accurate registration. The specific process is as follows: S1: Select an initial transformation matrix R is a 3×3 rotation matrix, t is a 3×1 column vector, T (0) Set to the identity matrix; S2: For each point p in the ground point cloud P i , find the nearest point q in the target point cloud Q i , implemented through KD tree: q π(i) =argmin q∈Q ∥T (k) p i -q i ∥, π(i) represents the same as p i The index of the nearest point in Q, k is the number of iterations, T (k) represents the kth iteration transformation matrix; S3: Calculate a transformation including rotation and translation to minimize the registration error from P to Q: S4: Use singular value decomposition to solve the optimal rotation matrix R and translation variable t: S4.1: For the ground point cloud P and the target point cloud Q, first calculate their respective centroids. and Then center the point set at the origin: Among them, p i ′、q i ′ represents the original point p in the ground point cloud P and the target point cloud Q respectively. i ,q i The coordinates after decentralization are 1≤i≤n; S4.2: Construct the cross-covariance matrix H: T represents the transpose operation; S4.3: Apply singular value decomposition: H = UΣV T , U and V are orthogonal matrices, Σ is a diagonal matrix; S4.4: Calculate the rotation matrix R and translation variable t: R = VU T , Then the corresponding transformation matrix is ​​obtained; S5: Based on the transformation matrix T (k+1) Update point cloud to achieve registration: P←T (k+1) P; S6: Set the number of iterations, relative error, absolute error, and calculate the average error and variance after each iteration for quality control of the entire registration process; The specific process of step 4 is as follows: First, the entire point cloud space is divided into three-dimensional cube grids, and the grid size is defined according to the amount of point cloud data. Then, each point is traversed and the relationship between the point coordinates and the grid boundary is calculated to determine which grid the coordinates belong to. In the grid, the center P of all points is calculated. center As a representative point: m is the number of points in the grid, x i ,y i 、z i are the coordinates of each point in the grid; The specific process of step 6 is as follows: Step 6.1: Initialize the particle swarm: Each particle represents a point X in three-dimensional space i , whose position is X i =(x i ,y i ,z i ), its initial position and speed They are all randomly generated within the boundary of the point cloud data layer and obey a uniform distribution: Among them, U(bounds min ,bounds max ) indicates that the bounds min 、bounds max Uniformly randomly distributed between min and bounds max are the minimum and maximum bounds of the search space, respectively; U(-|bounds max -bounds min |,|bounds max -bounds min |) means in -|bounds max -bounds min |,|bounds max -bounds min |Uniformly randomly distributed between; Step 6.2: Evaluation function f(x i ) is particle X i The sum of the Euclidean distances to all points in the point cloud: Among them, P j is the position of the jth point in the point cloud, M is the total number of points in the point cloud, and ∥.∥2 represents the Euclidean distance. Step 6.3: Particle update: At each iteration t, the velocity of particle i Updated as follows: Among them, w is the inertia weight; c1 and c2 are acceleration coefficients; r1 and r2 are random numbers between [0,1]; p best,i is the best position experienced by the ith particle; g best is the global optimal position; At each iteration t, the position of particle i is Updated as follows: Step 6.4: Individual optimal and global optimal updates: like Update like For any i, update Step 6.5: Set the number of particles and the maximum number of iterations, repeat the update operation until the maximum number of iterations is reached, and finally return to the global optimal solution g best , which is used as the representative center point of the point cloud layer.

2. The method for identifying the tilt state of a hyperbolic cooling tower based on laser point cloud data according to claim 1, characterized in that: The specific process of step 1 is as follows: Firstly, an integrated application drone equipped with multiple lenses was used to obtain image data of the upper and peripheral structures of the cooling tower. The cooling tower was divided into 24 sections for up and down flight missions. The horizontal overlap was greater than 70%, and the vertical overlap was set by the time interval of the up and down flights of the integrated application drone. Then, a ground laser scanner was used to scan the point cloud data of the cooling tower. With the cooling tower to be tested as the center, multiple stations were set up around the cooling tower. The inside of the cooling tower was also scanned using a ground laser scanner.

3. The method for identifying the tilt state of a hyperbolic cooling tower based on laser point cloud data according to claim 1, characterized in that: The specific process of layering in step 5 is as follows: Traverse the entire point cloud data and calculate the lowest point Z min and the highest point Z max , determine the vertical range of the cooling tower point cloud |Z max -Z min |; Establish a hierarchical structure: set the layer height h and calculate the total number of layers N: Z min to Z max The range is evenly divided into N levels. For all point cloud data, according to the Z-axis coordinate value z of the point i Divide into corresponding levels, and for each level n, select points that meet the following conditions: Z min +nh≤z i <Z min +(n+1)h, n=0,1,2…N-1.

4. The method for identifying the tilt state of a hyperbolic cooling tower based on laser point cloud data according to claim 1, characterized in that: The specific process of step 7 is as follows: Calculate the representative center point (x i ,y i ,z i ) The coordinates of the center point are transformed into a coordinate system with this mean as the origin. The principal component analysis method is applied to the transformed data to solve the eigenvector v corresponding to the maximum eigenvalue of the covariance matrix. This eigenvector is the main direction of the data, that is, the first principal component, v = (v x ,v y ,v z ) is the direction vector of the fitted line; define the Z-axis vector as z(0, 0, 1), and use the vector dot product formula and vector length to calculate the cosine value of the angle between the fitted line and the Z-axis: Among them, v x 、v y 、v z The vector coordinates representing the main direction of the point cloud data; And the angle θ is obtained by arc cosine calculation: The angle θ represents the degree of deviation of the cooling tower from the vertical direction.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program data, and when the program data is executed by the processor, the steps of the method for identifying the inclination state of a hyperbolic cooling tower based on laser point cloud data described in claim 1 are implemented.