Damaged cantilever structure model obtaining method based on point cloud processing

Abstract

The invention discloses a damaged cantilever structure model obtaining method based on point cloud processing, and relates to the technical field of model construction. Scanning the damaged cantilever structure model to obtain two cantilever structure point clouds with different spatial positions and background point clouds as original point clouds; filtering the original point cloud to obtain an effective point cloud, and performing ICP registration on the effective point cloud based on the feature points to obtain registered point cloud data; and carrying out patch processing and smoothing processing on the registered point cloud data to obtain processed point cloud data, extracting a middle cross section point cloud outer contour curve of the processed point cloud data, and carrying out reverse modeling in combination with a cross section boundary curve and a boundary surface to obtain a damaged cantilever structure model. Through precise point cloud registration and optimization processing, the method can generate a high-precision damaged cantilever structure model, and provides accurate basic data for subsequent analysis and repair work.

Description

Technical Field

[0001] The present invention relates to the technical field of model building, and in particular to a method for acquiring a damaged cantilever structure model through point cloud processing. Background Art

[0002] Cantilever beam damage detection technology is mainly used to identify and evaluate defects in cantilever structures. Since cantilever beams are widely used in bridges, buildings and mechanical equipment, damage detection is crucial to ensure structural safety. Commonly used methods include vibration analysis, strain measurement and acoustic testing.

[0003] Vibration analysis identifies the location and severity of damage by monitoring the natural frequency and vibration mode changes of the beam. Vibration analysis does not require destruction of the structure and is suitable for real-time monitoring. It can detect various types of damage, such as cracks and corrosion. However, it has limited sensitivity to small-scale or minor damage and is greatly affected by factors such as environmental noise and temperature changes.

[0004] Strain measurement uses strain sensors to monitor stress changes in beams to help detect potential cracks or deformations. It can provide detailed strain data and accurately locate damage. Strain measurement requires the installation of strain sensors on the structure, which may affect the structure. In addition, the processing and analysis of strain data is relatively complex and requires professional equipment and technology.

[0005] Acoustic testing uses ultrasound or other sound wave technologies to detect internal defects. It is applicable to a variety of materials and structures. The damage location can be located through the sound wave propagation characteristics. However, acoustic testing may be affected by environmental noise and other interference factors, and requires professional equipment and technicians to operate, and the data analysis is complex.

[0006] In summary, the existing technology mainly requires actual survey of the derrick and damage detection based on the survey data. The survey data and analysis process that need to be processed are very complicated, and the data is easily affected by environmental interference, which has an adverse effect on the final damage detection results. Summary of the invention

[0007] In order to overcome the shortcomings of the above-mentioned prior art for damage detection of derricks, that the survey data and analysis process to be processed are very complicated, and the data are easily disturbed by the environment, which has an adverse effect on the final damage detection results, the main purpose of the present invention is to provide a method for obtaining a damaged cantilever structure model by point cloud processing.

[0008] To achieve the above object, the present invention adopts the following technical solution, a method for obtaining a damaged cantilever structure model by point cloud processing, comprising the following steps:

[0009] By scanning the damaged cantilever structure model, two cantilever structure point clouds and background point clouds with different spatial positions are obtained as original point clouds;

[0010] The original point cloud is filtered to obtain a valid point cloud, and the valid point cloud is ICP registered based on the feature points to obtain the registered point cloud data;

[0011] The registered point cloud data is processed by patching and smoothing to obtain the processed point cloud data. The outer contour curve of the middle section point cloud of the processed point cloud data is extracted, and the damaged cantilever structure model is obtained by combining the section boundary curve and boundary surface reverse modeling.

[0012] The step of obtaining a valid point cloud comprises the following steps:

[0013] Flip the original point cloud data along a certain axis to obtain flipped point cloud data;

[0014] A simulation algorithm is introduced to set a virtual cloth to fall from the top of the flipped point cloud until it completely covers the flipped point cloud surface. The flipped point cloud data is divided into ground point cloud and non-ground point cloud by distinguishing the shape of the cloth after the interaction between the cloth and the point cloud surface.

[0015] Smoothing the ground point cloud to obtain a continuous and smooth reference plane; and setting the normal direction of the ground point cloud as the z-axis;

[0016] Get the elevation difference between each point in the non-ground point cloud and the reference plane;

[0017] Assign the elevation difference to the z-axis coordinate of the non-ground point to complete the elevation normalization of the point cloud;

[0018] By setting the elevation value range, outliers with abnormal elevation are identified and removed to obtain a valid point cloud.

[0019] The step of obtaining the registered point cloud data comprises the following steps:

[0020] According to the quantity, quality and completeness of the point cloud, the valid point cloud is divided into reference point cloud and target point cloud;

[0021] Extracting feature points of the reference point cloud and the target point cloud, matching the feature points of the reference point cloud and the target point cloud, and obtaining a preliminary matching result of the point cloud;

[0022] Obtain three groups of non-contributing feature points for initial transformation estimation, and obtain the initial rotation matrix and initial translation vector;

[0023] Based on the preliminary matching results of the point cloud, combined with the initial rotation matrix and the initial translation vector, the ICP algorithm is used for precise registration to obtain the matching results and the registered point cloud data.

[0024] The obtaining of the matching result specifically includes the following steps:

[0025] For each point in the target point cloud, find the closest point in the reference point cloud as the corresponding point.

[0026] Get the error between the corresponding points minimized and update the rotation matrix and translation vector;

[0027] The initial rotation matrix and the initial translation vector are applied to the target point cloud until the error change is less than the set threshold, and a fine matching result is obtained.

[0028] A system for obtaining a damaged cantilever structure model by point cloud processing, comprising:

[0029] A point cloud acquisition module is used to obtain two cantilever structure point clouds with different spatial positions and a background point cloud as original point clouds by scanning the damaged cantilever structure;

[0030] The point cloud registration module is used to filter the original point cloud to obtain a valid point cloud, perform ICP registration on the valid point cloud based on feature points, and obtain the registered point cloud data;

[0031] The model building module is used to obtain the processed point cloud data by faceting and smoothing the registered point cloud data, extract the outer contour curve of the middle section point cloud of the processed point cloud data, and obtain the damaged cantilever structure model by combining the section boundary curve and boundary surface for reverse modeling.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] 1. Through precise point cloud registration and optimization processing, the present invention can generate a high-precision damaged cantilever structure model, providing accurate basic data for subsequent analysis and repair work. Through filtering and outlier removal, the error and noise in the point cloud data are effectively reduced, and the data quality is improved.

[0034] 2. Multiple steps in the present invention, such as point cloud segmentation, feature point matching and ICP registration, can quickly extract useful structural information from the original scan data through efficient algorithms and data processing processes, speeding up the entire modeling process, realizing automated processing, reducing manual operations, and improving processing efficiency, thereby reducing overall project costs.

[0035] 3. In the present invention, since the accuracy of data processing is improved, the generated model is more reliable and can be used for more accurate engineering analysis and design. It can handle complex and irregular cantilever structures and obtain an accurate model even in the case of damage or deformation.

[0036] 4. The present invention also provides an effective tool for reverse engineering, which can be used for the repair and transformation of old structures. By regularly obtaining structural models, the health of the structure can be more effectively monitored and potential problems can be discovered in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings described herein are used to provide further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute improper limitations on the present application.

[0038] Figure 1 The present invention Figure 1 Schematic diagram of the cantilever beam damage rapid detection process based on point cloud processing;

[0039] Figure 2 is a schematic diagram of CSF of the present invention;

[0040] Figure 3 It is a schematic diagram of coordinate axis transformation of the present invention;

[0041] Figure 4 It is a schematic diagram of the rotation and translation of the target point cloud to the reference point cloud of the present invention;

[0042] Figure 5 It is the flow chart of ICP algorithm of the present invention;

[0043] Figure 6 It is the reverse modeling flow chart of the present invention;

[0044] Figure 7 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0045] Cantilever beam damage detection technology is mainly used to identify and evaluate defects in cantilever structures. Since cantilever beams are widely used in bridges, buildings and mechanical equipment, damage detection is crucial to ensure structural safety. Existing technologies mainly require actual surveys of the derrick and damage detection based on survey data. The survey data and analysis process that need to be processed are very complex, and the data is easily affected by environmental interference, which has an adverse effect on the final damage detection results.

[0046] In order to overcome the shortcomings of the above-mentioned prior art in performing damage detection on the derrick, the survey data and analysis process to be processed are very complicated, and the data is easily disturbed by the environment, which has an adverse effect on the final damage detection result, the main purpose of the present invention is to provide a method for obtaining a damaged cantilever structure model by point cloud processing, comprising the following steps:

[0047] By scanning the damaged cantilever structure model, two cantilever structure point clouds with different spatial positions and background point clouds are obtained as original point clouds;

[0048] The original point cloud is filtered to obtain a valid point cloud, and the valid point cloud is ICP-registered based on the feature points to obtain the registered point cloud data;

[0049] The registered point cloud data is processed by patching and smoothing to obtain the processed point cloud data. The outer contour curve of the middle section point cloud of the processed point cloud data is extracted, and the damaged cantilever structure model is obtained by combining the section boundary curve and boundary surface reverse modeling.

[0050] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0051] Embodiment 1:

[0052] The method for obtaining the damaged cantilever structure model based on the point cloud processing of the present application first obtains the point cloud data by scanning the damaged cantilever structure with a laser device to obtain two cantilever structure point clouds with different spatial positions and a background point cloud. The process of this method is as follows Figure 1 and Figure 7 , as shown:

[0053] After obtaining the point cloud data, it is processed. The filtering process in the specific processing process is as follows:

[0054] The original point cloud is flipped, and a piece of cloth is used to simulate the falling process from above the point cloud to completely cover the flipped point cloud surface. The interaction between the points on the simulated cloth and the target point cloud is analyzed to determine the cloth shape, which is used as the basis for classifying ground points and non-ground points, such as Figure 2 As shown. The obtained point cloud is segmented to obtain ground points and non-ground points. The filtered ground points are smoothed, and the ground point cloud is used as the reference plane. The normal direction of the ground point cloud is set to the z-axis, and the elevation between the non-ground point and the ground is calculated. The calculated elevation value is assigned to the z-axis. Thus, the point cloud elevation normalization is completed. At this time, the elevation value of the point cloud is consistent with the positive direction of the z-axis. Finally, the outliers are removed by setting the elevation value range to extract the required point cloud.

[0055] Register the processed point clouds: Multi-view point cloud registration requires adjusting the two point clouds to the same global coordinate system through rotation and translation of coordinates.

[0056] The transformation of points p and p' in the two coordinate systems can be performed using the following formula:

[0057] p′=Rp+T (1);

[0058] Where R represents the rotation matrix, expressed as:

[0059]

[0060] In the matrix R, α, β, and γ represent the rotation angles of the transformation point around the three coordinate axes x, y, and z, respectively, and T is the translation vector T = [t x ,t y ,t z ] T .t x ,t y ,t z Respectively represent the distances of translation of the transformation point along the three coordinate axes, such as Figure 3 shown.

[0061] Determine the target point cloud and the reference point cloud. The selection criteria for the two are generally as follows: the point cloud with more points is used as the reference point cloud, and the point cloud with less points is used as the target point cloud; the point cloud with good data quality is used as the reference point cloud, and the point cloud with poor data quality is used as the target point cloud; the point cloud with more complete scans is used as the reference point cloud, and the point cloud with more missing parts is used as the target point cloud. In simple terms, the target point cloud is used as a supplement to the reference point cloud.

[0062] Define a point cloud set P = {p1, p2, ..., p n}, is the reference point cloud set, point cloud set Q = {q1,q2,...,q n}, is the target point cloud set. Studying the problem of point cloud Q approaching point cloud P is equivalent to rotating and translating point cloud Q a certain distance to point cloud P. Figure 4 As shown, the error term for each pair of points can be defined as:

[0063] e i =p i -(Rq i +t) (3);

[0064] It can be further described as the problem of optimizing the least squares objective function error, R and t when the sum of squares of the following objective function errors is extremely small:

[0065]

[0066] Among them, the functional relationship between R and t; assuming that R is a constant, let get:

[0067]

[0068] Calculate the centroid of points P and Q:

[0069]

[0070] The above formula can be used to calculate the translation difference of point sets P and Q relative to the center of mass:

[0071]

[0072] The translation vector can be obtained:

[0073] μ p -Rμ q =t (8);

[0074] Substituting the above relationship into equation (4), we can obtain:

[0075]

[0076] That is, find R when the following objective function value is minimized:

[0077]

[0078] The expanded terms in the formula are as follows:

[0079]

[0080] Except for the second term, the rest of the formula is independent of the parameter R and can be ignored when solving the extreme value. Instead, the R value at the minimum value of the following formula is calculated:

[0081]

[0082] make Let S = QWP T Perform SVD decomposition on S:

[0083] S=U∑V T (13);

[0084] U and V are orthogonal matrices, and ∑ is a diagonal matrix composed of singular values.

[0085] The rotation matrix R is obtained as follows:

[0086] R=VU T (14);

[0087] Use three sets of feature points to roughly align the point cloud. The distance between the three sets of feature points is relatively far, and try to ensure that they are not distributed in a straight line. The distance between the three sets of feature points in the z direction can be relatively large. Based on the rough alignment, the ICP algorithm is used to make the two point clouds highly consistent. The ICP algorithm flow chart is as follows Figure 5 As shown in Figure 2. Local details after registration.

[0088] The extracted point cloud data is combined with traditional modeling and rapid modeling methods to complete the 3D model reconstruction of the target object. First, the point cloud is processed by patching to provide clear topological information; secondly, the patched point cloud is smoothed; then the outer contour curve of the middle section point cloud is extracted; the 3D reverse model reconstruction is completed through the section boundary curve and boundary surface. Figure 6 The reconstructed model is then subjected to damage detection and size measurement. Figure 7 .

[0089] Embodiment 2:

[0090] In order to accurately detect the damage of the cantilever structure, laser scanning technology is used to obtain high-precision point cloud data, and the three-dimensional model is reconstructed through the reverse modeling method, and damage detection and modal analysis are performed. This embodiment shows how to accurately identify the damage of the cantilever beam and calculate the natural frequency of the structure through point cloud data processing and analysis. Specifically, it includes the following contents:

[0091] The laser scanning instrument is used to scan the damaged cantilever structure and obtain the point cloud data of two cantilever beams with different spatial positions and the background point cloud. Assuming that the damaged area of ​​the cantilever beam is located in the middle of the beam, the scanning range covers the entire cantilever structure and its surrounding environment. When scanning, the scanning accuracy is 0.1mm; the scanning range is 0-2m, which is the full length of the cantilever beam; the scanning time is about 5 minutes per scan.

[0092] After acquiring the point cloud data, data preprocessing is first performed. The specific processing process is: flip the point cloud and simulate the falling of cloth, flip the point cloud data upside down, and cover the point cloud surface with simulated cloth to simulate the falling process. This process helps to determine the classification of ground points and non-ground points. Then the point cloud is segmented and smoothed, and the ground points and non-ground points are classified by analyzing the interaction between the point cloud and the cloth. The ground points are smoothed, and the normal direction of the ground point cloud is set to the Z axis. The elevation between the non-ground points and the ground is calculated and normalized. After removing outliers from the normalized point cloud data, outliers are removed according to the preset elevation value range, and valid point clouds are extracted. The filtered point cloud data includes the number of ground points: approximately 20,000 points and the number of non-ground points: approximately 80,000 points

[0093] The point cloud data of the two scanning perspectives are registered to the same global coordinate system using the multi-view point cloud registration method. The target point cloud QQ is aligned with the reference point cloud PP using the rotation matrix RR and the translation vector TT. The rotation matrix and the translation vector are adjusted by optimizing the least squares objective function to minimize the error between the point pairs. The coordinate system is aligned, and the error between the target point cloud QQ and the reference point cloud PP is achieved by minimizing the objective function. The rotation matrix RR and the translation vector TT are solved by SVD singular value decomposition. After the rough registration, the ICP algorithm is used for further precise alignment. After the error is minimized, the coincidence degree between the target point cloud and the reference point cloud reaches 97%.

[0094] After the point cloud data is aligned, reverse modeling is performed to reconstruct the three-dimensional model of the cantilever beam. The specific processing process is: the point cloud data is processed into patches to provide topological information. The patched point cloud data is smoothed to remove noise points. The modeling process is further optimized by extracting the outer contour curve of the cross-sectional point cloud. The geometric dimensions of the cantilever beam are confirmed, including a length of 801mm, a width of 39.90mm, and a height of 8.00mm. The error between the model and the actual size is within 0.1%. The point cloud model is used to calculate the damaged area of ​​the cantilever beam, and the damage is identified by comparing the geometric changes of the model with the known damaged area.

[0095] By performing modal analysis on the reconstructed 3D model, the natural frequency of the cantilever beam is calculated and compared with the frequency measured in the experiment. Comparison of the natural frequency of the reconstructed model with the natural frequency of the physical experiment:

[0096] It can be seen from the above table that the natural frequency reconstructed based on the point cloud has a high consistency with the physical experimental results, and the error is controlled within 8%, indicating that this method can effectively reflect the dynamic characteristics of the damaged cantilever beam in practical applications.

[0097] 6. Dimension detection and error analysis

[0098] The reconstructed 3D model is dimensionally inspected and the actual measured values ​​are compared with the dimensions obtained based on the point cloud model. The results are as follows:

[0099] Table 1 shows the actual measured values ​​and the specific size data obtained based on the point cloud model.

[0100] Geometric parameters Actual measurement (mm) Measurement based on point cloud model (mm) error(%) length 801.00 800.97 0.00375 width 39.90 39.41 1.23 high 8.00 7.95 0.60 Slot width 50.66 50.47 0.38 Groove depth 1.70 1.69 0.59

[0101] 7. Conclusion

[0102] The point cloud data processing and reverse modeling method based on laser scanning has significant advantages in the damage detection of cantilever beams. This method can not only accurately obtain the three-dimensional shape of the structure, but also calculate the damage location and size change through a high-precision model. The accuracy of damage analysis is ensured through efficient data processing and registration algorithms. Finally, the reliability of the point cloud model in dynamic characteristics analysis is verified by comparing the natural frequencies.

[0103] Advantages of this approach include:

[0104] High precision: The point cloud scanning accuracy reaches the millimeter level, which can capture tiny damages.

[0105] Fast acquisition: Scanning of large structures can be completed in a short time.

[0106] Completeness: Covering complex areas and reducing data missing.

[0107] Visualization: The generated 3D model can be intuitively displayed through VR and AR technology.

[0108] The error between the reconstructed model size and the actual measured size is shown in Table 1.

[0109] The natural frequency and mode of the damaged cantilever beam are calculated by reconstructing the model. The error between the natural frequency of the point cloud reconstruction model and the natural frequency obtained by the test is shown in Table 2.

[0110] The point cloud reverse modeling method based on laser scanning shows significant advantages in cantilever beam damage detection. When acquiring data, it can accurately capture tiny damage details with millimeter or even sub-millimeter scanning accuracy to ensure the accuracy of damage reproduction. The fast data acquisition capability enables the scanning of large scenes or objects to be completed within a few minutes to a few hours, greatly improving work efficiency. The comprehensive coverage feature ensures the scanning integrity of complex or inaccessible areas and reduces data loss. The efficient data processing software simplifies the post-process and supports point cloud processing and model generation. In addition, the generated three-dimensional model is intuitively displayed through VR and AR technology, which enhances the visualization effect.

[0111] Embodiment 2:

[0112] The following is a specific embodiment of the method for obtaining a damaged cantilever structure model based on the above point cloud processing. This embodiment provides specific data and steps. The method for obtaining a damaged cantilever structure model based on point cloud processing specifically includes the following steps:

[0113] First, the original point cloud data is obtained;

[0114] A damaged cantilever structure is scanned using a laser scanner to obtain original point cloud data.

[0115] Assume that the acquired point cloud contains 10,000,000 points, and the spatial distribution of the point cloud includes the cantilever structure itself and the background environment, such as the ground, walls, etc.

[0116] Cantilever structure point cloud: contains 2,000,000 points and is located above in space.

[0117] Background point cloud: includes background ground point cloud and other environmental elements.

[0118] Then the original point cloud data is filtered.

[0119] Use a common point cloud filtering algorithm such as voxel grid filtering to downsample the original point cloud data, reduce the number of point clouds, and improve processing efficiency. Assume that after the original point cloud is filtered, the number of points is reduced to 2,000,000 points.

[0120] According to the filtering results, the background noise was removed, the main cantilever structure part was retained, and the effective point cloud data was selected.

[0121] Then ICP is used for coarse and fine registration. The specific coarse registration is as follows:

[0122] Point cloud segmentation: divide the valid point cloud into reference point cloud and target point cloud.

[0123] The reference point cloud contains the main part of the cantilever structure and contains 1,500,000 points.

[0124] The target point cloud, including the cantilever structure and its surroundings, contains 500,000 points.

[0125] Feature point extraction: extract feature points of the reference point cloud and the target point cloud. Assuming that the normal vector feature descriptor is used, 1,000 key points are extracted as feature points through the normal information of the point cloud.

[0126] Preliminary matching, using a feature point-based matching algorithm, such as a nearest neighbor search-based algorithm, to perform point cloud matching and obtain preliminary registration results.

[0127] The specific details of fine registration are:

[0128] Initial transformation estimation: Select three groups of non-contributing feature points from the initial matched feature points, and make preliminary rotation matrix and translation vector estimates. Assume that the rotation matrix is:

[0129]

[0130] The initial translation vector is:

[0131] The ICP algorithm is used to align the reference point cloud and the target point cloud, and the rotation matrix and displacement vector are iteratively updated until the error change is less than the set threshold, such as 0.001. After the ICP algorithm is iterated 20 times,

[0132] The registration results are as follows:

[0133]

[0134] Final rotation matrix:

[0135]

[0136] Final translation vector:

[0137]

[0138] Through this precise registration, the error between the target point cloud and the reference point cloud is significantly reduced.

[0139] Then the registered electric cloud is processed into patches and smoothed, specifically:

[0140] The patch processing is to convert the registered point cloud into triangular patches through the triangulation algorithm to construct the surface model of the cantilever structure.

[0141] Assume that this process converts the point cloud into 1,000,000 triangle patches.

[0142] Smoothing is to use Gaussian filtering to smooth the point cloud surface, remove small fluctuations caused by noise and scanning errors, and finally obtain a smooth point cloud surface.

[0143] The outer contour extraction of the cross-section point cloud is to extract the outer contour curve of the middle cross-section point cloud of the cantilever structure. Assume that the extracted outer contour curve contains 200 key points.

[0144] Reverse modeling combines the cross-sectional outer contour curve and boundary surface and uses reverse modeling techniques such as NURBS modeling or Bezier curve fitting to reconstruct the geometric model of the damaged cantilever structure.

[0145] Through the above steps, a 3D model of the damaged cantilever structure is finally generated. The 3D model includes surface details, features of the damaged part, and its corresponding geometric shape, which can provide a basis for subsequent analysis and simulation.

[0146] In this embodiment, through detailed point cloud processing steps, including filtering, ICP registration, patching, smoothing, cross-section extraction and reverse modeling, the 3D model of the damaged cantilever structure was successfully reconstructed. This model can not only accurately reflect the geometric shape of the cantilever structure, but also provide strong support for subsequent structural analysis, damage assessment and repair.

[0147] It should be noted that, in the present invention, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0148] The above embodiments are merely examples of the present invention and do not limit the protection scope of the present invention. All designs that are the same or similar to the present invention fall within the protection scope of the present invention.

Claims

1. A method for obtaining a damaged cantilever structure model by point cloud processing, characterized in that: The following steps are involved: By scanning the damaged cantilever structure model, two cantilever structure point clouds and background point clouds with different spatial positions are obtained as original point clouds; The original point cloud is filtered to obtain a valid point cloud. Based on the feature points of the valid point cloud, the valid point cloud is ICP-registered to obtain the registered point cloud data. The registered point cloud data is processed by patching and smoothing to obtain the processed point cloud data. The outer contour curve of the middle section point cloud of the processed point cloud data is extracted, and the damaged cantilever structure model is obtained by combining the section boundary curve and boundary surface reverse modeling.

2. The method for obtaining a damaged cantilever structure model by point cloud processing according to claim 1, characterized in that: The step of obtaining a valid point cloud comprises the following steps: Flip the original point cloud data along the set axis to obtain flipped point cloud data; By setting the virtual cloth to fall from the top of the flipped point cloud until it completely covers the flipped point cloud surface, the flipped point cloud data is divided into ground point cloud and non-ground point cloud by distinguishing the cloth shape after the cloth interacts with the point cloud surface; Smoothing the ground point cloud to obtain a continuous and smooth reference plane; and setting the normal direction of the ground point cloud as the z-axis; Get the elevation difference between each point in the non-ground point cloud and the reference plane; Assign the elevation difference to the z-axis coordinate of the non-ground point to complete the elevation normalization of the point cloud; By setting the elevation value range, outliers with abnormal elevation are identified and removed to obtain a valid point cloud.

3. The method for obtaining a damaged cantilever structure model by point cloud processing according to claim 1, characterized in that: The step of obtaining the registered point cloud data comprises the following steps: According to the attributes of the point cloud, the valid point cloud is divided into reference point cloud and target point cloud; Extracting feature points of the reference point cloud and the target point cloud, matching the feature points of the reference point cloud and the target point cloud, and obtaining a preliminary matching result of the point cloud; Obtain three groups of non-contributing feature points for initial transformation estimation, and obtain the initial rotation matrix and initial translation vector; Based on the preliminary matching results of the point cloud, combined with the initial rotation matrix and the initial translation vector, the ICP algorithm is used for precise registration to obtain the matching results, that is, the registered point cloud data.

4. The method for acquiring a damaged cantilever structure model by point cloud processing according to claim 3, characterized in that: The obtaining of the matching result specifically includes the following steps: For each point in the target point cloud, find the nearest point in the reference point cloud as the corresponding point; Get the error between the corresponding points minimized and update the rotation matrix and translation vector; The initial rotation matrix and the initial translation vector are applied to the target point cloud until the error change is less than the set threshold, and a fine matching result is obtained.

5. A system for obtaining a damaged cantilever structure model by point cloud processing, characterized in that: include: A point cloud acquisition module is used to obtain two cantilever structure point clouds and background point clouds at different spatial positions as original point clouds by scanning the damaged cantilever structure; The point cloud registration module is used to filter the original point cloud to obtain a valid point cloud, and perform ICP registration on the valid point cloud based on the feature points of the valid point cloud to obtain the registered point cloud data; The model building module is used to obtain the processed point cloud data by faceting and smoothing the registered point cloud data, extract the outer contour curve of the middle section point cloud of the processed point cloud data, and obtain the damaged cantilever structure model by combining the section boundary curve and boundary surface for reverse modeling.

6. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

7. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method described in any one of claims 1 to 4 is implemented.

Images