Non-contact deformation monitoring device and monitoring method for large-span spatial steel structure

Abstract

The present application relates to the technical field of steel structure deformation monitoring, and specifically to a non-contact deformation monitoring device and monitoring method for large-span space steel structures. The method includes: collecting building point cloud data and building surface grayscale images of large-span space steel structures, dividing supervoxels, marking target supervoxels, and determining the normal vector deformation degree and building deformation degree of the target supervoxel; determining the fitting surface distance of the point cloud data, and determining the data smoothness of the point cloud data; denoising the building point cloud data, obtaining the denoised building point cloud data, and obtaining the non-contact deformation monitoring results for large-span space steel structures based on the denoised building point cloud data and the building surface grayscale image. The present application solves the problem that the selection of the filter radius of the denoising algorithm is unreasonable, resulting in poor denoising effect of the point cloud data, and insufficient accuracy of the deformation monitoring results of large-span space steel structures.

Description

Technical Field

[0001] The present application relates to the technical field of steel structure deformation monitoring, and in particular to a non-contact deformation monitoring device and a monitoring method for large-span spatial steel structures. Background Art

[0002] During the use of large-span steel structures such as bridges and buildings, they are often affected by factors such as the external environment and load changes. Real-time deformation monitoring is required to ensure the structural stability and safety of large-span steel structures. Generally, 3D laser scanning technology is used to monitor the deformation of large-span steel structures, but laser scanners are often affected by instrument errors, errors caused by the reflection characteristics of the target object, and external environmental errors. The collected point cloud data will contain noise data, which needs to be denoised.

[0003] Generally, denoising algorithms such as Gaussian filtering, median filtering and mean filtering are used to denoise point cloud data. However, the denoising effect is often affected by the value of the filter radius of the denoising algorithm. When the filter radius of the denoising algorithm is too large, the point cloud data will be over-smoothed, resulting in information loss of large-span spatial steel structures; when the filter radius of the denoising algorithm is too small, the noise data in the point cloud data cannot be completely removed, resulting in information distortion of large-span spatial steel structures. The denoising effect of point cloud data will affect the accuracy of deformation monitoring results of large-span spatial steel structures. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a non-contact deformation monitoring device and monitoring method for a large-span spatial steel structure. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present application provides a non-contact deformation monitoring method for a large-span spatial steel structure, the method comprising the following steps:

[0006] Collecting building point cloud data and building surface grayscale images of a large-span spatial steel structure building, and dividing the building point cloud data into a first preset number of supervoxels;

[0007] Any supervoxel divided from the building point cloud data is recorded as a target supervoxel, and a point cloud surface is fitted to the target supervoxel to obtain a fitting surface of the target supervoxel. The normal vector deformation degree of the target supervoxel is determined according to the difference between the normal vectors of the projection points of the point cloud data contained in the target supervoxel and other supervoxels at the corresponding fitting surface. The building deformation degree of the target supervoxel is determined by combining the difference between the Gaussian curvatures of all the point cloud data contained in the target supervoxel at the projection points of the fitting surface of the target supervoxel;

[0008] Determine the fitting surface distance of the point cloud data contained in the target supervoxel according to the relative position of the point cloud data contained in the target supervoxel and the fitting surface of the target supervoxel, and determine the data smoothness of the point cloud data contained in the target supervoxel according to the difference between the building deformation degree of the supervoxel where the point cloud data is located and the fitting surface distance of all the point cloud data contained in the supervoxel where the point cloud data is located, as well as the normal vector deformation degree of the target supervoxel;

[0009] The building point cloud data is denoised according to the data smoothness of the point cloud data to obtain the denoised building point cloud data, and the non-contact deformation monitoring results for large-span spatial steel structures are obtained based on the denoised building point cloud data and the building surface grayscale image.

[0010] Furthermore, the process of determining the normal vector deformation degree of the target supervoxel is as follows:

[0011] Determine the projection point of the centroid of all point cloud data contained in the target supervoxel on the fitting surface of the target supervoxel, and record the normal vector of the projection point on the fitting surface of the target supervoxel as the normal vector of the target supervoxel;

[0012] The average of the similarities between the normal vector of the target supervoxel and the normal vector of each supervoxel divided by the building point cloud data is recorded as the normal vector similarity of the target supervoxel;

[0013] According to the normal vector similarity of the target supervoxel, the normal vector deformation degree of the target supervoxel is determined, and the normal vector deformation degree of the target supervoxel is negatively correlated with the normal vector similarity of the target supervoxel.

[0014] Furthermore, the building deformation degree of the target supervoxel is determined by combining the difference between the Gaussian curvatures of the projection points of the fitting surface of the target supervoxel with all the point cloud data contained in the target supervoxel. The specific process of obtaining the difference is as follows:

[0015] The linear normalized value of the standard deviation of the Gaussian curvature of the projection points of all point cloud data contained in the target supervoxel on the fitting surface of the target supervoxel is recorded as the curvature difference of the target supervoxel;

[0016] The building deformation degree of the target supervoxel is obtained according to the normal vector deformation degree and the curvature difference of the target supervoxel.

[0017] Furthermore, the building deformation degree of the target supervoxel is obtained according to the normal vector deformation degree and curvature difference of the target supervoxel. The specific process of obtaining is:

[0018] The average of the normal vector deformation and curvature difference of the target supervoxel is recorded as the building deformation of the target supervoxel.

[0019] Furthermore, the construction process of the fitting surface distance of the point cloud data contained in the target supervoxel is as follows:

[0020] Any point cloud data contained in the target supervoxel is recorded as the target point cloud data, the projection point of the target point cloud data on the fitting surface of the target supervoxel is recorded as the projection point of the target point cloud data, and the Euclidean distance between the projection point of the target point cloud data and the target point cloud data is recorded as the fitting surface distance of the target point cloud data.

[0021] Furthermore, the data smoothness of the point cloud data contained in the target supervoxel is determined according to the difference between the building deformation degree of the supervoxel where the point cloud data is located and the fitting surface distance of all the point cloud data contained in the supervoxel where the point cloud data is located, and the normal vector deformation degree of the target supervoxel. The specific process of obtaining the smoothness of the point cloud data contained in the target supervoxel is as follows:

[0022] The difference between the fitting surface distances of the target point cloud data and other point cloud data contained in the target supervoxel is recorded as the fitting surface distance difference between the target point cloud data and other point cloud data contained in the target supervoxel, and the linear normalized value of the mean of the fitting surface distance differences between the target point cloud data and all other point cloud data contained in the target supervoxel is recorded as the noise difference distance of the target point cloud data;

[0023] The data smoothness of the target point cloud data is determined according to the noise difference distance of the target point cloud data and the building deformation degree of the supervoxel where the target point cloud data is located.

[0024] Furthermore, the data smoothness of the target point cloud data is determined according to the noise difference distance of the target point cloud data and the building deformation degree of the supervoxel where the target point cloud data is located. The specific process of obtaining the smoothness of the target point cloud data is as follows:

[0025] The linear normalized value of the ratio of the noise difference distance of the target point cloud data to the building deformation degree of the supervoxel where the target point cloud data is located is recorded as the data smoothness of the target point cloud data.

[0026] Further, the building point cloud data is denoised according to the data smoothness of the point cloud data to obtain the denoised building point cloud data. The specific process of obtaining is:

[0027] The rounded value of the product of the data smoothness of the point cloud data and the second adjustment parameter is recorded as the first product of the point cloud data, and the sum of the first product of the point cloud data and the first adjustment parameter is recorded as the adjustment filter radius of the point cloud data;

[0028] The adjusted filter radius of the point cloud data is used as the value of the radius of the filter window of the point cloud data, and the building point cloud data is denoised using mean filtering to obtain the denoised building point cloud data.

[0029] Further, the non-contact deformation monitoring result for the large-span space steel structure is obtained based on the denoised building point cloud data and the building surface grayscale image. The specific process of obtaining is:

[0030] Project the denoised building point cloud data onto the grayscale image of the building surface to obtain the building depth map;

[0031] Collecting building point cloud data and building surface grayscale images of the large-span space steel structure building before commissioning, obtaining a building depth map of the large-span space steel structure building before commissioning, and recording the obtained building depth map as a standard building depth map;

[0032] Subtract the building depth map from the standard building depth map to obtain a building deformation map;

[0033] When the pixel value of a pixel point in the building deformation map is not zero, it is determined that the pixel point is deformed at the corresponding position of the large-span steel structure building, and the deformation amount is the pixel value of the pixel point; when the pixel value of the pixel point in the building deformation map is zero, it is determined that the pixel point is not deformed at the corresponding position of the large-span steel structure building.

[0034] In a second aspect, an embodiment of the present application provides a non-contact deformation monitoring device for a large-span spatial steel structure, wherein the deformation monitoring device includes: a deformation data acquisition module, a deformation degree determination module, a data smoothness determination module, and a deformation monitoring result acquisition module.

[0035] A deformation data acquisition module, used to acquire building point cloud data and building surface grayscale images of a large-span spatial steel structure building, and divide the building point cloud data into a first preset number of supervoxels;

[0036] A deformation degree determination module is used to record any supervoxel divided from the building point cloud data as a target supervoxel, perform point cloud surface fitting on the target supervoxel, obtain the fitted surface of the target supervoxel, determine the deformation degree of the normal vector of the target supervoxel according to the difference between the normal vectors of the point cloud data contained in the target supervoxel and other supervoxels at the projection points of the corresponding fitted surface, and determine the building deformation degree of the target supervoxel in combination with the difference between the Gaussian curvatures of all the point cloud data contained in the target supervoxel at the projection points of the fitted surface of the target supervoxel;

[0037] a data smoothness determination module, for determining the fitting surface distance of the point cloud data contained in the target supervoxel according to the relative position of the point cloud data contained in the target supervoxel and the fitting surface of the target supervoxel, and determining the data smoothness of the point cloud data contained in the target supervoxel according to the difference between the building deformation degree of the supervoxel where the point cloud data is located and the fitting surface distance of all the point cloud data contained in the supervoxel where the point cloud data is located;

[0038] The deformation monitoring result acquisition module is used to denoise the building point cloud data according to the data smoothness of the point cloud data, obtain the denoised building point cloud data, and obtain the non-contact deformation monitoring results for large-span spatial steel structures based on the denoised building point cloud data and the grayscale image of the building surface.

[0039] It can be seen from the above embodiments that the non-contact deformation monitoring device and monitoring method for large-span spatial steel structures provided in the embodiments of the present application have at least the following beneficial effects:

[0040] The present application starts with deformation monitoring of large-span spatial steel structures, including overall deformation and local deformation. First, according to the characteristic that the direction of the normal of the surface at the position of overall deformation is greatly different from the direction of the normal of the surface at the position where no deformation occurs, the deformation degree of the normal vector of the target supervoxel is determined. Further, combined with the characteristic that the concavity and convexity of the surface at different positions within the local area of ​​the local deformation position are inconsistent, the possibility of deformation problems at the corresponding positions of the large-span spatial steel structure building is evaluated, and the building deformation degree of the supervoxel is obtained. Furthermore, according to the characteristic that the values ​​of the noise data contained in the building point cloud data are randomly distributed and cannot fit the fitting surface of the target supervoxel tightly, the fitting surface of the point cloud data is evaluated. The difference between the fitting surface distance and the fitting surface distance of other point cloud data contained in the supervoxel where the point cloud data is located is evaluated. The evaluation result is also an evaluation of the possibility that the point cloud data contained in the supervoxel is interfered by noise, and the data smoothness of the point cloud data contained in the target supervoxel is further obtained; finally, the building point cloud data is denoised according to the data smoothness of the point cloud data to obtain the denoised building point cloud data, and the non-contact deformation monitoring results of large-span spatial steel structures are obtained based on the denoised building point cloud data and the grayscale image of the building surface, so as to solve the problem that the unreasonable selection of the filter radius of the denoising algorithm makes the denoising effect of the point cloud data poor, resulting in insufficient accuracy of the deformation monitoring results of the large-span spatial steel structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0042] Figure 1 A flowchart of the steps of a non-contact deformation monitoring method for a large-span spatial steel structure provided in one embodiment of the present application;

[0043] Figure 2A schematic diagram of an overall deformation provided for an embodiment of the present application;

[0044] Figure 3 A schematic diagram of local deformation provided for an embodiment of the present application;

[0045] Figure 4 A schematic structural diagram of a non-contact deformation monitoring device for a large-span spatial steel structure provided in one embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to further explain the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the non-contact deformation monitoring device and monitoring method for large-span spatial steel structures proposed in the present application, in combination with the accompanying drawings and preferred embodiments, are described in detail as follows, including their specific implementation methods, structures, features and effects. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0047] Unless otherwise specified and limited, terms such as "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a circuit structure, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such articles or devices. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the article or device including the element. In addition, the term "and\or" used herein includes any and all combinations of one or more related listed items. All technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of this application.

[0048] The specific scheme of the non-contact deformation monitoring device and monitoring method for large-span spatial steel structures provided by the present application is described in detail below with reference to the accompanying drawings.

[0049] See also Figure 1 , which shows a flowchart of a non-contact deformation monitoring method for a large-span spatial steel structure provided by an embodiment of the present application, the method comprising the following steps:

[0050] S001: Collect building point cloud data and building surface grayscale images of a large-span spatial steel structure building, and divide the building point cloud data into a first preset number of supervoxels.

[0051] A laser scanner and a high-definition camera are fixed at the same position on a large-span steel structure building. The laser scanner is used to collect the three-dimensional point cloud data of the large-span steel structure building to obtain the point cloud data of the building. The high-definition camera is used to collect the image of the large-span steel structure building to obtain the surface image of the building. The building image is an RGB image, and the building image is converted into a grayscale image to obtain the grayscale image of the building surface.

[0052] The deformation of large-span space steel structures includes two types: overall deformation and local deformation. Overall deformation refers to the deformation of the overall shape of the large-span space steel structure. The normal line of the surface at the position where the deformation occurs is significantly different from the normal line of the surface at the position where the deformation does not occur. The overall deformation diagram is shown in the figure below. Figure 2 As shown in the diagram of overall deformation, ① is a steel structure without deformation, and ② is a steel structure with overall deformation. Local deformation is the deformation of a local area of ​​a large-span spatial steel structure. Only the concavity and convexity of the curved surface at different positions in the local area are inconsistent. The diagram of local deformation is shown in Figure 3 In the local deformation schematic diagram, ③ is a steel structure without deformation, and ④ is a steel structure with local deformation.

[0053] The building point cloud data is divided using a supervoxel clustering point cloud segmentation algorithm PCL to obtain a first preset number of supervoxels.

[0054] It can be understood that each supervoxel corresponds to a part of a large-span spatial steel structure building, and each supervoxel is point cloud data.

[0055] In this embodiment, the value of the first preset number is 20; using the supervoxel clustering point cloud segmentation algorithm PCL to divide the point cloud data into the first preset number of supervoxels is a well-known technology and will not be described in detail.

[0056] At this point, the supervoxel and building surface grayscale images are obtained.

[0057] S002: any supervoxel divided from the building point cloud data is recorded as a target supervoxel, and a point cloud surface is fitted for the target supervoxel to obtain a fitted surface of the target supervoxel. The normal vector deformation degree of the target supervoxel is determined based on the difference between the normal vectors of the point cloud data contained in the target supervoxel and other supervoxels at the projection points of the corresponding fitted surfaces. The building deformation degree of the target supervoxel is determined by combining the difference between the Gaussian curvatures of all the point cloud data contained in the target supervoxel at the projection points of the fitted surface of the target supervoxel.

[0058] Any supervoxel divided from the building point cloud data is recorded as the target supervoxel, and the target supervoxel is fitted with a point cloud surface using a NURBS curve to obtain a fitting surface of the target supervoxel. The fitting surface of the target supervoxel can represent the surface geometric structure distribution of the building position corresponding to the target supervoxel.

[0059] Among them, using NURBS curves to fit the point cloud data to the point cloud surface and obtaining the fitting surface of the point cloud data is a well-known technology and will not be described in detail.

[0060] The normal vector deformation degree of the target supervoxel is determined according to the difference between the normal vectors of the projection points of the point cloud data included in the target supervoxel and other supervoxels at the corresponding fitting surfaces.

[0061] Determine the projection point of the centroid of all point cloud data contained in the target supervoxel on the fitting surface of the target supervoxel, and record the normal vector of the projection point on the fitting surface of the target supervoxel as the normal vector of the target supervoxel. The normal vector of any supervoxel divided by the building point cloud data can be obtained in the same way.

[0062] The average of the similarities between the normal vector of the target supervoxel and the normal vector of each supervoxel divided from the building point cloud data is recorded as the normal vector similarity of the target supervoxel.

[0063] Among them, this embodiment selects cosine similarity as the similarity measure between normal vectors. The implementer can use other methods in the prior art such as Pearson correlation coefficient to measure the similarity between normal vectors, and this application does not make any special restrictions. In the process of calculating the normal vector similarity of the target supervoxel, it is also necessary to calculate the similarity between the normal vector of the target supervoxel and the target supervoxel itself.

[0064] According to the normal vector similarity of the target supervoxel, the normal vector deformation degree of the target supervoxel is determined. The normal vector deformation degree of the target supervoxel is negatively correlated with the normal vector similarity of the target supervoxel.

[0065] Preferably, as an embodiment of the present application, a linearly normalized value of the ratio of the number 1 to the normal vector similarity of the target supervoxel is recorded as the normal vector deformation degree of the target supervoxel.

[0066] It should be noted that, in the process of calculating the ratio, in order to avoid the situation where the denominator is zero, a preset value needs to be added to the denominator, and an embodiment of the preset value is 0.01.

[0067] In actual application, as another implementation method, the opposite number of the normal vector similarity of the target supervoxel is used as the exponent, and the natural constant is used as the power of the base, which is recorded as the first power value of the target supervoxel, and the difference between the number 1 and the first power value of the target supervoxel is recorded as the normal vector deformation degree of the target supervoxel.

[0068] The normal vector deformation degree of the target supervoxel is used to evaluate the possibility of overall deformation at the corresponding position of the large-span spatial steel structure building. When the difference between the normal vectors of the projection points of the point cloud data contained in the target supervoxel and other supervoxels at the corresponding fitting surface is greater, the normal vector deformation degree of the target supervoxel is greater. At this time, the possibility of overall deformation of the target supervoxel at the corresponding position of the large-span spatial steel structure building is greater.

[0069] The architectural deformation degree of the target supervoxel is determined according to the difference between the Gaussian curvatures of the projection points of all point cloud data contained in the target supervoxel on the fitting surface of the target supervoxel and the deformation degree of the normal vector of the target supervoxel.

[0070] The linear normalized value of the standard deviation of the Gaussian curvature of the projection points of all point cloud data contained in the target supervoxel on the fitting surface of the target supervoxel is recorded as the curvature difference of the target supervoxel. The average of the normal vector deformation degree and the curvature difference of the target supervoxel is recorded as the architectural deformation degree of the target supervoxel.

[0071] In actual application, as another implementation method, the product of the normal vector deformation degree and the curvature difference of the target supervoxel is recorded as the building deformation degree of the target supervoxel.

[0072] It can be understood that the curvature difference of the target supervoxel is used to evaluate the possibility of local deformation at the corresponding position of the large-span space steel structure building. When the difference between the Gaussian curvatures of the projection points of the fitting surface of the target supervoxel of all point cloud data contained in the target supervoxel is greater, the curvature difference of the target supervoxel is greater, the surface of the target supervoxel at the corresponding position of the large-span space steel structure building is greater, and the possibility of local deformation is greater.

[0073] Finally, the architectural deformation degree of the target supervoxel is used to more accurately evaluate the possibility of deformation problems at the corresponding position of the large-span spatial steel structure building. When the difference between the normal vector deformation degree and the curvature of the target supervoxel is greater, the architectural deformation degree of the target supervoxel is greater. At this time, the possibility of overall deformation or local deformation problems on the surface of the target supervoxel at the corresponding position of the large-span spatial steel structure building is greater.

[0074] The building deformation degree of any supervoxel divided from the building point cloud data can be obtained in the same way.

[0075] At this point, the building deformation degree of any supervoxel divided by the building point cloud data is obtained.

[0076] S003: Determine the fitting surface distance of the point cloud data contained in the target supervoxel according to the relative position of the point cloud data contained in the target supervoxel and the fitting surface of the target supervoxel, and determine the data smoothness of the point cloud data contained in the target supervoxel according to the difference between the building deformation degree of the supervoxel where the point cloud data is located and the fitting surface distance of all point cloud data contained in the supervoxel where the point cloud data is located, as well as the normal vector deformation degree of the target supervoxel.

[0077] The values ​​of the noise data contained in the building point cloud data are randomly distributed, which results in the inability of the noise data to fit tightly to the fitting surface of the target supervoxel. That is to say, compared with normal point cloud data, the distance between the noise data contained in the target supervoxel and the fitting surface of the target supervoxel is larger.

[0078] The projection point of any point cloud data included in the target supervoxel on the fitting surface of the target supervoxel is recorded as the projection point of the any point cloud data, and the Euclidean distance between the projection point of the any point cloud data and the any point cloud data is recorded as the fitting surface distance of the any point cloud data. Calculating the Euclidean distance between two point cloud data is a well-known technology and will not be described in detail.

[0079] According to the difference between the fitting surface distances of the point cloud data included in the target supervoxel, the noise difference distance of each point cloud data included in the target supervoxel is determined respectively.

[0080] Preferably, as an embodiment of the present application, any point cloud data contained in the target supervoxel is recorded as the target point cloud data, the difference in fitting surface distance between the target point cloud data and other point cloud data contained in the target supervoxel is recorded as the fitting surface distance difference between the target point cloud data and other point cloud data contained in the target supervoxel, and the linear normalized value of the mean of the difference in fitting surface distance between the target point cloud data and all other point cloud data contained in the target supervoxel is recorded as the noise difference distance of the target point cloud data.

[0081] Among them, this embodiment uses the maximum and minimum normalization method to calculate the linear normalization value. In actual application, the implementer can use other methods in the prior art such as the Z-Score standard normalization method to calculate the linear normalization value, which is not limited here.

[0082] The noise difference distance of any point cloud data contained in any supervoxel of the building point cloud data can be obtained in the same way.

[0083] Specifically, the calculation formula for the noise difference distance of any point cloud data contained in any supervoxel of the building point cloud data is:

[0084]

[0085] In the formula, Represents the building point cloud data The supervoxel contains The noise difference distance of point cloud data; Represents the building point cloud data The supervoxel contains The fitting surface distance of point cloud data; Represents the building point cloud data The supervoxel contains The fitting surface distance of point cloud data; Represents a linear normalization function, which is used to take the linear normalized value of the value in brackets; Represents the building point cloud data The number of point cloud data contained in a supervoxel.

[0086] The noise difference distance of the point cloud data is used to evaluate the possibility that the point cloud data contained in the supervoxel is interfered by noise. The greater the difference between the fitting surface distance of the point cloud data and the fitting surface distance of other point cloud data contained in the supervoxel where the point cloud data is located, the greater the distance between the point cloud data and the fitting surface of the supervoxel where the point cloud data is located, and the greater the possibility that the point cloud data is point cloud data interfered by the noise data contained in the target supervoxel.

[0087] The data smoothness of the point cloud data is determined according to the noise difference distance of the point cloud data and the building deformation degree of the supervoxel where the point cloud data is located.

[0088] Preferably, as an embodiment of the present application, a linear normalized value of the ratio of the noise difference distance of the point cloud data to the building deformation degree of the supervoxel where the point cloud data is located is recorded as the data smoothness of the point cloud data.

[0089] In the process of calculating the ratio, in order to avoid the situation where the denominator is zero, a preset value needs to be added to the denominator, and an embodiment of the preset value is 0.01.

[0090] The data smoothness of any point cloud data contained in any supervoxel of the building point cloud data can be obtained in the same way.

[0091] Specifically, the calculation formula for the data smoothness of any point cloud data contained in any supervoxel of the building point cloud data is:

[0092]

[0093] In the formula, Represents the building point cloud data The supervoxel contains The data smoothness of the point cloud data; Represents the building point cloud data The supervoxel contains The noise difference distance of point cloud data; represents a preset constant. In this embodiment, the value of the preset constant is 0.01. The function of the preset constant is to prevent the denominator from being 0. Represents a linear normalization function, which is used to take the linear normalized value of the value in brackets.

[0094] When the noise difference distance of the point cloud data is larger and the building deformation degree of the supervoxel where the point cloud data is located is smaller, the data smoothness of the point cloud data is greater. At this time, the point cloud data is more likely to be interfered by noise. When denoising the building point cloud data, a larger smoothness should be selected for the point cloud data to ensure that the denoising result of the acquired point cloud data is more accurate.

[0095] At this point, the data smoothness of any point cloud data contained in any supervoxel of the building point cloud data is obtained.

[0096] S004: De-noise the building point cloud data according to the data smoothness of the point cloud data to obtain the denoised building point cloud data, and obtain non-contact deformation monitoring results for large-span spatial steel structures based on the denoised building point cloud data and the grayscale image of the building surface.

[0097] The adjustment filter radius of the point cloud data is determined according to the data smoothness of the point cloud data.

[0098] Preferably, as an embodiment of the present application, the rounded value of the product of the data smoothness of the point cloud data and the second adjustment parameter is recorded as the first product of the point cloud data, and the sum of the first product of the point cloud data and the first adjustment parameter is recorded as the adjustment filter radius of the point cloud data.

[0099] Specifically, the calculation formula for adjusting the filter radius of any point cloud data contained in any supervoxel of the building point cloud data is:

[0100]

[0101] In the formula, Represents the building point cloud data The supervoxel contains The data smoothness of the point cloud data; represents a first adjustment parameter. In this embodiment, the value of the first adjustment parameter is 3; represents the second adjustment parameter. In this embodiment, the value of the first adjustment parameter is 6; Represents the rounding function, which is used to round the value in brackets.

[0102] When the data smoothness of the point cloud data is greater, the adjustment filter radius of the point cloud data is larger, and the degree of smoothing of the point cloud data in the building point cloud data is more obvious.

[0103] The adjusted filter radius of the point cloud data is used as the value of the radius of the filter window of the point cloud data, and the building point cloud data is denoised using mean filtering to obtain the denoised building point cloud data.

[0104] This embodiment uses mean filtering to denoise the building point cloud data. As other implementation methods, on the basis of achieving the purpose of denoising the building point cloud data, the implementer can use other methods in the existing technology such as Gaussian filtering, median filtering, etc. to denoise the building point cloud data, and this application does not impose any special restrictions.

[0105] The denoised building point cloud data is projected onto the grayscale image of the building surface to obtain the building depth map. The building depth map can provide the depth of each position on the surface of a large-span steel structure building.

[0106] Projecting the point cloud data onto a grayscale image to obtain a depth map is a well-known technique and will not be described in detail.

[0107] Before the large-span spatial steel structure building is put into use, the building point cloud data and the building surface grayscale image of the large-span spatial steel structure building before the use are collected according to the same method, and the building depth map of the large-span spatial steel structure building before the use is obtained, and the obtained building depth map is recorded as the standard building depth map.

[0108] The building depth map is subtracted from the standard building depth map to obtain the building deformation map. The pixel value of the pixel point in the building deformation map is the depth change in the vertical direction of the corresponding position of the pixel point in the large-span space steel structure building during use. When the depth change is greater, the deformation of the corresponding position of the pixel point in the building deformation map in the large-span space steel structure building is more obvious.

[0109] When the pixel value of a pixel point in the building deformation map is not zero, it is determined that the pixel point is deformed at the corresponding position of the large-span steel structure building, and the deformation amount is the pixel value of the pixel point. When the pixel value of a pixel point in the building deformation map is zero, it is determined that the pixel point is not deformed at the corresponding position of the large-span steel structure building.

[0110] At this point, the non-contact deformation monitoring results for large-span spatial steel structures are obtained.

[0111] See also Figure 4 , Figure 4 1 is a schematic diagram of the structure of a non-contact deformation monitoring device for a large-span space steel structure provided by an embodiment of the present application. In this embodiment, the device includes various units for executing various steps in the corresponding embodiment of the non-contact deformation monitoring method for a large-span space steel structure. Figure 4 The deformation monitoring device includes: a deformation data acquisition module, a deformation degree determination module, a data smoothness determination module, and a deformation monitoring result acquisition module.

[0112] A deformation data acquisition module, used to acquire building point cloud data and building surface grayscale images of a large-span spatial steel structure building, and divide the building point cloud data into a first preset number of supervoxels;

[0113] A deformation degree determination module is used to record any supervoxel divided from the building point cloud data as a target supervoxel, perform point cloud surface fitting on the target supervoxel, obtain the fitted surface of the target supervoxel, determine the deformation degree of the normal vector of the target supervoxel according to the difference between the normal vectors of the point cloud data contained in the target supervoxel and other supervoxels at the projection points of the corresponding fitted surface, and determine the building deformation degree of the target supervoxel in combination with the difference between the Gaussian curvatures of all the point cloud data contained in the target supervoxel at the projection points of the fitted surface of the target supervoxel;

[0114] a data smoothness determination module, for determining the fitting surface distance of the point cloud data contained in the target supervoxel according to the relative position of the point cloud data contained in the target supervoxel and the fitting surface of the target supervoxel, and determining the data smoothness of the point cloud data contained in the target supervoxel according to the difference between the building deformation degree of the supervoxel where the point cloud data is located and the fitting surface distance of all the point cloud data contained in the supervoxel where the point cloud data is located;

[0115] The deformation monitoring result acquisition module is used to denoise the building point cloud data according to the data smoothness of the point cloud data, obtain the denoised building point cloud data, and obtain the non-contact deformation monitoring results for large-span spatial steel structures based on the denoised building point cloud data and the grayscale image of the building surface.

[0116] It is to be understood that the above-mentioned sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-tasking and parallel processing are also possible or may be advantageous. The various embodiments in this specification are described in a progressive manner, and the same and similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments.

[0117] The above contents are only implementation methods of the present application and are not intended to limit the scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the protection scope of the present application.

Claims

1. A non-contact deformation monitoring method for large-span spatial steel structures, characterized in that: The method comprises the following steps: Collecting building point cloud data and building surface grayscale images of a large-span spatial steel structure building, and dividing the building point cloud data into a first preset number of supervoxels; Any supervoxel divided from the building point cloud data is recorded as a target supervoxel, and a point cloud surface is fitted to the target supervoxel to obtain a fitting surface of the target supervoxel. The normal vector deformation degree of the target supervoxel is determined according to the difference between the normal vectors of the projection points of the point cloud data contained in the target supervoxel and other supervoxels at the corresponding fitting surface. The building deformation degree of the target supervoxel is determined by combining the difference between the Gaussian curvatures of all the point cloud data contained in the target supervoxel at the projection points of the fitting surface of the target supervoxel; Determine the fitting surface distance of the point cloud data contained in the target supervoxel according to the relative position of the point cloud data contained in the target supervoxel and the fitting surface of the target supervoxel, and determine the data smoothness of the point cloud data contained in the target supervoxel according to the difference between the building deformation degree of the supervoxel where the point cloud data is located and the fitting surface distance of all the point cloud data contained in the supervoxel where the point cloud data is located, as well as the normal vector deformation degree of the target supervoxel; The building point cloud data is denoised according to the data smoothness of the point cloud data to obtain the denoised building point cloud data, and the non-contact deformation monitoring results for large-span spatial steel structures are obtained based on the denoised building point cloud data and the building surface grayscale image.

2. The non-contact deformation monitoring method for a large-span spatial steel structure according to claim 1, characterized in that: The process of determining the normal vector deformation degree of the target supervoxel is as follows: Determine the projection point of the centroid of all point cloud data contained in the target supervoxel on the fitting surface of the target supervoxel, and record the normal vector of the projection point on the fitting surface of the target supervoxel as the normal vector of the target supervoxel; The average of the similarities between the normal vector of the target supervoxel and the normal vector of each supervoxel divided by the building point cloud data is recorded as the normal vector similarity of the target supervoxel; According to the normal vector similarity of the target supervoxel, the normal vector deformation degree of the target supervoxel is determined, and the normal vector deformation degree of the target supervoxel is negatively correlated with the normal vector similarity of the target supervoxel.

3. The non-contact deformation monitoring method for a large-span spatial steel structure according to claim 1, characterized in that: The specific process of determining the building deformation degree of the target supervoxel by combining the difference between the Gaussian curvatures of the projection points of the fitting surface of the target supervoxel with all the point cloud data contained in the target supervoxel is as follows: The linear normalized value of the standard deviation of the Gaussian curvature of the projection points of all point cloud data contained in the target supervoxel on the fitting surface of the target supervoxel is recorded as the curvature difference of the target supervoxel; The building deformation degree of the target supervoxel is obtained according to the normal vector deformation degree and the curvature difference of the target supervoxel.

4. The non-contact deformation monitoring method for a large-span space steel structure according to claim 3, characterized in that: The building deformation degree of the target supervoxel is obtained according to the normal vector deformation degree and curvature difference of the target supervoxel. The specific process of obtaining the building deformation degree of the target supervoxel is as follows: The average of the normal vector deformation and curvature difference of the target supervoxel is recorded as the building deformation of the target supervoxel.

5. The non-contact deformation monitoring method for a large-span spatial steel structure according to claim 1, characterized in that: The construction process of the fitting surface distance of the point cloud data contained in the target supervoxel is: Any point cloud data contained in the target supervoxel is recorded as the target point cloud data, the projection point of the target point cloud data on the fitting surface of the target supervoxel is recorded as the projection point of the target point cloud data, and the Euclidean distance between the projection point of the target point cloud data and the target point cloud data is recorded as the fitting surface distance of the target point cloud data.

6. The non-contact deformation monitoring method for a large-span space steel structure according to claim 5, characterized in that: The specific process of determining the data smoothness of the point cloud data contained in the target supervoxel according to the difference between the building deformation degree of the supervoxel where the point cloud data is located and the fitting surface distance of all the point cloud data contained in the supervoxel where the point cloud data is located, and the normal vector deformation degree of the target supervoxel is as follows: The difference between the fitting surface distances of the target point cloud data and other point cloud data contained in the target supervoxel is recorded as the fitting surface distance difference between the target point cloud data and other point cloud data contained in the target supervoxel, and the linear normalized value of the mean of the fitting surface distance differences between the target point cloud data and all other point cloud data contained in the target supervoxel is recorded as the noise difference distance of the target point cloud data; The data smoothness of the target point cloud data is determined according to the noise difference distance of the target point cloud data and the building deformation degree of the supervoxel where the target point cloud data is located.

7. The non-contact deformation monitoring method for a large-span space steel structure according to claim 6, characterized in that: The data smoothness of the target point cloud data is determined according to the noise difference distance of the target point cloud data and the building deformation degree of the supervoxel where the target point cloud data is located. The specific process of obtaining the data smoothness is as follows: The linear normalized value of the ratio of the noise difference distance of the target point cloud data to the building deformation degree of the supervoxel where the target point cloud data is located is recorded as the data smoothness of the target point cloud data.

8. The non-contact deformation monitoring method for a large-span spatial steel structure according to claim 1, characterized in that: The building point cloud data is denoised according to the data smoothness of the point cloud data to obtain the denoised building point cloud data. The specific process of obtaining the denoised building point cloud data is as follows: The rounded value of the product of the data smoothness of the point cloud data and the second adjustment parameter is recorded as the first product of the point cloud data, and the sum of the first product of the point cloud data and the first adjustment parameter is recorded as the adjustment filter radius of the point cloud data; The adjusted filter radius of the point cloud data is used as the value of the radius of the filter window of the point cloud data, and the building point cloud data is denoised using mean filtering to obtain the denoised building point cloud data, wherein the value of the first adjustment parameter is 3 and the value of the second adjustment parameter is 6.

9. The non-contact deformation monitoring method for a large-span spatial steel structure according to claim 1, characterized in that: The non-contact deformation monitoring result for the large-span space steel structure is obtained based on the denoised building point cloud data and the building surface grayscale image. The specific process of obtaining is as follows: Project the denoised building point cloud data onto the grayscale image of the building surface to obtain the building depth map; Collecting building point cloud data and building surface grayscale images of the large-span space steel structure building before commissioning, obtaining a building depth map of the large-span space steel structure building before commissioning, and recording the building depth map of the large-span space steel structure building before commissioning as a standard building depth map; Subtract the building depth map from the standard building depth map to obtain a building deformation map; When the pixel value of a pixel point in the building deformation map is not zero, it is determined that the pixel point is deformed at the corresponding position of the large-span steel structure building, and the deformation amount is the pixel value of the pixel point; When the pixel value of a pixel point in the building deformation map is zero, it is determined that no deformation occurs at the corresponding position of the pixel point in the large-span steel structure building.

10. Non-contact deformation monitoring device for large-span space steel structure, characterized in that: The deformation monitoring device comprises: A deformation data acquisition module, used to acquire building point cloud data and building surface grayscale images of a large-span spatial steel structure building, and divide the building point cloud data into a first preset number of supervoxels; A deformation degree determination module is used to record any supervoxel divided from the building point cloud data as a target supervoxel, perform point cloud surface fitting on the target supervoxel, obtain the fitted surface of the target supervoxel, determine the deformation degree of the normal vector of the target supervoxel according to the difference between the normal vectors of the point cloud data contained in the target supervoxel and other supervoxels at the projection points of the corresponding fitted surface, and determine the building deformation degree of the target supervoxel in combination with the difference between the Gaussian curvatures of all the point cloud data contained in the target supervoxel at the projection points of the fitted surface of the target supervoxel; a data smoothness determination module, for determining the fitting surface distance of the point cloud data contained in the target supervoxel according to the relative position of the point cloud data contained in the target supervoxel and the fitting surface of the target supervoxel, and determining the data smoothness of the point cloud data contained in the target supervoxel according to the difference between the building deformation degree of the supervoxel where the point cloud data is located and the fitting surface distance of all the point cloud data contained in the supervoxel where the point cloud data is located; The deformation monitoring result acquisition module is used to denoise the building point cloud data according to the data smoothness of the point cloud data, obtain the denoised building point cloud data, and obtain the non-contact deformation monitoring results for large-span spatial steel structures based on the denoised building point cloud data and the grayscale image of the building surface.

Images