Tower-shaped building offset detection method and device

Abstract

The invention relates to a tower-shaped building offset detection method, and relates to the constructional engineering detection technology, and the method comprises the steps: collecting the complete point cloud data of a target tower body; registering the point cloud data of different detection stations to obtain an overall point cloud model of the target tower body; screening and clearing noisy points in the overall point cloud model to obtain a pure point cloud model; performing point cloud thinning on the pure point cloud model to obtain a lightweight point cloud model; performing hierarchical slicing processing on the lightweight point cloud model to obtain a plurality of point cloud slices; and performing centroid fitting on each layer of point cloud slice through a least square method to obtain a circle center of each layer of point cloud slice, connecting the circle centers of each layer of point cloud slice from bottom to top to obtain an off-axis line of the target tower body, and analyzing the inclination condition of the target tower body according to the off-axis line. According to the invention, the problems of high difficulty and low precision of ancient tower detection are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of building engineering inspection, and in particular to a method and device for detecting the offset of a tower-shaped building. Background Art

[0002] Tower-shaped structures have a large aspect ratio of length to diameter, require high self-rigidity, and are prone to various damages caused by inclination. For ancient towers, it is difficult to avoid inclination over time. However, regular inspections and repairs in the initial stage of inclination can provide as much protection as possible. Traditional verticality detection methods mostly involve single-point data collection, which can only make a qualitative assessment of the overall tower body. In addition, when measuring various data inside the tower, the space is relatively narrow, there is line-of-sight obstruction, and it is difficult to transfer control points. If quantitative analysis of the verticality of each level of the tower body can be carried out using 3D laser scanning, it will be of great help to the protection work of the building. Adopting 3D laser scanning technology is a very necessary task for optimizing the offset detection of tower-shaped structures.

[0003] Common methods for detecting the verticality of ancient tower buildings mainly include the plumb line method, the laser plumb method, the theodolite stake-out method, and the total station prism-free method. For the plumb line method, its principle is to utilize the action of gravity and judge the verticality by measuring the inclination angle of the hanging rope inside the tower body. This method has poor detection accuracy, great operation difficulty when the storey height is high, and cumulative errors between layers, and is not applicable to solid ancient towers or ancient tower buildings that are unstable due to age and are not suitable for climbing. The laser plumb method has a principle similar to the plumb line method and uses a laser to measure the relative deviation in the vertical direction. However, visible lasers are greatly affected by light illumination and cannot guarantee the accuracy of the offset transfer of each layer. Therefore, this method is not applicable to ancient towers with too high tower bodies. For the theodolite stake-out method, the basic principle is to measure the offset of the center points at the top and bottom of the ancient tower to judge the overall verticality of the building. This method requires an open space around the ancient tower, and there are certain limitations for the detection of ancient towers located in urban scenic areas. For the total station prism-free method, control points are arranged around the ancient tower, the point coordinates of the tower body are collected, and the coordinate sum and height difference between the vertex and the base center are calculated to obtain the verticality of the ancient tower. The prism-free measurement method has high requirements for the engineering environment. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and device for detecting the offset of a tower-shaped building, which can efficiently, accurately, and intelligently obtain the overall offset situation of the tower-shaped structure.

[0005] In a first aspect, the present invention provides a method for detecting the offset of a tower-shaped building and adopts the following technical solutions: A method for detecting the offset of a tower-shaped building includes: Collecting the complete point cloud data of the target tower body; Registering the point cloud data of different detection stations to obtain the overall point cloud model of the target tower body; Filter out and remove the noise points in the overall point cloud model to obtain a pure point cloud model; Downsample the pure point cloud model to obtain a lightweight point cloud model; Perform layered slicing on the lightweight point cloud model to obtain multiple point cloud slices; Use the least squares method to perform centroid fitting on each layer of the point cloud slices to obtain the center points of each layer of point cloud slices. Connect the center points of each layer of point cloud slices from bottom to top to obtain the offset axis of the target tower body, and analyze the inclination of the target tower body according to the offset axis.

[0006] A further technical solution of the present invention is that: the acquisition of the complete point cloud data of the target tower body specifically includes: Obtain the top point cloud data of the target tower body through a drone; Obtain the side point cloud data of the target tower body through a ground detection station; Combine the top point cloud data and the side point cloud data to obtain the complete point cloud data of the target tower body.

[0007] A further technical solution of the present invention is that: the registration of the point cloud data from different detection stations to obtain the overall point cloud model of the target tower body specifically includes: Establish a unified coordinate system. The point cloud data is located in the unified coordinate system, and the point cloud data is rigidly transformed to the global coordinate system; Perform optimization fitting on the overlapping areas of the point cloud data collected from different detection stations, and initially fix these overlapping areas to form a plane with similar characteristics; The registered point cloud model is tested with different evaluation algorithms to ensure the accuracy of the registration.

[0008] A further technical solution of the present invention is that: the filtering out and removing the noise points in the overall point cloud model to obtain a pure point cloud model specifically includes: Screen out the noise point information of typical objects that are not the target tower body structure in the overall point cloud model and perform rough denoising; After rough denoising of the overall point cloud model, screen out the noise point information of non-typical objects that are not the target tower body structure and perform fine denoising to obtain a pure point cloud model.

[0009] A further technical solution of the present invention is that: the screening out and removing the noise point information of typical objects that are not the target tower body structure in the overall point cloud model and performing rough denoising specifically includes: Denoise according to the gradient of the color or intensity information of the points in the overall point cloud model; or based on the method of geometric gradient, screen and remove the noise point areas with large curvature changes in the overall point cloud model.

[0010] A further technical solution of the present invention is that thinning the point cloud of the pure point cloud model to obtain a lightweight point cloud model specifically includes: Evaluating the point cloud characteristics in the pure point cloud model, and selecting a thinning algorithm according to the point cloud characteristics to thin the pure point cloud model to obtain a lightweight point cloud model.

[0011] A further technical solution of the present invention is that performing hierarchical slicing processing on the lightweight point cloud model to obtain a plurality of point cloud slices, specifically including: Selecting the center at each layer height of the target tower body as the point selection for the section plane, and determining the height of the slicing part; Determining the thickness of the slice according to the point cloud density at different positions of the target tower body.

[0012] A further technical solution of the present invention is that performing centroid fitting on each layer of the point cloud slices by the least squares method to obtain the center of the circle of each layer of point cloud slices, specifically including: Establishing an algebraic equation of a circle based on the point cloud slice and parameterizing it to form a parameterized circle equation; Substituting each point cloud coordinate in the point cloud slice into the parameterized circle equation to establish an overdetermined system of equations; Solving the overdetermined system of equations by the least squares method to obtain the radius and center of the circle of the point cloud slice.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. Non-contact detection: There is no need for operators to contact the target object, no need for high-altitude operations, and the overall data of the tower body can be obtained through long-distance scanning; at the same time, in the face of dangerous buildings with a large inclination degree, the detection personnel do not need to enter the dangerous area to ensure the safety of personnel.

[0014] 2. Do not need to enter the interior: Some tower-like structures are solid buildings or have a narrow internal space, and there is no working space for methods such as the hanging plumb line method and the laser plumb bob. The present invention does not need to enter the interior of the tower body to avoid damage and disturbance to cultural relic buildings; there is no need for segmented measurement, and the overall verticality can be directly calculated through the point cloud model to avoid cumulative errors.

[0015] 3. Facilitate the preservation of detection data: Paper records are easily lost and the state during detection cannot be traced back. Point cloud data can be retained throughout the life cycle, and at the same time, artificial errors generated during data recording are avoided.

[0016] In the second aspect, the present invention also discloses a tower-like building offset detection device, including: A data acquisition module for acquiring the point cloud data of the target tower body; A model establishment module for registering the point cloud data of different detection stations to obtain an overall point cloud model of the target tower body; A model lightweight module is used to filter out and remove noise points in the overall point cloud model to obtain a pure point cloud model, and perform point cloud thinning on the pure point cloud model to obtain a lightweight point cloud model; A slicing module is used to perform hierarchical slicing on the lightweight point cloud model to obtain a plurality of point cloud slices; An analysis module is used to perform centroid fitting on each layer of the point cloud slices by the least squares method to obtain the centers of the circles of each layer of the point cloud slices, connect the centers of the circles of each layer of the point cloud slices from bottom to top to obtain the offset axis of the target tower body, and analyze the inclination of the target tower body according to the offset axis.

[0017] Compared with the prior art, the beneficial effects of a tower-shaped building offset detection device provided by the present invention are the same as those of the tower-shaped building offset detection method described in the first aspect above, and will not be elaborated here. Description of the Drawings

[0018] Figure 1 is a schematic flowchart of a method provided by an embodiment of the present invention; Figure 2 is a diagram of the installation position of a scanner and the position of a target ball provided by an embodiment of the present invention; Figure 3a is an effect diagram of point cloud model registration provided by an embodiment of the present invention; Figure 3b is another effect diagram of point cloud model registration provided by an embodiment of the present invention; Figure 4 is an effect diagram of denoising of a point cloud model by gradient and curvature methods provided by an embodiment of the present invention; Figure 5 is a model diagram of the denoised point cloud model thinned by voxel and statistical filtering provided by an embodiment of the present invention; Figure 6 is a diagram of each slice of the target from a 45° bird's-eye view provided by an embodiment of the present invention; Figure 7a is a top view of the point cloud of the first layer slice provided by an embodiment of the present invention; Figure 7b is a fitting slice and a fitting centroid of the point cloud of the first layer slice provided by an embodiment of the present invention; Figure 8 is an offset axis formed by connecting the fitting centroids of each layer provided by an embodiment of the present invention. Detailed Embodiments

[0019] For the convenience of clearly describing the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, terms such as "first" and "second" are used to distinguish identical or similar items with basically the same functions and roles. For example, the first threshold and the second threshold are only used to distinguish different thresholds, and do not limit their sequence. Those skilled in the art can understand that terms such as "first" and "second" do not limit the quantity and execution order, and the terms "first" and "second" do not necessarily limit being different.

[0020] It should be noted that in the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, using words such as "exemplary" or "for example" aims to present relevant concepts in a specific manner.

[0021] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following items" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can represent: a, b, c, the combination of a and b, the combination of a and c, the combination of b and c, or the combination of a, b and c, where a, b, c can be single or multiple.

[0022] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings of the specification.

[0023] The embodiments of the present invention provide a method for detecting the offset of a tower-shaped building. The main process of the method is described as follows.

[0024] As Figure 1 shown: Step S1: Collect the complete point cloud data of the target tower body.

[0025] Among them, the target tower body is the ancient tower building to be detected. The point cloud data of the target tower body is collected by using a three-dimensional laser scanner. The on-site scanning work should set up appropriate stations according to the field of view of the three-dimensional laser scanner and the obstacles of the target tower body.

[0026] For large tower-like buildings, multiple stations usually need to be set up. Considering the field of view of the 3D laser scanner, select an appropriate distance between the instrument and the target tower body: on the one hand, it should not be too close to avoid the upper tower body being blocked by the tower eaves due to too large an elevation angle; on the other hand, it should not be too far, resulting in sparse point clouds. During scanning, attention should be paid to the environment. Scanning should be carried out under good visibility and less vibration conditions. At the same time, pay attention to the humidity of the air. Humid air will hinder the propagation of laser. When there is water vapor on the surface of the target tower body, specular reflection will occur, affecting the accuracy of the data. In the example, the positions of the scanner and the target balls are as Figure 2 shown.

[0027] The method for collecting complete point cloud data of the target tower body specifically includes: obtaining the top point cloud data of the target tower body through a drone, obtaining the side point cloud data of the target tower body through a ground detection station, and combining the top point cloud data and the side point cloud data to obtain the complete point cloud data of the target tower body.

[0028] Among them, before setting up the 3D laser scanner, conduct a site survey of the on-site environment. According to the different structures and construction characteristics of the tower body, select the positions and sequences of the measuring stations. For special structural parts such as the internal and external corners of the building and the parts with structural damage on the facade, detection stations can be densely arranged. Arrange special personnel according to the detection tasks, and prepare a scanning plan and conduct a technical disclosure. For relatively high tower bodies, use a drone for top-down scanning, combined with a ground 3D laser scanner, to obtain complete point cloud data. Before the drone scans, preview the entire site through the camera carried by the drone to determine the high-altitude scanning route. Regarding the station spacing of the detection stations, arrange the stations in sequence according to the principle of appropriate overlap of the detection areas. The overlap area between the front and rear stations is between 30% and 40%, which can not only ensure smooth registration but also avoid redundant calculations.

[0029] Step S2: Register the point cloud data of each detection station to obtain the overall point cloud model of the target tower body.

[0030] Among them, point cloud registration is a technology that aligns multiple point cloud data sets to the same coordinate system, aiming to accurately overlap them in the same spatial reference system. Registering the point cloud data of each detection station specifically includes: Step S21: Establish a unified coordinate system. The point cloud data is located in the unified coordinate system, and the rigid body coordinates of the point cloud data are transformed into the global coordinate system.

[0031] Among them, rigid body coordinate transformation is the basic transformation form that describes the change of the position and direction of an object or point cloud data in space only through rotation and translation. The unified coordinate system is equivalent to the local coordinate system, which is a coordinate system established based on the object itself or a specific reference point; the global coordinate system is a fixed 3D space coordinate system where the target tower body is located.

[0032] Step S22: Optimize and fit the overlapping regions of the point cloud data collected by different detection stations, and preliminarily fix these overlapping regions to form a plane with similar features.

[0033] First, it is necessary to detect the overlapping regions, which can be achieved by using the octree detection method or the voxel filtering algorithm; then, the fitting of the overlapping regions of the point cloud can be realized by using the polynomial regression algorithm. The nonlinear surface is fitted by polynomial feature expansion, and regularization is combined to prevent overfitting, and finally a plane with similar features is obtained. After the registration is completed, the overall point cloud model is obtained, and different algorithms are used for verification to ensure the accuracy of the registration.

[0034] The accuracy of the registration verification can be specifically verified by using the following three evaluation algorithms: Evaluation based on distance metrics: It is implemented by using the mean squared error (MSE). First, the source point cloud is transformed to the target point cloud coordinate system through the estimated transformation matrix, and then for each point in the transformed source point cloud, the KD tree is used to accelerate the search to find the nearest neighbor point in the target point cloud. Finally, the mean squared error formula is used to calculate the squared distance of all corresponding point pairs and take the average. The mean squared error formula is: .

[0035] Evaluation based on overlapping regions: First, a distance threshold is defined, and the ratio of the points in the transformed source point cloud whose distance from the target point cloud is less than the distance threshold is calculated. When using this algorithm, only the median of the point cloud distances in the overlapping region is calculated to reduce the interference of the non-overlapping region.

[0036] Evaluation based on feature matching degree: First, the point cloud features are extracted. For the rough registration model, the fast point feature histogram is used to calculate three geometric parameters of the point cloud, namely the angle between the point normal vectors, the angle between the point tangent directions, and the Euclidean distance between two points. These parameters are discretized into the histogram to form a high-dimensional feature vector; for the fine registration model, the direction histogram signature is used. First, a local reference system is constructed, the covariance matrix of the neighborhood points is calculated, the main direction of the eigenvector is obtained by decomposition, the coordinate system direction is adjusted according to the consistency of the main direction symbols, and then the neighborhood space is divided into multiple spherical shells and azimuth regions. In each sub-region, the change of the normal vector or the point density direction is statistically analyzed by the cosine value of the angle, and the statistical results are spliced into a histogram feature vector. Then the feature descriptors are matched, and the number of correct matching pairs is counted. Finally, the distance error (feature similarity) of the matching pairs is calculated.

[0037] Step S3: Screen out and remove the noise points in the overall point cloud model to obtain a pure point cloud model.

[0038] Among them, screening out and removing the noise points in the overall point cloud model to obtain a pure point cloud model specifically includes: Step S31: Filter out the typical object noise information of the non-target tower structure in the overall point cloud model and perform rough denoising processing.

[0039] Among them, for typical objects, such as large-scale classification and screening of trees, rivers, vehicles, etc. around the target tower body, it can be achieved through the following methods: Method 1: Denoise according to the gradient of color or intensity information in the point cloud. Calculate the change in the color or intensity gradient of adjacent point clouds to identify noise points. The noise points have a large difference in color or intensity from the surrounding point clouds and a high gradient. By setting a gradient threshold, the point clouds exceeding the gradient threshold are regarded as noise points and removed. Combining spatial information, calculate the gradient within the local neighborhood and then perform filtering in combination with the spatial distance.

[0040] Method 2: If true color scanning is not used and the target only has point cloud coordinates, use a geometric gradient-based method to determine whether the area with large curvature changes is a noise point or a feature point. Calculate the normal or curvature of each point. The places with large and uniform gradients are edges, and the places with large but non-uniform gradients are noise points.

[0041] Step S32: Filter out the non-typical object noise information of the non-target tower structure in the overall point cloud model and perform fine manual denoising processing to obtain a pure point cloud model.

[0042] Step S4: Perform point cloud thinning on the pure point cloud model to obtain a lightweight point cloud model, which specifically includes: Step S41: Evaluate the characteristics of the point cloud, count the number of point clouds in the pure point cloud model, and calculate the point cloud density according to the volume of the target tower body.

[0043] Among them, calculating the point cloud density can provide a basis for the thinning algorithm. Dynamically set the thinning rejection threshold through the average value of the point cloud spacing, increase the rejection ratio in the dense point cloud area, and reduce the rejection ratio in the sparse point cloud area, so as to reasonably compress the data volume while retaining features.

[0044] When calculating the point cloud density of the target tower body, first find the nearest neighbor point of each point cloud, construct a data structure for spatial partitioning, search for neighboring points in multi-dimensional space, calculate the distances of these nearest neighbor point clouds, and then take the average value. The approximate value of the point cloud density can be indirectly deduced through the average value of the "nearest neighbor distance". The formula is , directly calculating the point cloud density by dividing the number of point clouds by the volume will be affected by local discrete points or non-uniform distributions. The average nearest neighbor distance can reflect the relationship of the local neighborhood and has better spatial distribution characteristics. The specific steps to find the average value of the nearest neighbor point cloud distance are as follows: Step S411: Determine the point cloud data set P = {p1, p2,..., pN} of the target tower body, and each point cloud pi contains three-dimensional coordinates (x, y, z); Step S412: Select the dimension with the largest variance as the splitting axis, that is, the axis with the most dispersed data distribution among X, Y, and Z. Use the median of all points on the selected splitting axis as the splitting plane to divide the point set into left and right subtrees. Recursively process the left and right subtrees until the number of points contained in the leaf nodes is less than the set elimination threshold, which is 1 point in this example; Step S413: For each point , query its nearest neighbor point , that is, the point with the smallest Euclidean distance. Starting from the root node, according to the splitting axis and splitting value, gradually enter the subtree that may contain the nearest neighbor. Backtrack and check the unvisited branches to ensure finding the global nearest neighbor. The Euclidean distance formula is: ; Step S413: Traverse all point clouds. For each point , perform a nearest neighbor search once to obtain the distance . Exclude duplicate calculations. Since each element in the matrix represents the distance from to , so = , that is , so only calculate the upper triangular part of the matrix (that is, the point pairs that satisfy ). However, in this example, all point clouds are traversed, sacrificing some efficiency for simplicity. Calculate the mean value: The average distance of the nearest neighbor point cloud = .

[0045] Step S42: According to the point cloud density, geometric characteristics, curvature sensitivity of lines and planes, and different reflection colors of the target tower body, select a single thinning algorithm or a combination of multiple thinning algorithms.

[0046] Among them, when the point cloud is evenly distributed, use voxels for uniform sampling in space; when the curvature dominates the features, use curvature-sensitive filtering and retain points according to the curvature weight; when both geometry and curvature account for a relatively large proportion, use voxel filtering for the planar geometric region and sample the high-curvature region to retain details; when both reflection color and geometry account for a relatively large proportion, use curvature-protected thinning for the high-reflection region (such as metal) and perform aggressive elimination on the low-reflection region (such as green plants). The specific steps for combining multiple algorithms for thinning are as follows: Step S421: Transform the smallest point cloud unit from a single point cloud into a spatial cube, that is, a voxel, and retain a representative point for each voxel. Define the voxel side length voxel_size (5mm is taken in this example). Traverse all points, map them to the voxel grid they belong to, take the mean value of the points within each voxel and retain the center point, and first downsample in the voxel to reduce 70% of the data.

[0047] Step S422: The processed voxels will lose local details, and further retain local regions with high curvature and rich features. Calculate the curvature of each point: ① Use PCA to analyze the local neighborhood (radius or K-nearest neighbors) to obtain the geometric distribution of the local neighborhood; ② Perform eigenvalue decomposition of the covariance matrix and curvature calculation based on the geometric distribution of the local neighborhood, curvature = λ3 / (λ1 + λ2 + λ3); ③ Sort by curvature, retain high-curvature points, and randomly sample low-curvature regions. Calculate the curvature of the remaining points and retain the top 20% of the points by curvature.

[0048] Step S423: Use statistical filtering to remove outliers.

[0049] Among them, during the thinning process, due to the change in the point cloud density in some regions, it is inevitable to generate new noise points, and these points will float isolated in space, that is, outliers. It is necessary to analyze the statistical characteristics of the point cloud neighborhood density through statistical filtering and remove discrete noise points that do not conform to the main distribution.

[0050] Through this step, the memory occupancy and geometric feature retention can be balanced. The specific parameters are adjusted according to the characteristics of the point cloud (such as the accuracy of the scanning device, the complexity of the scene). After the above thinning steps, the finally obtained point cloud model is as Figure 5 shown.

[0051] Step S5: Perform hierarchical slicing on the lightweight point cloud model, fit the centroid of each layer slice of the tower body by the least squares method to obtain the centroid of each level, and connect the centroids of each level from bottom to top to obtain the offset axis of the tower body.

[0052] Among them, performing hierarchical slicing on the lightweight point cloud model specifically includes: Step S511: Select the center at each layer height of the target tower body as the slice selection point to determine the height of the point cloud slice part.

[0053] Among them, determine the height of the point cloud slice part so that the formed offset axis can represent the offset situation of the overall tower body. In this example, it is a nine-level tower body, with a total of 12 slices including the tower bottom, tower top, and pedestal, as Figure 6 shown.

[0054] Step S512: Determine the thickness of the point cloud slice according to the point cloud density at different positions of the target tower body. Reduce the slice thickness in the point cloud dense area (such as the pedestal, tower bottom), and increase the slice thickness in the point cloud sparse area (such as the tower top), with the number of points in each slice being preferably 1000, so as to determine the thickness of the slice. As thin a slice as possible can reduce the calculation amount and avoid fitting errors caused by too many calculation points.

[0055] When the sliced point cloud cannot form a relatively complete circle, or when the distribution of obstacles is uneven, the algebraic fitting is not accurate enough. In this case, the least squares method of geometric distance is used to solve the center coordinates of the circle. and the radius , that is, to minimize the square of the actual distance from each point to the circle, which is . At this time, the algebraic solution needs to be used as the initial value for optimization.

[0056] Thus, the centroid fitting of each layer of the point cloud slices is performed by the least squares method to obtain the center of the circle of each layer of the point cloud slices, and the centers of the circle of each point cloud slice are connected from bottom to top to obtain the offset axis of the tower body, which specifically includes: Step S521: Establish an algebraic equation of the circle based on the point cloud slice and parameterize it to form a parameterized circle equation.

[0057] Among them, the algebraic equation of the circle is: ; Among them, the coordinates is the center of the circle, is the radius. Parameterize it and organize it into a parameterized circle equation: Make up ; Among them: , , .

[0058] Step S522: Substitute each point cloud coordinate in the point cloud slice into the parameterized circle equation to establish an overdetermined system of equations.

[0059] For each point cloud coordinate , substitute it into the equation to get: , where is the residual. Minimize the sum of the squares of the residuals , to obtain an overdetermined system of equations: .

[0060] Step S523: Use the least squares method to solve the overdetermined system of equations to obtain the radius and center of the point cloud slice.

[0061] Write the overdetermined system of equations in matrix form . The dimension of the point cloud design matrix is . The parameter vector in the matrix is , where each row is , and the dimension is . The solution of the least squares method is . Through the parameter vector Obtain the center and radius of the fitted circle, and the center coordinates are , , and the radius is .

[0062] The center coordinates of the fitted circles of the tower body slices at all levels are shown in Figure 7.

[0063] Step S6: Analyze the inclination of the target tower body according to the offset axis.

[0064] Specifically, according to the bottommost slice circle and the corresponding center coordinates, mark the central axis of the slice circle; the angle between the central axis and the offset axis is the offset angle of the target tower body.

[0065] The embodiment of the present application also discloses a tower-shaped building offset detection device, including: A data acquisition module for acquiring the point cloud data of the target tower body; A model establishment module for registering the point cloud data of different detection stations to obtain an overall point cloud model of the target tower body; A model lightweight module for screening and removing the noise points in the overall point cloud model to obtain a pure point cloud model, and thinning the point cloud of the pure point cloud model to obtain a lightweight point cloud model; A slicing module for performing hierarchical slicing on the lightweight point cloud model according to the detection focus and detection accuracy; An analysis module for performing centroid fitting on each layer slice of the tower body by the least squares method to obtain the centroids of each level, connecting the centroids of each level from bottom to top to obtain the offset axis of the tower body, and analyzing the inclination of the target tower body according to the offset axis.

[0066] Although the present invention has been described in conjunction with specific features and their embodiments, it is obvious that various modifications and combinations can be made without departing from the spirit and scope of the present invention. Accordingly, this specification and the accompanying drawings are merely exemplary descriptions of the present invention defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present invention. Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. A method for detecting the offset of a tower-like building, characterized in that, Including: Collecting the complete point cloud data of the target tower body; Registering the point cloud data of different detection stations to obtain the overall point cloud model of the target tower body; Screening out and removing the noise points in the overall point cloud model to obtain a pure point cloud model; Thinning the point cloud of the pure point cloud model to obtain a lightweight point cloud model; Performing layered slicing on the lightweight point cloud model to obtain multiple point cloud slices; Performing centroid fitting on each layer of the point cloud slices by the least squares method to obtain the centers of the circles of each layer of the point cloud slices, connecting the centers of the circles of each layer of the point cloud slices from bottom to top to obtain the offset axis of the target tower body, and analyzing the inclination of the target tower body according to the offset axis.

2. The tower-shaped building offset detection method according to claim 1, characterized in that The collecting the complete point cloud data of the target tower body specifically includes: Obtaining the top point cloud data of the target tower body by using an unmanned aerial vehicle; Obtaining the side point cloud data of the target tower body by using a ground detection station; Combining the top point cloud data and the side point cloud data to obtain the complete point cloud data of the target tower body.

3. The tower-shaped building offset detection method according to claim 1, characterized in that The registering the point cloud data of different detection stations to obtain the overall point cloud model of the target tower body specifically includes: Establishing a unified coordinate system, the point cloud data is located in the unified coordinate system, and rigid body coordinate transformation of the point cloud data is performed into the global coordinate system; Performing optimization fitting on the overlapping regions of the point cloud data collected by different detection stations and initially fixing these overlapping regions to form a plane with similar characteristics; The point cloud model after registration is tested by different evaluation algorithms to ensure the accuracy of registration.

4. A method for detecting the offset of a tower-like building according to claim 1, characterized in that, The screening out and removing the noise points in the overall point cloud model to obtain a pure point cloud model specifically includes: Screening out the noise point information of typical objects of non-target tower body structures in the overall point cloud model for rough noise removal; After rough noise removal of the overall point cloud model, screening out the noise point information of non-typical objects of non-target tower body structures for fine noise removal to obtain a pure point cloud model.

5. A method for detecting the offset of a tower-shaped building according to claim 4, characterized in that, The screening out the noise point information of typical objects of non-target tower body structures in the overall point cloud model for rough noise removal specifically includes: Performing noise removal according to the gradient of the color or intensity information of the point cloud in the overall point cloud model; or based on the method of geometric gradient, screening and removing the noise point regions with large curvature changes in the overall point cloud model.

6. A method for detecting the offset of a tower-like building according to claim 1, characterized in that, The thinning the point cloud of the pure point cloud model to obtain a lightweight point cloud model specifically includes: Evaluating the point cloud characteristics in the pure point cloud model, and selecting a thinning algorithm according to the point cloud characteristics to thin the pure point cloud model to obtain a lightweight point cloud model.

7. A method for detecting the offset of a tower-like building according to claim 1, characterized in that, The performing layered slicing on the lightweight point cloud model to obtain multiple point cloud slices specifically includes: Selecting the center at each layer height of the target tower body as the section selection point to determine the height of the section part; Determining the thickness of the slice according to the point cloud density at different positions of the target tower body.

8. A method for detecting the offset of a tower-shaped building according to claim 1, characterized in that The performing centroid fitting on each layer of the point cloud slices by the least squares method to obtain the centers of the circles of each layer of the point cloud slices specifically includes: Establishing and parameterizing the algebraic equation of a circle according to the point cloud slice to form a parameterized circle equation; Substituting each point cloud coordinate in the point cloud slice into the parameterized circle equation to establish an overdetermined system of equations; Solve the overdetermined equations by the least squares method to obtain the radius and center of the point cloud slice.

9. A tower-shaped building offset detection device, characterized in that, It includes: A data acquisition module for acquiring the point cloud data of the target tower body; A model establishment module for registering the point cloud data of different detection stations to obtain the overall point cloud model of the target tower body; A model lightweight module for screening out and removing the noise points in the overall point cloud model to obtain a pure point cloud model, and thinning the pure point cloud model to obtain a lightweight point cloud model; A slicing module for performing hierarchical slicing on the lightweight point cloud model according to the detection key points and detection accuracy; An analysis module for performing centroid fitting on each layer slice of the tower body by the least squares method to obtain the centroids of each level, connecting the centroids of each level from bottom to top to obtain the offset axis of the tower body, and analyzing the inclination of the target tower body according to the offset axis.

Images