Window and wall turning angle positioning method for high-rise residential facade point cloud

Through the projection, clustering and skeleton extraction methods of point clouds on the facade of high-rise residential buildings, combined with Freud's algorithm and planar projection processing, the precise positioning of the turning angles between windows and walls is achieved, and the existing domestic and foreign countries are solved The positioning problems under complex structures improve the accuracy of automated processing.

CN118691846BActive Publication Date: 2024-11-26ZHEJIANG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411191417.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-28
Publication Date
2024-11-26
Estimated Expiration
2044-08-28

AI Technical Summary

Technical Problem

The point cloud data structure of the exterior facade of high-rise residential buildings is complex, which makes it difficult to automatically handle the positioning of the turning angles of windows and walls, and the existing technology cannot effectively and accurately position.

Method used

By projecting the three-dimensional point clouds and counting the density distribution, the single-floor point clouds are segmented using the non-maximum suppression method, noise is removed based on DBSCAN clustering, skeleton points are extracted, and the wall corners are determined using the Freud algorithm, combining planar projection and Rasterizes the positioning of window points.

Benefits of technology

It realizes the automated processing of the point cloud of the facade of high-rise residential buildings, accurately locates the turning angles of windows and walls, solves the problem of difficult to extract the features of the facade, and improves the accuracy of the positioning of the facade of the building.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118691846B_ABST
    Figure CN118691846B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for locating the turning angle of a window and a wall applicable to a point cloud of a high-rise residential building facade, the method comprising: projecting a three-dimensional point cloud in a vertical direction, and statistically analyzing the density distribution, determining the floor interval points by means of a non-maximum suppression method, and segmenting the point cloud into a single floor; projecting the single floor point cloud to a horizontal plane and rasterizing it, clustering the grid points based on DBSCAN, and obtaining a grid point group of a single wall; for each grid point group of the wall, determining the two endpoints of the skeleton based on the Floyd algorithm, and determining the turning angle based on the direction change of the point; determining the rough plane of the window and rasterizing it, expanding and corroding the grid map, determining the bounding box of the blank area in the grid map, and using the points selected by the box as window points. The present invention only needs to collect the point cloud data of the facade of a high-rise residential building for analysis, can perform layered processing on the residential facade point cloud, and further segment the window and turning angle parts, solving the problem that the facade features are difficult to extract.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of high-rise residential building detection, and in particular to a method for locating turning angles of windows and walls applicable to a point cloud of a facade of a high-rise residential building. Background Art

[0002] At present, in the spraying work of the exterior walls of high-rise residential buildings, it is necessary to locate the turning angles of the wall to determine the boundaries of a single spraying area, and it is also necessary to locate the windows on the exterior facade of the building to determine the areas that need to be avoided during spraying. The positioning of the spraying target in traditional spraying operations relies on the workers' experience and habits to perceive and locate based on the naked eye. With the gradual application of exterior wall spraying robots, the realization of non-contact perception and positioning of the exterior walls of high-rise residential buildings has become a very important part of robot intelligence.

[0003] The point cloud data obtained using 3D LiDAR can accurately reflect the geometric shape of the building facade. However, the point cloud of the high-rise residential facade has problems such as complex structure, many messy points, and uneven distribution. Current technology faces some challenges in processing point cloud data. Due to the complexity of the wall, it cannot be simply processed as a plane. Therefore, it is necessary to design a method for locating the turning angles of windows and walls suitable for the point cloud of the high-rise residential facade, realize the automated processing of the point cloud of the high-rise residential facade, and realize the positioning of the turning angles of windows and walls. Summary of the invention

[0004] The purpose of the present invention is to provide a method for locating window and wall turning angles suitable for high-rise residential facade point clouds in view of the shortcomings of the prior art. The present invention only needs to obtain the high-rise residential facade point cloud for processing, and can realize the accurate positioning of window and wall turning angles.

[0005] The object of the present invention is achieved through the following technical scheme: A method for locating the turning angles of windows and walls applicable to the point cloud of the facade of a high-rise residential building, comprising the following steps:

[0006] (1) The 3D point cloud is projected in a direction perpendicular to the high-rise residential building, and its density distribution is counted. With the help of the non-maximum suppression method, the floor interval points are determined according to the maximum value of the density distribution curve to segment the 3D point cloud into single-floor point clouds.

[0007] (2) Project the single-floor point cloud onto the horizontal plane and rasterize it. Cluster the grid points based on the DBSCAN clustering method and remove noise to obtain the grid point group of a single wall.

[0008] (3) For each grid point group corresponding to the wall, extract the skeleton of the grid image, then determine the two endpoints of the line connecting the skeleton points based on the Floyd algorithm, traverse the skeleton points in sequence, and determine the wall corner based on the change of the direction vector of the point;

[0009] (4) Determine the point plane where the window point is located based on the close endpoints of the grid point groups on the two walls, and use the points within a preset distance threshold from the point plane as coarse screening points. Project the coarse screening points onto the window plane and rasterize them. Dilate and erode the projected grid image to remove noise points. Then find the minimum bounding box of the blank area in the grid image, set the border width based on the minimum bounding box, and use the points selected by the box in the coarse screening points as window points.

[0010] Furthermore, the step (1) specifically includes:

[0011] The three-dimensional point cloud is projected in a direction perpendicular to the height of the high-rise residential building, and the one-dimensional density distribution of the point cloud is statistically analyzed to obtain the density distribution curve of the point cloud; the maximum point of the density distribution curve is found, the floor interval length is set, and each maximum point within the floor interval length is traversed using the non-maximum suppression method. For each maximum point, if there are other maximum points larger than the maximum point within the floor interval length, the maximum point is suppressed. After the traversal is completed, the remaining maximum points are the floor interval points; the three-dimensional point cloud of the floor is segmented according to the coordinates of the floor interval points to obtain a single floor point cloud.

[0012] Furthermore, the step (2) specifically includes:

[0013] The single-floor point cloud is projected onto the horizontal plane and rasterized. The grid points are clustered using the DBSCAN clustering method to obtain multiple cluster point groups. If the number of grid points in a single cluster point group is less than the preset grid point number threshold, all grid points in the cluster point group are considered to be noise points, and the noise points are removed. The noise point removal process is repeated until the first two largest cluster point groups remain. This cluster point group is the clustering of the wall grid point group, and the grid point group of a single wall is obtained based on this.

[0014] Furthermore, the step (3) specifically includes:

[0015] For each grid point group corresponding to the wall, the skeleton of each cluster point in its grid map is extracted; for all grid point pairs in a single cluster, the Floyd algorithm is used to determine its shortest path; the grid point pair corresponding to the maximum value of the shortest path is the two endpoints of the wall; starting from any endpoint, its neighboring points are continuously added to the list as the next point in order, so as to reorder the points in the cluster in order and obtain an ordered point set; let the ordered point set be , set the point interval to n, set the angle threshold to AT, and take the endpoint as the starting point , traverse backward until you find the point Make , then Mark the turning point and use it as the new starting point , continue to traverse the points in the ordered point set until the end, where Representation vector and vector All the turning points obtained are the turning angles of the wall. In the three-dimensional point cloud, the points within the preset radius at the corresponding position of the turning angle are taken as the turning angle of the wall to achieve the segmentation of the turning angle of the wall.

[0016] Furthermore, the step (4) specifically includes:

[0017] According to the two pairs of wall endpoints corresponding to the two walls determined in step (3), find the two endpoints with the shortest distance between different wall endpoint pairs, and determine the projection of the rough plane of the window according to the connection line between the two endpoint pairs; map the projection of the rough plane of the window back to the original three-dimensional point cloud of the single floor to determine the point plane where the window point cloud is located, and take all three-dimensional point clouds within a preset distance threshold from the point plane as coarse screening points; project the coarse screening points onto the rough plane of the window and perform rasterization processing, and use a 3×3 core to expand and erode the raster image to remove noise points; find the minimum and maximum values ​​of the two-dimensional coordinates of the blank area in the raster image, and determine the minimum bounding box of the window with the maximum value; set a border width for the minimum bounding box, expand the size of the bounding box based on the border width, and use the points selected by the bounding box as window points for two-dimensional segmentation, and map the results of the two-dimensional segmentation back to the three-dimensional point cloud to achieve segmentation of the window points.

[0018] The beneficial effects of the present invention are that the present invention only needs to collect point cloud data of the facade of a high-rise residential building for analysis, and can perform layered processing on the point cloud of the residential facade, and further segment the windows and turning corners, thereby solving the problem of difficulty in extracting facade features; the present invention realizes the positioning of the wall of the building facade through projection, clustering and other methods, and realizes the positioning of the wall turning corner by analyzing the change in the direction vector between the wall points; and locates the window point cloud in the coarse screening points through the plane projection rasterization method, thereby realizing the positioning of the window. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of the method for locating the turning angles of windows and walls applicable to the point cloud of the facade of a high-rise residential building of the present invention;

[0020] Figure 2 It is a schematic diagram of point cloud density distribution after one-dimensional projection of the facade of a high-rise residential building in the height direction of the present invention;

[0021] Figure 3 It is a grid image of a single-layer facade point cloud projected on a horizontal plane and a result image after removing noise points;

[0022] Figure 4 This is a schematic diagram of the single-layer facade wall skeleton extraction effect of the present invention;

[0023] Figure 5 It is the external wall skeleton endpoint search result diagram of the present invention;

[0024] Figure 6 It is the result diagram of the external wall segmentation of the present invention;

[0025] Figure 7 It is a schematic diagram of the window grid image processing effect and the bounding box positioning result of the present invention. DETAILED DESCRIPTION

[0026] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims. It should be understood that the above general description and the detailed description that follows are merely exemplary and explanatory and cannot limit the present application.

[0027] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0028] It should be understood that, although the term first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of..." or "when..." or "in response to determination". Moreover, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process and method including a series of elements include not only those elements, but also include other elements that are not clearly listed, or also include elements inherent to such process and method. In the absence of more restrictions, the elements limited by the statement "including one..." do not exclude the existence of other identical elements in the process, method, article or equipment including the elements.

[0029] The present invention is described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the features of the following embodiments and implementations can be combined with each other.

[0030] See also Figure 1 The window and wall turning angle positioning method applicable to the point cloud of the facade of a high-rise residential building of the present invention specifically comprises the following steps:

[0031] (1) The 3D point cloud is projected in a direction perpendicular to the high-rise residential building, and its density distribution is counted. With the help of the non-maximum suppression method, the floor interval points are determined according to the maximum value of the density distribution curve to segment the 3D point cloud into single-floor point clouds.

[0032] Specifically, the three-dimensional point cloud is projected along the direction perpendicular to the height of the high-rise residential building, and the one-dimensional density distribution of the point cloud is statistically analyzed to obtain the density distribution curve of the point cloud, such as Figure 2 As shown; find the maximum point of the density distribution curve, set the floor interval length, and use the non-maximum suppression method to traverse each maximum point within the floor interval length. For each maximum point, if there are other maximum points larger than the maximum point within the floor interval length, the maximum point is suppressed. After the traversal is completed, the remaining maximum points are the floor interval points; the three-dimensional point cloud of the floor is segmented according to the coordinates of the floor interval points to obtain a single floor point cloud. Among them, the floor interval length can be selected according to the general height of the actual floor, and the following conditions are met: floor interval length*grid length<common height of the actual floor.

[0033] It should be understood that due to the volatility of data, e.g. Figure 2 As shown in the figure, there are multiple density peaks, so there may be many maximum points in the density distribution curve, so it is necessary to merge the data within a certain range. Since the floor interval length is set according to the general height of the floor, there can only be one maximum point in a floor interval length, and other maximum points will be suppressed. Due to the changing characteristics of the density distribution curve, the location of the maximum point is the interval between floors, and the remaining points after filtering out the redundant maximum points are the floor interval points.

[0034] (2) The single-floor point cloud is projected onto a horizontal plane and rasterized. The grid points are clustered and noise is removed based on the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering method to obtain the grid point group of a single wall.

[0035] Specifically, the single-floor point cloud is projected onto the horizontal plane and rasterized. The grid points are clustered using the DBSCAN clustering method to obtain multiple cluster point groups. If the number of grid points in a single cluster point group is less than the preset grid point number threshold, all grid points in the cluster point group are considered to be noise points, and the noise points are removed. The noise point removal process is repeated until the first two largest cluster point groups remain. This cluster point group is the cluster of the wall grid point group, and the grid point group of a single wall is obtained, such as Figure 3 As shown in the figure, the result of rasterization after projecting the single-floor point cloud to the horizontal plane is shown. If there is a point in the grid, it is set to 1 (displayed in white). The horizontal and vertical axes are the coordinates of the grid points, which are on the same horizontal plane as the original point cloud, but there is a certain proportional relationship in the values.

[0036] It should be understood that the DBSCAN clustering method is an unsupervised density-based noise application spatial clustering method, which can determine the category of points based on the density of neighborhood points without setting the number of clusters in advance.

[0037] (3) For each grid point group corresponding to the wall, extract the skeleton of the grid image, and then determine the two endpoints of the line connecting the skeleton points based on the Floyd algorithm. Traverse the skeleton points in sequence and determine the wall corner based on the change of the direction vector of the point.

[0038] Specifically, for each grid point group corresponding to each wall, the skeleton of each cluster point in its grid map is extracted, such as Figure 4 As shown in the figure, the data can be simplified to speed up the processing. To determine the topological relationship of the grid points in each cluster, the Floyd algorithm is used to determine the shortest path for all grid point pairs in a single cluster. The grid point pair corresponding to the maximum value of the shortest path is the two endpoints of the wall, as shown in Figure 5 As shown in , the horizontal and vertical axes represent the grid coordinates of the grid map; starting from any endpoint, its neighboring points are continuously added to the list as the next point in order, so as to reorder the points in the cluster in order and obtain an ordered point set; let the ordered point set be , set the point interval to n, set the angle threshold to AT, and take the endpoint as the starting point , traverse backward until you find the point Make , then Mark the turning point and use it as the new starting point , continue to traverse the points in the ordered point set until the end, where Representation vector and vector All the turning points obtained are the turning angles of the wall. In the three-dimensional point cloud, the points within a preset radius at the corresponding position of the turning angle are taken as the turning angle of the wall to achieve the segmentation of the turning angle of the wall. The corresponding wall segmentation results are as follows: Figure 6 As shown, the intersection of each segment is the location of the turning angle of the exterior wall, and its horizontal and vertical coordinates respectively represent the results of the rasterization of the x and y coordinates of the origin cloud.

[0039] It should be understood that the Floyd algorithm, also known as the interpolation method, is an algorithm for finding the shortest path between multiple source points in a given weighted graph. The Floyd algorithm is based on the distance matrix between points, and through triple traversal, confirms that each point is the middle point of the path to find the shortest distance of all grid point pairs.

[0040] (4) The point plane where the window point is located is determined based on the close endpoints of the grid point groups on the two walls. Points within a preset distance threshold from the point plane are used as coarse screening points. The coarse screening points are projected onto the window plane and rasterized. The projected grid image is dilated and eroded to remove noise points. The minimum bounding box of the blank area in the grid image is then found. A certain border width is set based on the minimum bounding box. The points selected in the box are used as window points in the coarse screening points.

[0041] Specifically, according to the two pairs of wall endpoints corresponding to the two walls determined in step (3), find the two endpoints with the shortest distance between different wall endpoint pairs, and determine the projection of the rough plane of the window according to the connection line between the two endpoint pairs; map the projection of the rough plane of the window back to the original three-dimensional point cloud of the single floor to determine the point plane where the window point cloud is located, and take all three-dimensional point clouds within a preset certain distance threshold from the point plane as coarse screening points; project the coarse screening points onto the rough plane of the window and perform rasterization processing, and use a 3×3 kernel to dilate and erode the raster image to remove noise points, such as Figure 7 As shown; find the minimum and maximum values ​​of the two-dimensional coordinates of the blank area in the grid image, and use the maximum values ​​to determine the minimum bounding box of the window; set a certain border width for the minimum bounding box, expand the size of the bounding box based on the border width, and use the points selected by the bounding box as window points for two-dimensional segmentation, and map the results of the two-dimensional segmentation back to the three-dimensional point cloud to achieve segmentation of the window points, which can avoid omissions in segmentation.

[0042] It should be understood that by finding the two endpoints that are closer to each other between different pairs of wall endpoints, the two endpoints determine a line segment; since the window is perpendicular to the horizontal plane (i.e. the ground), the projection of the window on the horizontal plane is also a line segment, therefore, the line segment determined by the two endpoints is the projection of the rough plane of the window.

[0043] Furthermore, dilation and erosion are common processing methods for image pixels. In terms of implementation, the grid image is regarded as a matrix, and the matrix elements are 0 or 1, indicating whether there is a point in the grid. In the case of dilation, all 3×3 positions around all pixels with values ​​of 1 are set to 1; in the case of erosion, a 3×3 matrix of all 1s is used as the core, and the grid image is traversed from the upper left corner of the grid image. Only when the elements at the traversed position are completely consistent with the core (that is, the element and its surrounding elements also form a 3×3 matrix of all 1s), the corresponding core position of the new grid image is set to 1, and the corresponding core positions in other cases are set to 0. Through a single dilation and erosion, the scattered holes and noise points in the grid image can be effectively reduced, making the result more stable.

[0044] In summary, the present invention only needs to collect point cloud data of the facade of a high-rise residential building for analysis, can perform layered processing on the point cloud of the residential facade, and further segment the windows and turning corners, thereby solving the problem of difficulty in extracting facade features; the present invention realizes the positioning of the building facade wall through projection, clustering and other methods, and realizes the positioning of the wall turning corner by analyzing the change in the direction vector between the wall points; and locates the window point cloud in the coarse screening points through the plane projection rasterization method, thereby realizing the positioning of the window.

[0045] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for locating the turning angles of windows and walls for point clouds of high-rise residential facades. It is characterized in that The following steps are involved: (1) The 3D point cloud is projected in a direction perpendicular to the high-rise residential building, and its density distribution is counted. With the help of the non-maximum suppression method, the floor interval points are determined according to the maximum value of the density distribution curve to segment the 3D point cloud into single-floor point clouds. (2) Project the single-floor point cloud onto the horizontal plane and rasterize it. Cluster the grid points based on the DBSCAN clustering method and remove noise to obtain the grid point group of a single wall. (3) For each grid point group corresponding to the wall, extract the skeleton of the grid image, then determine the two endpoints of the line connecting the skeleton points based on the Floyd algorithm, traverse the skeleton points in sequence, and determine the wall corner based on the change of the direction vector of the point; (4) Determine the point plane where the window point is located based on the close endpoints of the grid point groups on the two walls, and use the points within a preset distance threshold from the point plane as coarse screening points. Project the coarse screening points onto the window plane and rasterize them. Dilate and erode the projected grid image to remove noise points. Then find the minimum bounding box of the blank area in the grid image, set the border width based on the minimum bounding box, and use the points selected by the box in the coarse screening points as window points.

2. According to claim 1, the method for locating the turning angles of windows and walls applicable to the point cloud of the facade of a high-rise residential building, It is characterized in that The step (1) specifically includes: The three-dimensional point cloud is projected in a direction perpendicular to the height of the high-rise residential building, and the one-dimensional density distribution of the point cloud is statistically analyzed to obtain the density distribution curve of the point cloud; the maximum point of the density distribution curve is found, the floor interval length is set, and each maximum point within the floor interval length is traversed using the non-maximum suppression method. For each maximum point, if there are other maximum points larger than the maximum point within the floor interval length, the maximum point is suppressed. After the traversal is completed, the remaining maximum points are the floor interval points; the three-dimensional point cloud of the floor is segmented according to the coordinates of the floor interval points to obtain a single floor point cloud.

3. According to claim 1, the method for locating the turning angles of windows and walls applicable to the point cloud of the facade of a high-rise residential building, It is characterized in that The step (2) specifically includes: The single-floor point cloud is projected onto the horizontal plane and rasterized. The grid points are clustered using the DBSCAN clustering method to obtain multiple cluster point groups. If the number of grid points in a single cluster point group is less than the preset grid point number threshold, all grid points in the cluster point group are considered to be noise points, and the noise points are removed. The noise point removal process is repeated until the first two largest cluster point groups remain. This cluster point group is the clustering of the wall grid point group, and the grid point group of a single wall is obtained based on this.

4. According to claim 1, the method for locating the turning angles of windows and walls applicable to the point cloud of the facade of a high-rise residential building, It is characterized in that The step (3) specifically includes: For each grid point group corresponding to the wall, the skeleton of each cluster point in its grid map is extracted; for all grid point pairs in a single cluster, the Floyd algorithm is used to determine its shortest path; the grid point pair corresponding to the maximum value of the shortest path is the two endpoints of the wall; starting from any endpoint, its neighboring points are continuously added to the list as the next point in order, so as to reorder the points in the cluster in order and obtain an ordered point set; let the ordered point set be , set the point interval to n, set the angle threshold to AT, and take the endpoint as the starting point , traverse backward until you find the point Make , then Mark the turning point and use it as the new starting point , continue to traverse the points in the ordered point set until the end, where Representation vector and vector All the turning points obtained are the turning angles of the wall. In the three-dimensional point cloud, the points within the preset radius at the corresponding position of the turning angle are taken as the turning angle of the wall to achieve the segmentation of the turning angle of the wall.

5. According to claim 1, the method for locating the turning angles of windows and walls applicable to the point cloud of the facade of a high-rise residential building, It is characterized in that The step (4) specifically includes: According to the two pairs of wall endpoints corresponding to the two walls determined in step (3), find the two endpoints with the shortest distance between different wall endpoint pairs, and determine the projection of the rough plane of the window according to the connection line between the two endpoint pairs; map the projection of the rough plane of the window back to the original three-dimensional point cloud of the single floor to determine the point plane where the window point cloud is located, and take all three-dimensional point clouds within a preset distance threshold from the point plane as coarse screening points; project the coarse screening points onto the rough plane of the window and perform rasterization processing, and use a 3×3 core to expand and erode the raster image to remove noise points; find the minimum and maximum values ​​of the two-dimensional coordinates of the blank area in the raster image, and determine the minimum bounding box of the window with the maximum value; set a border width for the minimum bounding box, expand the size of the bounding box based on the border width, and use the points selected by the bounding box as window points for two-dimensional segmentation, and map the results of the two-dimensional segmentation back to the three-dimensional point cloud to achieve segmentation of the window points.