A power distribution room point cloud fine segmentation method considering boundary features in panoramic images

Abstract

The application discloses a power distribution room point cloud fine segmentation method considering boundary features in panoramic images and belongs to the technical field of three-dimensional modeling. The method steps are as follows: S1, using a laser scanner to perform three-dimensional scanning on equipment in a power distribution room to obtain point cloud data; S2, according to the point cloud data obtained in step S1, the point cloud data is divided into a plurality of grid units by using a three-dimensional regular grid, and the plurality of grid units jointly construct a three-dimensional regular grid; S3, according to the three-dimensional regular grid constructed in the step, coarse segmentation and fine segmentation of the point cloud data are performed; and S4, merging and establishing of the point cloud data, taking a layer with the highest point cloud data integrity as a reference layer, setting a height difference threshold, calculating coordinate points of equipment in the power distribution room according to the threshold and a projection curve, and importing the coordinate points obtained after the calculation into three-dimensional design software to perform lofting processing, so that the correction effect of discrete data can be effectively utilized while the discrete data is removed, and the accuracy of model establishment is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of modeling, and in particular relates to a power distribution room point cloud fine segmentation method considering boundary features in panoramic images. BACKGROUND

[0002] Effective segmentation of point cloud data is the basis and premise of data processing, and has extremely wide applications in 3D printing, virtual reality, augmented reality, smart city, intelligent transportation and other fields. Common point cloud data segmentation techniques mainly include RanSaC algorithm, Euclidean algorithm, minimum cut algorithm and super body clustering segmentation algorithm, and the segmentation of point cloud data mainly includes the process of dividing point cloud data into several mutually disjoint subsets according to certain attributes or rules.

[0003] In the existing power distribution room point cloud data segmentation method, since there are many devices in the power distribution room, the point cloud data obtained by three-dimensional scanning has many discrete data, and it is difficult to remove the discrete points in the point cloud data obtained by scanning in the power distribution room. Therefore, a power distribution room point cloud data segmentation method based on device features is disclosed in Chinese patent application No. 202110152838.2, which can effectively segment the point cloud data constructed by scanning. However, this method has certain limitations in use, such as the formation of discrete point cloud after point cloud data segmentation. The point cloud can correct the fitted curve during model construction, but the above method cannot use the discrete point cloud to correct the fitted curve, resulting in that the accuracy of the fitted curve cannot be guaranteed. Therefore, it is necessary to design a method for guaranteeing high accuracy of curve fitting. SUMMARY

[0004] In view of the problem that the existing point cloud data segmentation method cannot effectively remove discrete data, the application provides a power distribution room point cloud fine segmentation method considering boundary features in panoramic images. The segmentation method is mainly for point cloud data obtained by three-dimensional scanning in the early stage, and determines the fitting points by establishing a seed grid, determining the fitting points and direction constraint growth. The method effectively utilizes the correction effect of discrete point cloud to guarantee the accuracy of the fitted curve and effectively solves the problem proposed in the background art. To solve the above problems, the application adopts the following technical solutions.

[0005] A power distribution room point cloud fine segmentation method considering boundary features in panoramic images, comprising the following steps:

[0006] S1, using a laser scanner to perform three-dimensional scanning on the equipment in the power distribution room to obtain a three-dimensional digital model, i.e., point cloud data of the power distribution room equipment;

[0007] S2, the point cloud data obtained in step S1 is divided into a plurality of grid units by using a three-dimensional regular grid, and the plurality of grid units jointly construct a three-dimensional regular grid; the three-dimensional regular grid is divided into a plurality of layers through layering processing;

[0008] S3, the three-dimensional regular grid constructed in step is used for point cloud data segmentation;

[0009] S4, merging and establishing of point cloud data, taking the layer with the highest point cloud data integrity as a reference layer, setting a height difference threshold, calculating the coordinate points of the equipment in the power distribution room according to the threshold and the projection curve, importing the coordinate points obtained after the calculation into a three-dimensional design software for lofting processing, and generating a three-dimensional model of the equipment in the power distribution room;

[0010] In step S3, the point cloud data segmentation includes seed grid establishment, fitting point determination, and direction constraint growth, wherein the fitting point determination includes coarse segmentation of point cloud data and fine segmentation of point cloud data.

[0011] Preferably, the point cloud data segmentation in step S3 includes the following steps:

[0012] S31, seed grid establishment: constructing a seed grid for each equipment according to the distribution of the equipment in the power distribution room;

[0013] S32, fitting point determination: projecting the point cloud in the constructed three-dimensional regular grid into an xoy coordinate system, coarsely segmenting the point cloud data and determining the fitting points after segmentation, and fitting the projection lines in the three-dimensional regular grid with the seed grid constructed in step S31 as the center by using the least square method to determine the growth direction;

[0014] S33, direction constraint growth: after the fitting points are determined, the projection line fitting is performed in the xoy plane by using the least square method to determine the growth direction, and the specific steps are as follows: taking the seed grid as the center, clustering and growing the point cloud of the feature equipment in the power distribution room along the growth direction with the seed grid as the center, thereby completing the segmentation.

[0015] Preferably, in step S3, the fine segmentation method of the point cloud data is as follows:

[0016] In the set Q i , any point q i (x i ,y i ) is selected, and any point p j (x j ,y j ) is selected in the set P j , the distance d ij between the two points is calculated based on the coordinates of the two points, and the distance d ijThe minimum value set P j Point p in j (x j ,y j ) is close to the set Q i Point q in i (x i ,y i ), determine q i (x i ,y i ) plane grid, fitting p j (x j ,y j ) in q i (x i ,y i ) within the plane grid.

[0017] Preferably, q i (x i ,y i ) and p j (x j ,y j )The distance between two points

[0018] Where: (x i ,y i ) is q i coordinates of

[0019] (x j ,y j ) is p j 's coordinates.

[0020] In order to further improve the accuracy of model building, in the above point cloud data segmentation process, the seed grid in step S31 is characterized by at least 2 point densities within the grid, and in addition, the grids adjacent to the seed grid have at least 4 point densities.

[0021] Preferably, in step S32, in order to ensure the accuracy of the fitted projection line, the projection center point (x, y) of the point cloud in the plane grid is selected as the fitting point of the projection line, and the fitting calculation formula is as follows:

[0022]

[0023] Where: x i 、y i is the plane coordinate of the i-th projection point in the plane grid, and n is the number of projection points in the plane grid.

[0024] Preferably, in step S33, the calculation formula for projection line fitting is as follows:

[0025] y=kx+b

[0026]

[0027] In the formula: k is the slope of the projection line, b is the solution of the projection line, x i , y i is the center of gravity coordinate point of each projection grid inside the point cloud, and n is the total number of grids participating in fitting.

[0028] Advantages

[0029] Compared with the prior art, the advantages of the present application are:

[0030] (1) The power distribution room point cloud data segmentation method in the present application includes seed grid determination, fitting point determination and direction constraint growth. Through the above segmentation method, the discrete data in the point cloud data obtained in three-dimensional scanning can be effectively removed, so that the effective data can be quickly fitted to form a projection curve. In addition, while removing the discrete data, the correction effect according to the discrete point cloud is utilized, effectively ensuring the accuracy of the fitted curve. Therefore, the coordinate points of the equipment in the power distribution room can be accurately calculated according to the projection curve and the threshold value, and a model is established after lofting processing, thereby improving the precision of model establishment and avoiding the problem of low resolution after traditional model establishment.

[0031] (2) The seed grid feature selected in the present application is that there are at least 2 point densities in the grid, and there are at least 4 point densities in the adjacent grid of the seed grid, which improves the accuracy of seed grid selection. In addition, the correction effect of discrete point data in the process of fitting the curve ensures the accuracy of the fitted curve, thereby further improving the accuracy of model establishment after the model is generated. The model has high recognition degree and the model modeling speed is fast, which reduces the workload of people and improves the work efficiency of people. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 is a point cloud data segmentation process diagram in the present application;

[0033] Figure 2 is a principle block diagram of point cloud data segmentation in the present application;

[0034] Figure 3 is a rough segmentation and fine segmentation of point cloud data in the present application;

[0035] Figure 4 is a structure diagram of point cloud data layer construction coordinate system in the present application. DETAILED DESCRIPTION

[0036] The present application will be further described below in combination with specific application.

[0037] Example 1

[0038] As Figure 1 shown, it is a power distribution room point cloud fine segmentation method considering the boundary features in panoramic image, the segmentation method of the embodiment includes the following steps:

[0039] S1, using a laser scanner to perform three-dimensional scanning on the equipment in the power distribution room to obtain a three-dimensional digital model, i.e. point cloud data of the power distribution room equipment;

[0040] S2, according to the point cloud data obtained in step S1, divide into multiple grid units using a three-dimensional regular grid, and the multiple grid units jointly build a three-dimensional regular grid; the point cloud data in the three-dimensional regular grid has the spatial distribution characteristics of large height difference, planar distribution and high density in the unit area; according to the hierarchical processing of the grid units, the three-dimensional regular grid is divided into multiple layers;

[0041] S3, according to the three-dimensional regular grid constructed in the step, perform point cloud data segmentation;

[0042] S31, seed grid establishment: according to the distribution of the equipment in the power distribution room, construct a seed grid for each equipment, and the characteristics of the seed grid are that there are at least 2 point densities in the grid, and at least 4 point densities in the adjacent grid of the seed grid;

[0043] S32, determination of fitting points: determine the fitting points of the point cloud after coarse segmentation and fine segmentation of the point cloud data, in the embodiment, the point cloud data segmentation in the above S3 includes coarse segmentation of the point cloud data and fine segmentation of the point cloud data, wherein the coarse segmentation method of the point cloud data includes the following steps:

[0044] S321, coarse segmentation of point cloud data: project the point cloud in the constructed three-dimensional regular grid into the xoy coordinate system, and use the least square method to fit the projection line in the three-dimensional regular grid with the seed grid constructed in step S31 as the center to determine the growth direction; in step S321, in order to ensure the accuracy of the fitted projection line, the projection center point (x, y) of the point cloud in the plane grid is selected as the fitting point of the projection line, and the calculation formula of the fitting is as follows:

[0045]

[0046] In the formula: x i , y i are the plane coordinates of the i-th projection point in the plane grid, and n is the number of projection points in the plane grid;

[0047] In the embodiment, after coarse segmentation of the point cloud data, the point cloud data formed after fitting constructs a set Q iWherein i = 1, 2, 3 ··· n, in this embodiment, the point cloud data is roughly segmented, and the remaining point cloud data constructs a set P j Wherein j = 1, 2, 3 ··· m.

[0048] S322, fine segmentation of point cloud data: in set Q i Any point q i (x i ,y i ) is selected, and any point p j (x j ,y j ) is selected in set P j The distance d ij Between two points is calculated based on the distance calculation method of two point coordinates. The minimum value of distance d ij The point p j (x j ,y j ) in set P j Is close to the point q i (x i ,y i ) in set Q i , determine the plane grid of q i (x i ,y i ), fit p j (x j ,y j ) in the plane grid of q i (x i ,y i ).

[0049] S33, direction constraint growth: after determining the fitting point, in the xoy plane, the projection line fitting is determined by using the least square method to determine the growth direction, the specific steps are as follows: taking the seed grid as the center, the feature equipment point cloud in the power distribution room is clustered and grown along the growth direction with the seed grid as the center, so as to complete the segmentation.

[0050] In the above step S33, the calculation formula of projection line fitting is as follows:

[0051] y = kx + b

[0052]

[0053] In the formula: k is the slope of the projection line, b is the solution of the projection line, x i , y i The center of gravity coordinate point of point cloud in each projection grid, n is the total number of grids participating in fitting.

[0054] S4, merging and establishing of the point cloud data, taking a layer with the highest point cloud data integrity as a reference layer, setting a height difference threshold, calculating coordinate points of the equipment in the power distribution room according to the threshold and the projection curve, importing the coordinate points obtained after the calculation into a three-dimensional design software to perform lofting processing, and generating a three-dimensional model of the equipment in the power distribution room.

[0055] The point cloud data segmentation method can effectively remove discrete data in the point cloud data obtained through three-dimensional scanning, can quickly fit the effective data to form a projection curve, can accurately calculate the coordinate points of the equipment in the power distribution room according to the projection curve and the threshold, can establish a model after lofting processing, and can improve the precision of model establishment.

[0056] The above is a further detailed description of the present application in combination with the specific embodiments, and cannot be considered as limiting the specific embodiments of the present application to these descriptions. For ordinary skilled persons in the technical field to which the present application belongs, some simple deductions or replacements can be made without departing from the concept of the present application, and all of them should be considered as falling within the protection scope of the claims of the present application.

Claims

1. A method for fine segmentation of distribution room point clouds taking into account boundary features in panoramic images, comprising the following steps: S1. Use a laser scanner to perform a three-dimensional scan of the equipment in the power distribution room to obtain a three-dimensional digital model, i.e., point cloud data of the equipment in the power distribution room; S2. Dividing the point cloud data obtained in step S1 into a plurality of grid units using a three-dimensional regular grid, and constructing a three-dimensional regular grid with the plurality of grid units; performing layered processing on the grid units, and dividing the constructed three-dimensional regular grid into multiple layers; S3. Segment the point cloud data according to the three-dimensional regular grid constructed in step S3; S4. Merge and create point cloud data. Use the layer with the highest point cloud data integrity as the base layer, set a height difference threshold, and calculate the coordinate points of the equipment in the power distribution room based on the threshold and the projection curve. Import the calculated coordinate points into the 3D design software for layout processing to generate a 3D model of the power distribution room equipment. Its characteristics are: In step S3, the point cloud data segmentation includes the establishment of a seed grid, the determination of fitting points, and direction-constrained growth. When the fitting points are determined, the coarse segmentation of the point cloud data and the fine segmentation of the point cloud data are performed; The point cloud data segmentation in step S3 includes the following steps: S31. Establishing a seed grid: Build a seed grid for each device based on the distribution of devices in the power distribution room; S32, determining fitting points: projecting the point cloud within the constructed three-dimensional regular grid into the Xoy coordinate system, roughly segmenting the point cloud data and determining fitting points after segmentation, fitting the projection lines within the three-dimensional regular grid using the least squares method with the seed grid constructed in step S31 as the center to determine the growth direction; S33, Direction Constrained Growth: After the fitting points are determined, the least squares method is used to perform projection line fitting in the xoy plane to determine the growth direction. The specific steps are as follows: With the seed grid as the center, the feature equipment point cloud in the distribution room is clustered and grown along the growth direction with the seed grid as the center, thereby completing the segmentation; After the point cloud data is roughly segmented, the point cloud data formed after fitting constructs the set Q i , where i = 1, 2, 3······n, the point cloud data is roughly segmented, and the remaining point cloud data constructs the set P j , where j = 1, 2, 3······m; In step S3, the fine segmentation method of point cloud data is as follows: In the set Q i Choose any point as q i (x i ,y i ), in the set P j Pick any point p j (x j ,y j ), calculate the distance d between two points based on the distance calculation method of the two point coordinates ij , distance d ij The minimum value set P j Point p in j (x j ,y j ) is close to the set Q i Point q in i (x i ,y i ), determine q i (x i ,y i ) plane grid, fitting p j (x j ,y j ) in q i (x i ,y i ) within the plane grid; The characteristic of the seed grid in step S31 is that there are at least 2 point densities in the grid, and in addition, the grids adjacent to the seed grid have at least 4 point densities; In step S32, in order to ensure the accuracy of the fitted projection line, the projection center point (x, y) of the point cloud in the plane grid is selected as the fitting point of the projection line. The fitting calculation formula is as follows: Where: x i 、y i is the plane coordinate of the i-th projection point in the plane grid, and n is the number of projection points in the plane grid; In step S33, the calculation formula for projection line fitting is as follows: y=kx+b Where: k is the slope of the projection line, b is the solution of the projection line, x i ,y i is the centroid coordinate point of the point cloud in each projected grid, and n is the total number of grids involved in the fitting.

2. The method for fine segmentation of distribution room point cloud taking into account boundary features in panoramic images according to claim 1 is characterized in that: q i (x i ,y i ) and p j (x j ,y j )The distance between two points Where: (x i ,y i ) is q i coordinates; (x j ,y j ) is p j 's coordinates.

Images