A pipe network refined modeling method and system

Abstract

The present application discloses a method for fine-grained modeling of a pipe network, including: providing a mobile ARM three-dimensional laser scanning device, a base station laser scanning device and a surveying-level UAV; the mobile ARM three-dimensional laser scanning device scans the connection parts of the pipe network to generate a first point cloud data set; the base station laser scanning device scans the pipe network within the line-of-sight and / or non-line-of-sight range to generate a second point cloud data set; the surveying-level UAV scans the pipe network within the line-of-sight range to generate a third point cloud data set; the first point cloud data set, the second point cloud data set and the third point cloud data set are serially and parallelly connected to achieve data fusion of the three point cloud data sets to obtain the complete model point cloud data of the pipe network; the complete model point cloud data of the pipe network is rendered to obtain the modeling model of the pipe network. The present application can realize the fine-grained modeling of the pipe network and improve the accuracy and integrity of the pipe network model.

Description

Technical Field

[0001] The present application relates to the technical field of simulation modeling, and in particular to a method and system for fine-grained modeling of a pipeline network. Background Art

[0002] With the accelerated construction of oil and gas pipeline infrastructure in my country, the traditional two-dimensional pipeline management method is difficult to accurately describe the spatial position relationship between pipelines, and the rough three-dimensional pipeline model cannot meet the accuracy requirements of pipeline modeling; moreover, in the existing pipeline modeling methods, most of the straight segments and curved segments of each pipeline are processed separately, especially the connecting segments such as pipeline inflection points and diameter changes, which need to be spliced ​​manually, including a series of tedious operations such as part position selection, angle adjustment, and scaling. The modeling work efficiency is low, and the accuracy of the model is difficult to guarantee. Summary of the invention

[0003] In order to solve the deficiencies of the prior art, this application adopts the following technical solutions:

[0004] In a first aspect, the present application provides a method for fine-grained modeling of a pipe network, the method comprising the following steps:

[0005] Provide mobile ARM 3D laser scanning equipment, base station laser scanning equipment and surveying-level UAVs;

[0006] The mobile ARM three-dimensional laser scanning device is used to scan the connection part of the pipe network to generate a first point cloud data set; the base station laser scanning device is used to scan the pipe network within the line of sight and / or non-line of sight range to generate a second point cloud data set; the surveying and mapping-level unmanned aerial vehicle is used to scan the pipe network within the line of sight range to generate a third point cloud data set; the base station laser scanning device and the surveying and mapping-level unmanned aerial vehicle have at least partially different scanning ranges for the pipe network;

[0007] The first point cloud data set, the second point cloud data set and the third point cloud data set are serially and parallelly combined to achieve data fusion of the three point cloud data sets, so as to obtain complete model point cloud data of the pipe network;

[0008] The complete model point cloud data of the pipe network is rendered to obtain a modeling model of the pipe network.

[0009] In summary, the present application provides a method for refined pipe network modeling, which uses different scanning devices to perform targeted scanning on different parts of the pipe network, and obtains high-precision point cloud data sets from different parts of the pipe network, and fuses the point cloud data collected by the three devices to obtain high-precision point cloud data of the complete model of the pipe network. Based on the high-precision point cloud data of the complete model of the pipe network, through data rendering, high-precision and complete pipe network models are automatically and widely generated, which can reduce the work of manual model splicing and improve the efficiency of pipe network modeling.

[0010] Furthermore, the data fusion of the three point cloud data sets includes the following steps:

[0011] Step S101, comparing the three point cloud data sets, and selecting the point cloud data sets representing the same part of the pipe network from the three point cloud data sets;

[0012] Step S102, in the same origin coordinate system, weighting the three selected point cloud data sets representing the same part of the pipe network respectively to obtain a learning set;

[0013] Step S103, randomly selecting k samples from the learning set as initial center points;

[0014] Step S104, for each sample point in the learning set, calculating the distance between it and the k extracted center points, and dividing the sample point into the cluster corresponding to the center point closest to it;

[0015] Step S105, updating the center of each cluster to the mean value of all sample points belonging to the cluster;

[0016] Step S106, repeating steps S104 and S105 until the change in the center of the cluster is less than a preset threshold.

[0017] Furthermore, the step S101 also includes:

[0018] Based on the three-dimensional coordinates of the point cloud data in any point cloud data set, a number of specific points are selected, and the data set is divided into different point neighborhood ranges with the specific points as the center;

[0019] Obtain all points in the neighborhood of any point, calculate the covariance matrix of the point cloud in the neighborhood of the point, perform eigenvalue decomposition of the covariance matrix, and obtain several eigenvalues ​​of the covariance matrix and eigenvectors corresponding to the eigenvalues;

[0020] Select the eigenvector corresponding to the minimum eigenvalue, the eigenvector corresponding to the minimum eigenvalue represents the normal direction of the minimum curvature direction of the specific point in the point neighborhood on the local surface, calculate the surface curvature of the specific point, and use the surface curvature of the specific point as the corresponding point domain curvature; repeat this step to calculate the curvature of each point neighborhood;

[0021] The eigenvector corresponding to the minimum eigenvalue is selected as the normal vector of the point to confirm the direction of each point cloud, and the curvatures of each point cloud in different point cloud data sets are aligned. In response to the normal direction deviation of the point clouds in any two point cloud data sets being less than a preset angle, and the curvature value deviation being less than a preset range, the corresponding point cloud data are determined to represent the same part of the pipeline network; point cloud data sets representing the same part of the pipeline network are selected from the three point cloud data sets respectively, thereby realizing data fusion.

[0022] Furthermore, in step S102, weights are assigned to the three point cloud data sets respectively by the following formulas:

[0023] D= ;

[0024] In the formula, A represents the second point cloud dataset, B represents the third point cloud dataset, C represents the first point cloud dataset, and p n The sample points of the learning set, u, s, t represent the weight coefficients, u+s+t=1, u=s, u and s are random numbers greater than 0 and less than 0.5.

[0025] Furthermore, the method further comprises:

[0026] A color binocular camera is provided, and during the scanning process of the mobile ARM three-dimensional laser scanning device, the base station laser scanning device and the surveying-level UAV, the color binocular camera obtains RGB image data of the pipeline;

[0027] Extracting pipeline texture features based on the RGB image data, and obtaining pipeline material information at different parts of the pipeline network based on the extracted texture feature analysis;

[0028] The model is rendered in combination with the pipeline material information so that the pipeline material at the corresponding position can be reflected in the modeling model of the pipeline network.

[0029] Furthermore, the method further comprises:

[0030] Extracting point cloud information from the RGB image data, wherein the point cloud information extracted from the RGB image data includes one or more of the following information: point cloud curvature, eigenvalue, and eigenvector;

[0031] The point cloud information extracted from the RGB image data is aligned with the complete model point cloud data of the pipe network to determine the position of the pipe network part represented by the RGB image data in the complete model point cloud data of the pipe network.

[0032] Furthermore, the method further comprises:

[0033] Through the scene algorithm, the point cloud data after the fusion of the three point cloud data sets are subjected to point cloud filtering, point cloud sampling, point cloud feature extraction, model encapsulation, and texture extraction, so as to output a complete modeling model of the pipeline network and a modeling model of at least part of the pipeline network.

[0034] Furthermore, the method further comprises:

[0035] Build a model library, establish an access path under the model file system to access the model, and perform one or more of the following operations on the model in the model library: add, delete, modify, and query.

[0036] The present application also provides a pipe network refined modeling system, the system comprising:

[0037] Mobile ARM 3D laser scanning equipment is used to scan the connection parts of the pipe network and generate the first point cloud data set;

[0038] Base station laser scanning equipment and surveying-grade drones are used to scan the pipe network in line-of-sight and / or non-line-of-sight ranges to generate a second point cloud data set;

[0039] A surveying-level UAV is used to scan the pipe network within the line of sight to generate a third point cloud data set; the scanning ranges of the pipe network of the base station laser scanning device and the surveying-level UAV are at least partially different;

[0040] The processing module serially and parallelizes the first point cloud data set, the second point cloud data set and the third point cloud data set to achieve data fusion of the three point cloud data sets to obtain the complete model point cloud data of the pipe network; the processing module renders the complete model point cloud data of the pipe network to obtain the modeling model of the pipe network.

[0041] Furthermore, the mobile ARM three-dimensional laser scanning device, the base station laser scanning device and the surveying-level UAV are all equipped with a color binocular camera, and the color binocular camera is used to obtain RGB image data of the pipeline during the scanning process;

[0042] The processing module is also configured to: extract pipeline texture features based on the RGB image data, and obtain pipeline material information at different parts of the pipeline network based on the extracted texture feature analysis; the processing module performs model rendering in combination with the pipeline material information so that the pipeline material at the corresponding position is reflected in the modeling model of the pipeline network. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flowchart of the steps of constructing a pipe network model in a pipe network refined modeling method provided in an embodiment of the present application;

[0044] Figure 2 A flowchart of the steps of fusing point cloud data sets in a method for fine-grained modeling of a pipe network provided in one embodiment of the present application;

[0045] Figure 3 A flowchart of the steps of constructing corresponding pipeline material information in a pipeline network model in a pipeline network refined modeling method provided in an embodiment of the present application;

[0046] Figure 4 A flowchart of the overall steps of a pipe network refined modeling method provided in one embodiment of the present application;

[0047] Figure 5 A schematic diagram of the composition structure of a pipe network refined modeling system provided in one embodiment of the present application. DETAILED DESCRIPTION

[0048] The present application will be described in detail below in conjunction with the specific implementation modes shown in the accompanying drawings, but these implementation modes do not limit the present application. Structural, methodological, or functional changes made by ordinary technicians in the field based on these implementation modes are included in the protection scope of the present application.

[0049] In order to solve the deficiencies of the prior art, in the first aspect, the present application provides a method for fine-tuning pipe network modeling, such as Figure 1 As shown, the method comprises the following steps:

[0050] Step S11, providing a mobile ARM three-dimensional laser scanning device, a base station laser scanning device and a surveying and mapping-level drone;

[0051] Step S12, using a mobile ARM three-dimensional laser scanning device to scan the connection part of the pipe network to generate a first point cloud data set; using a base station laser scanning device to scan the pipe network within the line of sight and / or non-line of sight range to generate a second point cloud data set; using a surveying and mapping-level unmanned aerial vehicle to scan the pipe network within the line of sight range to generate a third point cloud data set; the base station laser scanning device and the surveying and mapping-level unmanned aerial vehicle have at least partially different scanning ranges for the pipe network;

[0052] Step S13, combining the first point cloud data set, the second point cloud data set and the third point cloud data set in series and in parallel to achieve data fusion of the three point cloud data sets to obtain complete model point cloud data of the pipe network;

[0053] Step S14, rendering the complete model point cloud data of the pipe network to obtain a model of the pipe network.

[0054] In step S11, three types of equipment, including mobile ARM 3D laser scanning equipment, base station laser scanning equipment and surveying-level UAV, are provided to collect point cloud data of the pipe section. Among them, for the mobile ARM 3D laser scanning equipment, a handheld type can be used to collect point cloud data of the selected area of ​​the pipeline outside the pipeline, and an ARM 3D laser scanning equipment that can move inside the pipeline can also be used to collect point cloud data of the selected area of ​​the pipeline from the inside of the pipeline, which can adapt to some occasions that are difficult to collect manually (such as underground, closed parks, etc.).

[0055] In step S12, three different scanning devices are used to obtain point cloud data of different parts of the pipeline network. Specifically, it includes: using a mobile ARM three-dimensional laser scanning device to scan the connection parts of the pipeline network, including scanning the connection parts such as the turning points, reducers, tees, crosses, elbows, etc. of the pipeline network, obtaining the scanning data of the above-mentioned connection parts of the pipeline network, and generating a first point cloud data set; using a base station laser scanning device to scan the pipeline network within the line of sight and / or non-line of sight range, including scanning the closed parts and semi-open parts of the pipeline network, such as scanning the pipeline network installed underground, in residential buildings, indoor partitions, courtyards and greenhouses, obtaining the scanning data of the above-mentioned pipeline network parts, and generating a second point cloud data set; using a surveying and mapping-level drone to scan the pipeline network within the line of sight range, including scanning the outdoor open pipeline network parts such as urban long-distance pipelines and urban gas receiving stations, obtaining the scanning data of the above-mentioned pipeline network parts, and generating a third point cloud data set.

[0056] It should be noted that the scanning ranges of pipeline networks by base-station laser scanning equipment and surveying-level drones are at least partially different. Base-station laser scanning equipment focuses on scanning complex and hidden pipeline network locations in closed and semi-closed areas within the city, while surveying-level drones focus on scanning pipeline network locations in open areas such as the edge of the city.

[0057] In step S13, the first point cloud data set, the second point cloud data set and the third point cloud data set are serially and parallelly connected through the point cloud particle recognition algorithm, and the point cloud normals, curvatures and point features of different pipe network parts collected by each device are identified. The second point cloud data set generated by the base station laser scanning device and the third point cloud data set generated by the surveying and mapping-level drone are matched and calibrated with the first point cloud data set generated by the mobile ARM three-dimensional laser scanning device. By serially and paralleling the three point cloud data sets, different weights are assigned to each point cloud data set, all data are converted into learning samples, and the weight of each point cloud data set is dynamically adjusted to realize the recognition and superposition of data learning samples to generate a learning set, and the data is automatically fused through point features, normals and curvature auxiliary corrections, so as to obtain the complete model point cloud data of the pipe network.

[0058] In step S14, based on the complete model point cloud data of the pipeline network, the modeling model of the pipeline network is quickly built through data rendering, and a high-precision complete model of the pipeline network is generated, which ensures the accuracy and integrity of the pipeline network model, reduces the workload of manual model construction, and improves work efficiency.

[0059] According to the above description, the present application provides a method for refined modeling of a pipeline network, which uses different scanning devices to perform targeted scanning on different parts of the pipeline network, and obtains high-precision point cloud data sets for different parts of the pipeline network. In particular, a mobile ARM three-dimensional laser scanning device is used to scan the connection parts of the pipeline network to obtain point cloud data of the connection parts of the pipeline network. The point cloud data collected by the three devices are fused through a point cloud data fusion algorithm to obtain high-precision point cloud data of a complete model of the pipeline network. Based on the high-precision point cloud data of the complete model of the pipeline network, a high-precision and complete pipeline network model is automatically generated through data rendering, which can reduce the work of manual model splicing and improve the efficiency of pipeline modeling.

[0060] It should be noted that the complete model point cloud data of the pipe network described in the embodiments of the present application refers to all the point cloud data of the pipe network within a certain area. Taking the urban pipe network as an example, it can be all the point cloud data of the pipe network in the entire city, or it can be the point cloud data of the pipe network in one or more areas divided in the urban pipe network. Through the complete model point cloud data of the pipe network, large-scale direct modeling of the pipe network can be realized, and the splicing operation of detailed parts (such as connecting parts) can be omitted or reduced, thereby improving the modeling efficiency of the pipe network.

[0061] As a way to achieve this, Figure 2 As shown in the figure, the data fusion of the three point cloud data sets includes the following steps:

[0062] Step S131, comparing the three point cloud data sets, and selecting point cloud data sets representing the same part of the pipe network from the three point cloud data sets;

[0063] Step S132, in the same origin coordinate system, weighting the three point cloud data sets respectively to obtain a learning set;

[0064] Step S133, randomly select k samples from the learning set as initial center points;

[0065] Step S134, for each sample point in the learning set, calculate its distance from the k extracted center points, and divide the sample point into the cluster corresponding to the center point closest to it;

[0066] Step S135, updating the center of each cluster to the mean of all sample points belonging to the cluster;

[0067] Step S136, repeating steps S134 and S135 until the change in the center of the cluster is less than a preset threshold.

[0068] Step S131 includes: based on the three-dimensional coordinates of the point cloud data in any point cloud data set, selecting a number of specific points (each point cloud can be used as a specific point, or a part of them can be selected), setting the size of the neighborhood, and dividing the data set into different point neighborhood ranges with the specific point as the center;

[0069] Get all points in any neighborhood, calculate the covariance matrix of the point cloud in the neighborhood, perform eigenvalue decomposition of the covariance matrix, obtain several eigenvalues ​​of the covariance matrix and eigenvectors corresponding to the eigenvalues; select the eigenvector corresponding to the minimum eigenvalue, which represents the normal direction of the minimum curvature direction of the point on the local surface. Calculate the surface curvature of the point.

[0070] Repeat this step to calculate the neighborhood curvature of each point.

[0071] The eigenvector corresponding to the minimum eigenvalue is selected as the normal vector of the point to confirm the direction of each point, and the curvatures of each point cloud in different point cloud data sets are aligned. In response to the normal direction deviation of the point clouds in any two point cloud data sets being less than a preset angle, and the curvature value deviation being less than a preset range, it is determined that the corresponding point cloud data represent the same part of the pipeline network, and the point cloud data sets representing the same part of the pipeline network are selected from the three point cloud data sets respectively, thereby realizing data fusion.

[0072] Specifically, each data set contains three-dimensional coordinates P(x, y, z);

[0073] A1=(x1,y1,z1),A={A1,A2,...,An};

[0074] B1=(x2,y2,z2),B={B1,B2,...,Bn};

[0075] C1=(x3,y3,z3),C={C1,C2,...,Cn};

[0076] In the formula, A represents the first point cloud dataset, B represents the second point cloud dataset, and C represents the third point cloud dataset.

[0077] For any point cloud data set, by selecting a local plane G, we can calculate:

[0078] For the above point cloud set, the covariance matrix is ​​a 3×3 symmetric matrix

[0079] ;

[0080] For each scan point P in the point cloud, search for its nearest neighbor m points and calculate the local plane G of these points. This plane can be expressed as:

[0081] ;

[0082] I is the normal vector in plane G, d is the distance from plane G to the origin of the coordinate system, P n Represents the nth point in the neighborhood. The normal vector is one of the features that describes the local geometric shape of the point cloud surface.

[0083] The neighborhood eigenvalue of each point is calculated by subtracting the coordinates of the center point of the neighborhood from the coordinates of the point cloud in the neighborhood. Specifically, the calculation formula of the eigenvalue M is as follows.

[0084] ;

[0085] ;

[0086] In the formula, m is The number of adjacent points selected in the neighborhood of is the center of the neighborhood, represents the j-th eigenvalue of the covariance matrix, represents the jth eigenvector.

[0087] If the eigenvalues ​​satisfy , then the surface curvature F of the point is:

[0088] ;

[0089] Specifically, in step S132, the three point cloud data sets are compared, and the point cloud data representing the same part of the pipeline network are selected for fusion. The first point cloud data set, the second point cloud data set and the third point cloud data set are placed in the same origin coordinate system, and each point cloud data set is assigned a weight to obtain a learning set.

[0090] It is easy to understand that some parts of the pipe network may only use one or two devices for data collection. Therefore, the fusion of the three point cloud data sets described in this application refers to: fusing the point cloud data sets representing the same part of the pipe network in the three point cloud data sets. That is, when fusing point cloud data, depending on the different parts, there may be a fusion of part of the point cloud data in the first point cloud data set with part of the point cloud data in the second point cloud data set, there may also be a fusion of part of the point cloud data in the first point cloud data set with part of the point cloud data in the third point cloud data set, and there may also be a fusion of part of the point cloud data in the three point cloud data sets.

[0091] Randomly extract K samples from the learning set and use K samples as the initial center points. The learning set is denoted as H. n ={S 1 , S 2 , ..., S K}, initialize the learning set.

[0092] For each sample point in the learning set, calculate the distance between each sample point and the K extracted center points, and divide each sample point into the cluster corresponding to the center point closest to it. Furthermore, update the center of each cluster to the mean of all sample points belonging to the cluster, which can be expressed by the following formula:

[0093] ;

[0094] ;

[0095] ;

[0096] In the formula, represents the learning set, represents the sampling point, represents the centroid of the learning set, represents the cluster whose center is closest to the learning set, n∈{1,2,…,k}.

[0097] Repeat step S134 and step S135, and continuously update the central average value of the sample center until the change in the center of the cluster is less than the preset threshold, thereby realizing the data learning sample recognition superposition to generate a learning set.

[0098] Further, as an implementation method, in step S132, weights are assigned to the three point cloud data sets respectively by the following formulas:

[0099] ;

[0100] In the formula, A represents the second point cloud dataset, B represents the third point cloud dataset, C represents the first point cloud dataset, and p nThe sample points of the learning set, u, s, t represent weight coefficients, u is the weight coefficient of the second point cloud dataset, s is the weight coefficient of the third point cloud dataset, and t is the weight coefficient of the first point cloud dataset. And, u+s+t=1, u=s, u and s are random numbers greater than 0 and less than 0.5.

[0101] As a way to achieve this, Figure 3 As shown, the pipe network refined modeling method provided by this application also includes:

[0102] Step S21, providing a color binocular camera, and in the scanning process of the mobile ARM 3D laser scanning device, the base station laser scanning device and the surveying and mapping-level UAV, the color binocular camera obtains RGB image data of the pipeline;

[0103] Step S22, extracting pipeline texture features based on the RGB image data, and obtaining pipeline material information at different parts of the pipeline network based on the extracted texture feature analysis;

[0104] Step S23, model rendering is performed in combination with the pipeline material information, so that the pipeline material at the corresponding position is reflected in the modeling model of the pipeline network.

[0105] Specifically, during the scanning process of the three devices on the pipe network, the color binocular camera follows the scanning action of the device to obtain the RGB image data of the pipe network. Based on the RGB image data of the pipe network, the texture features of the pipe network are extracted to obtain the pipe material information of different parts of the pipe network. When rendering and modeling the pipe network point cloud data, the pipe material information of different parts of the pipe network is combined to reflect the pipe material of the corresponding position in the generated pipe network model, thereby improving the accuracy and completeness of the pipe model.

[0106] It should be noted that the RGB image data of the pipe network is obtained synchronously by the color binocular camera when the device collects point cloud data of the pipe network within the line of sight. For point cloud data collection outside the line of sight, the color binocular camera cannot obtain the RGB image data of the pipe network, but the pipe network material in the area can be inferred from the RGB image data of the nearby pipe network. For example, for a pipe section buried underground, the material of the middle pipe section can be inferred from the pipe materials at both ends.

[0107] As an optional implementation, the method further includes: extracting point cloud information from the RGB image data, wherein the point cloud information extracted from the RGB image data includes one or more of the following information: point cloud curvature, eigenvalue, and eigenvector;

[0108] The point cloud information extracted from the RGB image data is aligned with the complete model point cloud data of the pipe network to determine the position of the pipe network part represented by the RGB image data in the complete model point cloud data of the pipe network.

[0109] In this way, the corresponding position of the pipeline network represented by the RGB image data can be found in the complete model point cloud data, so that the grayscale features carried by the RGB image data can be used to determine the pipeline material at that location and reflect it in the modeling model.

[0110] As an implementation method, the pipeline network refined modeling method provided in the present application also includes: using a scene algorithm to perform point cloud filtering, point cloud sampling, point cloud feature extraction, model encapsulation, and texture extraction on the point cloud data after the fusion of the three point cloud data sets, thereby outputting a complete modeling model of the pipeline network and a modeling model of at least part of the pipeline network.

[0111] Specifically, based on the scene algorithm, the point cloud filtering operation includes: removing hash points and isolated points in the point cloud data through bilateral filtering, Gaussian filtering, conditional filtering, through filtering, etc.

[0112] Point cloud sampling operations include: converting all point cloud data operations to key points through uniform sampling, geometric sampling, upsampling, and sliding least squares sampling, and performing the nearest interpolation of point cloud data, thereby efficiently calculating point cloud features.

[0113] The point cloud feature extraction operation includes: implementing a point cloud clustering method based on partitioning, dividing the data into M data sets, then dividing the data into S groups, and then dividing the groups into X piles, each pile represents a cluster, and X≤S≤M; segmenting the evenly distributed point cloud data by dividing the layer grid to achieve rapid clustering.

[0114] Model packaging operations include: implementing point cloud triangular patch packaging through geometric and texture features.

[0115] The texture extraction operation includes: the image texture feature reflects the grayscale distribution of a pixel and its surrounding spatial neighborhood pixels, and the pixel is extracted separately from the grayscale distribution of the surrounding spatial neighborhood pixels through geometric and signal processing methods.

[0116] Furthermore, as an implementation method, the method also includes: building a model library, establishing an access path under the model file system to access the model, and performing one or more of the following operations on the model in the model library: adding, deleting, modifying, and querying.

[0117] Specifically, after obtaining the complete modeling model of the pipeline network and the modeling model of at least part of the pipeline network, a model library of the pipeline network is built, and the access path under the model file system is restored to automatically access the model to improve the modeling speed. In the model library, operations such as adding, deleting, modifying and querying the model are performed, and the model file is edited and compiled. Combined with Dynamo parametric modeling, operations on any node of the model are realized, including stretching, rotating, translating, enlarging, reducing, curvature collapse, peeling and deleting, so as to build a more precise and accurate pipeline network model.

[0118] According to the above description, if Figure 4 As shown, the present application provides a method for fine-grained modeling of a pipe network, which adopts a mobile ARM three-dimensional laser scanning device, a base station laser scanning device and a surveying-level UAV to scan the connection parts of the pipe network, the pipe network parts in closed and semi-open areas, and the pipe network parts in outdoor areas, respectively, to obtain point cloud data of different parts of the pipe network, and to fuse the point cloud data generated by scanning of different devices through a point cloud fusion algorithm to generate point cloud data of a complete model of the pipe network, and to achieve large-scale direct modeling of the pipe network through data rendering, thereby omitting or reducing the splicing operation of detailed parts and improving the speed of pipe network modeling; and, during the scanning process, the RGB image data of the pipeline is collected through a color binocular camera, and based on the RGB image data of the pipeline, the pipe material information of different parts of the pipe network is obtained through methods such as texture feature extraction; when rendering and modeling the pipe network point cloud data, the pipe material information of different parts of the pipe network is combined, and the pipe material of the corresponding position is reflected in the generated pipe network model, thereby improving the accuracy and completeness of the pipe model.

[0119] In the second aspect, the present application also provides a pipe network refined modeling system, such as Figure 5 As shown, the system includes: a mobile ARM three-dimensional laser scanning device, a base station laser scanning device, a surveying and mapping-level UAV, and a processing module.

[0120] The mobile ARM three-dimensional laser scanning device is used to scan the connection parts of the pipe network to generate a first point cloud data set; the base station laser scanning device and the surveying and mapping-level drone are used to scan the pipe network within the line of sight and / or non-line of sight range to generate a second point cloud data set; and the surveying and mapping-level drone is also used to scan the pipe network within the line of sight range to generate a third point cloud data set;

[0121] It should be noted that the scanning ranges of pipeline networks by base-station laser scanning equipment and surveying-level drones are at least partially different. Base-station laser scanning equipment focuses on scanning complex and hidden pipeline network locations in closed and semi-closed areas within the city, while surveying-level drones focus on scanning pipeline network locations in open areas such as the edge of the city.

[0122] The processing module is configured to serially and parallelly combine the first point cloud data set, the second point cloud data set and the third point cloud data set to achieve data fusion of the three point cloud data sets to obtain the complete model point cloud data of the pipeline network; the processing module renders the complete model point cloud data of the pipeline network to obtain the modeling model of the pipeline network.

[0123] According to the above description, the present application provides a refined pipeline modeling system, which uses different scanning devices to perform targeted scanning on different parts of the pipeline network, and obtains high-precision point cloud data sets for different parts of the pipeline network. In particular, a mobile ARM three-dimensional laser scanning device is used to scan the connection parts of the pipeline network to obtain point cloud data of the connection parts of the pipeline network. The point cloud data collected by the three devices are fused through a point cloud data fusion algorithm to obtain high-precision point cloud data of a complete model of the pipeline network. Based on the high-precision point cloud data of the complete model of the pipeline network, a high-precision and complete pipeline network model is automatically generated through data rendering, which can reduce the work of manual model splicing and improve the efficiency of pipeline modeling.

[0124] Further, as an implementation method, the mobile ARM 3D laser scanning device, the base station laser scanning device and the surveying-level UAV are all equipped with a color binocular camera, which is used to obtain RGB image data of the pipeline during the scanning process;

[0125] The processing module is also configured to: extract pipeline texture features based on RGB image data, and obtain pipeline material information at different parts of the pipeline network based on the extracted texture feature analysis; the processing module performs model rendering in combination with the pipeline material information so that the pipeline material at the corresponding position is reflected in the modeling model of the pipeline network.

[0126] In summary, the present application provides a pipe network refined modeling system, which uses three scanning devices to perform targeted scanning on different parts of the pipe network, and fuses the point cloud data collected by the three devices to obtain high-precision point cloud data of the complete model of the pipe network. Based on the high-precision point cloud data of the complete model of the pipe network, through data rendering, high-precision and complete pipe network models are automatically and widely generated, which can reduce the work of manually splicing models and improve the efficiency of pipe network modeling. In addition, during the scanning process, the RGB image data of the pipeline is collected by a color binocular camera. Based on the RGB image data of the pipeline, the processing module obtains the pipe material information of different parts of the pipe network through texture feature extraction and other methods; when rendering and modeling the pipe network point cloud data, the processing module combines the pipe material information of different parts of the pipe network, and reflects the pipe material of the corresponding position in the generated pipe network model, thereby improving the accuracy and completeness of the pipe model.

[0127] It will be appreciated that the word "exemplary" as used herein means "serving as an example, instance, or illustration". Any embodiment described as "exemplary" is not necessarily preferred or superior to other embodiments and / or does not exclude the combination of features of other embodiments. It will be appreciated that certain features of the present application described in the context of separate embodiments for the sake of clarity may also be provided in a single embodiment by combination. Conversely, various features of the present application described in the context of a single embodiment for the sake of clarity may also be provided individually or in any suitable combination or as any other described embodiment of the present application.

[0128] The above disclosure is only the preferred embodiment of the present application, but it is not intended to limit the scope of rights of the present application. A person of ordinary skill in the art can understand that without departing from the spirit and scope of the present application and the appended claims, changes, modifications, substitutions, combinations, and simplifications should all be equivalent replacement methods and still fall within the scope of the invention.

Claims

1. A method for fine-tuning pipe network modeling, characterized in that: The method comprises the following steps: Provide mobile ARM 3D laser scanning equipment, base station laser scanning equipment and surveying-level UAVs; The mobile ARM three-dimensional laser scanning device is used to scan the connection part of the pipe network to generate a first point cloud data set; the base station laser scanning device is used to scan the pipe network within the line of sight and / or non-line of sight range to generate a second point cloud data set; the surveying and mapping-level unmanned aerial vehicle is used to scan the pipe network within the line of sight range to generate a third point cloud data set; the base station laser scanning device and the surveying and mapping-level unmanned aerial vehicle have at least partially different scanning ranges for the pipe network; The first point cloud data set, the second point cloud data set and the third point cloud data set are serially and parallelly connected to achieve data fusion of the three point cloud data sets, so as to obtain complete model point cloud data of the pipe network; Rendering the complete model point cloud data of the pipe network to obtain a modeling model of the pipe network; The data fusion of the three point cloud data sets comprises the following steps: Step S101, comparing the three point cloud data sets, and selecting the point cloud data sets representing the same part of the pipe network from the three point cloud data sets; Step S102, in the same origin coordinate system, weighting the three selected point cloud data sets representing the same part of the pipe network respectively to obtain a learning set; Step S103, randomly selecting k samples from the learning set as initial center points; Step S104, for each sample point in the learning set, calculating the distance between it and the k extracted center points, and dividing the sample point into the cluster corresponding to the center point closest to it; Step S105, updating the center of each cluster to the mean value of all sample points belonging to the cluster; Step S106, repeating steps S104 and S105 until the change in the center of the cluster is less than a preset threshold.

2. The pipe network refined modeling method according to claim 1, characterized in that: The step S101 also includes: Based on the three-dimensional coordinates of the point cloud data in any point cloud data set, a number of specific points are selected, and the data set is divided into different point neighborhood ranges with the specific points as the center; Obtain all points in the neighborhood of any point, calculate the covariance matrix of the point cloud in the neighborhood of the point, perform eigenvalue decomposition of the covariance matrix, and obtain several eigenvalues ​​of the covariance matrix and eigenvectors corresponding to the eigenvalues; Select the eigenvector corresponding to the minimum eigenvalue, the eigenvector corresponding to the minimum eigenvalue represents the normal direction of the minimum curvature direction of the specific point in the point neighborhood on the local surface, calculate the surface curvature of the specific point, and use the surface curvature of the specific point as the corresponding point domain curvature; repeat this step to calculate the curvature of each point neighborhood; The eigenvector corresponding to the minimum eigenvalue is selected as the normal vector of the point to confirm the direction of each point cloud, and the curvatures of each point cloud in different point cloud data sets are aligned. In response to the normal direction deviation of the point clouds in any two point cloud data sets being less than a preset angle, and the curvature value deviation being less than a preset range, the corresponding point cloud data are determined to represent the same part of the pipeline network; point cloud data sets representing the same part of the pipeline network are selected from the three point cloud data sets respectively, thereby realizing data fusion.

3. The pipe network refined modeling method according to claim 2 is characterized in that: In step S102, weights are assigned to the three point cloud data sets respectively by the following formulas: D= ； In the formula, A represents the second point cloud dataset, B represents the third point cloud dataset, C represents the first point cloud dataset, and p n The sample points of the learning set, u, s, t represent the weight coefficients, u+s+t=1, u=s, u and s are random numbers greater than 0 and less than 0.

5.

4. The pipe network refined modeling method according to claim 1, characterized in that: The method further comprises: A color binocular camera is provided, and during the scanning process of the mobile ARM three-dimensional laser scanning device, the base station laser scanning device and the surveying-level UAV, the color binocular camera obtains RGB image data of the pipeline; Extracting pipeline texture features based on the RGB image data, and obtaining pipeline material information at different parts of the pipeline network based on the extracted texture feature analysis; The model is rendered in combination with the pipeline material information so that the pipeline material at the corresponding position can be reflected in the modeling model of the pipeline network.

5. The pipe network refined modeling method according to claim 4 is characterized in that: The method further comprises: Extracting point cloud information from the RGB image data, wherein the point cloud information extracted from the RGB image data includes one or more of the following information: point cloud curvature, eigenvalue, and eigenvector; The point cloud information extracted from the RGB image data is aligned with the complete model point cloud data of the pipe network to determine the position of the pipe network part represented by the RGB image data in the complete model point cloud data of the pipe network.

6. The pipe network refined modeling method according to claim 1, characterized in that: The method further comprises: Through the scene algorithm, the point cloud data after the fusion of the three point cloud data sets are subjected to point cloud filtering, point cloud sampling, point cloud feature extraction, model encapsulation, and texture extraction, so as to output a complete modeling model of the pipeline network and a modeling model of at least part of the pipeline network.

7. The pipe network refined modeling method according to claim 6, characterized in that: The method further comprises: Build a model library, establish an access path under the model file system to access the model, and perform one or more of the following operations on the model in the model library: add, delete, modify, and query.

8. A pipe network refined modeling system, characterized in that: The system comprises: Mobile ARM 3D laser scanning equipment is used to scan the connection parts of the pipe network and generate the first point cloud data set; Base station laser scanning equipment and surveying-grade drones are used to scan the pipe network in line-of-sight and / or non-line-of-sight ranges to generate a second point cloud data set; A surveying-level UAV is used to scan the pipe network within the line of sight to generate a third point cloud data set; the scanning ranges of the pipe network of the base station laser scanning device and the surveying-level UAV are at least partially different; A processing module is used to perform serial and parallel processing on the first point cloud data set, the second point cloud data set and the third point cloud data set to achieve data fusion of the three point cloud data sets, so as to obtain the complete model point cloud data of the pipe network; the processing module renders the complete model point cloud data of the pipe network, so as to obtain the modeling model of the pipe network; The data fusion of the three point cloud data sets comprises the following steps: Step S101, comparing the three point cloud data sets, and selecting point cloud data sets representing the same part of the pipe network from the three point cloud data sets; Step S102, in the same origin coordinate system, weighting the three selected point cloud data sets representing the same part of the pipe network respectively to obtain a learning set; Step S103, randomly selecting k samples from the learning set as initial center points; Step S104, for each sample point in the learning set, calculating the distance between it and the k extracted center points, and dividing the sample point into the cluster corresponding to the center point closest to it; Step S105, updating the center of each cluster to the mean value of all sample points belonging to the cluster; Step S106, repeating steps S104 and S105 until the change in the center of the cluster is less than a preset threshold.

9. The pipe network refined modeling system according to claim 8, characterized in that: The mobile ARM three-dimensional laser scanning device, the base station laser scanning device and the surveying-level UAV are all equipped with a color binocular camera, and the color binocular camera is used to obtain RGB image data of the pipeline during the scanning process; The processing module is also configured to: extract pipeline texture features based on the RGB image data, and obtain pipeline material information at different parts of the pipeline network based on the extracted texture feature analysis; the processing module performs model rendering in combination with the pipeline material information so that the pipeline material at the corresponding position is reflected in the modeling model of the pipeline network.

Images