Total element tower model automatic matching method and system based on multi-source data, computer readable storage medium and electronic equipment

Abstract

The invention provides a multi-source data-based total element tower model automatic matching method and system, a computer readable storage medium and electronic equipment, and relates to the technical field of tower modeling. Comprising the steps of obtaining point cloud tower data, and extracting tower contour features as first contour features; obtaining visible light tower data, and extracting tower contour features as second contour features; fusing the first contour feature and the second contour feature to obtain tower feature information, and establishing a tower attribute file; the method comprises the following steps: obtaining tower drawings of different voltage grades and different types, manufacturing a total-factor tower model comprising a tower head, a tower body and tower legs according to the drawings, and constructing a total-factor tower model library; and matching tower elements in the total element tower model library based on the tower attribute file information, and realizing three-dimensional vectorization modeling of the tower based on the successfully matched tower elements. According to the method, the tower three-dimensional vectorization modeling can be finely realized, and the precision and efficiency of tower three-dimensional vectorization modeling are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of pole tower modeling and analysis, and in particular relates to a method, system, computer-readable storage medium and electronic equipment for automatic matching of a full-element pole tower model based on multi-source data. Background Art

[0002] Power towers are very important structures in high-voltage transmission lines. Power towers are mainly used to erect high-voltage power transmission lines and are important infrastructure projects in the power transmission network. When lifting the conductors, the towers must not only bear the vertical loads of themselves and the conductors, but also the tension loads after the power conductors are suspended. Under the action of natural forces, gravity, and stress, the foundation of the towers is prone to deformation, causing the towers to tilt, even collapse, and disconnect the lines, paralyzing the power supply lines and causing significant economic losses.

[0003] Since the problem of transmission line tower tilt is difficult to detect in time, when the tower tilt is serious, it will cause tower disconnection, tower collapse and other accidents, seriously affecting the safe and stable operation of the power grid. It is very important to realize the three-dimensional structural modeling of the tower and apply it to the analysis of the tower tilt state.

[0004] At present, there are still some technical obstacles in the three-dimensional structural modeling of towers. The main problems are:

[0005] (1) Point cloud missing phenomenon. The pole tower is a hollow three-dimensional structure. Due to the interference of the pole tower itself and warning signs, the point cloud data has a certain degree of missing phenomenon. At the same time, the density of airborne point cloud data is not enough to accurately express the detailed features of the pole tower, especially for the pole tower models with complex structures such as cat head tower and wine glass tower. When the pole tower model is finely reconstructed, it often cannot meet the requirements.

[0006] (2) The structure of the modeling object is complex. According to the shape of the tower, it can be divided into drum-shaped tower, dry-shaped tower, cat-head tower, sheep-horn tower, T-shaped tower, wine glass-shaped tower, etc. The spatial structure of different tower types is complex and diverse, which makes it difficult to identify the geometric constraints during the tower modeling process;

[0007] (3) Manual modeling has a long cycle and high cost. Summary of the invention

[0008] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a method, system, computer-readable storage medium and electronic device for automatic matching of full-factor tower models based on multi-source data, which can refine the three-dimensional vector modeling of towers, improve the accuracy and efficiency of three-dimensional vector modeling of towers, and realize scene applications such as tower simulation tilt analysis.

[0009] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0010] A first aspect of the present invention provides a method for automatic matching of full-element tower models based on multi-source data.

[0011] The method for automatic matching of full-factor tower models based on multi-source data includes the following steps:

[0012] Obtain point cloud tower data, and use a deep learning algorithm to extract tower contour features as the first contour feature;

[0013] Obtaining visible light tower data, and extracting tower contour features of the visible light data as second contour features;

[0014] The first contour feature and the second contour feature are integrated to obtain the pole tower feature information, and a pole tower attribute file is established;

[0015] Obtain drawings of towers of different voltage levels and types, and make full-element tower models including the tower head, tower body and tower legs according to the drawings, and build a full-element tower model library;

[0016] The tower elements in the full-element tower model library are matched based on the tower attribute file information, and the three-dimensional vector modeling of the tower is realized based on the successfully matched tower elements.

[0017] As an optional technical solution, obtain point cloud tower data and use deep learning algorithms to extract tower contour features, including:

[0018] Acquire 3D LiDAR data;

[0019] Solve the 3D laser radar data to obtain point cloud data;

[0020] De-noise the point cloud data, distinguish the category of each point according to the color and elevation information of each point in the point cloud, and obtain the point cloud tower data through classification and segmentation;

[0021] Preprocess the point cloud tower data, including data completion;

[0022] The PointNet++ deep learning algorithm is used to process the point cloud tower data, sample the point cloud in layers, and extract features from the local area around each sampling point to obtain the three-dimensional point cloud tower contour features.

[0023] As an optional technical solution, obtain visible light tower data and extract tower contour features from the visible light data, specifically including:

[0024] Acquire two-dimensional visible light data;

[0025] Preprocess the two-dimensional visible light data, including binarization and denoising;

[0026] Use the Canny edge detection algorithm to detect the edge of the tower;

[0027] Harris algorithm is used to extract tower feature points;

[0028] The extracted tower contour edge and tower feature point information are combined to obtain the visible light tower contour features.

[0029] As an optional technical solution, the first contour feature and the second contour feature are integrated to obtain the tower feature information, and a tower attribute file is established, which specifically includes:

[0030] Select the point cloud tower data and visible light tower data of the same tower;

[0031] Reduce the dimension of the 3D point cloud tower data, project the reduced point cloud tower data onto the 2D visible light data, and establish a fine-grained correspondence relationship;

[0032] Based on the point cloud tower data and the visible light tower data with established corresponding relationships, feature matching is performed on the first contour feature and the second contour feature;

[0033] After feature matching is completed, the feature points in the two-dimensional visible light data are merged with the feature points in the corresponding three-dimensional point cloud tower data to form fused feature points with color and position information;

[0034] Based on the fused feature points, a fused contour feature is obtained, and then a pole tower attribute file is established, wherein the pole tower attribute file includes the height, geographical location, and pole tower grade information of the pole tower model;

[0035] As an optional technical solution, based on the point cloud tower data and the visible light tower data with established corresponding relationships, the first contour feature and the second contour feature are matched, specifically including:

[0036] Use the nearest neighbor search algorithm to find the similarity between the descriptors and perform initial feature matching between the feature points of the first contour feature and the second contour feature;

[0037] Then, the position and orientation information of the feature points in three-dimensional space are considered to further filter the matches;

[0038] Finally, the matches are further screened by using the information geometric constraints in the 3D point cloud. If the distance between two matching feature points in 3D space exceeds a set threshold, the two matches are considered to be wrong.

[0039] As an optional technical solution, after feature matching is completed, the feature points in the two-dimensional visible light data are merged with the feature points in the corresponding three-dimensional point cloud tower data to form fused feature points with color and position information, specifically including:

[0040] The weighted fusion method is used to merge feature points;

[0041] In specific weighted fusion, feature points in two-dimensional visible light data are set to have a greater weight in terms of color information, while feature points in three-dimensional point cloud tower data are set to have a greater weight in terms of positioning and depth description;

[0042] The fused feature points are used to train the classifier to identify the tower contour information, and the accuracy and reliability of the fused feature points are verified. The loss function of the classifier selects the boundary loss function.

[0043] As an optional technical solution, a full-element tower model library is constructed, including:

[0044] Obtain drawings of towers of different voltage levels and types, including wine glass towers, dry-shaped towers, drum towers, cathead towers, sheep horn towers and steel pipe towers;

[0045] The tower is divided into three parts: tower head, tower body and tower legs, and the overall tower modeling is carried out to achieve full-factor modeling;

[0046] Use symmetry to model;

[0047] Separate the various parts and detailed components of the model to meet the visual effects and overall scene smoothness under different scene conditions.

[0048] As an optional technical solution, the tower elements in the full-element tower model library are matched based on the tower attribute file information, and the three-dimensional vector modeling of the tower is realized based on the successfully matched tower elements, which specifically includes:

[0049] Obtaining the pole tower grade information in the pole tower attribute file, searching the full-element pole tower model library for full-element components and pole tower models that match the pole tower grade, wherein the full-element components include tower head, tower body and tower leg components;

[0050] Based on the tower attribute file information, the geometric structure and topological structure information of the tower head, tower body and tower leg components searched out are matched in sequence;

[0051] The tower head, tower body and tower leg components with the highest overlap are superimposed and matched, and the scaling coefficients of the three are adjusted to obtain a tower model with completely consistent tower head, height and profile;

[0052] As an optional technical solution, based on the tower attribute file information, the geometric structure and topological structure information of the tower head, tower body, and tower leg components searched out are matched in sequence, specifically including:

[0053] When matching tower heads, the type of tower head is determined based on the geometric structure, and all tower head components found out are traversed in sequence. The tower head with a higher matching degree is more likely to meet the modeling requirements.

[0054] When matching the tower body, the main material and the diagonal material characteristics are matched. After the tower body component structure meets the requirements, the telescopic tower body component structure is calculated to coincide with the tower body structure characteristics of the tower outline. The higher the coincidence, the more it meets the requirements.

[0055] The tower legs are matched in the same way as the tower body.

[0056] As an optional technical solution, it also includes obtaining the coordinates of the tower head and tower leg center points of the point cloud tower according to the tower attribute file, and calculating the tower inclination angle:

[0057] q=S / H,

[0058] Among them, q is the inclination angle of the tower; S is the inclination value of the tower, and H is the actual height of the tower. The laser radar is used to determine the bottom center point O, the top center point O1 and its projection point O ′ 1, then the distance from the center point of the tower top to its bottom projection point is the actual height H of the tower, and the distance from the projection point to the center point of the tower bottom is the tower inclination value S.

[0059] A second aspect of the present invention provides a full-element tower model automatic matching system based on multi-source data.

[0060] The full-factor tower model automatic matching system based on multi-source data includes:

[0061] The first contour feature extraction module is configured to: obtain point cloud tower data, and use a deep learning algorithm to extract tower contour features as first contour features;

[0062] The second contour feature extraction module is configured to: obtain visible light tower data, and extract tower contour features of the visible light data as second contour features;

[0063] The tower attribute file establishment module is configured to: fuse the first contour feature and the second contour feature to obtain tower feature information, and establish a tower attribute file;

[0064] The tower model library establishment module is configured to: obtain tower drawings of different types and different voltage levels, make a full-element tower model including three tower elements, namely, tower head, tower body and tower leg, according to the drawings, and build a full-element tower model library;

[0065] The matching modeling module is configured to: match the pole tower elements in the full-element pole tower model library based on the pole tower attribute file information, and realize the three-dimensional vector modeling of the pole tower based on the successfully matched pole tower elements.

[0066] The third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the method for automatic matching of full-element tower models based on multi-source data as described in the first aspect of the present invention.

[0067] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for automatic matching of full-element tower models based on multi-source data as described in the first aspect of the present invention are implemented.

[0068] One or more of the above technical solutions have the following beneficial effects:

[0069] The present invention provides a method, system, computer-readable storage medium and electronic device for automatic matching of full-element pole tower models based on multi-source data, which obtain point cloud data, obtain point cloud pole tower data after denoising and manual classification and segmentation, and extract pole tower contour features by using a deep learning algorithm after preprocessing the data; obtain visible light data, and extract pole tower contour features of the visible light data by using a Canny edge detection algorithm and a Harris feature extraction algorithm respectively through preprocessing; obtain high-precision pole tower feature information by fusing the pole tower contour features extracted from the two types of data, and statistically calculate tower type position, height and other information to establish a pole tower data attribute file; obtain pole tower drawings of different types and different voltage levels, and prepare full-element pole tower models including a tower head, a tower body and a tower leg according to the drawings to obtain a complete pole tower model library; query the pole tower elements in the full-element pole tower model library through attribute file information to automatically match feature points, rotate and scale the model proportionally, and realize three-dimensional vectorized pole tower modeling. The present invention obtains a three-dimensional vectorized tower model by matching the tower model in the model library with the tower attribute file and realizing componentized modeling. The model is rotated and translated to facilitate the analysis of the tower inclination in the later stage.

[0070] In the related art, there is no modeling of feature matching methods for extracting features from multi-source data of pole towers, resulting in low accuracy in extracting pole tower feature points. The present invention combines the pole tower contour features extracted from point cloud pole tower data and visible light pole tower data, and performs high-precision matching and fusion of the features, thereby improving the accuracy of pole tower feature points and facilitating subsequent more accurate matching of the full-element pole tower model library.

[0071] In the process of feature matching between the first contour feature and the second contour feature, the present invention first uses the nearest neighbor search algorithm to find the similarity between the descriptors, and performs initial feature matching between the feature points of the first contour feature and the second contour feature; then, the position and direction information of the feature points in the three-dimensional space are considered to further screen the matching; finally, the geometric constraints of the information in the 3D point cloud are used to further screen the matching. If the distance between the two matching feature points in the 3D space exceeds the set threshold, the two matches are considered to be wrong. Through three matches, high-precision matching of contour features is achieved.

[0072] After feature matching is completed, the present invention merges the feature points in the two-dimensional visible light data with the feature points in the corresponding three-dimensional point cloud tower data to form fused feature points with color and position information. The weighted fusion method is used to merge the feature points. In the specific weighted fusion, the feature points in the two-dimensional visible light data are set to have a greater weight in terms of color information, while the feature points in the three-dimensional point cloud tower data are set to have a greater weight in terms of positioning and depth description. The fused feature points are used to train a classifier to identify the tower contour information, and the accuracy and reliability of the fused feature points are verified. The above process can ensure the accuracy of the fused feature points.

[0073] The present invention extracts features from multi-source data of pole towers and performs full-factor automatic matching modeling, thereby achieving high-precision combined modeling of tower heads, tower bodies, and tower legs, improving the reusability of the model. At the same time, the modeling efficiency and accuracy are greatly improved, making it easier to implement analysis of different pole tower tilt scenarios.

[0074] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0076] Figure 1 It is an overall flow chart of the first embodiment of the present invention.

[0077] Figure 2 This is a flow chart of first contour feature extraction in Embodiment 1 of the present invention.

[0078] Figure 3 This is a flow chart of the second contour feature extraction according to the first embodiment of the present invention.

[0079] Figure 4 The present invention is a flowchart of a tower attribute file establishment process according to an embodiment of the present invention.

[0080] Figure 5This is a flow chart of matching a tower attribute file with a full-element tower model library in accordance with an embodiment of the present invention.

[0081] Figure 6 It is a schematic diagram comparing the structures of a drum-shaped tower head and a ram-horn-shaped tower head according to an embodiment of the present invention.

[0082] Figure 7 This is a system structure diagram of embodiment 2 of the present invention.

[0083] Figure 8 This is a schematic diagram of an electronic device provided in Embodiment 4 of the present invention.

[0084] In the accompanying drawings, the components represented by the reference numerals are listed as follows:

[0085] 1 tower head, 2 tower body, 3 tower legs, 4 foundation, 5 seat plate, 6 boot plate, 7 main material, 8 diagonal material. DETAILED DESCRIPTION

[0086] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0087] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0088] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0089] Overall concept:

[0090] The present invention discloses a method and system for matching and generating full-element towers based on multi-source data feature extraction. The method comprises the following steps:

[0091] First, collect 3D lidar data, obtain point cloud data through solution, and then denoise and manually classify and segment the point cloud data to obtain point cloud tower data; secondly, collect visible light image data, and obtain corrected visible light data through image preprocessing; then, according to the 2D visible light feature extraction algorithm and the 3D point cloud deep learning algorithm, extract the tower feature points with high precision to obtain the outer contour features of the tower model, match the feature points in the 2D image with the feature points in the 3D point cloud, find the corresponding feature points as the contour points of the tower data, and establish the tower attribute file; further, collect tower drawings, draw the tower head, tower body, and tower leg models, and establish a full-factor tower model library; finally, automatically match the tower outer contour feature information attribute table with the full-factor tower model to obtain the 3D vector tower model, and realize the automatic modeling of the tower model.

[0092] The present invention collects three-dimensional laser radar data, and after preprocessing the data, identifies tower type feature information and detects key point information through a deep learning algorithm; collects two-dimensional visible light data, which not only has high resolution and rich texture information, but can also extract two-dimensional pole tower contour features as a supplement to the missing feature information of point cloud data, and the obtained pole tower contour features have higher accuracy; establishes a pole tower attribute file, obtains drawings of different types of pole towers with different voltage levels, uses blender three-dimensional modeling software to achieve three-dimensional modeling of the pole tower head, tower body, and tower legs, and establishes a full-element pole tower model library; queries the pole tower model components in the full-element pole tower model library through the pole tower attribute file, determines the spatial position and posture of the pole tower in the transmission line through feature matching and parameter setting proportional stretching, completes the matching between the point cloud and the pole tower model in the model library, and generates a corresponding three-dimensional structured model of the pole tower. It has the following advantages:

[0093] 1) The tower feature points obtained based on multi-source data are highly accurate;

[0094] 2) Model all the elements of the tower according to the drawings to improve modeling efficiency;

[0095] 3) Automatically match the tower model and feature information to achieve high-precision three-dimensional vector modeling;

[0096] 4) Simulate the tilt of pole towers to provide a rehearsal basis for studying the collapse and tilt of pole towers.

[0097] Embodiment 1

[0098] The method for automatic matching of full-factor tower models based on multi-source data provided in this embodiment specifically includes:

[0099] Collect point cloud data, and obtain point cloud tower data through denoising and manual classification;

[0100] Collect two-dimensional visible light data, pre-process the data, and obtain corrected two-dimensional tower data;

[0101] Use deep learning algorithms to extract characteristic contour points of point cloud tower data;

[0102] The Canny edge detection algorithm and Harris feature extraction algorithm are used to extract the contour features of the two-dimensional visible light tower data;

[0103] The extracted 3D point cloud tower features and the extracted 2D tower visible light features are matched and fused to obtain high-precision tower contour features;

[0104] Obtain the geographical location, tower elevation, and tower feature points of the three-dimensional point cloud tower data, and establish a point cloud tower attribute file;

[0105] Obtain tower drawings, establish full-factor tower models of different types and voltage levels, and obtain a tower model library;

[0106] The tower model library is queried through the point cloud tower attribute file to automatically match the tower to achieve three-dimensional vector modeling of the tower.

[0107] In some embodiments, the original three-dimensional point cloud data is obtained. In order to improve the data quality and avoid the influence of noise on the extracted data features, the data is first denoised. Then, the category of each point in the point cloud is manually distinguished according to the color and elevation information of each point. Through classification and segmentation, the point cloud tower data is obtained. Before using the deep learning algorithm to train the tower, the point cloud tower data needs to be preprocessed. Due to the occlusion of trees and other reasons, the extracted tower point cloud data may be incomplete. Since the tower is generally a symmetrical structure, the point cloud data can be completed. Then, the PointNet++ deep learning processing network is used to process the point cloud tower data.

[0108] In some embodiments, two-dimensional visible light data is obtained. Since there are various noises in real images, in order to reduce the impact on edge contours. First, the data is preprocessed, the image is converted into a binary image, the data calculation dimension is reduced, the algorithm calculation efficiency is improved, and Gaussian filtering is used for denoising. Secondly, the Canny edge detection algorithm is used to detect the edge of the tower, and the gradient and direction of each pixel point are calculated. For each pixel point, it is compared with the two adjacent pixels according to its gradient direction. If the gradient intensity of the pixel point is not the maximum value along the gradient direction, it is considered that the point is not an edge point, and its intensity is set to 0. The purpose is to retain the strongest edge response and eliminate weak edges. Then the edge points are connected to obtain the edge contour features of the tower.

[0109] Then, the Harris feature extraction algorithm is used to extract the tower contour feature points. Finally, by merging the feature information of the two, high-precision two-dimensional tower contour features are obtained.

[0110] In some embodiments, according to the feature matching and feature fusion methods, the two-dimensional visible light tower data features and the three-dimensional point cloud tower data features are matched and fused to obtain a high-precision tower outline.

[0111] The feature points in the two-dimensional image are fused and matched with the feature points in the three-dimensional point cloud, and the corresponding feature points are found to be the contour features of the number of towers that need to be matched.

[0112] The fusion of two-dimensional tower contour features and three-dimensional tower contour features obtains high-precision tower contour features through feature symmetry, feature overlap, and extraction of lines and points with similar features. The tower attribute file is established by measuring the point cloud tower height and position.

[0113] Furthermore, the height and position information of the point cloud tower data are obtained, and a tower attribute file is established with the fused tower wheel information.

[0114] In some embodiments, establishing a tower model library includes:

[0115] In order to establish a complete tower model, it is necessary to obtain sufficient drawings of tower voltage levels and types, including wine glass tower, dry tower, drum tower, cat head tower, sheep horn tower, and steel pipe tower. At the same time, in order to improve the reusability of the tower model and realize full-factor modeling, this method divides the tower into three parts: tower head, tower body, and tower leg, as well as overall tower modeling.

[0116] Due to the large number of models, in order to improve modeling efficiency, we try to use a symmetrical method to make the models. In addition, the various parts and detailed components of the model can be split to meet the visual effects and overall scene smoothness under different scenes.

[0117] In some embodiments, the matching and assembly of relevant model components in the model library is searched according to the pole tower attribute file, including: matching and fusion of the pole tower characteristic contour points in the pole tower attribute file and the tower head characteristic points in the model library, and calculating the matching accuracy. The higher the accuracy, the closer the construction model is. The tower body parameterization method is studied, and the model components that meet the requirements are constructed layer by layer in a building block style, and the parameters are adjusted iteratively. While keeping the tower head characteristic points coincident, the telescopic parameters are determined. The tower body height can also be changed by optimizing the parameters by the above method to complete the modeling of pole towers with the same tower head and different heights. This modeling method is highly flexible.

[0118] Match the tower height, obtain the telescopic parameters of the three components by matching the tower head, tower legs and tower body, and then stack and assemble the tower components to obtain a tower model with consistent height and outline.

[0119] In some embodiments, according to the selected tower model, by setting the rotation and expansion coefficients of the tower, the position of the tower model is calculated to achieve three-dimensional vector modeling of the tower.

[0120] After obtaining the tower model, it is necessary to adjust the tower's rotation scaling factor and the tower's longitude and latitude according to the tower attribute file to achieve tower model positioning.

[0121] In some embodiments, after obtaining the optimal tower model, the method further includes: calculating the inclination angle of the tower. The inclination of a power tower is generally defined as q=S / H, where S is the inclination value of the tower and H is the actual height of the tower. The tower is generally considered to be a rigid whole, and its inclination can be abstracted as the ratio of the deviation between the top center of gravity and the tower base center of gravity to the tower height. The laser radar is used to determine the bottom center point O, the top center point O1 and its projection point O ′1, then the distance from the center point of the tower top to its bottom projection point is the actual height H of the tower, and the distance from the projection point to the center point of the tower bottom is the inclination S of the tower, which can realize the measurement of the inclination state of the power tower.

[0122] In some embodiments, obtaining the inclination angle of the tower includes:

[0123] Segment the towers of the power corridor; pre-process the 3D point cloud of the tower and extract key point information, roughly calculate the inclination state of the tower based on the original 3D point cloud and key point coordinates, and rotate the tower body of the point cloud tower data according to the inclination; then fit the tower feature plane and tower body according to the point cloud elevation density. Finally, calculate the final tower inclination angle based on the center point of the tower plane and the theoretical center point of the tower.

[0124] Next, the technical solution of this embodiment will be further explained in conjunction with the accompanying drawings.

[0125] Combination Figure 1 As shown, the method for automatic matching of all-element towers for multi-source data feature extraction provided by the embodiment of the present disclosure includes the following steps:

[0126] S11, obtaining three-dimensional point cloud data, two-dimensional visible light data and tower model drawings of different levels and categories.

[0127] S12, 3D point cloud data denoising and manual classification, using deep learning algorithm to extract tower contours, the specific steps are shown in Figure 2 .

[0128] S121, acquiring three-dimensional laser radar data.

[0129] S122, solve the three-dimensional laser radar data to obtain point cloud data; in order to improve data quality and avoid the influence of noise on the extraction of data features, denoise the data; then manually distinguish the category of each point in the point cloud based on the color and elevation information of each point, and obtain point cloud tower data through classification and segmentation.

[0130] S123, before using the deep learning algorithm to train the pole tower, the point cloud pole tower data needs to be preprocessed. Due to reasons such as occlusion by trees, the extracted point cloud data of the pole tower may be incomplete. Since the pole tower is generally a symmetrical structure, the point cloud data can be completed.

[0131] S124 uses the PointNet++ deep learning algorithm to process the point cloud tower data, samples the point cloud in layers, and extracts features from the local area around each sampling point.

[0132] S125, define the loss function, define the loss function, supervise the model learning process, compare with the manual classification truth value, calculate the accuracy, and obtain the three-dimensional point cloud tower contour features.

[0133] S13, preprocessing of two-dimensional visible light data, using Canny edge detection algorithm to extract the edge contour of the tower and Harris feature extraction algorithm to extract the tower contour feature points. For specific steps, see Figure 3 .

[0134] S131, acquiring two-dimensional visible light data.

[0135] S132, first pre-process the data. In order to reduce the data calculation dimension and improve the algorithm calculation efficiency, the visible light data is binarized. In order to improve the data accuracy, Gaussian filtering is used to denoise the image.

[0136] S133, using the Canny edge detection algorithm to detect the edge of the tower, emphasizing the position and strength of the boundary. Calculate the gradient and direction of each pixel point, and for each pixel point, compare it with the two adjacent pixels according to its gradient direction. If the gradient intensity of the pixel point is not the maximum value along the gradient direction, it is considered that the point is not an edge point, and its intensity is set to 0. The purpose is to retain the strongest edge response, eliminate weak edges, and then connect the edge points to obtain the edge contour features of the tower. The gradient M is calculated as:

[0137]

[0138] Direction D is:

[0139]

[0140] Among them G y Represents the vertical gradient, G x Represents the horizontal gradient.

[0141] S134, using Harris algorithm to extract tower feature points. Harris algorithm has a different purpose from Canny edge detection algorithm, and its main purpose is to extract feature points in the image.

[0142] S135, combining the extracted tower contour and tower feature point information to obtain visible light tower contour features.

[0143] S14, the three-dimensional and two-dimensional tower contour features are merged, and the tower height and tower position are obtained to establish the tower attribute file.

[0144] The specific information included in the tower attribute file includes: tower height, tower outline features, center coordinates of the tower head, center coordinates of the tower body, center coordinates of the tower legs, features of the tower body diagonal and main materials, geographical location of the tower center, and tower grade information.

[0145] For specific steps, see Figure 4 ,Since 2D visible light data has high resolution and rich texture information, but insufficient depth information, 3D point cloud tower spatial information, rich depth information, but sometimes due to occlusion and other reasons, the data contour features are obviously missing, therefore, the tower features extracted by the two are fused to make up for the missing, and high-precision feature information can be obtained. The specific steps include:

[0146] S141, selecting visible light tower data and point cloud tower data of the same tower.

[0147] S142, dimensionality reduction of the three-dimensional point cloud tower data, projecting the reduced-dimensional data onto the two-dimensional visible light data, and establishing a fine-grained correspondence.

[0148] S143,

[0149] Using a nearest neighbor search algorithm to find similarities between descriptors for initial feature matching, the descriptors are feature point descriptors of the first contour feature and the second contour feature;

[0150] Then the position and orientation information of the feature points in three-dimensional space are considered to further filter the matches;

[0151] Compare the distance between two feature descriptors, use the approximate nearest neighbor search algorithm to accelerate the matching process, and use the ratio test method proposed in SIFT to exclude unreasonable matching pairs.

[0152] Finally, the matches are further screened by using geometric constraints such as the position in the 3D point cloud. If the distance between two matching key points in 3D space exceeds a certain threshold, the two matches are considered to be likely to be wrong.

[0153] S144, merging the feature points in the two-dimensional image with the feature points in the corresponding three-dimensional point cloud to form fused feature points with color and position information.

[0154] Specifically, a weighted fusion method is used. Since the image has rich texture information and high resolution, the color information of the visible light data is more important when fusing information. In terms of positioning and depth information, the three-dimensional point cloud data has obvious advantages, and the point cloud data is more important in this regard.

[0155] The weighted feature vector calculation formula is defined as f=f1W1+f2W2, where f1 and f2 represent the feature vectors of 2D image and 3D point cloud respectively, and W1 and W2 correspond to the weights of 2D image and 3D point cloud respectively. Suppose we find that the contour integrity of the features extracted by visible light is better, while the features extracted by point cloud are stronger in positioning and depth description. Therefore, we can set W1=0.7 and W2=0.3 to weight the fusion features, and use the weighted feature vector to train a classifier (such as support vector machine SVM) to identify the tower contour information. Finally, the results of manual inspection are used as the reference standard to verify the accuracy and reliability of the weighted fusion features.

[0156] It can be understood that the above merging of the feature points in the two-dimensional image with the feature points in the corresponding three-dimensional point cloud to form fused feature points with color and position information requires weighted fusion of the feature points in the two-dimensional image and the feature points in the corresponding three-dimensional point cloud according to the color and position information, that is, fusion is performed once according to color, and then fusion is performed once according to position information, and finally summed up to obtain fused feature points with color and position information.

[0157] S145, during the fusion process, it is necessary to define a loss function to measure the gap between the predicted results and the actual results and back-propagate to update the model parameters.

[0158] It should be noted that the defined loss function is a boundary loss function, which can better capture the contour edge L B (y,y ’ )=‖yy ′ ‖ 2 , where y and y ′ Represent the positions of the actual boundary points and the predicted boundary points respectively.

[0159] S146, collating the contour feature information after the fusion of the two-dimensional and three-dimensional tower data and obtaining the height, geographical location, and tower grade information of the tower model, and establishing a tower attribute file.

[0160] S15, establish a full-element tower model library based on tower drawings of different levels and types.

[0161] Optionally, a tower model library is established, including:

[0162] In order to establish a complete tower model, it is necessary to obtain sufficient voltage level and type drawings of the tower, including wine glass tower, dry tower, drum tower, cat head tower, sheep horn tower, and steel pipe tower. At the same time, in order to improve the reusability of the tower model and realize full-factor modeling, this method divides the tower into three parts: tower head, tower body, tower leg, and overall tower modeling. Due to the large number of models, in order to improve modeling efficiency, try to use a symmetrical method to make the model. In addition, the various parts and detailed parts of the model can be split to meet the visual effects and overall scene fluency under different scenes.

[0163] S16, automatically searching for the tower head, tower body and tower leg in the matching model library according to the tower attribute file, obtaining the expansion coefficients of the tower head, tower body and tower leg respectively through the center point and feature contour matching and overlap, and then superimposing and assembling the tower model to obtain a highly consistent 3D vector structure model of the same type.

[0164] Specifically, match the point cloud tower with the model. For specific steps, see Figure 5 :

[0165] S161, obtaining the tower attribute file that needs to be modeled.

[0166] S162: According to the tower data file information, for example, the tower is a tower outline of a 220kv line, it is necessary to search all the full-factor components and tower models related to the 220kv tower in the full-factor tower database.

[0167] S163, sequentially matching the queried geometric structure and topological structure information of the tower head, tower body, and tower leg components.

[0168] When matching tower heads, the type of tower head can be determined based on the geometric structure. For example, a drum-type tower head generally has three or four parallel tower beams, while a ram-horn-type tower head has two corner points and two or three tower beams below. Therefore, to distinguish whether it is a drum-type tower or a ram-horn tower, it is only necessary to determine whether there are two corner points in the tower head and parallel tower beams below the corner points. All tower head components found out can be traversed in turn. The tower head with a higher matching degree can better meet the modeling requirements.

[0169] The selection of tower body and tower leg components is similar. Due to the different tower heights, the diagonal materials, main materials and node spacing of the tower body are also different. Figure 6 ,in Figure 6The left picture shows a U-shaped pole tower, and the right picture shows a wine glass-shaped pole tower. Both the U-shaped pole tower and the wine glass-shaped pole tower include a tower head 1, a tower body 2, a tower leg 3, a foundation 4, a seat plate 5, a boot plate 6, a main material 7, and an oblique material 8, but the specific structures of the two are quite different. Therefore, the selection of tower body components requires matching the main material and oblique material feature selection component model. After the structure meets the requirements, the telescopic component structure is calculated to coincide with the tower body structure features of the pole tower outline. The higher the coincidence, the more it meets the requirements.

[0170] The matching method of the tower legs is the same as that of the tower body, so I will not go into details.

[0171] S164, superimpose and splice the tower head, tower body and tower leg components with the highest overlap, and adjust the scaling coefficient of the three to superimpose and assemble them to complete the tower modeling with completely consistent tower head, height and outline.

[0172] S17, according to the point cloud tower attribute file, obtain the coordinates of the tower head and the center point of the tower leg of the point cloud tower, and calculate the inclination angle of the tower.

[0173] Specifically, after obtaining the optimal tower model, it also includes: calculating the inclination angle of the tower. The inclination of the power tower is generally defined as q = S / H, where S is the inclination value of the tower and H is the actual height of the tower. The tower is generally equivalent to a rigid whole, and its inclination can be abstracted as the ratio of the deviation value between the top center of gravity and the tower base center of gravity to the tower height. Use laser radar to determine the bottom center point O of the tower, the top center point O1 and its projection point O ′ 1, then the distance from the center point of the tower top to its bottom projection point is the actual height H of the tower, and the distance from the projection point to the center point of the tower bottom is the inclination S of the tower, which can realize the measurement of the inclination state of the power tower.

[0174] Optionally, obtaining the inclination angle of the pole tower includes: segmenting the pole tower of the power corridor; preprocessing the three-dimensional point cloud of the pole tower and extracting key point information, roughly calculating the inclination state of the pole tower based on the original three-dimensional point cloud and the key point coordinates, and rotating the tower body of the point cloud pole tower data according to the inclination. Then, fitting the characteristic plane of the pole tower and the tower body according to the point cloud elevation density. Finally, the final inclination angle of the pole tower is calculated according to the center point of the tower body plane and the theoretical center point of the pole tower.

[0175] In an embodiment of the present invention, three-dimensional point cloud data is obtained, and after denoising and manual segmentation, three-dimensional point cloud tower data is obtained, and the three-dimensional tower feature contour is extracted through the PointNet++ deep learning algorithm; two-dimensional visible light data is obtained, and after preprocessing and feature extraction algorithm, the two-dimensional tower feature contour is obtained; through feature matching and fusion of three-dimensional and two-dimensional data, high-precision tower contour features with texture and depth information are obtained, and a tower attribute file is established based on height, position information, etc.

[0176] Drawings of different types of pole towers with different voltage levels are collected, and the three-dimensional modeling software Blender is used to model the pole tower legs, tower body, tower head and the entire tower according to the pole tower drawings, so as to facilitate the overall scaling of the pole tower or adjust the superposition and assembly of all elements with different scaling ratios. The model components in the pole tower model library are queried according to the pole tower attribute file, and the three-dimensional vector reconstruction of the pole tower is automatically realized. The reconstructed model rotates and scales the pole tower model according to the pole tower attribute file to realize the positioning of the pole tower model, and obtains the center point coordinates of the tower head and tower legs through the attribute file, calculates the inclination angle of the pole tower, and performs inclination analysis on the pole tower model. Through model matching, the disclosed embodiment makes the acquired three-dimensional vector model more accurate and efficient, and can quickly realize the inclination warning of the power corridor pole tower.

[0177] In the related art, multi-source data is not used to extract the contour features of the tower, which cannot make up for the loss of the extracted point cloud tower features and the low accuracy, resulting in large errors in modeling. When establishing the model, there is no adaptive adjustment of the proportions of the parameters of the tower head, tower body, and tower legs, but a unified telescopic transformation is performed, resulting in deformation problems in the tower model and large errors in the simulated tower tilt analysis. The present invention uses multi-source data feature extraction for full-factor modeling, uses multi-source data to extract tower contour features, makes full use of data information, makes up for the lack of data, and improves data accuracy at the same time. According to the tower attribute file, the tower model established by matching the corresponding tower components and adjusting the proportions of the tower head, tower body, and tower legs is more accurate. The efficiency and accuracy of modeling are improved.

[0178] Embodiment 2

[0179] This embodiment discloses a full-element tower model automatic matching system based on multi-source data.

[0180] The full-factor tower model automatic matching system based on multi-source data includes:

[0181] The first contour feature extraction module is configured to: obtain point cloud tower data, and use a deep learning algorithm to extract tower contour features as first contour features;

[0182] The second contour feature extraction module is configured to: obtain visible light tower data, and extract tower contour features of the visible light data as second contour features;

[0183] The tower attribute file establishment module is configured to: fuse the first contour feature and the second contour feature to obtain tower feature information, and establish a tower attribute file;

[0184] The tower model library establishment module is configured to: obtain tower drawings of different types and different voltage levels, make a full-element tower model including three tower elements, namely, tower head, tower body and tower leg, according to the drawings, and build a full-element tower model library;

[0185] The matching modeling module is configured to: match the pole tower elements in the full-element pole tower model library based on the pole tower attribute file information, and realize the three-dimensional vector modeling of the pole tower based on the successfully matched pole tower elements.

[0186] It can be understood that this embodiment discloses a full-element tower model automatic matching system based on multi-source data, which can also be divided into the following modules, including:

[0187] A collection module, configured to collect three-dimensional lidar data and two-dimensional visible light image data of the power corridor and to purchase and collect drawings of towers of different types and at different voltage levels;

[0188] The processing module is configured to process point cloud data, remove noise points and manual classification and segmentation, extract the tower type feature points, position and height information of the point cloud tower; use the point cloud algorithm to extract the three-dimensional point cloud tower feature points and the two-dimensional feature extraction algorithm to extract the tower visible light feature points, and use the three-dimensional modeling software blender to obtain full-factor tower models of different types and voltage levels according to the drawings to obtain the tower model library.

[0189] The modeling module is configured to establish a pole tower attribute file to match the point cloud pole tower model, and to establish a three-dimensional vector structure by matching the pole tower contour feature points with each other and adjusting the height parameters.

[0190] The positioning module is configured to be used for positioning the three-dimensional vectorized tower model and realizes model positioning through rotation and scaling.

[0191] Combination Figure 7 As shown, the embodiment of the present disclosure also provides a full-element pole tower automatic matching device 400 for multi-source data feature extraction, which is used to execute the full-element pole tower automatic matching method provided in the above content of the embodiment of the present invention. The full-element pole tower automatic matching device includes: an acquisition module 410, a processing module 420, a modeling module 430 and a positioning module 440.

[0192] The acquisition module 410 acquires three-dimensional point cloud data, two-dimensional visible light data, and collects CAD drawings of towers of different types and at different voltage levels.

[0193] The processing module 420 performs denoising and artificial human segmentation on the 3D point cloud data, obtains the 3D point cloud tower data, pre-processes the 2D visible light data, extracts the 2D tower contour features using a feature extraction algorithm, obtains high-precision tower contour features with color and depth information through feature matching and feature fusion, and establishes a tower attribute file based on tower height, location, grade and other information. According to different grades and types of CAD drawings, a full-factor model of the tower head, tower body and tower legs is established.

[0194] The modeling module 430 queries the assembly parameters of the tower head, tower body and tower legs in the full-element tower model library according to the tower attribute file, automatically matches the tower components, and establishes a tower vectorization model.

[0195] The positioning module 440 is used for positioning the three-dimensional vectorized tower model and realizes model positioning through rotation and scaling.

[0196] Embodiment 3

[0197] The purpose of this embodiment is to provide a computer-readable storage medium.

[0198] A computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps in the method for automatic matching of full-element tower models based on multi-source data as described in Example 1 of the present disclosure.

[0199] The embodiment of the present disclosure provides a computer-readable storage medium storing computer executable instructions, wherein the computer executable instructions are configured to execute the above-mentioned method for automatic matching of element tower models for multi-source data feature extraction. The technical solution of the embodiment of the present disclosure can be embodied in the form of a software product, which is stored in a storage medium and includes one or more instructions for enabling a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in the embodiment of the present disclosure. The aforementioned storage medium may be a non-transitory storage medium, such as a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a disk or an optical disk, and other media that can store program codes. The above description and the accompanying drawings fully illustrate the embodiments of the present disclosure so that those skilled in the art can practice them. Other embodiments may include structural, logical, electrical, process and other changes. The embodiments represent only possible changes. Unless explicitly required, individual components and functions are optional, and the order of operations may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the words used in this application are only used to describe the embodiments and are not used to limit the claims. As used in the description of the embodiments and claims, unless the context clearly indicates, the singular forms of "a", "an" and "the" are intended to include the plural forms as well. Similarly, the term "and / or" as used in this application refers to any and all possible combinations of one or more associated listings. In addition, when used in this application, the term "comprise" and its variants "comprises" and / or including (comprising) refer to the existence of stated features, wholes, steps, operations, elements, and / or components, but do not exclude the existence or addition of one or more other features, wholes, steps, operations, elements, components and / or these groups. In the absence of more restrictions, the elements defined by the statement "including one..." do not exclude the existence of other identical elements in the process, method or device including the elements. In this article, each embodiment may focus on the differences from other embodiments, and the same and similar parts between the embodiments may refer to each other. For the methods, products, etc. disclosed in the embodiments, if they correspond to the method part disclosed in the embodiments, then the relevant parts can be found in the description of the method part.

[0200] Embodiment 4

[0201] The purpose of this embodiment is to provide an electronic device.

[0202] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for automatic matching of full-element tower models based on multi-source data as described in Example 1 of the present disclosure are implemented.

[0203] The disclosed embodiment provides an electronic device, including: an electronic device body, and the above-mentioned device 100 for automatic matching of full-element pole tower models for multi-source data feature extraction. The device 100 for automatic matching of full-element pole tower models for multi-source data feature extraction is installed on the electronic device body. The installation relationship described here is not limited to placement inside the electronic device body, but also includes installation connections with other components of the electronic device, including but not limited to physical connections, electrical connections or signal transmission connections, etc. Those skilled in the art can understand that the device 100 for automatic matching of element pole tower models can be adapted to a feasible electronic device body, thereby realizing other feasible embodiments.

[0204] Combination Figure 8 As shown, the electronic device 100 provided in the embodiment of the present disclosure includes a processor 50 and a memory 51. Optionally, the electronic device 100 may also include a communication interface 52 and a bus 53. The processor 50, the communication interface 52, and the memory 51 may communicate with each other through the bus 53. The communication interface 52 may be used for information transmission. The processor 50 may call the logic instructions in the memory 51 to execute the method for automatic matching of the full-element tower model for multi-source data feature extraction of the above embodiment.

[0205] In addition, the logic instructions in the above-mentioned memory 51 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. The memory 51, as a computer-readable storage medium, can be used to store software programs, computer executable programs, such as program instructions / modules corresponding to the method in the embodiment of the present disclosure. The processor 50 executes functional applications and data processing by running the program instructions / modules stored in the memory 51, that is, the method for automatic matching of element tower models for multi-source data feature extraction in the above-mentioned embodiment is implemented.

[0206] The memory 51 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function; the data storage area may store data created according to the use of the terminal device, etc. In addition, the memory 51 may include a high-speed random access memory and may also include a non-volatile memory.

[0207] The steps involved in the apparatuses of the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0208] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0209] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A full-factor tower model automatic matching method based on multi-source data, characterized in that: The following steps are involved: Obtain point cloud tower data, and use a deep learning algorithm to extract tower contour features as the first contour feature; Obtaining visible light tower data, and extracting tower contour features of the visible light data as second contour features; The first contour feature and the second contour feature are integrated to obtain the pole tower feature information, and a pole tower attribute file is established; Obtain drawings of towers of different voltage levels and types, and make full-element tower models including the tower head, tower body and tower legs according to the drawings, and build a full-element tower model library; The tower elements in the full-element tower model library are matched based on the tower attribute file information, and the three-dimensional vector modeling of the tower is realized based on the successfully matched tower elements.

2. The method for automatically matching a full-factor tower model based on multi-source data according to claim 1, characterized in that: Obtain point cloud tower data and use deep learning algorithms to extract tower contour features, including: Acquire 3D LiDAR data; Solve the 3D laser radar data to obtain point cloud data; De-noise the point cloud data, distinguish the category of each point according to the color and elevation information of each point in the point cloud, and obtain the point cloud tower data through classification and segmentation; Preprocess the point cloud tower data, including data completion; The PointNet++ deep learning algorithm is used to process the point cloud tower data, sample the point cloud in layers, and extract features from the local area around each sampling point to obtain the three-dimensional point cloud tower contour features.

3. The method for automatic matching of all-element tower models based on multi-source data according to claim 1, characterized in that: Obtain visible light tower data and extract tower profile features from the visible light data, including: Acquire two-dimensional visible light data; Preprocess the two-dimensional visible light data, including binarization and denoising; Use the Canny edge detection algorithm to detect the edge of the tower; Use Harris algorithm to extract tower feature points; The extracted tower contour edge and tower feature point information are combined to obtain the visible light tower contour features.

4. The method for automatically matching a full-element tower model based on multi-source data according to claim 1, characterized in that: The first contour feature and the second contour feature are integrated to obtain the tower feature information and establish a tower attribute file, which specifically includes: Select the point cloud tower data and visible light tower data of the same tower; Reduce the dimension of the 3D point cloud tower data, project the reduced point cloud tower data onto the 2D visible light data, and establish a fine-grained correspondence relationship; Based on the point cloud tower data and the visible light tower data with established corresponding relationships, feature matching is performed on the first contour feature and the second contour feature; After feature matching is completed, the feature points in the two-dimensional visible light data are merged with the feature points in the corresponding three-dimensional point cloud tower data to form fused feature points with color and position information; Based on the fused feature points, the fused contour features are obtained, and then a pole tower attribute file is established. The information in the pole tower attribute file includes the pole tower height, pole tower contour features, the center coordinates of the tower head, the center coordinates of the tower body, the center coordinates of the tower legs, the tower body oblique material and main material features, the pole tower center geographical location and the pole tower grade information; or, Based on the point cloud tower data and the visible light tower data with established corresponding relationships, feature matching is performed on the first contour feature and the second contour feature, specifically including: Use the nearest neighbor search algorithm to find the similarity between the descriptors and perform initial feature matching between the feature points of the first contour feature and the second contour feature; Then, the position and orientation information of the feature points in three-dimensional space are considered to further filter the matches; Finally, the matches are further screened by using the information geometric constraints in the 3D point cloud. If the distance between two matching feature points in 3D space exceeds a set threshold, the two matches are considered to be wrong. or, After feature matching is completed, the feature points in the two-dimensional visible light data are merged with the feature points in the corresponding three-dimensional point cloud tower data to form fused feature points with color and position information, including: The weighted fusion method is used to merge feature points; In specific weighted fusion, feature points in two-dimensional visible light data are set to have a greater weight in terms of color information, while feature points in three-dimensional point cloud tower data are set to have a greater weight in terms of positioning and depth description; The fused feature points are used to train the classifier to identify the tower contour information, and the accuracy and reliability of the fused feature points are verified. The loss function of the classifier selects the boundary loss function.

5. The method for automatic matching of all-element tower models based on multi-source data according to claim 1, characterized in that: Construct a full-element tower model library, including: Obtain drawings of towers of different voltage levels and types, including wine glass towers, dry-shaped towers, drum towers, cathead towers, sheep horn towers and steel pipe towers; The tower is divided into three parts: tower head, tower body and tower legs, and the overall tower modeling is carried out to achieve full-factor modeling; Use symmetry to model; Separate the various parts and detailed components of the model to meet the visual effects and overall scene smoothness under different scene conditions.

6. The method for automatic matching of all-element tower models based on multi-source data according to claim 1, characterized in that: Based on the tower attribute file information, the tower elements in the full-element tower model library are matched, and based on the successfully matched tower elements, the three-dimensional vector modeling of the tower is realized, which includes: Obtaining the pole tower grade information in the pole tower attribute file, searching the full-element pole tower model library for full-element components and pole tower models that match the pole tower grade, wherein the full-element components include tower head, tower body and tower leg components; Based on the tower attribute file information, the geometric structure and topological structure information of the tower head, tower body and tower leg components searched out are matched in sequence; The tower head, tower body and tower leg components with the highest overlap are superimposed and matched, and the scaling coefficients of the three are adjusted to obtain a tower model with completely consistent tower head, height and profile; or, Based on the tower attribute file information, the geometric structure and topological structure information of the tower head, tower body, and tower leg components searched out are matched in sequence, including: When matching tower heads, the type of tower head is determined based on the geometric structure, and all tower head components found out are traversed in sequence. The tower head with a higher matching degree is more likely to meet the modeling requirements. When matching the tower body, the main material and the diagonal material characteristics are matched. After the tower body component structure meets the requirements, the telescopic tower body component structure is calculated to coincide with the tower body structure characteristics of the tower outline. The higher the coincidence, the more it meets the requirements. The tower legs are matched in the same way as the tower body.

7. The method for automatic matching of all-element tower models based on multi-source data according to claim 1, characterized in that: It also includes obtaining the tower head and tower leg center point coordinates of the point cloud tower according to the tower attribute file, and calculating the tower inclination angle: q=S / H, Among them, q is the inclination angle of the tower; S is the inclination value of the tower, and H is the actual height of the tower. The laser radar is used to determine the bottom center point O, the top center point O1 and its projection point O ′ 1, then the distance from the center point of the tower top to its bottom projection point is the actual height H of the tower, and the distance from the projection point to the center point of the tower bottom is the tower inclination value S.

8. The full-factor tower model automatic matching system based on multi-source data is characterized by: include: The first contour feature extraction module is configured to: obtain point cloud tower data, and use a deep learning algorithm to extract tower contour features as first contour features; The second contour feature extraction module is configured to: obtain visible light tower data, and extract tower contour features of the visible light data as second contour features; The tower attribute file establishment module is configured to: fuse the first contour feature and the second contour feature to obtain tower feature information, and establish a tower attribute file; The tower model library establishment module is configured to: obtain tower drawings of different types and different voltage levels, make a full-element tower model including three tower elements, namely, tower head, tower body and tower leg, according to the drawings, and build a full-element tower model library; The matching modeling module is configured to: match the pole tower elements in the full-element pole tower model library based on the pole tower attribute file information, and realize the three-dimensional vector modeling of the pole tower based on the successfully matched pole tower elements.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps in the method for automatic matching of full-element tower models based on multi-source data as described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the method for automatic matching of full-element tower models based on multi-source data as described in any one of claims 1 to 7 are implemented.

Images