Power transmission line three-dimensional reconstruction method and system based on edge side calculation

Abstract

The invention discloses a power transmission line three-dimensional reconstruction method and system based on edge side calculation, and the method comprises the steps: carrying out the feature extraction of collected image data of a power transmission line through a preset priori constraint and a preset structural feature extraction model based on an edge side environment, and obtaining key feature information; performing three-dimensional point cloud mapping in the edge side environment according to the key feature information, and constructing an initial three-dimensional model; and transmitting the initial three-dimensional model to a cloud for model optimization to obtain a three-dimensional model of the power transmission line. According to the method, a more accurate and effective feature extraction strategy is provided for feature extraction by using the prior constraint related to the geometric structure of the power transmission line, and feature extraction and initial model construction are directly performed in an edge side environment without transmitting a large amount of data to a cloud for processing; the network bandwidth pressure is reduced; and the construction speed of the three-dimensional model is improved.

Description

Technical Field

[0001] The present application belongs to the technical field of spatial data processing of power transmission lines, and specifically relates to a three-dimensional reconstruction method and system for power transmission lines based on edge-side computing. Background Art

[0002] With the development of smart grid and drone inspection technology, more accurate transmission line structure data collected by drones can be obtained to support the management and maintenance of power facilities in the power system. Among them, three-dimensional reconstruction of transmission lines is an important means of power facility management and maintenance. Building a transmission line model based on transmission line structure data can effectively check the nodes and lines that need to be optimized in the transmission line, and promptly discover power facilities with failure risks or failures, thereby improving the safety and operational stability of the power system.

[0003] Traditional 3D reconstruction methods often face the problems of large amount of calculation and long time consumption when constructing large-scale, high-precision transmission line models. Especially when processing massive amounts of data in a centralized manner on the cloud, network transmission delays and bandwidth limitations further increase the difficulty of calculation, resulting in high data transmission pressure and slow response speed. In addition, since centralized computing on the cloud lacks an effective feature extraction strategy before model construction, a large amount of data resources are wasted, further reducing the accuracy of model construction. Summary of the invention

[0004] The present application proposes a method and system for three-dimensional reconstruction of transmission lines based on edge-side computing, which uses prior constraints related to the geometric structure of the transmission lines to provide a more accurate and effective feature extraction strategy for feature extraction, and directly performs feature extraction and initial model construction in the edge-side environment. There is no need to transmit large amounts of data to the cloud for processing, which reduces network bandwidth pressure and improves the speed of building three-dimensional models.

[0005] A first aspect of the present application provides a method for three-dimensional reconstruction of a power transmission line based on edge-side computing, the method comprising:

[0006] Based on the edge side environment, the collected transmission line image data is subjected to feature extraction through preset prior constraints and preset structural feature extraction models to obtain key feature information;

[0007] Based on the key feature information, three-dimensional point cloud mapping is performed in the edge environment to build an initial three-dimensional model;

[0008] The initial three-dimensional model is transmitted to the cloud for model optimization to obtain a three-dimensional model of the transmission line.

[0009] The above scheme uses prior constraints to introduce additional information into the structural feature extraction model for feature extraction, which improves the generalization ability and performance of the model, and can extract more accurate key feature information of the transmission line that meets the requirements, and then use the key feature information to build the initial three-dimensional model. Because feature extraction and model construction are both performed on the edge side, there is no need to transmit data to the cloud for processing, which not only reduces dependence on cloud computing resources, but also reduces network transmission pressure, and will not cause the model to be built slowly due to network bandwidth limitations; in addition, model optimization in the cloud can further improve the accuracy of the three-dimensional model of the transmission line, making the reconstruction of the transmission line more standardized.

[0010] In a possible implementation method of the first aspect, based on the edge side environment, feature extraction is performed on the collected transmission line image data through preset prior constraints and a preset structural feature extraction model to obtain key feature point information, specifically:

[0011] Loading the prior constraints and the structural feature extraction model in the edge side environment;

[0012] Determining a feature extraction strategy of the structural feature extraction model according to the prior constraints;

[0013] Based on the feature extraction strategy, the feature extraction of the power transmission line image data is performed through the structural feature extraction model to obtain key feature information.

[0014] The above scheme directly processes the collected transmission line image data in real time in the edge environment, saving the time of uploading the data to the cloud; and uses prior constraints to obtain a more accurate feature extraction strategy, so as to more quickly locate and extract features of the transmission line image data, further saving data processing time.

[0015] In a possible implementation method of the first aspect, the priori constraint is specifically:

[0016] According to the preset geometric design of the transmission line, the geometric relationship of the power equipment is detected by using the preset prior model on the historical structural data of the transmission line to obtain several geometric key points of the transmission line;

[0017] The priori constraints are generated according to the geometric key points of the transmission line.

[0018] The above scheme is based on the geometric design of transmission lines. First, the geometric key points of the specified power equipment are identified from the historical structure data of the transmission lines, and then the prior constraints are generated based on the geometric key points of the transmission lines. In the subsequent feature extraction, the target features can be quickly identified and located from the data based on the prior constraints, thereby improving the efficiency of feature extraction.

[0019] In a possible implementation method of the first aspect, the prior model is specifically:

[0020] The specific formula of the prior model is:

[0021] P=F prior (I)→(x i ,y i ,z i ,n i ,θ i );

[0022] Where P is the set of key points of the transmission line extracted from the historical structure data I of the transmission line, and F prior (I) is the deep neural network based on which the prior model is based, (x i ,y i ,z i ) is the spatial coordinate of the key geometric points of the transmission line, n i is the normal vector of the key geometric point of the transmission line, θ i is the angle parameter of the geometric key point of the transmission line, and i is the i-th data in the historical structure data I of the transmission line.

[0023] In a possible implementation method of the first aspect, three-dimensional point cloud mapping is performed in an edge environment according to key feature information to construct an initial three-dimensional model, specifically:

[0024] Based on a preset spatial perspective, extracting a set of key feature points of the current transmission line from the key feature information;

[0025] The key feature point set is optimized and mapped into three-dimensional space through a multi-view stereo matching algorithm to obtain an initial three-dimensional model.

[0026] The above scheme performs a preliminary three-dimensional model construction in the edge environment; first, a key feature point set is extracted from the key feature information, and then the points in the point set are optimized through a multi-view stereo matching algorithm to reduce the noise in the point set and improve the accuracy of the data as much as possible. Then, the optimized point set is mapped into the three-dimensional space to construct an initial three-dimensional model.

[0027] In a possible implementation method of the first aspect, the initial three-dimensional model is transmitted to the cloud for model optimization to obtain a three-dimensional model of the transmission line, specifically:

[0028] After the initial three-dimensional model is transmitted to the cloud, the initial three-dimensional model is locally optimized and globally integrated according to the preset transmission line structure standard to obtain a three-dimensional model of the transmission line.

[0029] The above solution performs local optimization and global fusion of the initial 3D model in the cloud, corrects the errors in the model, and improves the details of the model. Because the cloud only needs to optimize the model and does not need to process a large amount of transmission line image data, this greatly reduces the computing load of the cloud and speeds up the entire transmission line reconstruction process.

[0030] A second aspect of the present application provides an error correction system for a current transformer with an air gap, the system comprising: a feature extraction module, a model construction module and a model optimization module;

[0031] The feature extraction module is used to extract features from the collected power transmission line image data based on the edge side environment through preset prior constraints and preset structural feature extraction models to obtain key feature information;

[0032] The model building module is used to perform three-dimensional point cloud mapping in the edge side environment according to key feature information to build an initial three-dimensional model;

[0033] The model optimization module is used to transmit the initial three-dimensional model to the cloud for model optimization to obtain a three-dimensional model of the transmission line

[0034] In a possible implementation of the second aspect, the feature extraction module includes: a key feature information extraction unit;

[0035] Among them, the key feature information extraction unit is used to load the prior constraints and the structural feature extraction model in the edge side environment; determine the feature extraction strategy of the structural feature extraction model according to the prior constraints; based on the feature extraction strategy, perform feature extraction on the transmission line image data through the structural feature extraction model to obtain key feature information.

[0036] In a possible implementation of the second aspect, the feature extraction module includes: a priori constraint construction unit;

[0037] Among them, the prior constraint construction unit is used to detect the geometric relationship of power equipment on the historical structure data of the transmission line according to the preset transmission line geometry design, using the preset prior model, to obtain a number of transmission line geometry key points; based on the transmission line geometry key points, the prior constraints are generated.

[0038] In a possible implementation of the second aspect, the model construction module includes: an initial model construction unit;

[0039] Among them, the initial model construction unit is used to extract the key feature point set of the current transmission line from the key feature information based on a preset spatial perspective; map the key feature point set to three-dimensional space through a multi-view stereo matching algorithm to obtain an initial three-dimensional model.

[0040] In a possible implementation of the second aspect, the model optimization module includes: a model verification optimization unit;

[0041] Among them, the model verification and optimization unit is used to transmit the initial three-dimensional model to the cloud, and then perform local optimization and global fusion on the initial three-dimensional model according to a preset transmission line structure standard to obtain a three-dimensional model of the transmission line. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the implementation manner will be briefly introduced below. Obviously, the drawings described below are only some implementation manners of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 It is a specific flow chart of a method for three-dimensional reconstruction of a power transmission line based on edge-side computing provided in a certain embodiment of the present application;

[0044] Figure 2 This is a structural diagram of a three-dimensional reconstruction system for power transmission lines based on edge-side computing provided in a certain embodiment of the present application. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0046] It should be understood that the step numbers used in this article are only for the convenience of description and are not intended to limit the order in which the steps are executed.

[0047] First embodiment

[0048] In order to more conveniently monitor the operation of transmission lines and power equipment, it is necessary to reconstruct the three-dimensional model of the transmission line and build a more detailed working condition of the electrical equipment. However, the workload of reconstructing the transmission line model is generally large and requires processing massive amounts of data. However, if all this data is placed in the cloud for centralized computing, it will increase the pressure on cloud computing. In addition, transmitting massive amounts of data to the cloud will also take a long time due to bandwidth limitations and network delays, further reducing the efficiency of model building.

[0049] Therefore, instead of processing data on the cloud, the data is directly processed on the edge side and the initial three-dimensional model of the power transmission line is constructed, which can save the time of transmitting massive data and relieve the pressure of cloud computing. In addition, the accuracy of model reconstruction can be further improved by adjusting the feature extraction strategy on the edge side.

[0050] like Figure 1 As shown, Figure 1 A specific flow chart of a method for three-dimensional reconstruction of a power transmission line based on edge-side computing is provided for a certain embodiment of the present application. The method for three-dimensional reconstruction of a power transmission line based on edge-side computing of this embodiment includes steps S1 to S3, which are described in detail as follows:

[0051] Step S1, based on the edge side environment, feature extraction is performed on the collected transmission line image data through preset prior constraints and a preset structural feature extraction model to obtain key feature information.

[0052] In an embodiment of the present application, a large amount of historical structure data of transmission lines needs to be collected as samples, and then a target detection network that can capture the characteristics of transmission lines is constructed based on a convolutional neural network, that is, a priori model, and the historical structure data of the transmission lines is input into this priori model for geometric feature extraction to obtain prior constraints.

[0053] Furthermore, in the process of geometric feature extraction, the spatial relationship and size regularity of the electrical facilities in the transmission line are extracted and processed based on the preset transmission line geometry design to obtain the corresponding transmission line geometry key points, wherein each transmission line geometry key point contains spatial coordinates, normal vectors and angle parameters. Based on the transmission line geometry key points, the key structure of the transmission line at the branch and node can be determined, which provides important data support for reconstructing the three-dimensional model of the transmission line.

[0054] Optionally, the embodiment of the present application uses YOLOv5 as the target detection network, which can be used for real-time object detection with excellent data processing speed and accuracy. The geometric feature extraction is mainly for important components such as towers, insulator strings and conductors on the transmission line.

[0055] The specific formula of the prior model is:

[0056] P=F prior (I)→(x i ,y i ,z i ,n i ,θ i );

[0057] Where P is the set of key points of the transmission line extracted from the historical structure data I of the transmission line, and Fprior (I) is the deep neural network based on which the prior model is based, (x i ,y i ,z i ) is the spatial coordinate of the key geometric points of the transmission line, n i is the normal vector of the key geometric point of the transmission line, θ i is the angle parameter of the geometric key point of the transmission line, and i is the i-th data in the historical structure data I of the transmission line.

[0058] Because these transmission line geometric key points can very typically indicate the characteristics of the transmission line in a specific physical structure, they can be used as prior constraints to identify and locate key features in other transmission line images. In the embodiment of the present application, the transmission line geometric key points are used as prior constraints for the structural feature extraction model to improve the generalization ability of the model, impose certain restrictions on the model parameters or prediction results, and reduce the prediction error of the model; secondly, the prior constraints can be used to reduce data processing time and computing resources in the process of model feature extraction.

[0059] First, after the transmission line image data is collected by the drone or other equipment, the prior constraints and structural feature extraction model loaded in the edge environment are used to extract features from the transmission line image data on the spot, so as to use edge computing to complete feature extraction from massive data, instead of transmitting the collected transmission line image data to the cloud for processing. The beneficial effect of using edge computing in the embodiment of the present application is that the application is initiated on the edge side, which can generate a faster network service response, and can directly process data locally without transmitting a large amount of data to the cloud, saving data transmission time, especially when the amount of data to be transmitted is very large, the time saved is more objective.

[0060] Among them, edge computing refers to providing the closest service on the side close to the object or data source, which can realize real-time data processing and reduce data transmission delay.

[0061] Optionally, in some embodiments, an open platform such as an unmanned aerial vehicle computer or an on-site intelligent terminal that integrates network, computing, storage, and application core capabilities can be used as the edge environment.

[0062] Before feature extraction, the feature extraction strategy of the structural feature extraction model is adjusted according to the prior constraints and the data noise is reduced.

[0063] Exemplarily, the structural feature extraction model described in the embodiment of the present application is generated based on ResNet50, which is widely used in image classification, target detection, image segmentation, etc. Therefore, by using prior constraints containing geometric key points of the transmission line, the geometric space knowledge about the transmission line is introduced into the structural feature extraction model, and a feature extraction strategy can be generated that uses the prior constraints to make assumptions and constraints on the targets in the transmission line image data.

[0064] Based on this feature extraction strategy, if there is missing data in the transmission line image data, the missing values ​​can be filled according to the prior constraints, and the missing values ​​will not have too much impact on the results; with the help of prior constraints, the structural feature extraction model can quickly identify and locate the spatial information of the reconstructed transmission line model in the transmission line image data, obtain key feature information, and improve the efficiency of feature extraction.

[0065] In the embodiment of the present application, the use of a priori constraints is equivalent to a data preprocessing of the power transmission line image data, which can avoid a large amount of computing resources being required to standardize the data before feature extraction, resulting in a waste of computing resources.

[0066] Step S2: Perform three-dimensional point cloud mapping in the edge environment based on key feature information to construct an initial three-dimensional model.

[0067] In the embodiment of the present application, the initial three-dimensional model is still constructed based on the key feature information in the original edge side environment.

[0068] Firstly, based on the preset spatial perspective, a set of key feature points with spatial association of the current power transmission line extracted from images of different spatial perspectives is extracted from the key feature information, and then the set of key feature points is optimized and mapped into three-dimensional space through a multi-view stereo matching algorithm.

[0069] Optionally, in order to adapt to the computing power of the edge environment, the embodiment of the present application adopts a lightweight 3D reconstruction algorithm, which is mainly a simplified multi-view stereo matching algorithm that simplifies the 3D reconstruction steps into two steps: 3D point set optimization and point cloud stitching and densification, thereby reducing the computing pressure on the edge side.

[0070] Furthermore, the three-dimensional point set optimization is to globally optimize the key feature point set through bundle adjustment, construct the Jacobian matrix about the camera parameters and key feature points, and then use the nonlinear optimization method Gauss-Newton method for optimization to complete the iterative optimization of the key feature point set and reduce data noise. Among them, the bundle adjustment is a method of global optimization with the goal of minimizing the coordinate error of the three-dimensional point cloud coordinates obtained by the reduction algorithm and the traditional three-dimensional point cloud reconstruction algorithm through the full amount of images.

[0071] Among them, the iterative optimization formula used for three-dimensional point set optimization is:

[0072]

[0073] Wherein, x is the camera parameter and key feature point, the camera parameter is the device parameter of the camera that shoots the transmission line image data, J is the Jacobian matrix, r is the reprojection error, and Δx is the camera parameter and key feature point that minimize the residual.

[0074] Furthermore, after the three-dimensional point set optimization is completed, the optimized key feature point set is subjected to point cloud splicing and densification to map the key feature points into the three-dimensional space and form an initial three-dimensional model.

[0075] The specific formula of the initial three-dimensional model is:

[0076] M edge =G(P1,P2,...,P n );

[0077] Where M edge is the initial 3D model constructed in the edge environment, P1, P2, ..., P n is the key feature point set, and G is the simplified multi-view stereo matching algorithm used in the embodiment of the present application.

[0078] Among them, point cloud stitching and densification are performed through the ICP method, which has the least computational effort. Since the prior constraints and three-dimensional point set optimization used previously reduce data noise, accurate and efficient three-dimensional mapping can be performed by using the ICP method on a high-precision key feature point set.

[0079] Step S3, transmitting the initial three-dimensional model to the cloud for model optimization to obtain a three-dimensional model of the transmission line.

[0080] In an embodiment of the present application, the constructed initial three-dimensional model is transmitted to the cloud, and then the initial three-dimensional model is locally optimized and globally integrated according to the preset transmission line structure standard to correct the errors in the model and improve the model details.

[0081] Because the previous feature extraction and preliminary model construction are performed on the edge side, and the subsequent model optimization is performed on the cloud, this division of labor and cooperation method places the part that requires processing massive data on the edge side, and the part that requires refined processing on the cloud, which greatly reduces the computing load on the cloud and speeds up the process of reconstructing the three-dimensional model of the entire transmission line.

[0082] Implementing the embodiments of the present application has the following beneficial effects:

[0083] The embodiment of the present application uses a priori constraints to introduce additional information about the specific structure of the transmission line into the structural feature extraction model for feature extraction, thereby improving the generalization ability and performance of the model, and can extract more accurate key feature information of the transmission line that meets the requirements, and then use the key feature information to build the initial three-dimensional model. Because feature extraction and model building are both performed on the edge side, there is no need to transmit data to the cloud for processing, which not only reduces dependence on cloud computing resources, but also reduces network transmission pressure, and will not cause the model to be built at a slow rate due to network bandwidth limitations; in addition, fine model optimization in the cloud can further improve the accuracy of the three-dimensional model of the transmission line, making the reconstruction of the transmission line more standardized.

[0084] Second embodiment

[0085] Furthermore, in order to execute the power transmission line 3D reconstruction system based on edge-side computing corresponding to the above method embodiment to achieve the corresponding functions and technical effects, Figure 2 A structural diagram of a power transmission line 3D reconstruction system based on edge side computing is provided. For ease of description, only the parts related to this embodiment are shown. The power transmission line 3D reconstruction system based on edge side computing provided by the embodiment of the present application includes:

[0086] The feature extraction module 201 is used to extract features from the collected power transmission line image data based on the edge side environment through preset prior constraints and a preset structural feature extraction model to obtain key feature information.

[0087] In an embodiment of the present application, a large amount of historical structure data of transmission lines needs to be collected as samples, and then a target detection network that can capture the characteristics of transmission lines is constructed based on a convolutional neural network, that is, a priori model, and the historical structure data of the transmission lines is input into this priori model for geometric feature extraction to obtain prior constraints.

[0088] Furthermore, in the process of geometric feature extraction, the spatial relationship and size regularity of the electrical facilities in the transmission line are extracted and processed based on the preset transmission line geometry design to obtain the corresponding transmission line geometry key points, wherein each transmission line geometry key point contains spatial coordinates, normal vectors and angle parameters. Based on the transmission line geometry key points, the key structure of the transmission line at the branch and node can be determined, which provides important data support for reconstructing the three-dimensional model of the transmission line.

[0089] Optionally, the embodiment of the present application uses YOLOv5 as the target detection network, which can be used for real-time object detection with excellent data processing speed and accuracy. The geometric feature extraction is mainly for important components such as towers, insulator strings and conductors on the transmission line.

[0090] Then, before feature extraction, the feature extraction strategy of the structural feature extraction model is adjusted according to the prior constraints and the data noise is reduced.

[0091] Exemplarily, the structural feature extraction model described in the embodiment of the present application is generated based on ResNet50, which is widely used in image classification, target detection, image segmentation, etc. Therefore, by using prior constraints containing geometric key points of the transmission line, the geometric space knowledge about the transmission line is introduced into the structural feature extraction model, and a feature extraction strategy can be generated that uses the prior constraints to make assumptions and constraints on the targets in the transmission line image data.

[0092] Finally, based on the feature extraction strategy, the structural feature extraction model is used to extract features from the transmission line image data to obtain key feature information.

[0093] The model building module 202 is used to perform three-dimensional point cloud mapping in the edge side environment according to key feature information to build an initial three-dimensional model.

[0094] In the embodiment of the present application, in the original edge side environment, firstly, based on the preset spatial perspective, a set of key feature points with spatial association of the current power transmission line extracted from images of different spatial perspectives is extracted from the key feature information. Then, the set of key feature points is optimized and mapped into the three-dimensional space through a multi-view stereo matching algorithm to construct an initial three-dimensional model.

[0095] The model optimization module 203 is used to transmit the initial three-dimensional model to the cloud for model optimization to obtain a three-dimensional model of the transmission line.

[0096] In an embodiment of the present application, the constructed initial three-dimensional model is transmitted to the cloud, and then the initial three-dimensional model is locally optimized and globally integrated according to the preset transmission line structure standard to correct the errors in the model and improve the model details.

[0097] Because the previous feature extraction and preliminary model construction are performed on the edge side, and the subsequent model optimization is performed on the cloud, this division of labor and cooperation method places the part that requires processing massive data on the edge side, and the part that requires refined processing on the cloud, which greatly reduces the computing load on the cloud and speeds up the process of reconstructing the three-dimensional model of the entire transmission line.

[0098] In some embodiments, the feature extraction module 201 further includes:

[0099] A key feature information extraction unit is used to load the prior constraints and the structural feature extraction model in the edge side environment; determine the feature extraction strategy of the structural feature extraction model according to the prior constraints; based on the feature extraction strategy, extract features of the transmission line image data through the structural feature extraction model to obtain key feature information.

[0100] The prior constraint construction unit is used to detect the geometric relationship of power equipment on the historical structure data of the transmission line according to the preset transmission line geometry design using the preset prior model to obtain a number of transmission line geometry key points; and generate the prior constraints according to the transmission line geometry key points.

[0101] The specific formula of the prior model is:

[0102] P=F prior (I)→(x i ,y i ,z i ,n i ,θ i );

[0103] Where P is the set of key points of the transmission line extracted from the historical structure data I of the transmission line, and F prior (I) is the deep neural network based on which the prior model is based, (x i ,y i ,z i ) is the spatial coordinate of the key geometric points of the transmission line, n i is the normal vector of the key geometric point of the transmission line, θ i is the angle parameter of the geometric key point of the transmission line, and i is the i-th data in the historical structure data I of the transmission line.

[0104] Because these transmission line geometric key points can very typically indicate the characteristics of the transmission line in a specific physical structure, they can be used as prior constraints to identify and locate key features in other transmission line images. In the embodiment of the present application, the transmission line geometric key points are used as prior constraints for the structural feature extraction model to improve the generalization ability of the model, impose certain restrictions on the model parameters or prediction results, and reduce the prediction error of the model; secondly, the prior constraints can be used to reduce data processing time and computing resources in the process of model feature extraction.

[0105] In some embodiments, the model building module 202 further includes:

[0106] The initial model construction unit is used to extract the key feature point set of the current transmission line from the key feature information based on a preset spatial perspective; map the key feature point set to a three-dimensional space through a multi-view stereo matching algorithm to obtain an initial three-dimensional model.

[0107] In an embodiment of the present application, in order to adapt to the computing power of the edge environment, a lightweight three-dimensional reconstruction algorithm is adopted. It is mainly a simplified multi-view stereo matching algorithm that simplifies the three-dimensional reconstruction steps into two steps: three-dimensional point set optimization and point cloud stitching and densification, thereby reducing the computing pressure on the edge side.

[0108] Furthermore, the three-dimensional point set optimization is to globally optimize the key feature point set through bundle adjustment, construct the Jacobian matrix about the camera parameters and key feature points, and then use the nonlinear optimization method Gauss-Newton method for optimization to complete the iterative optimization of the key feature point set and reduce data noise. Among them, the bundle adjustment is a method of global optimization with the goal of minimizing the coordinate error of the three-dimensional point cloud coordinates obtained by the reduction algorithm and the traditional three-dimensional point cloud reconstruction algorithm through the full amount of images.

[0109] Among them, the iterative optimization formula used for three-dimensional point set optimization is:

[0110]

[0111] Wherein, x is the camera parameter and key feature point, the camera parameter is the device parameter of the camera that shoots the transmission line image data, J is the Jacobian matrix, r is the reprojection error, and Δx is the camera parameter and key feature point that minimize the residual.

[0112] Furthermore, after the three-dimensional point set optimization is completed, the optimized key feature point set is subjected to point cloud splicing and densification to map the key feature points into the three-dimensional space and form an initial three-dimensional model.

[0113] The specific formula of the initial three-dimensional model is:

[0114] M edge =G(P1,P2,...,P n );

[0115] Where M edge is the initial 3D model constructed in the edge environment, P1, P2, ..., P n is the key feature point set, and G is the simplified multi-view stereo matching algorithm used in the embodiment of the present application.

[0116] In some embodiments, the model optimization module 203 further includes:

[0117] The model verification and optimization unit is used to transmit the initial three-dimensional model to the cloud, and then perform local optimization and global fusion on the initial three-dimensional model according to a preset transmission line structure standard to obtain a three-dimensional model of the transmission line.

[0118] Implementing the embodiments of the present application has the following beneficial effects:

[0119] The embodiment of the present application uses the prior constraints related to the geometric structure of the transmission line to introduce prior knowledge into the structural feature extraction model, providing a more accurate and effective feature extraction strategy for feature extraction, which can quickly identify and locate key feature points in the transmission line image data, and improve the efficiency of data processing; then use a simplified multi-view stereo matching algorithm to simplify the steps of three-dimensional reconstruction into two steps: three-dimensional point set optimization, point cloud splicing and densification, reduce the computing pressure on the edge side, optimize the key feature points and map them to the three-dimensional space, and obtain the initial three-dimensional model; feature extraction and initial model construction are performed directly in the edge side environment, without the need to transmit a large amount of data to the cloud for processing, reducing the network bandwidth pressure and improving the construction speed of the three-dimensional model. Finally, the model is refined and optimized in the cloud, the errors in the model are corrected, and the model details are improved.

[0120] The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A three-dimensional reconstruction method for power transmission lines based on edge-side computing, characterized in that: include: Based on the edge side environment, the collected transmission line image data is subjected to feature extraction through preset prior constraints and preset structural feature extraction models to obtain key feature information; Based on the key feature information, three-dimensional point cloud mapping is performed in the edge environment to build an initial three-dimensional model; The initial three-dimensional model is transmitted to the cloud for model optimization to obtain a three-dimensional model of the transmission line.

2. The method for three-dimensional reconstruction of power transmission lines based on edge-side computing according to claim 1, characterized in that: Based on the edge side environment, the collected transmission line image data is subjected to feature extraction through preset prior constraints and a preset structural feature extraction model to obtain key feature point information, specifically: Loading the prior constraints and the structural feature extraction model in the edge side environment; Determining a feature extraction strategy of the structural feature extraction model according to the prior constraints; Based on the feature extraction strategy, the feature extraction of the power transmission line image data is performed through the structural feature extraction model to obtain key feature information.

3. The method for three-dimensional reconstruction of power transmission lines based on edge-side computing according to claim 1, characterized in that: The a priori constraints are specifically: According to the preset geometric design of the transmission line, the geometric relationship of the power equipment is detected by using the preset prior model on the historical structural data of the transmission line to obtain several geometric key points of the transmission line; The priori constraints are generated according to the geometric key points of the transmission line.

4. The method for three-dimensional reconstruction of power transmission lines based on edge-side computing according to claim 3, characterized in that: The prior model is specifically: The specific formula of the prior model is: P=F prior (I)→(x i ,y i ,z i ,n i ,θ i ): Where P is the set of key points of the transmission line extracted from the historical structure data I of the transmission line, and F prior (I) is the deep neural network based on which the prior model is based, (x i ,y i ,z i ) is the spatial coordinate of the key geometric points of the transmission line, n i is the normal vector of the key geometric point of the transmission line, θ i is the angle parameter of the geometric key point of the transmission line, and i is the i-th data in the historical structure data I of the transmission line.

5. The method for three-dimensional reconstruction of power transmission lines based on edge-side computing according to claim 1, characterized in that: According to the key feature information, three-dimensional point cloud mapping is performed in the edge environment to construct an initial three-dimensional model, specifically: Based on a preset spatial perspective, extracting a set of key feature points of the current transmission line from the key feature information; The key feature point set is optimized and mapped into three-dimensional space through a multi-view stereo matching algorithm to obtain an initial three-dimensional model.

6. The method for three-dimensional reconstruction of power transmission lines based on edge-side computing according to claim 1, characterized in that: The initial three-dimensional model is transmitted to the cloud for model optimization to obtain a three-dimensional model of the transmission line, specifically: After the initial three-dimensional model is transmitted to the cloud, the initial three-dimensional model is locally optimized and globally integrated according to the preset transmission line structure standard to obtain a three-dimensional model of the transmission line.

7. A three-dimensional reconstruction system for power transmission lines based on edge computing, characterized in that: include: Feature extraction module, model building module and model optimization module; The feature extraction module is used to extract features from the collected power transmission line image data based on the edge side environment through preset prior constraints and preset structural feature extraction models to obtain key feature information; The model building module is used to perform three-dimensional point cloud mapping in the edge side environment according to key feature information to build an initial three-dimensional model; The model optimization module is used to transmit the initial three-dimensional model to the cloud for model optimization to obtain a three-dimensional model of the transmission line.

8. The power transmission line 3D reconstruction system based on edge-side computing according to claim 7, characterized in that: The feature extraction module includes: a key feature information extraction unit; Among them, the key feature information extraction unit is used to load the prior constraints and the structural feature extraction model in the edge side environment; determine the feature extraction strategy of the structural feature extraction model according to the prior constraints; based on the feature extraction strategy, perform feature extraction on the transmission line image data through the structural feature extraction model to obtain key feature information.

9. The power transmission line 3D reconstruction system based on edge-side computing according to claim 7, characterized in that: The feature extraction module includes: a priori constraint construction unit; Among them, the prior constraint construction unit is used to detect the geometric relationship of power equipment on the historical structure data of the transmission line according to the preset transmission line geometry design, using the preset prior model, to obtain a number of transmission line geometry key points; based on the transmission line geometry key points, the prior constraints are generated.

10. The power transmission line 3D reconstruction system based on edge-side computing according to claim 7, characterized in that: The model building module includes: an initial model building unit; Among them, the initial model construction unit is used to extract the key feature point set of the current transmission line from the key feature information based on a preset spatial perspective; map the key feature point set to three-dimensional space through a multi-view stereo matching algorithm to obtain an initial three-dimensional model.

11. The power transmission line 3D reconstruction system based on edge-side computing according to claim 7, characterized in that: The model optimization module includes: a model verification optimization unit; Among them, the model verification and optimization unit is used to transmit the initial three-dimensional model to the cloud, and then perform local optimization and global fusion on the initial three-dimensional model according to a preset transmission line structure standard to obtain a three-dimensional model of the transmission line.

Images