Distribution network line wire three-dimensional construction method and device, terminal equipment and storage medium

Abstract

The invention discloses a three-dimensional construction method and device for a distribution network line wire, terminal equipment and a storage medium. The method comprises the following steps: acquiring a to-be-processed distribution network line image and internal and external parameters of a camera; performing image preprocessing according to the to-be-processed distribution network line image and a preset image enhancement mode to obtain a target distribution network line image; calculating a structure tensor according to the to-be-processed distribution network line image to obtain an image structure tensor of the to-be-processed distribution network line image; inputting the to-be-processed distribution network line image, the target distribution network line image and the image structure tensor into a preset image feature extraction neural network for feature fusion to obtain conductor feature point positions and conductor feature vectors; and constructing a three-dimensional model according to the wire feature point positions, the wire feature vectors and the internal and external parameters of the camera to obtain a distribution network line wire three-dimensional model of the to-be-processed distribution network line image. According to the invention, the accuracy of three-dimensional construction of the distribution network line lead is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional construction of power grids, and in particular to a three-dimensional construction method, device, terminal equipment and storage medium for distribution network line conductors. Background Art

[0002] Due to the low voltage level of the distribution network, its conductors are smaller than those of the transmission lines. When using drones to inspect distribution lines, in order to prevent the drones from touching obstacles in the complex environment of the distribution network, the drones often maintain a high altitude when flying, which results in a very low pixel ratio of the distribution network line conductors in the captured image. The distribution network conductors themselves are slender and thin in diameter, occupying a small proportion in the image, and their visual features are not obvious, so they are easily overlooked, resulting in the inability of ordinary visual 3D reconstruction algorithms to extract the effective features of the distribution network conductors in the image, which in turn results in the reconstructed 3D point cloud having intermittent conductors, which seriously affects the accuracy of the 3D reconstruction of the distribution network lines. Summary of the invention

[0003] The embodiments of the present invention provide a method, apparatus, terminal device and storage medium for three-dimensional construction of distribution network line conductors, which can effectively solve the problem that the prior art cannot extract effective features of distribution network conductors in images, which leads to discontinuous conductors in the reconstructed three-dimensional point cloud, seriously affecting the accuracy of three-dimensional reconstruction of distribution network lines.

[0004] An embodiment of the present invention provides a method for three-dimensionally constructing a distribution network line conductor, comprising:

[0005] Obtain the image of the distribution network line to be processed and the internal and external parameters of the camera;

[0006] Perform image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image;

[0007] Calculate the structure tensor according to the distribution network line image to be processed, and obtain the image structure tensor of the distribution network line image to be processed;

[0008] Input the to-be-processed distribution network line image, the target distribution network line image and the image structure tensor into a preset image feature extraction neural network for feature fusion to obtain the conductor feature point position and the conductor feature vector;

[0009] A three-dimensional model is constructed according to the positions of the conductor feature points, the conductor feature vectors and the internal and external parameters of the camera to obtain a three-dimensional model of the distribution network line conductor of the distribution network line image to be processed.

[0010] Furthermore, the image enhancement method includes: multi-scale edge detection and local contrast enhancement;

[0011] Performing image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image includes:

[0012] Performing several Gaussian blur and downsampling operations on the distribution network line image to be processed to generate several image layers with different resolutions;

[0013] At each resolution, edge detection is performed on the corresponding image layer to extract image edge information;

[0014] Fusing edge information of several images with different resolutions to obtain a first feature image;

[0015] Dividing the first feature image into a plurality of feature small blocks, and equalizing the brightness histogram of each small block to obtain a plurality of target feature small blocks;

[0016] Each target feature block is smoothly transitioned to obtain the target distribution network line image.

[0017] Further, calculating the structure tensor according to the distribution network line image to be processed to obtain the image structure tensor of the distribution network line image to be processed includes:

[0018] According to the distribution network line image to be processed, determining the coordinates of each pixel point in the distribution network line image to be processed;

[0019] According to the coordinates of each pixel point, the horizontal gradient and the vertical gradient of the distribution network line image to be processed are calculated respectively;

[0020] According to the coordinates of each pixel point, a Gaussian weighting function is calculated;

[0021] An image structure tensor of the distribution network line image to be processed is calculated based on the horizontal gradient, the vertical gradient and the Gaussian weighting function.

[0022] Furthermore, the distribution network line image to be processed, the target distribution network line image and the image structure tensor are input into a preset image feature extraction neural network for feature fusion to obtain the conductor feature point position and the conductor feature vector, including:

[0023] Inputting the distribution network line image to be processed into a preset image feature extraction neural network, performing convolution calculation according to a preset first convolution layer number, and obtaining a first convolution feature;

[0024] Inputting the target distribution network line image into a preset image feature extraction neural network, performing convolution calculation according to a preset second convolution layer number, and obtaining a second convolution feature;

[0025] Inputting the image structure tensor into a preset image feature extraction neural network, performing convolution calculation according to a preset third convolution layer number, and obtaining a third convolution feature;

[0026] Performing weighted feature fusion according to the first convolution feature, the second convolution feature, the third convolution feature and a preset weight to obtain a wire feature point position and a wire feature vector;

[0027] Among them, the number of the first convolutional layers, the number of the second convolutional layers, and the number of the third convolutional layers are different.

[0028] Furthermore, the image feature extraction neural network also includes: a first multi-layer perceptron for representing low-level feature learning and a second multi-layer perceptron for representing high-level feature learning;

[0029] Performing weighted feature fusion according to the first convolution feature, the second convolution feature, the third convolution feature and a preset weight to obtain a wire feature point position and a wire feature vector, including:

[0030] Performing a dimension alignment operation on the first convolution feature, the second convolution feature, and the third convolution feature to obtain the first convolution feature, the second convolution feature, and the third convolution feature after dimension alignment;

[0031] According to the preset weights and the first convolution feature after the alignment dimension, the second convolution feature and the third convolution feature are fused by feature splicing to obtain the initial fused feature;

[0032] Inputting the initial fusion features into the first multi-layer perceptron to capture the edge features, corner features and texture features of the wire, and obtaining target fusion features according to the edge features, corner features and texture features of the wire;

[0033] The target fusion feature is input into the second multi-layer perceptron to capture the spatial position feature and spatial structure feature of the wire, and the final wire feature point position and wire feature vector are obtained according to the spatial position feature and spatial structure feature of the wire.

[0034] Furthermore, a three-dimensional model is constructed according to the positions of the conductor feature points, the conductor feature vectors, and the internal and external parameters of the camera to obtain a three-dimensional model of the distribution network line conductor of the distribution network line image to be processed, including:

[0035] Calculating the similarity of each wire feature point according to the wire feature vector;

[0036] A basic matrix is ​​obtained by solving the similarity of each wire feature point and the position of the wire feature point;

[0037] Perform matrix decomposition according to the basic matrix and the internal and external parameters of the camera to obtain the camera pose;

[0038] Point cloud fusion is performed according to the camera posture and the position of the conductor feature points to generate a three-dimensional model of the distribution network line conductor of the distribution network line image to be processed.

[0039] As an improvement of the above solution, another embodiment of the present invention provides a three-dimensional construction device for distribution network line conductors, including:

[0040] A conductor image acquisition module is used to acquire the distribution network line image to be processed;

[0041] An image preprocessing module, used to perform image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image;

[0042] A structure tensor calculation module, used to calculate the structure tensor according to the distribution network line image to be processed, so as to obtain the image structure tensor of the distribution network line image to be processed;

[0043] An image feature fusion module is used to input the distribution network line image to be processed, the target distribution network line image and the image structure tensor into a preset image feature extraction neural network for feature fusion to obtain the conductor feature point position and the conductor feature vector;

[0044] The wire model building module is used to build a three-dimensional model according to the wire feature point position, the wire feature vector and the internal and external parameters of the camera to obtain a three-dimensional model of the distribution network line wire of the distribution network line image to be processed.

[0045] Furthermore, the image enhancement method includes: multi-scale edge detection and local contrast enhancement;

[0046] The image preprocessing module is used to perform image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image;

[0047] Performing image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image includes:

[0048] Performing several Gaussian blur and downsampling operations on the distribution network line image to be processed to generate several image layers with different resolutions;

[0049] At each resolution, edge detection is performed on the corresponding image layer to extract image edge information;

[0050] Fusing edge information of several images with different resolutions to obtain a first feature image;

[0051] Dividing the first feature image into a plurality of feature small blocks, and equalizing the brightness histogram of each small block to obtain a plurality of target feature small blocks;

[0052] Each target feature block is smoothly transitioned to obtain the target distribution network line image.

[0053] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, a three-dimensional construction method for distribution network line conductors as described in the above embodiment is implemented.

[0054] Another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a three-dimensional construction method for distribution network line conductors described in the above embodiment.

[0055] By implementing the present invention, at least the following beneficial effects are achieved:

[0056] The present invention provides a three-dimensional construction method, device, terminal equipment and storage medium for distribution network line conductors. The method can obtain a distribution network line image to be processed and internal and external parameters of a camera; perform image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image; calculate a structure tensor according to the distribution network line image to be processed to obtain an image structure tensor of the distribution network line image to be processed; input the distribution network line image to be processed, the target distribution network line image and the image structure tensor into a preset image feature extraction neural network for feature fusion to obtain image feature point positions and image feature vectors; construct a three-dimensional model according to the image feature point positions, the image feature vectors and the internal and external parameters of the camera to obtain a three-dimensional model of the distribution network line conductor of the distribution network line image to be processed. The distribution network line image to be processed is preprocessed by image enhancement to enhance feature representation. Feature fusion is performed through the distribution network line image to be processed, the target distribution network line image and the image structure tensor to effectively capture the features of small conductors. The image structure tensor enhances the continuity of small conductors in the image, obtains the position of conductor feature points and conductor feature vectors, avoids the situation where the conductor is ignored, and thus improves the accuracy of three-dimensional construction of distribution network line conductors. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a flow chart of a method for three-dimensionally constructing a distribution network line conductor provided by an embodiment of the present invention;

[0058] Figure 2 This is a feature fusion schematic diagram of a method for constructing three-dimensional distribution network line conductors provided by an embodiment of the present invention;

[0059] Figure 3 It is a structural schematic diagram of a three-dimensional construction device for distribution network line conductors provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] See also Figure 1 , is a flow chart of a method for constructing three-dimensional distribution network line conductors provided by an embodiment of the present invention, comprising:

[0062] S1, obtaining the image of the distribution network line to be processed and the internal and external parameters of the camera;

[0063] S2, performing image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image;

[0064] S3, calculating a structure tensor according to the distribution network line image to be processed, and obtaining an image structure tensor of the distribution network line image to be processed;

[0065] S4, inputting the to-be-processed distribution network line image, the target distribution network line image and the image structure tensor into a preset image feature extraction neural network for feature fusion to obtain the conductor feature point position and the conductor feature vector;

[0066] S5. Construct a three-dimensional model according to the positions of the conductor feature points, the conductor feature vectors, and the internal and external parameters of the camera to obtain a three-dimensional model of the distribution network line conductor of the distribution network line image to be processed.

[0067] Specifically, the distribution network line image to be processed can be obtained through the image taken by the inspection drone; the internal and external parameters of the camera include focal length, principal point coordinates, rotation matrix and translation vector.

[0068] In a preferred embodiment of the present invention, the image of the distribution network line to be processed and the internal and external parameters of the camera are first obtained; then, the image is preprocessed according to the distribution network line image to be processed and the preset image enhancement method to obtain the target distribution network line image; then, the structure tensor is calculated according to the distribution network line image to be processed to obtain the image structure tensor of the distribution network line image to be processed; then, the distribution network line image to be processed, the target distribution network line image and the image structure tensor are input into the preset image feature extraction neural network for feature fusion to obtain the conductor feature point position and the conductor feature vector; finally, a three-dimensional model is constructed according to the conductor feature point position, the conductor feature vector and the internal and external parameters of the camera to obtain the distribution network line conductor three-dimensional model of the distribution network line image to be processed. Through the steps of image structure tensor analysis and conductor feature extraction, the features of small conductors can be captured more effectively, the situation where the conductors are ignored or misidentified is reduced, and the accuracy of the distribution network line conductor detail extraction is improved.

[0069] Preferably, the image enhancement method includes: multi-scale edge detection and local contrast enhancement;

[0070] Performing image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image includes:

[0071] Performing several Gaussian blur and downsampling operations on the distribution network line image to be processed to generate several image layers with different resolutions;

[0072] At each resolution, edge detection is performed on the corresponding image layer to extract image edge information;

[0073] Fusing edge information of several images with different resolutions to obtain a first feature image;

[0074] Dividing the first feature image into a plurality of feature small blocks, and equalizing the brightness histogram of each small block to obtain a plurality of target feature small blocks;

[0075] Each target feature block is smoothly transitioned to obtain the target distribution network line image.

[0076] Specifically, the image enhancement method includes: multi-scale edge detection and local contrast enhancement. The multi-scale edge detection detects edges at different resolutions through a multi-scale spatial pyramid structure to improve the edge detection accuracy of small wires. According to the distribution network line image to be processed, several Gaussian blur and downsampling operations are performed to generate several image layers with different resolutions, that is, image layers with different scale factors are obtained, and then at each resolution, edge detection is performed on the corresponding image layer to extract image edge information; and the edge information of several images with different resolutions is fused to obtain a first feature image; the first feature image is divided into several feature blocks, and the brightness histogram of each special block is equalized to obtain several target feature blocks; each target feature block is smoothly transitioned to achieve local contrast enhancement and obtain the target distribution network line image.

[0077] In a preferred embodiment of the present invention, different scale factors can capture features of different sizes, smaller scale factors help detect finer details, while larger scale factors can capture larger structural features. This multi-scale approach enables the algorithm to adapt to the different sizes and complexities of distribution network line conductors, thereby effectively enhancing visual features and extracting details in low-pixel ratio images taken by drones, ultimately improving the accuracy and completeness of three-dimensional reconstruction. For example, after selecting three different scale factors for image preprocessing, the three images of different scales are processed separately, and the images under each scale factor will undergo corresponding edge detection and local contrast enhancement processing, and these processed images each retain the feature information at different scales. Assume that the first feature image is: Where s is the scale factor. In this embodiment, 1 / 2, 1 / 4 and 1 / 8 are selected. G s and I s represents the Gaussian kernel and image at the corresponding scale factor, and * represents the convolution operation.

[0078] Specifically, calculating the structure tensor according to the distribution network line image to be processed to obtain the image structure tensor of the distribution network line image to be processed includes:

[0079] According to the distribution network line image to be processed, determining the coordinates of each pixel point in the distribution network line image to be processed;

[0080] According to the coordinates of each pixel point, the horizontal gradient and the vertical gradient of the distribution network line image to be processed are calculated respectively;

[0081] According to the coordinates of each pixel point, a Gaussian weighting function is calculated;

[0082] An image structure tensor of the distribution network line image to be processed is calculated based on the horizontal gradient, the vertical gradient and the Gaussian weighting function.

[0083] Preferably, the horizontal gradient and the vertical gradient are obtained by calculating the difference in pixel values ​​of adjacent pixels in the horizontal direction of the distribution network line image to be processed. For each pixel I (x, y) in the image, its gradient in the x direction (horizontal direction) is expressed as: Similarly, the gradient in the y direction (vertical direction) is expressed as: in and are the horizontal gradient of image I in the x direction and the vertical gradient in the y direction respectively; x and y are the horizontal and vertical coordinates of each pixel in the distribution network line image to be processed respectively. Then, according to the coordinates of each pixel, the Gaussian weighting function is calculated: Finally, the image structure tensor of the distribution network line image to be processed is calculated based on the horizontal gradient, the vertical gradient and the Gaussian weighting function. The structure tensor is used to calculate the local directional strength to enhance the coherence and detectability of small wires. The calculated structure tensor is retained as an image feature. The image structure tensor is shown in the following formula: in w(x,y) is the Gaussian weighting function.

[0084] Schematically, the distribution network line image to be processed, the target distribution network line image and the image structure tensor are input into a preset image feature extraction neural network for feature fusion to obtain the conductor feature point position and the conductor feature vector, including:

[0085] Inputting the distribution network line image to be processed into a preset image feature extraction neural network, performing convolution calculation according to a preset first convolution layer number, and obtaining a first convolution feature;

[0086] Inputting the target distribution network line image into a preset image feature extraction neural network, performing convolution calculation according to a preset second convolution layer number, and obtaining a second convolution feature;

[0087] Inputting the image structure tensor into a preset image feature extraction neural network, performing convolution calculation according to a preset third convolution layer number, and obtaining a third convolution feature;

[0088] Performing weighted feature fusion according to the first convolution feature, the second convolution feature, the third convolution feature and a preset weight to obtain a wire feature point position and a wire feature vector;

[0089] Among them, the number of the first convolutional layers, the number of the second convolutional layers, and the number of the third convolutional layers are different.

[0090] In a preferred embodiment of the present invention, the image feature extraction neural network has three input branches, each branch extracts different features, and finally the multi-branch feature fusion is realized through weighted feature fusion to realize wire image feature extraction. Figure 2 As shown, the first convolution layer has four layers, and the distribution network line image to be processed passes through four convolution layers to obtain the first convolution feature; the second convolution layer has two layers, and the target distribution network line image passes through two convolution layers to obtain the second convolution feature; the third convolution layer has three layers, and the image structure tensor passes through three convolution layers to obtain the third convolution feature. Then, weighted feature fusion is performed based on the first convolution feature, the second convolution feature, the third convolution feature and the preset weights to obtain the wire feature point position and the wire feature vector.

[0091] Specifically, a dimension alignment operation is first performed according to the first convolution feature, the second convolution feature, and the third convolution feature to obtain the first convolution feature, the second convolution feature, and the third convolution feature after the alignment dimension; then, according to the preset weight, the first convolution feature, the second convolution feature, and the third convolution feature after the alignment dimension are merged by feature splicing ( Figure 2 Concat) to obtain the initial fusion feature; then the initial fusion feature is input into the first multi-layer perceptron ( Figure 2 The first MLP after Concat in the process is used to capture the edge features, corner features and texture features of the wire, and the target fusion features are obtained according to the edge features, corner features and texture features of the wire; finally, the target fusion features are input into the second multi-layer perceptron ( Figure 2 In the last MLP in the process, the spatial position features and spatial structure features of the wire are captured, and the final wire feature point position and wire feature vector are obtained according to the spatial position features and spatial structure features of the wire. After two MLPs are used for deep feature extraction, the final wire feature point position and wire feature vector are obtained, realizing the fusion of multi-branch features.

[0092] Preferably, a three-dimensional model is constructed according to the positions of the conductor feature points, the conductor feature vectors and the internal and external parameters of the camera to obtain a three-dimensional model of the distribution network line conductor of the distribution network line image to be processed, including:

[0093] Calculating the similarity of each wire feature point according to the wire feature vector;

[0094] A basic matrix is ​​obtained by solving the similarity of each wire feature point and the position of the wire feature point;

[0095] Perform matrix decomposition according to the basic matrix and the internal and external parameters of the camera to obtain the camera pose;

[0096] Point cloud fusion is performed according to the camera posture and the position of the conductor feature points to generate a three-dimensional model of the distribution network line conductor of the distribution network line image to be processed.

[0097] In a preferred embodiment of the present invention, according to the obtained conductor feature point position and conductor feature vector, a general 3D reconstruction method such as the motion recovery structure algorithm SFM and the multi-view stereo matching algorithm MVS can be used to achieve a complete 3D construction of the distribution network line conductor. Through the steps of multi-scale edge detection, image structure tensor analysis and conductor feature point position extraction, the features of small conductors can be captured more effectively, the situation of conductors being ignored or misidentified is reduced, and the accuracy of detail extraction is improved; at the same time, through the optimization of image preprocessing and feature extraction, the complex environment and lighting changes of the distribution network line can be better coped with, the interference of these factors on conductor feature extraction and 3D construction is reduced, and the accuracy of conductor 3D construction is improved.

[0098] By implementing this embodiment, the distribution network line image to be processed and the internal and external parameters of the camera are obtained; the image is preprocessed according to the distribution network line image to be processed and the preset image enhancement method to obtain the target distribution network line image; the structure tensor is calculated according to the distribution network line image to be processed to obtain the image structure tensor of the distribution network line image to be processed; the distribution network line image to be processed, the target distribution network line image and the image structure tensor are input into the preset image feature extraction neural network for feature fusion to obtain the image feature point position and the image feature vector; a three-dimensional model is constructed according to the image feature point position, the image feature vector and the internal and external parameters of the camera to obtain the distribution network line conductor three-dimensional model of the distribution network line image to be processed. The distribution network line image to be processed is preprocessed by the image enhancement method to enhance the feature representation, and the feature fusion is performed by the distribution network line image to be processed, the target distribution network line image and the image structure tensor to effectively capture the features of the small wires. The image structure tensor enhances the coherence of the small wires in the image, obtains the wire feature point position and the wire feature vector, avoids the situation where the wire is ignored, and thus improves the accuracy of the three-dimensional construction of the distribution network line wire.

[0099] See also Figure 3 , is a schematic diagram of a structure of a three-dimensional construction device for a distribution network line conductor provided by an embodiment of the present invention, comprising:

[0100] A conductor image acquisition module is used to acquire the distribution network line image to be processed;

[0101] An image preprocessing module, used to perform image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image;

[0102] A structure tensor calculation module, used to calculate the structure tensor according to the distribution network line image to be processed, so as to obtain the image structure tensor of the distribution network line image to be processed;

[0103] An image feature fusion module is used to input the distribution network line image to be processed, the target distribution network line image and the image structure tensor into a preset image feature extraction neural network for feature fusion to obtain the conductor feature point position and the conductor feature vector;

[0104] The wire model building module is used to build a three-dimensional model according to the wire feature point position, the wire feature vector and the internal and external parameters of the camera to obtain a three-dimensional model of the distribution network line wire of the distribution network line image to be processed.

[0105] Specifically, the image enhancement method includes: multi-scale edge detection and local contrast enhancement;

[0106] The image preprocessing module is used to perform image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image;

[0107] Performing image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image includes:

[0108] Performing several Gaussian blur and downsampling operations on the distribution network line image to be processed to generate several image layers with different resolutions;

[0109] At each resolution, edge detection is performed on the corresponding image layer to extract image edge information;

[0110] Fusing edge information of several images with different resolutions to obtain a first feature image;

[0111] Dividing the first feature image into a plurality of feature small blocks, and equalizing the brightness histogram of each small block to obtain a plurality of target feature small blocks;

[0112] Each target feature block is smoothly transitioned to obtain the target distribution network line image.

[0113] The present invention provides a three-dimensional construction device for a distribution network line conductor. The device obtains a distribution network line image to be processed according to a conductor image acquisition module; in an image preprocessing module, image preprocessing is performed according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image; in a structure tensor calculation module, a structure tensor is calculated according to the distribution network line image to be processed to obtain an image structure tensor of the distribution network line image to be processed; through an image feature fusion module, the distribution network line image to be processed, the target distribution network line image and the image structure tensor are input into a preset image feature extraction neural network for feature fusion to obtain a conductor feature point position and a conductor feature vector; finally, in a conductor model construction module, a three-dimensional model is constructed according to the conductor feature point position, the conductor feature vector and the internal and external parameters of the camera to obtain a three-dimensional model of the distribution network line conductor of the distribution network line image to be processed. The distribution network line image to be processed is preprocessed by image enhancement to enhance feature representation. Feature fusion is performed through the distribution network line image to be processed, the target distribution network line image and the image structure tensor to effectively capture the features of small conductors. The image structure tensor enhances the continuity of small conductors in the image, obtains the position of conductor feature points and conductor feature vectors, avoids the situation where the conductor is ignored, and thus improves the accuracy of three-dimensional construction of distribution network line conductors.

[0114] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0115] Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0116] Another embodiment of the present invention further provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements a three-dimensional construction method of a distribution network line conductor as described in the above embodiment when executing the computer program. The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory.

[0117] The processor may be a central processing unit (CPU), or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and various interfaces and lines are used to connect various parts of the entire terminal device.

[0118] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Med i aCard, SMC), a secure digital (Secure Digital, SD) card, a flash card (F l ash Card), at least one disk storage device, a flash memory device or other volatile solid-state storage device.

[0119] Another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a three-dimensional construction method for distribution network line conductors described in the above embodiment.

[0120] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium.

[0121] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A three-dimensional construction method for distribution network line conductors, characterized in that: include: Obtain the image of the distribution network line to be processed and the internal and external parameters of the camera; Perform image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image; Calculate the structure tensor according to the distribution network line image to be processed, and obtain the image structure tensor of the distribution network line image to be processed; Input the to-be-processed distribution network line image, the target distribution network line image and the image structure tensor into a preset image feature extraction neural network for feature fusion to obtain the conductor feature point position and the conductor feature vector; A three-dimensional model is constructed according to the positions of the characteristic points of the conductors, the characteristic vectors of the conductors and the internal and external parameters of the camera to obtain a three-dimensional model of the distribution network line conductors of the distribution network line image to be processed.

2. A three-dimensional construction method for distribution network line conductors according to claim 1, characterized in that: The image enhancement method includes: multi-scale edge detection and local contrast enhancement; Performing image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image includes: Performing several Gaussian blur and downsampling operations on the distribution network line image to be processed to generate several image layers with different resolutions; At each resolution, edge detection is performed on the corresponding image layer to extract image edge information; Fusing edge information of several images with different resolutions to obtain a first feature image; Dividing the first feature image into a plurality of feature small blocks, and equalizing the brightness histogram of each small block to obtain a plurality of target feature small blocks; Each target feature block is smoothly transitioned to obtain the target distribution network line image.

3. A three-dimensional construction method for distribution network line conductors according to claim 1, characterized in that: Calculating a structure tensor according to the distribution network line image to be processed to obtain an image structure tensor of the distribution network line image to be processed includes: According to the distribution network line image to be processed, determining the coordinates of each pixel point in the distribution network line image to be processed; According to the coordinates of each pixel point, the horizontal gradient and the vertical gradient of the distribution network line image to be processed are calculated respectively; According to the coordinates of each pixel point, a Gaussian weighting function is calculated; An image structure tensor of the distribution network line image to be processed is calculated based on the horizontal gradient, the vertical gradient and the Gaussian weighting function.

4. A three-dimensional construction method for distribution network line conductors according to claim 1, characterized in that: The distribution network line image to be processed, the target distribution network line image and the image structure tensor are input into a preset image feature extraction neural network for feature fusion to obtain the conductor feature point position and the conductor feature vector, including: Inputting the distribution network line image to be processed into a preset image feature extraction neural network, performing convolution calculation according to a preset first convolution layer number, and obtaining a first convolution feature; Inputting the target distribution network line image into a preset image feature extraction neural network, performing convolution calculation according to a preset second convolution layer number, and obtaining a second convolution feature; Inputting the image structure tensor into a preset image feature extraction neural network, performing convolution calculation according to a preset third convolution layer number, and obtaining a third convolution feature; Performing weighted feature fusion according to the first convolution feature, the second convolution feature, the third convolution feature and a preset weight to obtain a wire feature point position and a wire feature vector; Among them, the number of the first convolutional layers, the number of the second convolutional layers, and the number of the third convolutional layers are different.

5. A three-dimensional construction method for distribution network line conductors according to claim 4, characterized in that: The image feature extraction neural network also includes: a first multi-layer perceptron for representing low-level feature learning and a second multi-layer perceptron for representing high-level feature learning; Performing weighted feature fusion according to the first convolution feature, the second convolution feature, the third convolution feature and a preset weight to obtain a wire feature point position and a wire feature vector, including: Performing a dimension alignment operation on the first convolution feature, the second convolution feature, and the third convolution feature to obtain the first convolution feature, the second convolution feature, and the third convolution feature after dimension alignment; According to the preset weights and the first convolution feature after the alignment dimension, the second convolution feature and the third convolution feature are fused by feature splicing to obtain the initial fused feature; Inputting the initial fusion features into the first multi-layer perceptron to capture the edge features, corner features and texture features of the wire, and obtaining target fusion features according to the edge features, corner features and texture features of the wire; The target fusion feature is input into the second multi-layer perceptron to capture the spatial position feature and spatial structure feature of the wire, and the final wire feature point position and wire feature vector are obtained according to the spatial position feature and spatial structure feature of the wire.

6. A three-dimensional construction method for distribution network line conductors according to claim 1, characterized in that: A three-dimensional model is constructed according to the conductor feature point position, the conductor feature vector and the internal and external parameters of the camera to obtain a three-dimensional model of the distribution network line conductor of the distribution network line image to be processed, including: Calculating the similarity of each wire feature point according to the wire feature vector; A basic matrix is ​​obtained by solving the similarity of each wire feature point and the position of the wire feature point; Perform matrix decomposition according to the basic matrix and the internal and external parameters of the camera to obtain the camera pose; Point cloud fusion is performed according to the camera posture and the position of the conductor feature points to generate a three-dimensional model of the distribution network line conductor of the distribution network line image to be processed.

7. A three-dimensional construction device for distribution network line conductors, characterized in that: include: A conductor image acquisition module is used to acquire the distribution network line image to be processed; An image preprocessing module, used to perform image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image; A structure tensor calculation module, used to calculate the structure tensor according to the distribution network line image to be processed, so as to obtain the image structure tensor of the distribution network line image to be processed; An image feature fusion module is used to input the distribution network line image to be processed, the target distribution network line image and the image structure tensor into a preset image feature extraction neural network for feature fusion to obtain the conductor feature point position and the conductor feature vector; The wire model construction module is used to construct a three-dimensional model according to the wire feature point position, the wire feature vector and the internal and external parameters of the camera to obtain a three-dimensional model of the distribution network line wire of the distribution network line image to be processed.

8. A three-dimensional construction device for distribution network line conductors as claimed in claim 7, characterized in that: The image enhancement method includes: multi-scale edge detection and local contrast enhancement; The image preprocessing module is used to perform image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image; Performing image preprocessing according to the distribution network line image to be processed and a preset image enhancement method to obtain a target distribution network line image includes: Performing several Gaussian blur and downsampling operations on the distribution network line image to be processed to generate several image layers with different resolutions; At each resolution, edge detection is performed on the corresponding image layer to extract image edge information; Fusing edge information of several images with different resolutions to obtain a first feature image; Dividing the first feature image into a plurality of feature small blocks, and equalizing the brightness histogram of each small block to obtain a plurality of target feature small blocks; Each target feature block is smoothly transitioned to obtain the target distribution network line image.

9. A terminal device, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a three-dimensional construction method for a distribution network line conductor is implemented as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute a three-dimensional construction method for distribution network line conductors as described in any one of claims 1 to 6.

Images