Power transmission bare conductor identification method, system and device, medium and program product

Abstract

The invention discloses a power transmission bare conductor identification method, system and device, a medium and a program product, and the method comprises the steps: obtaining a to-be-identified power transmission bare conductor image, and carrying out the preprocessing; inputting the preprocessed to-be-identified power transmission bare conductor image into a pre-trained power transmission bare conductor type identification model to obtain a power transmission bare conductor type in the to-be-identified power transmission bare conductor image; wherein the power transmission bare conductor type identification model is obtained by training a lightweight convolutional neural network embedded with a plurality of dynamic feature refiners. According to the method, a lightweight feature extraction module is utilized, model parameters and model complexity are greatly reduced, fine-grained feature extraction is performed on the difficulties of high inter-class similarity and high intra-class discrimination of the transmission bare conductor in combination with a dynamic feature refinement device, and the recognition capability of an artificial intelligence algorithm in a transmission bare conductor scene is improved. The model identification of the bare conductor can be completed under the condition of uninterruptible power supply, and the identification efficiency and precision are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid inspection, and in particular to a method, system, equipment, medium and program product for identifying bare power transmission wires. Background Art

[0002] Transmission lines are an important part of the electric power industry. Accurately identifying and counting the types of bare conductors in existing transmission systems is extremely critical for transmission line inspections and is vital to power security for social production and people's daily lives.

[0003] Existing bare wire identification relies on manual labor. Only when the line is out of power, the inspection personnel can climb the pole to measure the outer diameter of the bare wire with a vernier caliper and count the number of aluminum wires on the outer layer of the bare wire to determine the bare wire model. Manual pole climbing and naked eye identification is not only inefficient and has large errors, but also impossible to apply for line power outages in order to verify the bare wire model during actual project planning and design. The development of a technology that can be used for daily bare wire model identification is of great significance for circuit inspection work and power project planning.

[0004] Traditional computer vision recognition methods require artificial setting of feature extraction methods, and have the disadvantages of insufficient versatility and complex models. Emerging artificial intelligence target recognition methods are highly versatile and have high recognition accuracy. Target recognition technology based on artificial intelligence has become the mainstream direction in the field of target recognition. However, the similarity between different models and categories of overhead bare conductors is extremely large, while the imaging of overhead bare conductors of the same model category under different lighting conditions and camera angles is not very similar. The existing mainstream artificial intelligence algorithms are designed for daily scenarios and are not competent for the recognition of bare power transmission conductors with large differences in features within categories and high similarity between features between categories. In addition, the common mainstream artificial intelligence algorithms currently have huge parameters and high inference costs. They are only suitable for deployment on platforms with high computing power computing devices and cannot be deployed on general platforms. Summary of the invention

[0005] The present invention provides a method, system, device, medium and program product for identifying bare power transmission conductors, so as to solve the problem that the efficiency and accuracy of the existing bare power transmission conductors manually identified by climbing poles are difficult to meet the limitations of modern power transmission line inspections.

[0006] In a first aspect, a method for identifying a bare power transmission conductor is provided, comprising the following steps:

[0007] S1: Obtain the image of the bare power transmission wire to be identified and perform preprocessing;

[0008] S2: inputting the preprocessed bare power transmission wire image to be identified into a pre-trained bare power transmission wire type identification model to obtain the type of bare power transmission wire in the bare power transmission wire image to be identified;

[0009] The transmission bare wire type recognition model is obtained by training a lightweight convolutional neural network embedded with multiple dynamic feature refiners.

[0010] Furthermore, in step S1, the preprocessing process includes:

[0011] Resize the image to a preset size;

[0012] Normalization is performed along the R, G, and B channels of the resized image respectively.

[0013] Furthermore, the processing of the dynamic feature refiner includes:

[0014] The input image of the dynamic feature refiner is copied three times with the dimension of C×H×W, where C, H, and W are the number of channels, height, and width of the image respectively;

[0015] The three copied images are reshaped and their dimensions are adjusted to C×HW, HW×C, and C×HW respectively.

[0016] Perform matrix multiplication and Softmax operation on the image with dimension HW×C and one of the images with dimension C×HW at the same time, and then perform matrix multiplication operation with another image with dimension C×HW. After Reshape operation, adjust the dimension to C×H×W to obtain the changed feature image.

[0017] The changed feature image is added to the input image of the dynamic feature refiner and output as the feature image finally extracted by the dynamic feature refiner.

[0018] Furthermore, the lightweight convolutional neural network embedded with multiple dynamic feature refiners includes multiple stacked feature extraction modules and a Softmax module, wherein each feature extraction module includes a dynamic feature refiner and at least one lightweight feature extraction module connected in sequence.

[0019] Furthermore, the processing process of the lightweight feature extraction module includes:

[0020] The input of the lightweight feature extraction module is passed through two different Conv3×3 modules, partial convolution is performed on different channels, and multiple modes of feature extraction are performed on the input. Then, each module is connected to a 1×1 convolution to fuse information from different channels under the same mode, and then the features extracted from different modes are connected through the splicing operation to obtain the spliced ​​features.

[0021] The concatenated features are further transformed through batch normalization operations and activation functions, and then added to the input of the lightweight feature extraction module to obtain the final output features.

[0022] Furthermore, the transmission bare wire type identification model is obtained by the following method:

[0023] Obtain and preprocess the image data of various types of bare power transmission wires collected by drones, annotate the types of bare power transmission wires in the preprocessed bare power transmission wire images, and construct a sample data set;

[0024] Embed multiple dynamic feature refiners into a lightweight convolutional neural network to obtain a lightweight neural network model;

[0025] The preprocessed bare power transmission wire image is used as the input of the lightweight neural network model, and the labeled bare power transmission wire type is used as the output. The lightweight neural network model is trained using the sample data set to obtain the bare power transmission wire type recognition model.

[0026] In a second aspect, a power transmission bare wire identification system is provided, comprising:

[0027] An image acquisition module is used to acquire the image of the bare power transmission wire to be identified and perform preprocessing;

[0028] The transmission bare wire identification module is used to input the preprocessed transmission bare wire image to be identified into a pre-trained transmission bare wire type identification model to obtain the type of transmission bare wire in the transmission bare wire image to be identified; wherein the transmission bare wire type identification model is obtained by training a lightweight convolutional neural network embedded with multiple dynamic feature refiners.

[0029] In a third aspect, an electronic device is provided, including:

[0030] Memory on which computer programs or instructions are stored;

[0031] The processor is used to load and execute the computer program or instructions to implement the power transmission bare wire identification method as described above.

[0032] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program or instruction is stored, and when the computer program or instruction is executed by a processor, the method for identifying bare power transmission wires as described above is implemented.

[0033] In a fifth aspect, a computer program product is provided, wherein a computer program or instructions are stored in the computer program product, and when the computer program or instructions are executed by a processor, the method for identifying bare power transmission wires as described above is implemented.

[0034] The present invention proposes a method, system, device, medium and program product for identifying bare power transmission wires, which have the following beneficial effects:

[0035] 1. In view of the difficulty that mainstream artificial intelligence algorithms have in dealing with the large intra-class differences and high inter-class similarities that are unique to the bare power transmission conductor scenario, the present invention designs a dynamic feature refiner that dynamically generates kernels to classify the features with strong discriminability in the hierarchical structure, thereby enhancing the intra-class compactness and inter-class separability of the features, thereby achieving high-precision identification of the types of bare power transmission conductors;

[0036] 2. Aiming at the shortcomings of existing convolutional neural networks, such as large number of parameters and high computing power requirements, the present invention designs a novel lightweight feature extraction module as a feature extraction module of the convolutional neural network, which effectively reduces the model parameters and greatly speeds up the reasoning speed;

[0037] 3. The present invention can complete the bare wire model identification without power outage in the transmission line, which greatly saves manpower and improves the efficiency and accuracy of bare wire identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 This is a framework diagram of a transmission bare wire type identification model provided by an embodiment of the present invention;

[0040] Figure 2 is a schematic diagram of the structure of a dynamic feature refiner provided by an embodiment of the present invention;

[0041] Figure 3 is a schematic diagram of the structure of a lightweight feature extraction module provided by an embodiment of the present invention;

[0042] Figure 4 It is a flow chart of a method for obtaining a transmission bare conductor type identification model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0044] In order to solve the problem that the efficiency and accuracy of the existing bare power transmission conductor identification by manual climbing on poles are difficult to meet the limitations of modern power transmission line inspection, the present invention provides a bare power transmission conductor identification method, system, equipment, medium and program product. The technical solution of the present invention is described in detail below in conjunction with specific embodiments.

[0045] The embodiment of the present invention provides a method for identifying a bare power transmission conductor, comprising the following steps:

[0046] S1: Obtain the image of the bare power transmission wire to be identified and perform preprocessing.

[0047] During implementation, the preprocessing process includes resizing and normalization. Specifically, in this embodiment, the preprocessing process includes:

[0048] Resize the image to a preset size (e.g., resize both the height and width to 64 pixels);

[0049] The mean and variance of the pixel values ​​along the R, G, and B channels of the resized image are counted for normalization.

[0050] S2: Inputting the preprocessed bare power transmission wire image to be identified into a pre-trained bare power transmission wire type recognition model to obtain the type of bare power transmission wire in the bare power transmission wire image to be identified; wherein the bare power transmission wire type recognition model is obtained by training a lightweight convolutional neural network embedded with multiple dynamic feature refiners.

[0051] In specific implementation, before the transmission bare wire type identification is performed, a transmission bare wire type identification model needs to be pre-trained.

[0052] like Figure 4 As shown, in this embodiment, the transmission bare wire type identification model is obtained by the following method:

[0053] Step 1: Obtain the image data of various types of bare power transmission wires collected by drones and preprocess them, label the types of bare power transmission wires in the preprocessed bare power transmission wire images, and construct a sample data set.

[0054] The preprocessing process includes resizing and normalization. In the process of constructing the sample data set, the sample data can be enriched by data enhancement techniques (such as rotation, shearing, grayscale processing, etc.) to increase the sample data set.

[0055] Step 2: Embed multiple dynamic feature refiners into a lightweight convolutional neural network to obtain a lightweight neural network model.

[0056] Specifically, the lightweight neural network model includes a plurality of stacked feature extraction modules and a Softmax module, wherein each feature extraction module includes a dynamic feature refiner DHFR and at least one lightweight feature extraction module PC Block connected in sequence. Figure 1 As shown, in this embodiment, a feature extraction network including four feature extraction modules is provided, and the numbers of lightweight feature extraction modules included in the four feature extraction modules are 1, 2, 8, and 2 respectively.

[0057] like Figure 2 As shown, the processing process of the dynamic feature refiner DHFR includes:

[0058] The input image of the dynamic feature refiner is copied three times with the dimension of C×H×W, where C, H, and W are the number of channels, height, and width of the image respectively;

[0059] The three copied images are reshaped and their dimensions are adjusted to C×HW, HW×C, and C×HW respectively.

[0060] The image with the dimension of HW×C and one of the images with the dimension of C×HW are subjected to matrix multiplication (MauMul) operation and Softmax operation at the same time, and then matrix multiplication operation is performed with another image with the dimension of C×HW. After the Reshape operation, the dimension is readjusted to C×H×W to obtain the changed feature image;

[0061] The changed feature image is added to the input image of the dynamic feature refiner and output as the feature image finally extracted by the dynamic feature refiner.

[0062] like Figure 3 As shown, the processing process of the lightweight feature extraction module includes:

[0063] The input of the lightweight feature extraction module is passed through two different Conv3×3 modules, partial convolution is performed on different channels, and multiple modes of feature extraction are performed on the input. Then, each module is connected to a 1×1 convolution to fuse information from different channels under the same mode, and then the features extracted from different modes are connected through the splicing operation to obtain the spliced ​​features.

[0064] The concatenated features are further transformed through batch normalization (BN) operations and activation functions (ReLu) in turn, and then added to the input of the lightweight feature extraction module to obtain the final output features.

[0065] Step 3: Use the preprocessed bare power transmission wire image as the input of the lightweight neural network model, and the labeled bare power transmission wire type as the output. Use the sample data set to train the lightweight neural network model to obtain the bare power transmission wire type recognition model.

[0066] The method for identifying bare power transmission wires provided in the above embodiment has the following advantages:

[0067] (1) As mainstream artificial intelligence algorithms have difficulty in dealing with the unique difficulty of large intra-class differences and high inter-class similarities in the bare power transmission conductor scenario, the present invention designs a dynamic feature refiner that dynamically generates kernels to classify features with strong discriminability in the hierarchical structure, thereby enhancing the intra-class compactness and inter-class separability of features, thereby achieving high-precision identification of bare power transmission conductor types;

[0068] (2) In view of the shortcomings of existing convolutional neural networks, such as large number of parameters and high computing power requirements, the present invention designs a novel lightweight feature extraction module as a feature extraction module of the convolutional neural network, which effectively reduces the model parameters and greatly speeds up the reasoning speed;

[0069] (3) The present invention can be deployed on a general platform and can complete bare wire model identification without power outages on the transmission line, thereby greatly saving manpower and improving the efficiency and accuracy of bare wire identification.

[0070] The embodiment of the present invention further provides a transmission bare wire identification system, comprising:

[0071] An image acquisition module is used to acquire the image of the bare power transmission wire to be identified and perform preprocessing;

[0072] The transmission bare wire identification module is used to input the preprocessed transmission bare wire image to be identified into a pre-trained transmission bare wire type identification model to obtain the type of transmission bare wire in the transmission bare wire image to be identified; wherein the transmission bare wire type identification model is obtained by training a lightweight convolutional neural network embedded with multiple dynamic feature refiners.

[0073] It should be understood that the functional unit modules in various embodiments of the present invention may be concentrated in one processing unit, or each unit module may exist physically separately, or two or more unit modules may be integrated in one unit module, and may be implemented in the form of hardware or software.

[0074] An embodiment of the present invention further provides an electronic device, including:

[0075] Memory on which computer programs or instructions are stored;

[0076] The processor is used to load and execute the computer program or instructions to implement the method for identifying bare power transmission wires described in the above embodiment.

[0077] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program or instruction is stored. When the computer program or instruction is executed by a processor, the method for identifying bare power transmission wires described in the above embodiment is implemented.

[0078] An embodiment of the present invention further provides a computer program product, in which a computer program or instructions are stored. When the computer program or instructions are executed by a processor, the method for identifying bare power transmission wires described in the above embodiment is implemented.

[0079] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0080] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0081] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0082] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0083] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0084] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for identifying bare power transmission wires, characterized in that: The steps include: S1: Obtain the image of the bare power transmission wire to be identified and perform preprocessing; S2: inputting the preprocessed bare power transmission wire image to be identified into a pre-trained bare power transmission wire type identification model to obtain the type of bare power transmission wire in the bare power transmission wire image to be identified; The transmission bare wire type recognition model is obtained by training a lightweight convolutional neural network embedded with multiple dynamic feature refiners.

2. The method for identifying bare power transmission wires according to claim 1, characterized in that: In step S1, the preprocessing process includes: Resize the image to a preset size; Normalization is performed along the R, G, and B channels of the resized image respectively.

3. The method for identifying bare power transmission wires according to claim 1, characterized in that: The processing of the dynamic feature refiner includes: The input image of the dynamic feature refiner is copied three times with the dimension of C×H×W, where C, H, and W are the number of channels, height, and width of the image respectively; The three copied images are reshaped and their dimensions are adjusted to C×HW, HW×C, and C×HW respectively. Perform matrix multiplication and Softmax operation on the image with dimension HW×C and one of the images with dimension C×HW at the same time, and then perform matrix multiplication operation with another image with dimension C×HW. After Reshape operation, adjust the dimension to C×H×W to obtain the changed feature image. The changed feature image is added to the input image of the dynamic feature refiner and output as the feature image finally extracted by the dynamic feature refiner.

4. The method for identifying bare power transmission wires according to claim 1, characterized in that: The lightweight convolutional neural network embedded with multiple dynamic feature refiners includes multiple stacked feature extraction modules and a Softmax module, wherein each feature extraction module includes a dynamic feature refiner and at least one lightweight feature extraction module connected in sequence.

5. The method for identifying bare power transmission wires according to claim 4, characterized in that: The processing process of the lightweight feature extraction module includes: The input of the lightweight feature extraction module is passed through two different Conv3×3 modules, partial convolution is performed on different channels, and multiple modes of feature extraction are performed on the input. Then, each module is connected to a 1×1 convolution to fuse information from different channels under the same mode, and then the features extracted from different modes are connected through the splicing operation to obtain the spliced ​​features. The concatenated features are further transformed through batch normalization operations and activation functions, and then added to the input of the lightweight feature extraction module to obtain the final output features.

6. The method for identifying bare power transmission conductors according to claim 1, characterized in that: The transmission bare wire type identification model is obtained by the following method: Obtain and preprocess the image data of various types of bare power transmission wires collected by drones, annotate the types of bare power transmission wires in the preprocessed bare power transmission wire images, and construct a sample data set; Embed multiple dynamic feature refiners into a lightweight convolutional neural network to obtain a lightweight neural network model; The preprocessed bare power transmission wire image is used as the input of the lightweight neural network model, and the labeled bare power transmission wire type is used as the output. The lightweight neural network model is trained using the sample data set to obtain the bare power transmission wire type recognition model.

7. A power transmission bare wire identification system, characterized in that: include: An image acquisition module is used to acquire the image of the bare power transmission wire to be identified and perform preprocessing; The transmission bare wire identification module is used to input the preprocessed transmission bare wire image to be identified into a pre-trained transmission bare wire type identification model to obtain the type of transmission bare wire in the transmission bare wire image to be identified; wherein the transmission bare wire type identification model is obtained by training a lightweight convolutional neural network embedded with multiple dynamic feature refiners.

8. An electronic device, characterized in that: include: Memory on which computer programs or instructions are stored; A processor is used to load and execute the computer program or instructions to implement the method for identifying bare power transmission wires as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by a processor, the method for identifying bare power transmission wires as described in any one of claims 1 to 6 is implemented.

10. A computer program product, characterized in that The computer program product stores a computer program or instructions, and when the computer program or instructions are executed by a processor, the method for identifying bare power transmission wires as described in any one of claims 1 to 6 is implemented.

Images