Power transmission line small fitting defect detection method and device based on zero-order learning
Through the zero-learning method, the defect detection model is established and trained, and the problem of low accuracy and reliability of defect detection of small metal tools in transmission lines is solved, and efficient identification of small metal tools is achieved.
Patent Information
- Application Number
- CN202411687831.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-25
- Publication Date
- 2025-03-21
AI Technical Summary
In the prior art, the accuracy and reliability of defect detection of small metal tools on transmission lines are low, and mainly rely on human detection, resulting in insufficient detection efficiency and accuracy.
Using a zero-learning method, a label library connecting the defects of the metal and small metal tools is established by labeling the image of the transmission line, and an initial defect detection model is used for iterative training to generate a defect detection model to identify the defects of small metal tools in the transmission line.
It improves the accuracy and reliability of small metal defect detection, and realizes automatic, fast and accurate identification of small metal defects in transmission lines.
Smart Images
Figure CN119671952A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the field of electric power technology, and in particular, relates to a method and device for defect detection of small hardware of power transmission lines based on zero-shot learning. Background Art
[0002] With the rapid development of smart grids, the construction of power grids is constantly expanding. Transmission lines are an important part of smart grids. They are power lines used to transmit electric energy. They usually include infrastructure, ground wires, insulators, ancillary facilities, large-size hardware, small-size hardware, poles and towers, and channel environments. Among them, small-size hardware is also called small hardware. Small hardware includes but is not limited to bolts, nuts, and pins. Transmission lines will be affected by the external environment during long-term use, and various components of transmission lines may be damaged to varying degrees, which poses a huge challenge to the stable operation of the power system. Therefore, it is particularly important to detect transmission lines.
[0003] In the scenario of defect detection of transmission lines, the detection methods are generally aimed at the target shape features of the transmission lines, such as digital imaging methods, neural network technology, visual feature engineering, and fuzzy logic algorithms. These detection methods are mainly used to detect defects in infrastructure, ground wires, insulators, auxiliary facilities, large-size hardware, poles and towers, and channel environments in transmission lines. However, defect detection of small hardware is rarely involved. At present, defect detection of small hardware is often carried out manually, and its detection accuracy and reliability are relatively low. Summary of the invention
[0004] The purpose of the embodiments of the present application is to provide a method and device for defect detection of small metal fittings for transmission lines based on zero-shot learning, which can solve the problem of low detection accuracy and reliability of defect detection of small metal fittings for transmission lines.
[0005] In a first aspect, a method for defect detection of small fittings of a transmission line based on zero-shot learning is provided, which is executed by a terminal, and comprises: marking connecting fittings and marking small fitting defects in multiple first images of the transmission line to obtain a connecting fitting label library and a small fitting defect label library, wherein the connecting fitting label library comprises local images of connecting fittings containing small fitting defects, and the small fitting defect label library comprises first images containing small fitting defects; the local images in the connecting fitting label library, the first images in the small fitting defect label library, and the second images without small fitting defects are input into an initial defect detection model for iterative training to obtain a defect detection model; a third image of the target transmission line is obtained; the third image is input into the defect detection model for defect recognition detection to obtain a defect recognition result of the small fitting output by the defect detection model.
[0006] In a second aspect, a defect detection device for small fittings of a transmission line based on zero-shot learning is provided, comprising: a labeling module, used to label connecting fittings and mark small fitting defects in multiple first images of the transmission line, and obtain a connecting fitting label library and a small fitting defect label library, wherein the connecting fitting label library includes local images of connecting fittings containing small fitting defects, and the small fitting defect label library includes first images containing small fitting defects; a training module, used to input the local images in the connecting fitting label library, the first images in the small fitting defect label library, and the second image without small fitting defects into an initial defect detection model for iterative training to obtain a defect detection model; an acquisition module, used to acquire a third image of the target transmission line; and a detection module, used to input the third image into the defect detection model for defect recognition detection to obtain a defect recognition result of the small fitting output by the defect detection model.
[0007] In the third aspect, an embodiment of the present application provides a defect detection device for small hardware of a transmission line based on zero-shot learning, comprising: a memory, a processor, and computer executable instructions stored in the memory and executable on the processor, wherein when the computer executable instructions are executed by the processor, the steps of executing the method of the first aspect are implemented.
[0008] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which is used to store computer-executable instructions. When the computer-executable instructions are executed by a processor, the steps of the method of the first aspect are implemented.
[0009] In an embodiment of the present application, first, connecting fittings and small fitting defects are annotated in multiple first images of the transmission line to obtain a connecting fitting label library and a small fitting defect label library, wherein the connecting fitting label library includes local images of connecting fittings containing small fitting defects, and the small fitting defect label library includes first images containing small fitting defects; then, the local images in the connecting fitting label library, the first images in the small fitting defect label library, and the second image without small fitting defects are input into an initial defect detection model for iterative training to obtain a defect detection model; finally, a third image of the target transmission line is obtained; the third image is input into the defect detection model for defect recognition detection to obtain defect recognition results of small fittings output by the defect detection model.
[0010] In this way, the embodiment of the present application can use images with labeled defects of small metal fittings to train the initial defect detection model, and then use the trained defect detection model to detect defects of small metal fittings in the transmission line, thereby improving the detection accuracy and reliability of defect detection of small metal fittings. BRIEF DESCRIPTION OF THE DRAWINGS
[0011] In order to more clearly illustrate the embodiments of the present application 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 recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0012] Figure 1 It is a flowchart of a method for defect detection of small metal fittings for power transmission lines based on zero-shot learning provided by an exemplary embodiment of the present application.
[0013] Figure 2 It is one of the structural schematic diagrams of a defect detection device for small metal fittings of a power transmission line based on zero-shot learning provided by an exemplary embodiment of the present application.
[0014] Figure 3 It is one of the structural schematic diagrams of a defect detection device for small metal fittings of a power transmission line based on zero-shot learning provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION
[0015] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.
[0016] The technical solution provided by the embodiments of the present application is described in detail below through some embodiments and their application scenarios in combination with the accompanying drawings.
[0017] Figure 1 The flowchart of a method for defect detection of small metal fittings of power transmission lines based on zero-shot learning provided by an embodiment of the present application is shown. The method can be executed by an electronic device, such as a terminal device or a server device. In other words, the method can be executed by software or hardware installed in the terminal device or the server device. Figure 1 As shown, the method may include the following steps.
[0018] Step S101 : annotating connection fittings and marking small fitting defects in a plurality of first images of a transmission line to obtain a connection fitting label library and a small fitting defect label library.
[0019] The connection fitting label library includes partial images of connection fittings containing small fitting defects, and the small fitting defect label library includes a first image containing small fitting defects.
[0020] Specifically, the types of defects in transmission lines include but are not limited to eight categories, including small hardware, insulators, ancillary facilities, large hardware, ground wires, poles, channels and infrastructure. Among them, the number of small hardware defects accounts for 62%. The types of small hardware defects include but are not limited to improper pin installation, loose bolts, pin dislocation, pin missing, nut dislocation, nut missing, bolt missing, bolt rust, etc.
[0021] Furthermore, the embodiment of the present application collected more than 15,000 original images of drone inspections of transmission lines of different voltage levels of 35~500 kV. The image resolution is 2560×1440 pixels (16:9). In the embodiment of the present application, when annotating the first image, the connection hardware is first marked, and the connection hardware includes but is not limited to the line connection tower end, the insulator connection tower end, and the insulator connection line end. When marking the connection hardware, the embodiment of the present application adopts an iterative training model to assist in annotation. The specific process is as follows: First, a label library of about 5,000 connection hardware is constructed to train the initial model, and the original image of the transmission line is detected using the trained initial model, and the detection results of the initial model are added to the connection hardware label library. Then, the trained initial model is used to continue to detect the original images of other transmission lines. When the cumulative number of detected original images reaches 15,000, 30,000, 60,000, and 100,000, respectively, iterative training is performed to form models of 2, 3, 4, and 5 stages. This method eventually annotated 10,000 original images of power transmission lines. Then mark the small fitting defects. Since the targets of small fitting defects are relatively small, it is necessary to constantly zoom and drag the image when searching for small fitting defects on the original image. This is not only time-consuming and labor-intensive, but also easy to miss. In order to improve the efficiency of searching for small fitting defects, the embodiment of the present application is based on the method of connecting fitting thumbnails to quickly find and locate small fitting defects. First, mark the small fitting defects on the thumbnail of the marked connecting fittings, and then reversely convert it to the original image to complete the marking of small fitting defects.
[0022] After the final processed data accumulates to 10,000 images, all images and label data marked by the above method are summarized. A connection fitting label library and a small fitting defect label library are formed, and finally 63,000 images of connection fittings and 25,000 images of small fitting defects are annotated. Among them, the three types of small fitting defects, pin dislocation, pin missing and nut dislocation, account for the largest proportion, accounting for 80% of the total samples. The summarized connection fitting label library and small fitting defect label library are split into training set and test set. The ratio of training set to test set of connection fittings and small fittings can be 5:1.
[0023] Step S102 , inputting the local image in the connection fitting label library, the first image in the small fitting defect label library, and the second image without small fitting defects into the initial defect detection model for iterative training to obtain a defect detection model.
[0024] Specifically, the initial defect detection model in the embodiment of the present application can be a deep learning model, which includes three parts: a feature extraction network backbone, a network structure reorganization neck, and a prediction head. The initial defect detection model includes a generative adversarial network and a deep learning network. The generative adversarial network is used to generate defective samples using positive samples without small metal defects, thereby expanding the sample set of the deep learning network. The deep learning network is used to identify small metal defects in the sample based on the sample set of the expanded deep learning network.
[0025] Furthermore, when iteratively training the initial defect detection model, as an optional embodiment of the present application, the network parameters of the adversarial generative network and the deep learning network in the initial defect detection model are first initialized; then the first image in the small hardware defect label library and the second image without small hardware defects are input into the adversarial generative network for adversarial training, and the adversarial generative network is used to generate a defect image containing small hardware defects; then the defect image, the local image and the first image in the small hardware defect label library are input into the deep learning network for iterative training to obtain a defect detection model.
[0026] Specifically, the adversarial generative network includes a generator and a discriminator, and the training of the adversarial generative network includes generator training and discriminator training. When the adversarial generative network is subjected to adversarial training, the first image in the small fitting defect label library and the second image without small fitting defects are input into the generator of the adversarial generative network for training until the difference between the defect image generated by the generator and the first image reaches a preset accuracy. The generator is used to generate a defect image of the second image based on the first image; the defect image and the first image are input into the discriminator of the adversarial generative network for training until the error rate of the discriminator is minimized. The discriminator is used to distinguish between the first image and the defect image.
[0027] The purpose of generator training is to generate realistic defect samples. The generator tries to generate small metal defect samples that are as close to real defect samples as possible. The purpose of discriminator training is to distinguish between real defect samples and defect samples generated by the generator, and to update the parameters of the discriminator by minimizing the error rate of the discriminator. The generator and the discriminator are trained alternately until the generator generates highly realistic defect samples, while the discriminator has difficulty distinguishing between real and fake samples.
[0028] After the adversarial generative network generates the defect image, the defect image is expanded into the original data set. The expanded data set is used to train the deep learning network. When training the deep learning network, as an optional embodiment of the present invention, after loading the pre-trained weights in the deep learning network, the defect image, the local image, and the first image in the small metal fitting defect label library are input into the deep learning network for iterative training until the loss function of the deep learning network model converges to obtain a defect detection model.
[0029] Specifically, when training a deep learning network, a network structure is added to the deep learning network and pre-trained weights are loaded. During the training process, the Generalized Focal Loss loss function is used as the loss function of the deep learning network to solve the problem of imbalance between positive and negative samples.
[0030] Furthermore, the trained defect detection model can be applied to the test set for testing, which includes the original training set, the original image containing small fitting defects, and the partial image of the connection fitting containing small fitting defects. According to the test results of the test set, the defect detection model is optimized and adjusted, including adjusting the network structure, optimizing the loss function, etc., to improve the recognition accuracy and robustness of the defect detection model. The optimized defect detection model is applied to the actual inspection of the transmission line to realize the automatic, fast and accurate recognition of small fitting defects.
[0031] Step S103: acquiring a third image of the target power transmission line.
[0032] Specifically, the third image of the target transmission line includes an image of small hardware in the transmission line, and the small hardware corresponding to the image of the small hardware may have small hardware defects.
[0033] Step S104, inputting the third image into the defect detection model for defect recognition detection, and obtaining a defect recognition result of the small metal fitting output by the defect detection model.
[0034] Specifically, the defect detection model already has the ability to identify defects in small metal fittings, and the third image is input into the defect detection model for small metal fitting defect identification. The defect detection model outputs the small metal fitting defect identification result as the probability that the third image contains small metal fitting defects. If the probability is greater than the threshold, it means that there are defects in the small metal fittings in the target transmission line.
[0035] The embodiment of the present application can use images with labeled defects of small metal fittings to train an initial defect detection model, thereby using the trained defect detection model to detect defects in small metal fittings in transmission lines, thereby improving the detection accuracy and reliability of defect detection of small metal fittings.
[0036] Figure 2It is one of the structural schematic diagrams of a defect detection device for small hardware fittings of a power transmission line based on zero-shot learning provided by an exemplary embodiment of the present application. The device 200 includes: a labeling module 201, which is used to label the connection hardware and mark the defects of the small hardware in multiple first images of the power transmission line, and obtain a connection hardware label library and a small hardware defect label library, wherein the connection hardware label library includes a local image of the connection hardware containing the defects of the small hardware, and the small hardware defect label library includes a first image containing the defects of the small hardware; a training module 202, which is used to input the local image in the connection hardware label library, the first image in the small hardware defect label library, and the second image without the defects of the small hardware into the initial defect detection model for iterative training to obtain a defect detection model; an acquisition module 203, which is used to obtain a third image of the target power transmission line; a detection module 204, which is used to input the third image into the defect detection model for defect recognition detection, and obtain the defect recognition result of the small hardware output by the defect detection model.
[0037] The embodiment of the present application can use images with labeled defects of small metal fittings to train an initial defect detection model, thereby using the trained defect detection model to detect defects in small metal fittings in transmission lines, thereby improving the detection accuracy and reliability of defect detection of small metal fittings.
[0038] Optionally, the training module 202 is also used to initialize the network parameters of the adversarial generative network and the deep learning network in the initial defect detection model; input the first image in the small hardware defect label library and the second image without the small hardware defect into the adversarial generative network for adversarial training, and the adversarial generative network is used to generate a defect image containing small hardware defects; input the defect image, the local image and the first image in the small hardware defect label library into the deep learning network for iterative training to obtain a defect detection model.
[0039] Optionally, the training module 202 is also used to input the defect image, the local image and the first image in the small metal fitting defect label library into the deep learning network for iterative training after loading the pre-trained weights in the deep learning network, until the loss function of the deep learning network model converges to obtain a defect detection model.
[0040] Optionally, the training module 202 is also used to input the first image in the small hardware defect label library and the second image without the small hardware defect into the generator of the adversarial generative network for training until the difference between the defect image generated by the generator and the first image reaches a preset accuracy, and the generator is used to generate a defect image of the second image based on the first image; the defect image and the first image are input into the discriminator of the adversarial generative network for training until the error rate of the discriminator is minimized, and the discriminator is used to distinguish between the first image and the defect image.
[0041] The device 200 provided in the embodiment of the present application can execute each method in the foregoing method embodiment and realize the functions and beneficial effects of each method in the foregoing method embodiment, which will not be repeated here.
[0042] Figure 3 It is one of the structural schematic diagrams of a defect detection device for small hardware fittings of power transmission lines based on zero-shot learning provided by an exemplary embodiment of the present application. Referring to the figure, at the hardware level, the defect detection device for small hardware fittings of power transmission lines based on zero-shot learning includes a processor, and optionally, an internal bus, a network interface, and a memory. Among them, the memory may include a memory, such as a high-speed random access memory (Random-Access Memory, RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage device, etc. Of course, the defect detection device for small hardware fittings of power transmission lines based on zero-shot learning may also include hardware required for other services.
[0043] The processor, network interface and memory can be interconnected through an internal bus, which can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one bidirectional arrow is used in the figure, but it does not mean that there is only one bus or one type of bus.
[0044] The memory is used to store programs. Specifically, the program may include program code, and the program code includes computer operation instructions. The memory may include internal memory and non-volatile memory, and provides instructions and data to the processor.
[0045] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a device for locating the specified user at the logical level. The processor executes the program stored in the memory and is specifically used to execute: Figure 1 The method disclosed in the illustrated embodiment implements the functions and beneficial effects of each method in the previous method embodiments, which will not be described in detail here.
[0046] The above application Figure 1The method disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.
[0047] The defect detection device for small metal fittings of power transmission lines based on zero-shot learning can also execute each method in the foregoing method embodiments, and realize the functions and beneficial effects of each method in the foregoing method embodiments, which will not be repeated here.
[0048] Of course, in addition to software implementation methods, the defect detection device for small metal fittings of transmission lines based on zero-shot learning in the present application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the executor of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.
[0049] The present application also provides a computer-readable storage medium, which stores one or more programs. When the one or more programs are executed by an electronic device including multiple application programs, the electronic device executes Figure 1 The method disclosed in the illustrated embodiment implements the functions and beneficial effects of each method in the previous method embodiments, which will not be described in detail here.
[0050] Among them, the computer readable storage medium includes a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.
[0051] Furthermore, an embodiment of the present application also provides a computer program product, the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, and when the program instructions are executed by a computer, the following process is implemented: Figure 1 The method disclosed in the illustrated embodiment implements the functions and beneficial effects of each method in the previous method embodiments, which will not be described in detail here.
[0052] In short, the above are only preferred embodiments of the present application and are not intended to limit the protection scope of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.
[0053] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
[0054] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.
[0055] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0056] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
Claims
1. A method for defect detection of small metal fittings of power transmission lines based on zero-shot learning, characterized in that: include: Marking connection fittings and marking small fitting defects in multiple first images of the transmission line to obtain a connection fitting label library and a small fitting defect label library, wherein the connection fitting label library includes partial images of connection fittings containing small fitting defects, and the small fitting defect label library includes first images containing small fitting defects; Inputting the local image in the connection fitting label library, the first image in the small fitting defect label library, and the second image without small fitting defects into the initial defect detection model for iterative training to obtain a defect detection model; acquiring a third image of the target transmission line; The third image is input into the defect detection model for defect recognition detection, and a defect recognition result of the small metal fitting is obtained by output of the defect detection model.
2. The method for defect detection of small metal fittings for power transmission lines based on zero-shot learning according to claim 1, characterized in that: The partial image in the connection fitting label library, the first image in the small fitting defect label library, and the second image without small fitting defects are input into the initial defect detection model for iterative training to obtain the defect detection model, which includes: Initializing network parameters of the adversarial generation network and the deep learning network in the initial defect detection model; Inputting the first image in the small fitting defect label library and the second image without small fitting defects into the adversarial generation network for adversarial training, wherein the adversarial generation network is used to generate a defect image containing small fitting defects; The defect image, the local image and the first image in the small fitting defect label library are input into the deep learning network for iterative training to obtain the defect detection model.
3. The method for defect detection of small metal fittings for power transmission lines based on zero-shot learning according to claim 2, characterized in that: The step of inputting the second image without small metal fitting defects into the adversarial generation network for adversarial training comprises: Inputting the first image in the small fitting defect label library and the second image without small fitting defects into the generator of the generative adversarial network for training until the difference between the defect image generated by the generator and the first image reaches a preset accuracy, the generator is used to generate a defect image of the second image based on the first image; The defect image and the first image are input into a discriminator of the generative adversarial network for training until the error rate of the discriminator is minimized, and the discriminator is used to distinguish the first image from the defect image.
4. The method for defect detection of small metal fittings for power transmission lines based on zero-shot learning according to claim 2, characterized in that: The step of inputting the defect image, the local image, and the first image in the small fitting defect label library into the deep learning network for iterative training to obtain the defect detection model comprises: After loading the pre-trained weights in the deep learning network, the defect image, the local image and the first image in the small fitting defect label library are input into the deep learning network for iterative training until the loss function of the deep learning network model converges to obtain the defect detection model.
5. The method for defect detection of small metal fittings for power transmission lines based on zero-shot learning according to claim 4, characterized in that: The loss function includes a Generalized Focal Loss loss function.
6. A defect detection device for small metal fittings of power transmission lines based on zero-shot learning, characterized in that: include: A labeling module, used to label connection fittings and mark small fitting defects in a plurality of first images of the transmission line, and obtain a connection fitting label library and a small fitting defect label library, wherein the connection fitting label library includes partial images of connection fittings containing small fitting defects, and the small fitting defect label library includes first images containing small fitting defects; A training module, used for inputting the local image in the connection fitting label library, the first image in the small fitting defect label library, and the second image without small fitting defects into the initial defect detection model for iterative training to obtain a defect detection model; An acquisition module, used for acquiring a third image of the target power transmission line; The detection module is used to input the third image into a defect detection model for defect recognition detection, and obtain a defect recognition result of the small metal fitting output by the defect detection model.
7. The defect detection device for power transmission line small metal fittings based on zero-shot learning according to claim 6, characterized in that: The training module is also used to initialize the network parameters of the adversarial generation network and the deep learning network in the initial defect detection model; Inputting the first image in the small fitting defect label library and the second image without small fitting defects into the adversarial generation network for adversarial training, wherein the adversarial generation network is used to generate a defect image containing small fitting defects; The defect image, the local image and the first image in the small fitting defect label library are input into the deep learning network for iterative training to obtain the defect detection model.
8. The defect detection device for power transmission line small metal fittings based on zero-shot learning according to claim 7, characterized in that: The training module is also used to input the defect image, the local image and the first image in the small metal fitting defect label library into the deep learning network for iterative training after loading the pre-trained weights in the deep learning network, until the loss function of the deep learning network model converges to obtain the defect detection model.
9. The defect detection device for power transmission line small metal fittings based on zero-shot learning according to claim 7, characterized in that: The training module is further used to input the first image in the small fitting defect label library and the second image without small fitting defects into the generator of the adversarial network for training until the difference between the defect image generated by the generator and the first image reaches a preset accuracy, and the generator is used to generate a defect image of the second image based on the first image; The defect image and the first image are input into a discriminator of the generative adversarial network for training until the error rate of the discriminator is minimized, and the discriminator is used to distinguish the first image from the defect image.
10. A defect detection device for small metal fittings of power transmission lines based on zero-shot learning, characterized in that: The defect detection device for small hardware fittings for power transmission lines based on zero-shot learning comprises: a processor and a memory; wherein the memory is used to store computer programs that can be run on the processor; the processor is used to execute the program stored in the memory to implement the steps of the defect detection method for small hardware fittings for power transmission lines based on zero-shot learning as described in any one of claims 1-5.