Power transmission line insulator detection method and system based on YOLOv10

Abstract

The application discloses a power transmission line insulator detection method and system based on YOLOv10, wherein the method comprises the following steps: constructing an insulator dataset; constructing an insulator detection model based on a lightened YOLOv10n network, that is, replacing the original ordinary convolution network of a Backbone network with a GhostConv of a GhostBottleneck module stack; and replacing the original connection layer network of a Neck network with a feature enhancement module Concat_BiFPN, and setting an attention module ContextAggregation after a feature extraction module C2FCIB of the Neck network; finally, feeding the insulator dataset into the insulator detection model based on the lightened YOLOv10n for model training, obtaining the insulator detection model, and using the insulator detection model for insulator detection, which greatly improves the algorithm running efficiency and accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing and target detection technology in transmission line equipment detection technology, and particularly relates to a transmission line insulator detection method and system based on a light-weight improved YOLOv10 algorithm. BACKGROUND

[0002] High-voltage transmission lines are subjected to complex natural environments for a long time, such as lightning strikes and acid rain erosion, and thus are prone to failure. Therefore, it is crucial to regularly inspect the lines. In recent years, unmanned aerial vehicles (UAVs) have become the preferred tool for detecting transmission lines internationally. As a core component of transmission lines, regular inspection of insulators is particularly necessary. Traditional insulator inspection is performed manually, and although this method is simple, it is difficult to cover a large area, is inefficient, and is dangerous due to the fact that transmission lines are mostly located in sparsely populated areas. The commonly used inspection method at the present stage is to use a UAV to take inspection images at a specific position, and then a ground worker manually distinguishes and analyzes faults in the aerial inspection images. This method not only causes a delay in fault detection, but also places a heavy workload on power inspection personnel.

[0003] With the continuous development of deep learning, computer vision and other technologies, it has become possible to use deep learning and other technologies to detect insulator defects in aerial images in combination with existing hardware devices. However, due to the limited computing and storage capabilities of UAVs, the detection model is complex and has a large amount of computation, and the current UAV inspection method is inefficient and time-consuming, and cannot monitor the state of insulators in real time.

[0004] In order to accurately detect the state of insulators in real time, it is necessary to consider reducing the size and complexity of the detection model to meet the real-time detection requirements of UAV devices, while ensuring the detection accuracy of insulators and improving the efficiency and accuracy of UAV inspection of transmission line insulators. SUMMARY

[0005] The purpose of the present application is to solve the problem that the edge device is not suitable for carrying a heavy model in the task of unmanned aerial vehicle inspection of power transmission line insulators, resulting in the need for offline data processing, poor real-time information and low efficiency, and to provide a power transmission line insulator detection method and system based on YOLOv10. In the technical scheme of the present application, the structure of the Backbone network layer is optimized, that is, the original ordinary convolution network is replaced by a more lightweight GhostConv network; the connection layer network in the Neck network layer is replaced by a Concat_BiFPN module, which better balances accuracy and efficiency through top-down and bottom-up multi-scale feature fusion; and a ContextAggregation module is additionally added in the Neck network layer to improve attention and achieve better convergence speed and improve recognition efficiency.

[0006] To this end, the present application provides the following technical scheme:

[0007] A power transmission line insulator detection method based on YOLOv10, comprising the following steps:

[0008] S1: Obtain a real image of a power transmission line insulator and perform image labeling to obtain an insulator dataset;

[0009] Wherein, the insulator position in the real image is labeled;

[0010] S2: Construct an insulator detection model based on a lightweight YOLOv10n network, which comprises a Backbone network, a Neck network and a Head network connected in turn;

[0011] Wherein, the original ordinary convolution network of the Backbone network is replaced by a GhostConv layer based on a GhostBottleneck module stack; and the original connection layer network of the Neck network is replaced by a feature enhancement module Concat_BiFPN, and an attention module ContextAggregation is set after the feature extraction module C2FCIB of the Neck network;

[0012] S3: Send the insulator dataset into the insulator detection model based on the lightweight YOLOv10n for model training to obtain an insulator detection model;

[0013] Wherein, the trained insulator detection model is used for the detection of power transmission line insulators.

[0014] Optionally, the GhostConv layer is provided with 1 standard convolution module and 9 GhostBottleneck modules in turn;

[0015] The GhostBottleneck module is composed of two stacked Ghost modules, one of which is used as an expansion layer to increase the number of channels, and the second is used to reduce the number of channels and matches the shortcut path; the output of the former Ghost module is used as the input of the next Ghost module, and the shortcut is used to connect the input and output of the two Ghost modules.

[0016] Optionally, two ghost convolution layers GhostConv are arranged in the Backbone network.

[0017] The Backbone network comprises, in sequence, a standard convolution layer Conv_1, a ghost convolution layer GhostConv_1, a feature extraction module C2f_1, a ghost convolution layer GhostConv_2, a feature extraction module C2f_2, a down-sampling module SCDown_1, a feature extraction module C2f_3, a down-sampling module SCDown_2, a feature extraction module C2f_4, an attention module SPPF, and an attention module PSA.

[0018] The feature extraction module C2f_2, the down-sampling module SCDown_2, and the attention module PSA correspond to output feature maps P3, P4, and P5, respectively, and input the Neck network.

[0019] The Concat_BiFPN module is composed of a BiFPN network and is an existing module, and thus is not described in detail.

[0020] Optionally, after the feature maps P3, P4, and P5 are input into the Neck network, the feature maps P3, P4, and P5 are processed in the Neck network as follows:

[0021] The feature map P5 is fused with the feature map P4 in a feature enhancement module Concat_BiFPN_1 of the Neck network after being up-sampled by an up-sampling layer Upsample_1 of the Neck network.

[0022] The feature map output by the feature enhancement module Concat_BiFPN_1 is fused with the feature map P3 in a feature enhancement module Concat_BiFPN_2 after feature extraction by a feature extraction module C2f_5 and up-sampling by an up-sampling layer Upsample_2.

[0023] The feature map output by the feature enhancement module Concat_BiFPN_2 is extracted by the feature extraction module C2f_6, and then is input into the standard convolution layer Conv_2, and then is fused with the feature map P5 in the feature enhancement module Concat_BiFPN_3.

[0024] The feature map output by the feature enhancement module Concat_BiFPN_3 is extracted by the feature extraction module C2f_7, and then is input into the down-sampling module SCDown_3, and then is fused with the feature map P5 in the full connection layer Concat.

[0025] The feature map output by the full connection layer Concat is input into the feature extraction module C2FCIB and the attention module ContextAggregation for feature extraction.

[0026] The feature map output by the full connection layer Concat is input into the feature extraction module C2FCIB and the attention module ContextAggregation for feature extraction.

[0027] The feature extraction module C2f_6, the feature extraction module C2f_7 and the attention module ContextAggregation are all configured in the feature extraction module C2FCIB.

[0028] The feature map output by the full connection layer Concat is input into the feature extraction module C2FCIB and the attention module ContextAggregation for feature extraction.

[0029] Optionally, the scale of the standard convolution layer Conv of the Backbone network is P1 / 2, the scale of the ghost convolution layer GhostConv is P2 / 4, P3 / 8, P4 / 16 and P5 / 32, and the scale of the feature map is represented by Pn (n=1, 2, 3, 4, 5).

[0030] Optionally, the operation of the attention module ContextAggregation is represented as:

[0031] Y = ((alphaA(X) + betaA)V)W2 + X

[0032] In the formula, Y is the output of the attention module ContextAggregation, X is the input vector of the attention module ContextAggregation, alpha and beta are learnable parameters, A(X) is dynamically generated from X, A is an affinity matrix, V is a transformation of X obtained by linear projection, and W2 is a learnable parameter, i.e., a weight.

[0033] In addition, the present application also provides a detection system based on the above-mentioned detection method, comprising:

[0034] An insulator dataset construction module is configured to obtain a real image of an insulator of a power transmission line and perform image labeling to obtain an insulator dataset;

[0035] The insulator position in the real image is labeled.

[0036] A detection model construction module is configured to construct an insulator detection model based on a lightweight YOLOv10n network, wherein the lightweight YOLOv10n network comprises a Backbone network, a Neck network and a Head network connected in sequence.

[0037] The original ordinary convolution network of the Backbone network is replaced by a GhostConv layer based on a GhostBottleneck module stack, and the original connection layer network of the Neck network is replaced by a feature enhancement module Concat_BiFPN, and an attention module ContextAggregation is arranged after a feature extraction module C2FCIB of the Neck network.

[0038] A model training module is configured to send the insulator dataset into the insulator detection model based on the lightweight YOLOv10n for model training to obtain an insulator detection model.

[0039] The trained insulator detection model is used for detection of the insulator of the power transmission line.

[0040] The application further provides a computer device comprising at least:

[0041] One or more processors;

[0042] A memory storing one or more computer programs;

[0043] The processor calls the computer program to implement:

[0044] Steps of a power transmission line insulator detection method based on YOLOv10.

[0045] The application further provides a computer storage medium storing a computer program, wherein the computer program is called by a processor to implement:

[0046] Steps of a power transmission line insulator detection method based on YOLOv10.

[0047] Advantages

[0048] The power transmission line insulator detection method provided by the application adopts a deep learning model composed of a Backbone network layer, a Neck network layer and a Head network layer. Considering the limited computing resources of the embedded device, the structure of the Backbone network layer is optimized, and the original ordinary convolution network is replaced with a more lightweight GhostConv network, so that it occupies less computing resources, further ensures the real-time performance of information, and improves the efficiency of unmanned aerial vehicle inspection of power transmission line insulators. In addition, in order to ensure the lightness while improving the accuracy of target detection, the connection layer network in the Neck network layer is replaced with a Concat_BiFPN module, which better balances the accuracy and efficiency through the fusion of multi-scale features from top to bottom and from bottom to top. A ContextAggregation module is additionally added in the Neck network layer to improve attention, so as to achieve better convergence speed and improve recognition efficiency.

[0049] The essential difference between the existing YOLO application and the insulator detection field is that the technical scheme of the application innovatively balances the selection problem of accuracy and efficiency faced by lightweight, and has high accuracy while considering efficiency, reduces the size and complexity of the detection model. In addition, it has good robustness to the interference of complex background on insulator images, can effectively realize the positioning of insulator targets in the inspection image, provides a basis for subsequent defect detection, and can screen the images obtained by inspection, reduce the burden of manual screening, improve the inspection efficiency and reduce the work intensity of the inspection personnel, and escort the normal operation of the power system. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0051] Figure 1 The flow chart of the power transmission line insulator detection method of the application;

[0052] Figure 2 The Backbone network structure diagram in the detection model of the application;

[0053] Figure 3 The Neck network and Head network structure diagram in the detection model of the application;

[0054] Figure 4 The Ghost module network structure diagram in the Ghost convolution layer GhostConv in the detection model of the application;

[0055] Figure 5 This is the GhostBottleneck network structure diagram in the detection model of the present invention;

[0056] Figure 6 This is the network structure diagram of the ghost convolution layer GhostConv in the detection model of the present invention. DETAILED DESCRIPTION

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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 any creative work.

[0058] An embodiment of the present invention provides a transmission line insulator detection method based on YOLOv10, comprising the following steps:

[0059] Step 1: Obtain real-world images of transmission and transformation line insulators and annotate them to obtain an insulator dataset. This means constructing a training set and a test set for the insulators.

[0060] In this embodiment, 1500 images of insulators taken by drones are selected, and the image resolution is adjusted to 1152×864.

[0061] The LabelImg tool labels the images, obtains the corresponding XML files, and establishes the insulator dataset; 900 labeled insulator images are selected as the original training set, and the remaining 600 insulator images are used as the test set.

[0062] The training set is then preferably enhanced to 2,400 images using methods such as geometric transformation, color transformation, contrast and brightness adjustment, and simulation of complex weather, wherein the geometric transformation methods include but are not limited to rotation, translation, cropping, and mirroring; and the simulation of complex weather includes but is not limited to adding rain, snow, and fog.

[0063] It should be understood that the enhanced operation is a preferred mode of the embodiment of the present invention. In other feasible embodiments, selective operation can be performed according to actual needs.

[0064] Step 2: Construct an insulator detection model based on a lightweight YOLOv10n network. This paper uses YOLOv10n as the basic framework and improves the lightweight structure. The lightweight YOLOv10n network consists of a Backbone network, a Neck network, and a Head network connected in sequence.

[0065] like Figure 2As shown, the Backbone network includes a standard convolutional layer Conv_1, a ghost convolutional layer GhostConv_1, a feature extraction module C2f_1, a ghost convolutional layer GhostConv_2, a feature extraction module C2f_2, a down-sampling module SCDown_1, a feature extraction module C2f_3, a down-sampling module SCDown_2, a feature extraction module C2f_4, an attention module SPPF, and an attention module PSA connected in sequence. Based on the above network mechanism, the feature extraction module C2f_2, the down-sampling module SCDown_2, and the attention module PSA correspond to the output feature map P3, the feature map P4, and the feature map P5, respectively, and are input into the Neck network. In this embodiment, the scale of the standard convolutional layer Conv is P1 / 2, and the scale of the ghost convolutional layer GhostConv is P2 / 4, P3 / 8, P4 / 16, and P5 / 32.

[0066] As can be seen from the above network architecture, in order to realize network lightweight, the ghost convolutional layer GhostConv based on the GhostBottleneck module stack is used to replace the original ordinary convolutional network of the Backbone network. As shown in Figure 6 As shown, the ghost convolutional layer GhostConv is provided with one standard convolutional module and nine GhostBottleneck modules in sequence. Figure 5 As shown, the GhostBottleneck module is composed of two stacked Ghost modules (as shown in Figure 4 One of the Ghost modules is used as an expansion layer to increase the number of channels, and the second Ghost module is used to reduce the number of channels and match the shortcut path; the output of the first Ghost module is used as the input of the next Ghost module, and the shortcut is used to connect the input and output of the two Ghost modules. With the above structure, the network parameter quantity and the calculation quantity are greatly reduced while the size of the output feature map is unchanged.

[0067] In addition to optimizing the Backbone network, the Neck network is also optimized as follows: the original connection layer network of the Neck network is replaced by the feature enhancement module Concat_BiFPN, and the attention module ContextAggregation is arranged after the feature extraction module C2FCIB of the Neck network.

[0068] After the feature map P3, the feature map P4, and the feature map P5 are input into the Neck network, the feature map P3, the feature map P4, and the feature map P5 are processed in the Neck network as follows:

[0069] As shown in Figure 3As shown, the feature map P5 is fused with the feature map P4 in the feature enhancement module Concat_BiFPN_1 of the Neck network after being up-sampled by the up-sampling layer Upsample_1 of the Neck network; the feature map output by the feature enhancement module Concat_BiFPN_1 is fused with the feature map P3 in the feature enhancement module Concat_BiFPN_2 after being up-sampled by the up-sampling layer Upsample_2 and extracting features by the feature extraction module C2f_5; the feature map output by the feature enhancement module Concat_BiFPN_2 is fused with the feature map P2 in the feature enhancement module Concat_BiFPN_3 after extracting features by the feature extraction module C2f_6 and passing through the standard convolution layer Conv_2; the feature map output by the feature enhancement module Concat_BiFPN_3 is fused with the feature map P1 in the full connection layer Concat after extracting features by the feature extraction module C2f_7 and passing through the down-sampling module SCDown_3; the feature map output by the full connection layer Concat is fused with the feature map P4 in the feature extraction module C2FCIB and the attention module ContextAggregation after extracting features by the feature extraction module C2FCIB.

[0070] ContextAggregation.

[0071] ContextAggregation.

[0072] The feature maps output by the feature extraction module C2f_6, the feature extraction module C2f_7 and the attention module ContextAggregation are input into the Head network, and the Head network is correspondingly provided with three target detection modules Detect, and the results of the three target detection modules Detect are fused by convolution to obtain the insulator detection result.

[0073] The Concat_BiFPN module of the embodiment is composed of a BiFPN network, and a learnable weight is introduced on the basis of the FPN network to learn the importance of different input features, and the top-down and bottom-up multi-scale feature fusion is repeatedly applied, which has better accuracy and efficiency trade-off. Since the network architecture of the module is known, it will not be described in detail.

[0074] The ContextAggregation module of the embodiment is an additional attention mechanism module, which is a general construction module of multi-head context aggregation, which can utilize long-range interaction like Transformers, while still utilizing the inductive bias of local convolution operations, so as to realize faster convergence speed. The principle is as follows:

[0075] Y = ((aA(X) + bA)V)W2 + X (1)

[0076] Wherein, Y is the output, X is the input vector, a and b are learnable parameters, A(X) is dynamically generated from X, A is the affinity matrix, V is the transformation of X obtained by linear projection, and W2 is also a learnable parameter, i.e., the weight.

[0077] The Head network of the embodiment includes a target detection module Detect_1, a target detection module Detect_2, and a target detection module Detect_3; the target detection module Detect_1 performs target detection on the features extracted by the feature extraction module C2f_6 in the Neck network and outputs a target 1; the target detection module Detect_2 performs target detection on the features extracted by the feature extraction module C2f_7 and outputs a target 2; the target detection module Detect_3 performs target detection on the features extracted by the ContextAggregation module and outputs a target Detect_3; finally, the target 1, the target 2, and the target 3 are convoluted and fused, and the final insulator target is output.

[0078] In the embodiment, the input features of the insulator detection model are 640x640x3, the Backbone network layer is used to extract insulator image features, and the extracted image features have sizes of 160x160x64, 80x80x128, 40x40x256, 20x20x512, and 10x10x1024; the Neck network layer performs fusion processing on features of three scales of 40x40, 80x80, and 160x160, the three inputs of the feature fusion network layer are 40x40x256, 80x80x128, and 10x10x1024, and the three outputs of the feature fusion network layer are 40x40x256, 80x80x128, and 10x10x1024; the inputs of the target detection layer are connected with the outputs of the feature fusion network layer, and the target detection layer performs prediction on insulator images of three scales of 160x160, 80x80, and 40x40.

[0079] It should be understood that the above image sizes are examples of the embodiment, and in other feasible embodiments, adaptive adjustments can be made according to actual application requirements.

[0080] Step 3: The insulator dataset is input into the insulator detection model based on the lightweight YOLOv10n for model training, and an insulator detection model is obtained; wherein the trained insulator detection model is used for detection of insulators of a power transmission line.

[0081] In summary, the embodiment of the present application proposes a deep learning model for detecting insulators of power transmission lines, which is constructed based on a light-weight optimized YOLOv10 algorithm, greatly reducing the size and complexity of the network, occupying less computing resources while the accuracy is also improved to a certain extent. The YOLOv10 network architecture covers three layers of network, namely Backbone, Neck and Head. This model shows strong robustness to the interference of insulator images in complex background, and can accurately locate the insulator target in the inspection image, laying a solid foundation for subsequent defect detection. In addition, this model also has image screening function, automatically filtering non-key images, significantly reducing the workload of manual screening, indicating a wide application potential. The detection method of the present application reduces the model size and complexity, realizes efficient operation under low computing resource consumption, can quickly and accurately identify and locate the insulator from the complex background, not only improves the efficiency of inspection operation, but also effectively reduces the labor intensity of inspection personnel, providing strong support for the stable and efficient operation of the power system.

[0082] Embodiment 2:

[0083] The embodiment of the present application also provides a detection system based on the above insulator detection method, comprising: an insulator data set construction module, a detection model construction module and a model training module.

[0084] The insulator data set construction module is used to obtain real shot images of insulators of power transmission and transformation lines and perform image labeling to obtain an insulator data set, that is, label the positions of insulators in the real shot images.

[0085] The detection model construction module is used to construct an insulator detection model based on a light-weight YOLOv10n network, which comprises a Backbone network, a Neck network and a Head network connected in sequence.

[0086] The original ordinary convolution network of the Backbone network is replaced by a GhostConv layer based on a GhostBottleneck module stack, and the original connection layer network of the Neck network is replaced by a feature enhancement module Concat_BiFPN, and an attention module ContextAggregation is set after a feature extraction module C2FCIB of the Neck network.

[0087] The model training module is used to input the insulator data set into the insulator detection model based on the light-weight YOLOv10n for model training, to obtain an insulator detection model.

[0088] In some embodiments, a detection module is further included for insulator detection using the trained insulator detection model.

[0089] It should be understood that the specific implementation process of each module refers to the above method content, and the present application will not be repeated here, and the division of the above functional modules is only for illustration. In some embodiments, part of the functional modules can be combined, part of the functional modules can be split, each functional module can be implemented in software or hardware or a combination of software and hardware. The soft and hardware devices include but are not limited to general-purpose computer devices, programmable gate arrays, digital signal processors, microprocessors and their corresponding programming or burning software.

[0090] Embodiment 3:

[0091] The embodiment of the present application also provides a computer device, at least comprising: one or more processors; and a memory storing one or more computer programs.

[0092] The processor calls the computer program to implement the steps of the power transmission line insulator detection method based on YOLOv10.

[0093] That is, the specific implementation is:

[0094] Step 1: Obtain the actual shot image of the power transmission line insulator and perform image labeling to obtain an insulator data set.

[0095] Step 2: Construct an insulator detection model based on a lightweight YOLOv10n network.

[0096] Step 3: Send the insulator data set into the insulator detection model based on the lightweight YOLOv10n for model training to obtain an insulator detection model; wherein the trained insulator detection model is used for detection of the power transmission line insulator.

[0097] The specific implementation process of each step is described in the foregoing method.

[0098] It should be understood that in the embodiments of the present application, the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The memory can include read-only memory and random access memory, and provide instructions and data to the processor. Part of the memory can also include non-volatile random access memory. For example, the memory can also store device type information.

[0099] Embodiment 4:

[0100] The computer storage medium of the embodiment of the present application stores a computer program, and the computer program is called by a processor to implement the steps of a power transmission line insulator detection method based on YOLOv10.

[0101] That is, the specific implementation is:

[0102] Step 1: Obtain a real image of a power transmission line insulator and perform image labeling to obtain an insulator data set.

[0103] Step 2: Construct an insulator detection model based on a lightweight YOLOv10n network.

[0104] Step 3: Input the insulator data set into the insulator detection model based on the lightweight YOLOv10n to perform model training, and obtain an insulator detection model; wherein the trained insulator detection model is used for detection of power transmission line insulators.

[0105] The specific implementation process of each step is described in the foregoing method.

[0106] The readable storage medium is a computer readable storage medium, which can be an internal storage unit of the software and hardware device in any of the foregoing embodiments, such as a hard disk or a memory of the controller. The readable storage medium can also be an external storage device of the controller, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the readable storage medium can include both the internal storage unit and the external storage device of the controller. The readable storage medium is used to store the computer program and other programs and data required by the controller. The readable storage medium can also be used to temporarily store data that has been output or will be output.

[0107] Based on such understanding, the technical solutions of the present application, essentially or in the sense of contribution to the prior art, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The foregoing readable storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0108] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product embodied on one or more computer readable storage media (including but not limited to disk memory, CD-ROM, optical storage medium, etc.) having computer usable program code thereon. The present application is produced by referring to flowcharts of methods, devices (systems) and computer program products according to the embodiments of the present application, and instructions executed by a processor to implement functions specified in one or more flows of the flowcharts and / or one or more blocks of the block diagrams. These computer program instructions can also be stored in a computer readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer readable memory produce a product including instruction means, which implements the functions specified in one or more flows of the flowcharts and / or one or more blocks of the block diagrams. These computer program instructions can also be loaded into a computer or other programmable data processing device, so that a series of operation steps are performed on the computer or other programmable data processing device to produce a computer implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more flows of the flowcharts and / or one or more blocks of the block diagrams.

[0109] The above embodiments are only used to illustrate the technical solutions of the present application, rather than limit the present application; although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalent features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A transmission line insulator detection method based on YOLOv10, characterized by: The following steps are involved: S1: Obtain real-life images of transmission and transformation line insulators and annotate them to obtain an insulator dataset. Among them, the position of the insulator in the real-shot image is marked; S2: Build an insulator detection model based on a lightweight YOLOv10n network, where the lightweight YOLOv10n network includes a Backbone network, a Neck network, and a Head network connected in sequence; The original ordinary convolutional network of the Backbone network is replaced by the ghost convolution layer GhostConv based on the GhostBottleneck module stack; the original connection layer network of the Neck network is replaced by the feature enhancement module Concat_BiFPN, and the attention module ContextAggregation is set after the feature extraction module C2FCIB of the Neck network; the Concat_BiFPN module is composed of the BiFPN network; The Backbone network is provided with two ghost convolution layers GhostConv, and the Backbone network includes a standard convolution layer Conv_1, a ghost convolution layer GhostConv_1, a feature extraction module C2f_1, a ghost convolution layer GhostConv_2, a feature extraction module C2f_2, a downsampling module SCDown_1, a feature extraction module C2f_3, a downsampling module SCDown_2, a feature extraction module C2f_4, an attention module SPPF and an attention module PSA connected in sequence; The feature extraction module C2f_2, the downsampling module SCDown_2, and the attention module PSA respectively output feature maps P3, P4, and P5, and input them into the Neck network; After the feature maps P3, P4, and P5 are input into the Neck network, the processing of the feature maps P3, P4, and P5 in the Neck network is as follows: After the feature map P5 passes through the upsampling layer Upsample_1 of the Neck network, it is fused with the feature map P4 in the feature enhancement module Concat_BiFPN_1 of the Neck network; the feature map output by the feature enhancement module Concat_BiFPN_1 is subjected to feature extraction by the feature extraction module C2f_5 and passes through the upsampling layer Upsample_2, and is fused with the feature map P3 in the feature enhancement module Concat_BiFPN_2; the feature map output by the feature enhancement module Concat_BiFPN_2 is subjected to feature extraction by the feature extraction module C2f_6 and then passes through the standard convolution layer Conv_2, and is then fused with the feature map P5 in the feature enhancement module Concat_BiFPN_3; the feature map output by the feature enhancement module Concat_BiFPN_3 is subjected to feature extraction by the feature extraction module C2f_7, and then passes through the downsampling module SCDown_3 and is fused with the feature map P5 in the fully connected layer Concat; the feature map output by the fully connected layer Concat is subjected to feature extraction by the feature extraction module C2FCIB and the attention module ContextAggregation; The feature maps output by the feature extraction module C2f_6, the feature extraction module C2f_7, and the attention module ContextAggregation are all input into the Head network. The Head network is correspondingly provided with three target detection modules Detect. The results of the three target detection modules Detect are convolutionally fused to obtain the insulator detection result. The operation of the attention module ContextAggregation is expressed as: Y=((αA(X)+βA)V)W2+X Where Y is the output of the attention module ContextAggregation, X is the input vector of the attention module ContextAggregation, α and β are learnable parameters, A(X) is dynamically generated from X, A is the affinity matrix, V is the transformation of X obtained by linear projection, and W2 is a learnable parameter, i.e., weight; S3: Send the insulator dataset to an insulator detection model based on lightweight YOLOv10n for model training to obtain an insulator detection model; Among them, the trained insulator detection model is used to detect insulators of transmission and transformation lines.

2. The method according to claim 1, wherein: The ghost convolution layer GhostConv is sequentially provided with 1 standard convolution module and 9 GhostBottleneck modules; The GhostBottleneck module consists of two stacked Ghost modules, one of which is used as an expansion layer to increase the number of channels, and the second Ghost module is used to reduce the number of channels and match the shortcut path; the output of the previous Ghost module is used as the input of the next Ghost module, and then a shortcut is used to connect the input and output of the two Ghost modules.

3. A detection system based on the method according to any one of claims 1 to 2, characterized in that: include: The insulator dataset construction module is used to obtain real-time images of transmission and transformation line insulators and perform image annotation to obtain the insulator dataset; Among them, the position of the insulator in the real shot image is marked; A detection model construction module is used to build an insulator detection model based on a lightweight YOLOv10n network, wherein the lightweight YOLOv10n network includes a Backbone network, a Neck network, and a Head network connected in sequence; Among them, the original ordinary convolution network of the Backbone network is replaced by the ghost convolution layer GhostConv based on the GhostBottleneck module stack; and the original connection layer network of the Neck network is replaced by the feature enhancement module Concat_BiFPN, and the attention module ContextAggregation is set after the feature extraction module C2FCIB of the Neck network; A model training module is used to send the insulator dataset to an insulator detection model based on lightweight YOLOv10n for model training to obtain an insulator detection model; Among them, the trained insulator detection model is used to detect insulators of transmission and transformation lines.

4. A computer device, characterized in that: At least: one or more processors; a memory storing one or more computer programs; The processor calls the computer program to implement: The steps of the method according to any one of claims 1 to 2.

5. A computer storage medium, characterized in that: A computer program is stored, which is called by a processor to implement: The steps of the method according to any one of claims 1 to 2.

Images