A Transmission Line Fault Detection Method and System Based on YOLOX Network

Abstract

The present invention discloses a transmission line fault detection method and system based on a YOLOX network. The method obtains transmission line fault images in several real-world scenarios, expands and annotates the fault data set, and organizes the data set to form a transmission line fault detection data set containing five types of faults, including hanging foreign objects, defective insulators, bird's nests, broken strands, and ice coverage. Then, a transmission line fault detection network based on YOLOX is built, and the feature fusion method and loss function of the transmission line fault detection network are improved. The transmission line fault detection network is trained by adding a data enhancement method, and the trained YOLOX-based transmission line fault detection model is used to perform transmission line fault detection, completing YOLOX-based transmission line fault detection. The present invention improves the accuracy of transmission line fault detection, can effectively detect transmission line faults while ensuring real-time performance, and provides better technical support for transmission line fault detection and intelligent line patrol in practical applications.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent monitoring of power transmission lines, and in particular relates to a power transmission line fault detection method and system based on a YOLOX network. Background Art

[0002] Transmission lines are widely distributed and complex, making them susceptible to climate, topography, natural disasters, and human influence. Long-term exposure to harsh environments inevitably leads to conductor breakage, insulator defects, and foreign matter buildup, leading to transmission line stability failures. This directly impacts the reliable operation of the power system and, consequently, significantly impacts production and daily life. Therefore, real-time monitoring of transmission lines can provide timely information on the status of equipment along the lines and provide early warnings of potential faults, which is of great value in ensuring the stable operation of the power system.

[0003] In recent years, artificial intelligence technologies, particularly deep learning, have achieved tremendous success in image processing, particularly in tasks such as image restoration, object detection, and scene segmentation. This has provided a promising design approach for online monitoring of power transmission lines. Therefore, applying deep learning to transmission line fault detection and enabling real-time monitoring of transmission line faults can promptly identify faults and provide early warnings, which is of great significance to the safe and stable operation of power systems.

[0004] Transmission line fault detection based on deep learning has the following problems:

[0005] 1) Due to the particularity of the national power industry, there is a lack of public standard datasets in the field of transmission line fault detection, and most fault data types are single;

[0006] 2) Fault targets in power scenarios are complex and changeable. When fault targets of different scales exist simultaneously, the fault detection method based on deep learning has poor detection effect;

[0007] 3) It is often difficult to accurately locate faults using deep learning-based detection networks. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to address the deficiencies in the above-mentioned prior art and provide a transmission line fault detection method and system based on the YOLOX network, which is used to solve the technical problems of the lack of public standard data sets in the field of transmission line fault detection, the complexity and changeability of fault targets in power scenarios, and the poor detection effect when fault targets of different scales exist at the same time, thereby effectively improving the accuracy of transmission line fault detection.

[0009] The present invention adopts the following technical solutions:

[0010] A transmission line fault detection method based on a YOLOX network comprises the following steps:

[0011] S1. Use the transmission line fault image in the real scene as the original sample;

[0012] S2, expanding the original sample obtained in step S1;

[0013] S3. Use LabelImg to label the original samples expanded in step S2 and form a transmission line fault detection dataset;

[0014] S4. Build a YOLOX-based transmission line fault detection network and optimize it using an adaptive multi-scale fusion mechanism based on dilated convolution and a loss function based on polynomial IoU.

[0015] S5. Perform data enhancement on the transmission line fault detection dataset constructed in step S3, then train the YOLOX-based transmission line fault detection network optimized in step S4, and use the trained YOLOX-based transmission line fault detection network to implement transmission line fault detection.

[0016] Specifically, in step S2, the original sample obtained in step S1 is expanded through translation, rotation, scene simulation, and fault simulation operations.

[0017] Furthermore, the scenario simulation is as follows:

[0018] By simulating scenarios where conductors are loose or broken, and capturing images of conductors at different angles, under different lighting conditions, and with varying degrees of damage, we can expand the data on transmission lines.

[0019] The fault simulation is specifically as follows: using Photoshop software to simulate fault images of foreign objects hanging on the transmission line and insulator defects, and by splicing the fault images with the non-fault images, the fault data can be expanded.

[0020] Specifically, in step S3, the fault types of the transmission line fault detection dataset include hanging foreign objects, defective insulators, bird nests, loose strands, and ice coverage.

[0021] Specifically, in step S4, the loss function L is:

[0022] L=L IoU-p +FL(p t )+L cls

[0023] Among them, L IoU-p is the polynomial IoU loss function, FL(p t ) is the confidence loss function, L clsis the category loss function.

[0024] Furthermore, the confidence loss function FL(p t )for:

[0025] FL(p t )=-α t (1-p t )γlog((p t )

[0026] Among them, (1-p t )γ is the adjustment factor, γ is the adjustable focusing parameter, p t The predicted probability output by the YOLOX network.

[0027] Furthermore, the polynomial IoU loss L IoU-p for:

[0028]

[0029] Among them, x is the value of IoU, w i The coefficients of the polynomial are 1 or 0, x i is a polynomial, and n is 6.

[0030] Specifically, in step S5, Mosaic and Mixup are used to perform data enhancement on the transmission line fault detection dataset.

[0031] Furthermore, the specific process of Mixup is as follows:

[0032] New training samples and labels are constructed for data of different fault categories using linear interpolation. This means mixing two images proportionally and also mixing their labels.

[0033] The specific process of Mosaic is:

[0034] Randomly crop parts of four images and then stitch them together to form a sample to achieve data augmentation.

[0035] In a second aspect, an embodiment of the present invention provides a YOLOX network-based transmission line fault detection system, comprising:

[0036] The data module uses the transmission line fault images in real scenarios as original samples;

[0037] The expansion module expands the original samples obtained by the data module through translation, rotation, scene simulation, and fault simulation operations;

[0038] The labeling module uses LabelImg to label the original samples expanded by the expansion module, and organizes them into a transmission line fault detection dataset that includes five types of faults: foreign object hanging, insulator defect, bird's nest, loose strands, and icing;

[0039] Network module, builds a YOLOX-based transmission line fault detection network, and optimizes the YOLOX-based transmission line fault detection network using an adaptive multi-scale fusion mechanism based on dilated convolution and a loss function based on polynomial IoU;

[0040] The detection module performs data augmentation on the transmission line fault detection dataset constructed by the annotation module. It then trains the YOLOX-based transmission line fault detection network optimized by the network module and uses the trained YOLOX-based transmission line fault detection network to implement transmission line fault detection.

[0041] Compared with the prior art, the present invention has at least the following beneficial effects:

[0042] A transmission line fault detection method based on a YOLOX network, which obtains transmission line fault images in several real scenarios and expands the fault data set by means of scene simulation, image simulation, data enhancement, etc.; then LabelImg is used to annotate the expanded data to form a transmission line fault detection data set containing five types of faults, such as hanging foreign objects, defective insulators, bird's nests, loose strands, and ice; then a transmission line fault detection network based on YOLOX is built, and the feature fusion method and loss function of the network are improved; finally, data enhancement methods such as mosaic and mixup are added to train the network, and a transmission line fault detection model based on YOLOX is obtained to perform transmission line fault detection, thereby completing transmission line fault detection based on YOLOX. It should be noted that the present invention constructs an effective transmission line fault detection data set in the absence of a publicly available annotated data set for the power scenario, and realizes rapid and efficient detection of transmission line faults, providing better technical support for transmission line fault detection, intelligent line patrol, etc. in actual application scenarios.

[0043] Furthermore, data expansion can be achieved through operations such as translation, rotation, scene simulation, and fault simulation, thereby completing effective model training and improving the generalization performance of the algorithm.

[0044] Furthermore, scenario simulation and fault simulation can simulate real fault scenarios to achieve data expansion.

[0045] Furthermore, the fault types of the transmission line fault detection dataset include five types of data: hanging foreign objects, defective insulators, bird nests, loose strands and ice coverage, in preparation for training models that can detect multiple faults.

[0046] Furthermore, the loss function L improves the confidence loss and IoU loss, which can solve the problem of sample imbalance and insufficient optimization ability for high-quality detection frames.

[0047] Furthermore, in order to improve the detection effect of difficult samples, the present invention optimizes the loss function of YOLOX and replaces the error caused by the confidence L_conf with Focal Loss, which can effectively alleviate the sample imbalance problem.

[0048] Furthermore, by adjusting the coefficients, we can adjust the attention paid to detection frames of different qualities. When the IoU is small, the loss is basically the same as the original YOLOX loss. When the IoU is large, the polynomial weight becomes relatively large, causing the network to focus more attention on high-quality detection frames, thereby obtaining better regression results.

[0049] Furthermore, using Mosaic and Mixup to perform data augmentation on the transmission line fault detection dataset can improve the generalization ability of the model and further enhance the practicality of the model.

[0050] It can be understood that the beneficial effects of the second aspect mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0051] In summary, the present invention performs real-time detection of transmission line faults based on the improved YOLOX. By improving the multi-scale feature fusion method and loss function of YOLOX, the detection accuracy of the model is effectively improved. Experimental results show that the method of the present invention can achieve a high mAP while ensuring real-time performance. It can be effectively applied in transmission line fault detection scenarios, providing better technical support for transmission line fault detection and intelligent line patrol in practical applications.

[0052] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a flow chart of the present invention;

[0054] Figure 2 Some of the data graphs collected for this invention;

[0055] Figure 3 This is the improved YOLOX network structure diagram;

[0056] Figure 4 This is the adaptive multi-scale fusion module diagram based on dilated convolution proposed in the present invention;

[0057] Figure 5 Comparison chart of polynomial weights in polynomial loss;

[0058] Figure 6 Comparison of IoU loss with different coefficient polynomials;

[0059] Figure 7 Part of the test results. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0061] In the description of the present invention, it is to be understood that the terms “include” and “comprise” indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0062] It should also be understood that the terms used in the present specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0063] It should be further understood that the term "and / or" as used in the present specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items. For example, A and / or B may represent: A alone, A and B simultaneously, or B alone. In addition, the character " / " herein generally indicates that the associated items are in an "or" relationship.

[0064] It should be understood that although the terms "first," "second," and "third" may be used to describe preset ranges in embodiments of the present invention, these preset ranges should not be limited to these terms. These terms are merely used to distinguish one preset range from another. For example, without departing from the scope of embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0065] The word "if," as used herein, may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0066] The accompanying drawings illustrate various schematic diagrams of structures according to embodiments disclosed herein. These figures are not drawn to scale; for clarity, some details are exaggerated and some details may be omitted. The shapes of the various regions and layers shown in the figures, as well as their relative sizes and positional relationships, are merely exemplary and may deviate in practice due to manufacturing tolerances or technical limitations. Those skilled in the art may design regions / layers with different shapes, sizes, and relative positions as needed.

[0067] The present invention provides a method for detecting power transmission line faults based on the YOLOX network. The method obtains power transmission line fault images in several real scenarios, and expands the fault data set by using scene simulation, image simulation, data enhancement and other means. LabelImg is then used to label the expanded data, and a power transmission line fault detection data set is formed, which includes five types of faults, such as hanging foreign objects, defective insulators, bird's nests, broken strands, and ice. A power transmission line fault detection network based on YOLOX is then built, and the feature fusion method and loss function of the power transmission line fault detection network are improved. Finally, data enhancement methods such as Mosaic and Mixup are added to train the power transmission line fault detection network, and a power transmission line fault detection model based on YOLOX is obtained for power transmission line fault detection, thereby completing the power transmission line fault detection based on YOLOX. The present invention improves the accuracy of power transmission line fault detection, and can effectively detect power transmission line faults while ensuring real-time performance, providing better technical support for practical technologies such as power transmission line fault detection and intelligent line patrol in actual applications. Please refer to Figure 1 The present invention provides a transmission line fault detection method based on a YOLOX network, comprising the following steps:

[0068] S1. Obtain a transmission line fault image in a real scenario and use it as the original sample;

[0069] Install cameras at fixed locations on transmission line towers or use drones to collect power images while collecting relevant public images. Then screen the images and select those with faults.

[0070] S2. Expand the transmission line fault image through operations such as translation, rotation, scene simulation, and fault simulation;

[0071] Common data enhancement methods are used on the original image, including horizontal flipping, scaling, brightness and contrast transformation, to achieve the expansion of the transmission line data.

[0072] Simulating wire breakage and loose strand faults: Since the image data of loose and broken transmission line strands collected in real scenes is very limited, in addition to field collection and network collection, the present invention also simulates scenes of loose and broken wire strands to capture images of wires with different degrees of damage under different angles and lighting conditions.

[0073] Software simulation: Actual data collected on transmission line faults is limited. To expand the dataset, this paper utilizes Photoshop's image synthesis, special effects creation, image editing, and color correction capabilities to simulate two types of fault images: foreign objects hanging on transmission lines and insulator defects. By splicing fault images with non-fault images, the fault data can be expanded.

[0074] S3. Label the expanded data using LabelImg to form a transmission line fault detection dataset that includes five types of faults: hanging foreign objects, defective insulators, bird nests, loose strands, and icing.

[0075] The expanded data is labeled using LabelImg. Objects to be labeled include foreign objects hanging in transmission lines, bird nests, defective insulators, broken and loose strands of transmission lines, and ice-covered areas of transmission lines and related components. By labeling the above faults, a transmission line fault detection dataset is formed, which includes five types of faults: hanging foreign objects, defective insulators, bird nests, loose strands, and ice.

[0076] S4. Build a YOLOX-based transmission line fault detection network and propose to optimize the YOLOX network using an adaptive multi-scale fusion mechanism based on dilated convolution and a loss function based on polynomial IoU.

[0077] The PAFPN multi-scale fusion structure used in the original YOLOX network has been improved to an adaptive multi-scale fusion method based on dilated convolution. The IoU loss used in the original YOLOX network has been improved to a polynomial IoU loss that can guide the network to pay more attention to high-quality detection boxes. The detailed operation is as follows:

[0078] YOLOX uses the multi-scale fusion structure of PAFPN to leverage multi-scale features. This multi-scale feature fusion method is based on a feature pyramid. When using a feature pyramid for detection, large scales are often used to correspond to large objects, while small scales correspond to small objects. If an image contains objects of multiple scales, these objects will conflict in the feature pyramid, resulting in inconsistencies. This inconsistency at different scales can affect network training and, consequently, the performance of the feature pyramid, especially in single-stage object detection.

[0079] In power line inspection scenarios, multiple targets of different scales often appear in a single image. To address this issue, the present invention proposes introducing an adaptive feature fusion method (DASFF) based on dilated convolution in the multi-scale feature fusion stage of YOLOX. This method can adaptively fuse features of other scales and automatically obtain a suitable fusion parameter, thereby achieving the combination of useful features. Features of different scales in the network first need to be adjusted to match the resolution of the current fusion layer through upsampling or downsampling. Then, through continuous optimization through training, the optimal fusion parameters are obtained to achieve adaptive fusion of multiple scales.

[0080] In YOLOX, features of different scales can be represented by f l Indicates that l represents different scales (l∈({1,2,3}) of YOLOX). For a certain scale l, we first need to calculate the features f of other scales n (n≠l) n Adjust it so that its size is the same as f l In YOLOX, features at different scales have different sizes and numbers of channels, so they need to be adjusted through upsampling or downsampling strategies, and the number of channels needs to be adjusted through 1×1 convolution.

[0081] The feature fusion expression is expressed as formula (1):

[0082]

[0083] in, Represents the feature vector of the feature map position (i, j) from n to l, y l Represents the component of the (i, j)th vector output mapping feature map. and The parameters obtained through training represent the proportion that features of different scales should occupy at this scale. By optimizing these three parameters, the optimal fusion of information at multiple scales can be achieved; and The following conditions are met:

[0084]

[0085]

[0086] in, and It is defined in the form of a softmax function.

[0087] In order to better integrate context information into the weight coefficient, x 1→l 、x 2→l and x 3→l Perform 3×3 convolution with a void rate of 2 to obtain and By training the weights and The algorithm solves the problem to obtain the optimal fusion weights. The learned optimal weights enable adaptive fusion of features at different scales, filtering out useless information and enhancing useful information within the multi-scale information, thereby ensuring that useful information dominates the network. Using the multi-scale features obtained from adaptive feature fusion for detection effectively resolves conflicts between objects of different scales in the feature pyramid, resulting in better detection results.

[0088] The network structure designed by the present invention is as follows Figure 3 The DASFF structure designed by the present invention is shown in FIG. Figure 4 shown.

[0089] YOLOX's loss function L consists of three parts: position loss L IoU , confidence loss L conf , category loss L cls :

[0090] L=L Iou +L conf +L cls (4)

[0091] YOLOX uses the classic IoU loss to optimize the position, i.e. the detection frame. The calculation formula of IoU loss is as follows:

[0092] L IoU =1-IoU 2 (5)

[0093] The IoU loss used in YOLOX is simple and effective, but it ignores the imbalance of detection frames in the data, making it difficult to obtain high-quality regression results. In the process of target detection frame regression, the difficulty often lies in obtaining high-quality detection frames. However, existing IoU losses such as GIoU

[18] , DIoU

[19] , and CIoU

[20] often do not pay special attention to the regression of high-quality detection frames, but focus more on low-quality detection frames, making the network's ability to regress high-quality detection frames insufficient.

[0094] In the transmission line fault detection scenario, in order to accurately detect the specific location of the fault, it is necessary to accurately locate the fault. If the positioning is inaccurate, it is impossible to quickly and accurately point out the fault point to the maintenance personnel, making it difficult to achieve efficient transmission line fault inspection and maintenance, affecting the actual use value of the system.

[0095] This paper proposes a polynomial IoU loss to replace the original position loss L IoU , as shown in formula (6):

[0096]

[0097] Among them, x represents the value of IoU, L IOU-p represents the polynomial IoU loss, w i Indicates that the coefficient of the polynomial is 1 or 0, x i Represents a polynomial, and n is 6 in the present invention.

[0098] By adjusting the coefficients, we can adjust the attention of detection frames of different qualities. When the IoU is small, the loss is basically the same as the original YOLOX loss. When the IoU is large, the polynomial weight becomes relatively large, causing the network to focus more attention on high-quality detection frames, thereby obtaining better regression results.

[0099] See also Figure 5 By adjusting the polynomial weight coefficient, the polynomial weight increases with the increase of the IoU value, thereby relatively increasing the loss of the high IoU detection box on the basis of the original IoU loss.

[0100] See also Figure 6 ,By adjusting the polynomial weight coefficients, the monotonicity of the polynomial IoU loss is achieved, thereby ensuring that the optimal solution of IoU is obtained by minimizing the polynomial IoU loss.

[0101] YOLOX uses BCE Loss to calculate confidence loss. The penalty for each type of sample is the same, which is -log loss, and does not consider the sample imbalance problem.

[0102] In order to improve the detection effect of difficult samples, the present invention optimizes the loss function of YOLOX and sets the confidence L conf The error caused is replaced by Focal Loss.

[0103] The definition of Focal Loss is shown in formula (7):

[0104] FL(p t )=-α t (1-p t )γlog((pt ) (7)

[0105] (1-p t )γ is the adjustment factor, γ is the adjustable focusing parameter, p t is defined as:

[0106]

[0107] Where y∈{+1,-1}, p∈[0,1] is the predicted probability of the model when y = 1. Focal Loss is introduced to effectively alleviate the sample imbalance problem.

[0108] S5. Perform data enhancement such as Mosaic and Mixup on the constructed dataset, and train and test the improved YOLOX network to realize transmission line fault detection based on YOLOX.

[0109] The constructed transmission line fault detection dataset is enhanced by Mosaic and Mixup, and the improved YOLOX network is trained and tested. The model with the best detection effect is selected through experiments, thus realizing transmission line fault detection based on YOLOX.

[0110] The specific process of Mixup is as follows:

[0111] New training samples and labels are constructed for data of different fault categories by linear interpolation, that is, the two images are mixed in proportion and their labels are also mixed.

[0112] The specific process of Mosaic is:

[0113] Data augmentation is achieved by randomly cropping parts of four images and then stitching them together to form a single sample.

[0114] In another embodiment of the present invention, a transmission line fault detection system based on the YOLOX network is provided. The system can be used to implement the above-mentioned transmission line fault detection method based on the YOLOX network. Specifically, the transmission line fault detection system based on the YOLOX network includes a data module, an expansion module, a labeling module, a network module and a detection module.

[0115] Among them, the data module uses the transmission line fault images in real scenes as the original samples;

[0116] The expansion module expands the original samples obtained by the data module through translation, rotation, scene simulation, and fault simulation operations;

[0117] The labeling module uses LabelImg to label the original samples expanded by the expansion module, and organizes them into a transmission line fault detection dataset that includes five types of faults: foreign object hanging, insulator defect, bird's nest, loose strands, and icing;

[0118] Network module, builds a YOLOX-based transmission line fault detection network, and optimizes the YOLOX-based transmission line fault detection network using an adaptive multi-scale fusion mechanism based on dilated convolution and a loss function based on polynomial IoU;

[0119] The detection module performs data augmentation on the transmission line fault detection dataset constructed by the annotation module. It then trains the YOLOX-based transmission line fault detection network optimized by the network module and uses the trained YOLOX-based transmission line fault detection network to implement transmission line fault detection.

[0120] In another embodiment of the present invention, a terminal device is provided, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the transmission line fault detection method based on the YOLOX network, including:

[0121] The transmission line fault images in real scenarios are used as original samples; the original samples are expanded through translation, rotation, scene simulation, and fault simulation operations; LabelImg is used to annotate the expanded original samples, and a transmission line fault detection dataset containing five types of faults: foreign object hanging, insulator defect, bird's nest, loose strands, and icing is formed; a YOLOX-based transmission line fault detection network is built, and the YOLOX-based transmission line fault detection network is optimized using an adaptive multi-scale fusion mechanism based on void convolution and a loss function based on polynomial IoU; data enhancement is performed on the transmission line fault detection dataset, and then the optimized YOLOX-based transmission line fault detection network is trained, and the trained YOLOX-based transmission line fault detection network is used to realize transmission line fault detection.

[0122] In another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and, of course, the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory (Non-Volatile Memory), such as at least one disk memory.

[0123] The processor may load and execute one or more instructions stored in a computer-readable storage medium to implement the corresponding steps of the YOLOX network-based power transmission line fault detection method in the above embodiment; the processor may load and execute the following steps:

[0124] The transmission line fault images in real scenarios are used as original samples; the original samples are expanded through translation, rotation, scene simulation, and fault simulation operations; LabelImg is used to annotate the expanded original samples, and a transmission line fault detection dataset containing five types of faults: foreign object hanging, insulator defect, bird's nest, loose strands, and icing is formed; a YOLOX-based transmission line fault detection network is built, and the YOLOX-based transmission line fault detection network is optimized using an adaptive multi-scale fusion mechanism based on void convolution and a loss function based on polynomial IoU; data enhancement is performed on the transmission line fault detection dataset, and then the optimized YOLOX-based transmission line fault detection network is trained, and the trained YOLOX-based transmission line fault detection network is used to realize transmission line fault detection.

[0125] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0126] This paper collects a total of 3822 pieces of five types of fault data and conducts relevant experiments based on this data set.

[0127] The deep learning framework used in the experiment is PyTorch 1.7.1, the operating system is Windows 10, and the GPU is NVIDIA GeForce GTX 1070 Ti.

[0128] The YOLOX model was trained on the dataset proposed in this paper for 100 epochs and warmed up for 5 epochs. Stochastic gradient descent (SGD) was used for training, and a cosine learning mechanism was used.

[0129] The weight decay is 0.0005 and the SGD momentum is 0.9. Inference time is measured on a single GTX1070Ti.

[0130] The experimental results are shown in the table:

[0131] Table 1 Comparison between the improved YOLOX algorithm and the classic method

[0132]

[0133] From the experimental results in Table 1, we can see that the improved YOLOX network has a higher m〖AP〗_(50:95) than classic networks such as YOLOV5-S and YOLOV3-Tiny, and its inference time can also meet real-time requirements, which makes it suitable for application in the field of power transmission line fault detection.

[0134] See also Figure 7 ,From the detection results of the improved network, we can see that the ,improved network can better detect the hidden faults in the ,transmission lines.

[0135] To address the lack of publicly available standard datasets in the field of transmission line fault detection and the fact that most fault data fall into a single category, we collected a variety of fault data and compiled them into a transmission line fault detection dataset to provide data support for subsequent deep learning-based model training.

[0136] In response to the problem that fault targets in power scenarios are complex and changeable, and the detection effect is poor when fault targets of different scales coexist, an adaptive multi-scale fusion method based on void convolution is proposed. This method improves the feature fusion method of YOLOX and realizes the adaptive fusion of YOLOX multi-scale features. It fully utilizes features of different scales, filters out useless information in multi-scale information, and enhances useful information, thereby improving the detection effect.

[0137] In order to improve the network's optimization ability for high IoU detection frames, this paper proposes a new IoU loss function. The polynomial IoU loss designed by the present invention can enable the network to pay more attention to high-quality detection frames, thereby achieving high-precision positioning.

[0138] To alleviate the performance loss caused by sample imbalance, this paper introduces Focal Loss, which significantly improves detection accuracy and effectively enhances the accuracy of transmission line fault detection.

[0139] In summary, the present invention presents a YOLOX network-based power transmission line fault detection method and system. By improving YOLOX's multi-scale feature fusion method and loss function, the model's detection accuracy is effectively improved. Experimental results demonstrate that the proposed algorithm achieves a high mean average prediction (mAP) while ensuring real-time performance. It can be effectively applied in power transmission line fault detection scenarios, providing enhanced technical support for practical applications of power transmission line fault detection and intelligent line patrol.

[0140] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0141] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0142] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0143] In the embodiments provided by the present invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For example, the division of the modules or units is merely a logical functional division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interface, indirect coupling or communication connection of devices or units, and can be electrical, mechanical, or other forms.

[0144] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0145] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0146] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above-mentioned various method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0147] 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 produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. 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.

[0148] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 1a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0149] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device 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.

[0150] The above content is only for explaining the technical idea of the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A transmission line fault detection method based on YOLOX network, characterized in that: The following steps are involved: S1. Use the transmission line fault image in the real scene as the original sample; S2, expanding the original sample obtained in step S1; S3. Use LabelImg to label the original samples expanded in step S2 and form a transmission line fault detection dataset; S4. Build a YOLOX-based transmission line fault detection network, and optimize the YOLOX-based transmission line fault detection network using an adaptive multi-scale fusion mechanism based on dilated convolution and a loss function based on polynomial IoU. The loss function Specifically: in, is the polynomial IoU loss function, is the confidence loss function, is the category loss function; Confidence loss function for: in, is the regulating factor, is the adjustable focus parameter, The predicted probability output by the YOLOX network; Polynomial IoU loss for: in, is the value of IoU, The coefficients of the polynomial are 1 or 0. is a polynomial, is 6; S5. Perform data enhancement on the transmission line fault detection dataset constructed in step S3, then train the YOLOX-based transmission line fault detection network optimized in step S4, and use the trained YOLOX-based transmission line fault detection network to implement transmission line fault detection.

2. The power transmission line fault detection method based on the YOLOX network according to claim 1, characterized in that: In step S2, the original sample obtained in step S1 is expanded through translation, rotation, scene simulation, and fault simulation operations.

3. The power transmission line fault detection method based on the YOLOX network according to claim 2, characterized in that: The specific scenario simulation is as follows: By simulating scenarios where conductors are loose or broken, and capturing images of conductors at different angles, under different lighting conditions, and with varying degrees of damage, we can expand the data on transmission lines. The fault simulation is specifically as follows: using Photoshop software to simulate fault images of foreign objects hanging on the transmission line and insulator defects, and by splicing the fault images with the non-fault images, the fault data can be expanded.

4. The power transmission line fault detection method based on the YOLOX network according to claim 1, characterized in that: In step S3, the fault types of the transmission line fault detection dataset include hanging foreign objects, defective insulators, bird nests, loose strands, and ice coverage.

5. The power transmission line fault detection method based on YOLOX network according to claim 1, characterized in that: In step S5, Mosaic and Mixup are used to perform data enhancement on the transmission line fault detection dataset.

6. The power transmission line fault detection method based on the YOLOX network according to claim 5, characterized in that: The specific process of Mixup is as follows: New training samples and labels are constructed for data of different fault categories using linear interpolation. This means mixing two images proportionally and also mixing their labels. The specific process of Mosaic is: Randomly crop parts of four images and then stitch them together to form a sample to achieve data augmentation.

7. A power transmission line fault detection system based on YOLOX network, characterized in that: include: The data module uses the transmission line fault images in real scenarios as original samples; The expansion module expands the original samples obtained by the data module through translation, rotation, scene simulation, and fault simulation operations; The labeling module uses LabelImg to label the original samples expanded by the expansion module, and organizes them into a transmission line fault detection dataset that includes five types of faults: foreign object hanging, insulator defect, bird's nest, loose strands, and icing; Network module, build a YOLOX-based transmission line fault detection network, and optimize the YOLOX-based transmission line fault detection network using an adaptive multi-scale fusion mechanism based on void convolution and a loss function based on polynomial IoU. The loss function Specifically: in, is the polynomial IoU loss function, is the confidence loss function, is the category loss function; Confidence loss function for: in, is the regulating factor, is the adjustable focus parameter, The predicted probability output by the YOLOX network; Polynomial IoU loss for: in, is the value of IoU, The coefficients of the polynomial are 1 or 0. is a polynomial, is 6; The detection module performs data enhancement on the transmission line fault detection dataset constructed by the labeling module, and then trains the YOLOX-based transmission line fault detection network after optimization of the network module. The trained YOLOX-based transmission line fault detection network is used to realize transmission line fault detection.

Images