User side grounding wire hitching state supervision system and method based on machine vision
The user-side grounding wire connection status monitoring system uses machine vision and convolutional neural networks to automate and enhance the accuracy of grounding wire connection detection, addressing safety risks by reducing human error and ensuring correct connections.
Patent Information
- Application Number
- CN202311619631.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-29
- Publication Date
- 2024-12-31
AI Technical Summary
Current methods for monitoring the connection status of user-side grounding wires in power grids are unreliable and prone to human error, posing safety risks due to missing or incorrect connections, which are not adequately addressed by existing substation-side monitoring technologies.
A user-side grounding wire connection status monitoring system using machine vision, comprising an image acquisition device with a high-definition camera, support tube, and wireless communication, coupled with a convolutional neural network for image processing, to detect and guide correct connection of grounding wires.
Ensures accurate and automated detection of grounding wire connections, reducing human error and enhancing safety by promptly alerting operators to incorrect or missing connections, thereby ensuring personal and equipment safety.
Smart Images
Figure CN119228719A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of equipment status monitoring, and in particular to a system and method for supervising the hanging status of a user-side grounding wire based on machine vision. Background Art
[0002] The grounding wire is one of the essential safety tools in the process of power line and equipment maintenance. As the power grid structure becomes increasingly complex, there are more and more line segmentation, branching, and same-pole (tower) erection. During the construction, maintenance and accident handling of power lines, the correct use of the grounding wire can effectively prevent the occurrence of reverse power transmission, induced electricity, etc., which plays an extremely important role in maintaining the stable operation of the power grid and protecting the personal safety of the staff.
[0003] At present, the monitoring of the grounding wire connection status is mainly designed for the substation side. There are three solutions. One method is to set a pressure touch sensor or positioning signal in the wiring device to send the connection status and layout information of the grounding wire to the staff; one is to monitor the status data of the line through the line protection device and judge the status of the grounding device through the analysis of the server; one is to set a monitoring camera device in the substation to monitor the connection status of the grounding wire in real time. These three methods are not applicable to the user-side substation. The user-side grounding wire is only connected when necessary, the number of connections is small, and the cost of monitoring needs to be considered. At present, the supervision and management of the user-side grounding wire still adopts traditional telephone reporting, text records and other methods. There are many unreliable factors and high control risks. It is easy to cause malicious misoperation accidents such as missing or wrongly connecting the grounding wire or sending power with the grounding wire, which cannot meet the current safety production requirements. Therefore, it is necessary to propose a monitoring method for the connection status of the user-side grounding wire to detect and guide the connection status of the user-side grounding wire. Summary of the invention
[0004] The purpose of the present invention is to disclose a user-side grounding wire hanging status monitoring system based on machine vision, which captures the induced current through a non-contact detection device to monitor the user-side grounding wire hanging status, thereby solving the limitations of the user-side grounding wire hanging status monitoring in the prior art.
[0005] Another object of the present invention is to provide a method for monitoring the connection status of a user-side ground wire based on machine vision.
[0006] The technical solution of the present invention is: A user-side ground wire connection status monitoring system based on machine vision, comprising a working layer, a transmission layer and a processing layer; The working layer includes an image acquisition device and a communication terminal; the image acquisition device is used to obtain the hanging state information of the grounding wire and monitor the hanging state of the grounding wire; the communication terminal is used to feedback the state information of the grounding wire and guide the user to operate; The transmission layer includes a wireless communication device for transmitting the acquired hooking status information of the ground wire to the image information processing system, and feeding back the recognized processed information to the communication terminal; The processing layer includes an image information processing system, which is used for image processing and information feature extraction, and compares and analyzes the image information after the grounding wire is hung and removed with the collected standard image information to determine the hanging status of the grounding wire.
[0007] Furthermore, the image acquisition device includes a camera and a support tube, and the camera is installed on the support tube; the camera is composed of a lens group, a voice coil motor, a wireless communication device, a sensor, a circuit board and a 360-degree rotating ball camera connected in sequence, and the picture obtained by the camera is transmitted to the image information processing system via the wireless communication device, and the image features are extracted based on the convolutional neural network.
[0008] A method for monitoring the connection status of a user-side ground wire based on machine vision using the system as described above comprises the following steps: Step 1: Setting the image acquisition device and ground wire mounting position and properties; Step 2: Obtain the line-side image of the user's substation line switch, and transmit the image information to the image information processing system through the wireless network; Step 3: Extract image features based on convolutional neural network; Step 4: Compare and analyze the image information after the ground wire is hung and removed with the standard image information; Step 5: Determine the connection status of the grounding wire; if the image features collected by the grounding wire are within the set standard threshold, the grounding wire is assessed to be correctly connected or removed, otherwise, an alarm is issued to indicate that the grounding wire is missed or wrongly connected, and the grounding wire is re-connected, and then re-tested according to the steps.
[0009] In the step 1, the grounding wire is hung on the line side of the line switch, and the grounding rod is hung vertically downward at a distance of 0.5-1.5 m from the grounding switch; the image acquisition device is set at least three positions based on the grounding wire for image acquisition, and the three positions of the image acquisition device are respectively 5-7 m horizontally on both sides of the grounding wire and 0.5-0.8 m directly opposite to the grounding wire clamp, and the horizontal angle α between the image acquisition device and the grounding wire is 10°-30°, the elevation angle between the image acquisition device and the grounding wire is 45°-60°, and the vertical height of the image acquisition device from the ground is 1.5-2 m.
[0010] In the step 2, firstly, an image is acquired at the standard position where the grounding wire is hung and the image acquisition device is arranged, that is, a standard image; then, an image of the actual position of the grounding wire on the user side is acquired. The image acquisition device needs to acquire at least two pictures at each arrangement position, and transmit the image information to the image information processing system via a wireless network.
[0011] Furthermore, first, the grounding wire is hung at a standard position and a standard image is obtained. The grounding wire is hung at 0.5 m, 1.5 m, and 1 m from the center of the grounding switch. The image acquisition device is set at 5.5 m, 7 m from the left side of the grounding wire, 5.5 m, 7 m from the right side of the grounding wire, and 0.75 m directly opposite the grounding wire clamp. The height is 1.5 m and 2 m vertically from the ground. The high-definition camera has an elevation angle of 60° and 30° to obtain image information, that is, standard image information; Then, actual image acquisition is carried out on the user side. The image acquisition device is arranged between 5.5-7 m to the left of the actual location of the grounding wire, between 5.5-7 m to the right of the grounding wire, and 0.75 m opposite to the clamp position of the grounding wire. The vertical height of the image acquisition device from the ground is 1.5-2 m. The high-definition camera is automatically positioned to obtain image information.
[0012] Furthermore, the automatic positioning uses a correction algorithm based on contour extraction to obtain the rotation angle, and the angle error with the standard image is set to 1%.
[0013] Furthermore, the correction algorithm based on contour extraction includes: graying the image, threshold binarization, detecting contours, finding the contour's enclosing matrix, and obtaining the angle to give the camera rotation angle; Image grayscale: convert color images into grayscale images; Threshold binarization: Use a fixed threshold to perform binarization processing to achieve image segmentation; Detect contours: extract the contour of the target image and calculate the normal angle of each contour point, and then perform histogram statistics on the normal angles of the contour points; Find the bounding matrix of the contour and obtain the angle: The rotation angle of the two images can be quickly estimated by performing circular correlation calculation on the normal angle histograms of the contour points of the two images.
[0014] Furthermore, when determining the hanging position of the grounding wire, if the hanging position of the grounding wire does not meet the requirements, an alarm is given and the grounding wire is re-hanged; if the hanging position of the grounding wire meets the requirements, the position of the image acquisition device is determined again. If the arrangement of the image acquisition device does not meet the requirements, the position is re-selected. If the arrangement of the image acquisition device meets the requirements, the image deviation state information is extracted; the image rotation angle is calculated. If it does not meet the requirements, the camera is rotated and the line side image of the user substation line switch is acquired again. If it meets the requirements, image feature extraction is performed.
[0015] The step 3 comprises the following specific steps: Step 3.1: Extract standard image features based on convolutional neural network; Firstly, the collected ground wire hanging images are processed in advance by using digital image processing technology, and the color images collected on site are converted into grayscale images to enhance the contrast between image features and image background. Then, the noise in the image is reduced to make the image clearer, and grayscale features, color features and morphological features are selected for feature extraction. Then, the judgment standard of the ground wire hanging state is set as the images collected from the left, middle and right directions are all within the threshold range. The sample data is divided into three parts, and each part needs to be compared separately; the training set, validation set and test set are divided separately to ensure the measurement accuracy; Finally, a convolutional neural network model is established. The network structure includes convolutional layers, pooling layers, fully connected layers and soft-max classification layers. Soft-max is a multi-class classifier. Its output is an eigenvalue matrix between 0 and n. The transfer function in the network uses ReLU, and the stochastic gradient descent method is used to minimize the loss function to obtain the network parameters to obtain the network model. The network contains 3 convolutions, each with 1 convolution layer. Each convolution is followed by a maximum pooling layer to reduce the size and enhance the features, followed by a fully connected layer, and finally connected to the soft-max classification layer. The network will extract some edge features and texture features when the ground wire is hung. After multi-layer learning, the low-level features are converted into high-level features, the trained parameters are saved, and the trained model is tested using the test set.
[0016] Step 3.2: Setting the image information threshold; Step 3.3: User-side image extraction based on convolutional neural network; Firstly, two images of the left, middle and right positions of the ground wire hanging state are obtained for image preprocessing. Then, feature extraction of the images is performed based on the convolutional neural network.
[0017] The advantages and effects of the present invention are as follows: The system uses machine vision-based image recognition technology to implement the connection guidance and status detection of the grounding wire of the user-side substation. The image acquisition device collects and compares the images of the standard connection position and the actual connection position of the grounding wire. When the user does not use the grounding wire correctly according to the dispatcher's operating instructions, the relevant dispatcher will be promptly prompted to avoid accidents such as missing or wrong connection of the grounding wire, further ensuring the personal safety of maintenance personnel, the safety of equipment and the production safety of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 It is a flowchart of the supervision method of the present invention; Figure 2 It is a structural schematic diagram of the image acquisition device of the present invention; Figure 3 is a flow chart of a correction algorithm based on contour extraction of the present invention; Figure 4 It is a flow chart of the process of extracting standard image features based on convolutional neural network of the present invention; Figure 5 It is a schematic diagram of a convolutional neural network model of the present invention; Figure 6 is a schematic diagram of the threshold peak-valley comparison of the present invention; Figure 7 is a comparative schematic diagram of determining the hanging state of the ground wire of the present invention; Figure 8 It is a schematic block diagram of the supervision system structure of the present invention.
[0019] Fig. 9 It is a schematic diagram of the arrangement positions of the ground wire and the image acquisition device of the present invention.
[0020] in, Figure 2 Among them, 1. lens group, 2. voice coil motor, 3. wireless communication device, 4. sensor, 5. circuit board, 6. 360-degree rotating ball machine, 7. support tube; Fig. 9 Among them, 8. transmission line, 9. ground wire, 10. image acquisition device. DETAILED DESCRIPTION Example
[0021] like Figure 8 As shown, a user-side grounding wire connection status monitoring system based on machine vision is used. The system uses image recognition technology based on machine vision to achieve connection guidance and status detection of the user-side substation grounding wire. The entire system consists of a working layer, a transmission layer, and a processing layer. The working layer includes an image acquisition device and a communication terminal; the image acquisition device is used to obtain the hanging state information of the grounding wire and monitor the hanging state of the grounding wire; the communication terminal is used to feedback the state information of the grounding wire and guide the user to operate; The transmission layer includes a wireless communication device for transmitting the acquired hooking status information of the ground wire to the image information processing system, and feeding back the recognized processed information to the communication terminal; The processing layer includes an image information processing system, which is used for image processing and information feature extraction, and compares and analyzes the image information after the grounding wire is hung and removed with the collected standard image information to determine the hanging status of the grounding wire.
[0022] The image acquisition device is as follows Figure 2 As shown, it is composed of a high-definition camera and a support tube 7. The high-definition camera is composed of a lens group 1, a voice coil motor 2, a wireless communication device 3, a sensor 4, a circuit board 5 and a 360-degree rotating ball camera 6 connected in sequence. The picture obtained by the camera is transmitted to the image information processing system via the wireless communication device, and the image features are extracted based on the convolutional neural network. The image acquisition device can also use other commercially available products with image acquisition functions.
[0023] like Figure 1 As shown, a method for monitoring the connection status of a user-side ground wire based on machine vision using the system as described above comprises the following steps: Step 1: Setting the image acquisition device and ground wire mounting position and properties; Take 66kV transmission line as an example. Fig. 9 As shown, the grounding wire is hung on the line side of the line switch, and the grounding rod is hung vertically downward at a distance of 0.5-1.5 m from the grounding switch; the image acquisition device is set at least three positions based on the grounding wire for image acquisition, and the three positions of the image acquisition device are respectively 5-7 m horizontally on both sides of the grounding wire and 0.5-0.8 m directly opposite to the grounding wire clamp, and the horizontal angle α between the image acquisition device and the grounding wire is 10°-30°, the elevation angle β between the image acquisition device and the grounding wire is 45°-60°, and the vertical height of the image acquisition device from the ground is 1.5-2 m.
[0024] Step 2: Obtain the line-side image of the user's substation line switch, and transmit the image information to the image information processing system through the wireless network; The image is obtained at the standard position where the ground wire is hung and the image acquisition device is arranged, which is the standard image; then the image of the actual position of the ground wire on the user side is collected. The image acquisition device needs to obtain at least two pictures at each arranged position, and transmit the image information to the image information processing system through the wireless network, such as Figure 3 As shown; First, the grounding wire is hung at a standard position and a standard image is obtained. The grounding wire is hung at 0.5 m, 1.5 m, and 1 m from the center of the grounding switch. The image acquisition device is set at 5.5 m, 7 m on the left side of the grounding wire, 5.5 m, 7 m on the right side of the grounding wire, and 0.75 m opposite to the grounding wire clamp. The height is 1.5 m and 2 m vertically from the ground. The high-definition camera has an elevation angle of 60° and 30° to obtain image information, that is, standard image information. Then, actual image acquisition is carried out on the user side. The image acquisition device is arranged between 5.5-7 m to the left of the actual location of the grounding wire, between 5.5-7 m to the right of the grounding wire, and 0.75 m opposite to the clamp position of the grounding wire. The vertical height of the image acquisition device from the ground is 1.5-2 m. The high-definition camera is automatically positioned to obtain image information.
[0025] Among them, the automatic positioning adopts a correction algorithm based on contour extraction to obtain the rotation angle, and the angle error with the standard image is set to 1%; the correction algorithm based on contour extraction includes: image grayscale, threshold binarization, contour detection, finding the contour enclosing matrix, and obtaining the angle to give the camera rotation angle; Finally, the grounding wire hanging position is determined. If the grounding wire hanging position does not meet the requirements, an alarm is sounded and the grounding wire is re-hanged; if the grounding wire hanging position meets the requirements, the image acquisition device position is determined again. If the arrangement of the image acquisition device does not meet the requirements, the position is re-selected. If the arrangement of the image acquisition device meets the requirements, the image deviation state information is extracted; the image rotation angle is calculated. If it does not meet the requirements, the camera is rotated and the line side image of the user substation line switch is acquired again. If it meets the requirements, image feature extraction is performed; Step 3: Extract image features based on convolutional neural network; Step 3.1: Extract standard image features based on convolutional neural network The process of extracting standard image features based on convolutional neural networks is as follows: Figure 4 shown.
[0026] First, the collected grounding wire hanging images are processed in advance using digital image processing technology. The color images collected on site are converted into grayscale images to enhance the contrast between image features and image background. Then the noise in the image is reduced to make the image clearer. Grayscale features, color features, and morphological features are selected for feature extraction.
[0027] Then, the judgment standard for the grounding wire hanging status is set as the images collected from the left, middle and right directions are all within the threshold range. The sample data is divided into three parts, and each part needs to be compared separately; the measurement accuracy is guaranteed by dividing the training set, validation set and test set respectively.
[0028] Finally, a convolutional neural network model is established, such as Figure 5 As shown in the figure. The network structure includes convolution layer, pooling layer, fully connected layer and soft-max classification layer, where soft-max is a multi-class classifier, and its output is an eigenvalue matrix between 0 and n. The transfer function in the network uses ReLU, and the stochastic gradient descent method is used to minimize the loss function to obtain the network parameters and obtain the network model. The network contains 3 convolutions, each of which has 1 convolution layer. Each convolution is followed by a maximum pooling layer to reduce the size and enhance the features, followed by a fully connected layer, and finally connected to the soft-max classification layer. The network will extract some edge features and texture features when the ground wire is hung. After multi-layer learning, the low-level features are converted into high-level features, the parameters obtained by training are saved, and the trained model is tested using the test set.
[0029] Step 3.2: Set the image information threshold, such as Figure 6 shown.
[0030] Step 3.3: User-side image extraction based on convolutional neural network; Firstly, two images of the left, middle and right positions of the ground wire are acquired and preprocessed. Then, the features of the images are extracted based on the convolutional neural network. Step 4: Compare and analyze the image information after the ground wire is hung and removed with the standard image information; Compare the features of two pictures in the left, middle and right directions with the features of the standard picture. Step 5: Determine the connection status of the ground wire.
[0031] like Figure 7 If the image features collected by the grounding wire are within the threshold of the set standard, the grounding wire is assessed to be correctly hung or removed. Otherwise, an alarm is issued to indicate that the grounding wire is missed or wrongly hung, and the grounding wire is re-hanged, and then re-tested according to the steps.
[0032] It is known from common technical knowledge that the present invention can be implemented by other embodiments that do not deviate from its spirit or essential features. Therefore, the above disclosed embodiments are only illustrative in all respects and are not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are included in the present invention.
Claims
1. A user-side ground wire connection status monitoring system based on machine vision, characterized in that Includes working layer, transmission layer and processing layer; The working layer includes an image acquisition device and a communication terminal; the image acquisition device is used to obtain the hanging state information of the grounding wire and monitor the hanging state of the grounding wire; the communication terminal is used to feedback the state information of the grounding wire and guide the user to operate; The transmission layer includes a wireless communication device for transmitting the acquired hooking status information of the ground wire to the image information processing system, and feeding back the recognized processed information to the communication terminal; The processing layer includes an image information processing system, which is used for image processing and information feature extraction, and compares and analyzes the image information after the grounding wire is hung and removed with the collected standard image information to determine the hanging status of the grounding wire.
2. According to a machine vision-based user-side ground wire connection status monitoring system according to claim 1, it is characterized in that: The image acquisition device includes a camera and a support tube, and the camera is installed on the support tube; the camera is composed of a lens group, a voice coil motor, a wireless communication device, a sensor, a circuit board and a 360-degree rotating ball camera connected in sequence, and the picture obtained by the camera is transmitted to the image information processing system via the wireless communication device, and the image features are extracted based on the convolutional neural network.
3. According to claim 1, a monitoring method for a user-side ground wire connection status monitoring system based on machine vision is characterized in that The steps include: Step 1: Setting the image acquisition device and ground wire mounting position and properties; Step 2: Obtain the line-side image of the user's substation line switch, and transmit the image information to the image information processing system through the wireless network; Step 3: Extract image features based on convolutional neural network; Step 4: Compare and analyze the image information after the ground wire is hung and removed with the standard image information; Step 5: Determine the connection status of the grounding wire; if the image features collected by the grounding wire are within the set standard threshold, the grounding wire is assessed to be correctly connected or removed, otherwise, an alarm is issued to indicate that the grounding wire is missed or wrongly connected, and the grounding wire is re-connected, and then re-tested according to the steps.
4. According to claim 3, a method for monitoring the user-side ground wire connection status based on machine vision is characterized in that In the step 1, the grounding wire is hung on the line side of the line switch, and the grounding rod is hung vertically downward at a distance of 0.5-1.5 m from the grounding switch; the image acquisition device is set at least three positions based on the grounding wire for image acquisition, and the three positions of the image acquisition device are respectively 5-7 m horizontally on both sides of the grounding wire and 0.5-0.8 m directly opposite to the grounding wire clamp, and the horizontal angle α between the image acquisition device and the grounding wire is 10°-30°, the elevation angle between the image acquisition device and the grounding wire is 45°-60°, and the vertical height of the image acquisition device from the ground is 1.5-2 m.
5. According to claim 3, a method for monitoring the user-side ground wire connection status based on machine vision is characterized in that In the step 2, firstly, an image is acquired at the standard position where the grounding wire is hung and the image acquisition device is arranged, that is, a standard image; then, an image of the actual position of the grounding wire on the user side is acquired. The image acquisition device needs to acquire at least two pictures at each arrangement position, and transmit the image information to the image information processing system via a wireless network.
6. According to a method for monitoring the user-side ground wire connection status based on machine vision according to claim 5, it is characterized in that: First, the grounding wire is hung at a standard position and a standard image is obtained. The grounding wire is hung at 0.5 m, 1.5 m, and 1 m from the center of the grounding switch. The image acquisition device is set at 5.5 m, 7 m on the left side of the grounding wire, 5.5 m, 7 m on the right side of the grounding wire, and 0.75 m opposite to the grounding wire clamp. The height is 1.5 m and 2 m vertically from the ground. The high-definition camera has an elevation angle of 60° and 30° to obtain image information, that is, standard image information. Then, actual image acquisition is carried out on the user side. The image acquisition device is arranged between 5.5-7 m to the left of the actual location of the grounding wire, between 5.5-7 m to the right of the grounding wire, and 0.75 m opposite to the clamp position of the grounding wire. The vertical height of the image acquisition device from the ground is 1.5-2 m. The high-definition camera is automatically positioned to obtain image information.
7. A method for monitoring the user-side ground wire connection status based on machine vision according to claim 6, characterized in that: Automatic positioning uses a correction algorithm based on contour extraction to obtain the rotation angle, and the angle error with the standard image is set to 1%.
8. The method for monitoring the user-side ground wire connection status based on machine vision according to claim 7 is characterized in that: The correction algorithm based on contour extraction includes: graying the image, threshold binarization, detecting contours, finding the contour's enclosing matrix, and obtaining the angle to give the camera rotation angle; Image grayscale: convert color images into grayscale images; Threshold binarization: Use a fixed threshold to perform binarization processing to achieve image segmentation; Detect contours: extract the contour of the target image and calculate the normal angle of each contour point, and then perform histogram statistics on the normal angles of the contour points; Find the bounding matrix of the contour and obtain the angle: The rotation angle of the two images can be quickly estimated by performing circular correlation calculation on the normal angle histograms of the contour points of the two images.
9. The method for monitoring the user-side ground wire connection status based on machine vision according to claim 5 is characterized in that: When determining the hanging position of the grounding wire, if the hanging position of the grounding wire does not meet the requirements, an alarm is given and the grounding wire is re-hanged; if the hanging position of the grounding wire meets the requirements, the position of the image acquisition device is determined again. If the arrangement of the image acquisition device does not meet the requirements, the position is re-selected. If the arrangement of the image acquisition device meets the requirements, the image deviation state information is extracted; the image rotation angle is calculated. If it does not meet the requirements, the camera is rotated and the line side image of the user substation line switch is acquired again. If it meets the requirements, image feature extraction is performed.
10. The method for monitoring the user-side ground wire connection status based on machine vision according to claim 3 is characterized in that The step 3 comprises the following specific steps: Step 3.1: Extract standard image features based on convolutional neural network; Firstly, the collected ground wire hanging images are processed in advance by using digital image processing technology, and the color images collected on site are converted into grayscale images to enhance the contrast between image features and image background. Then, the noise in the image is reduced to make the image clearer, and grayscale features, color features and morphological features are selected for feature extraction. Then, the judgment standard of the ground wire hanging state is set as the images collected from the left, middle and right directions are all within the threshold range. The sample data is divided into three parts, and each part needs to be compared separately; the training set, validation set and test set are divided separately to ensure the measurement accuracy; Finally, a convolutional neural network model is established. The network structure includes convolution layer, pooling layer, fully connected layer and soft-max classification layer. Soft-max is a multi-class classifier. Its output is an eigenvalue matrix between 0 and n. The transfer function in the network uses ReLU. The stochastic gradient descent method is used to minimize the loss function to obtain the network parameters to obtain the network model. The network contains 3 convolutions, each with 1 convolution layer. Each convolution is followed by a maximum pooling layer to reduce the size and enhance the features, followed by a fully connected layer, and finally connected to the soft-max classification layer. The network will extract some edge features and texture features when the ground wire is hung. After multi-layer learning, the low-level features are converted into high-level features, the parameters obtained by training are saved, and the trained model is tested using the test set. Step 3.2: Setting the image information threshold; Step 3.3: User-side image extraction based on convolutional neural network; Firstly, two images of the left, middle and right positions of the ground wire hanging state are obtained for image preprocessing. Then, feature extraction of the images is performed based on the convolutional neural network.