A transmission line drone intelligent inspection system and method based on deep learning

Abstract

The present invention discloses a power transmission line UAV intelligent inspection system and method based on deep learning, and relates to the technical field of power transmission line inspection. The present invention performs electromagnetic environment modeling and path optimization, identifies high-interference sections and adjusts the flight path and altitude accordingly, reduces the impact of strong electromagnetic interference on sensors and communication systems, automatically optimizes the resolution and focusing parameters of the image when an abnormal temperature area is detected, obtains clearer high-resolution images, and focuses on the abnormal temperature area at the same time. After the abnormal temperature area is detected, the shooting distance and angle are automatically adjusted, and key components are photographed at close range and at multiple angles, so that tiny abnormalities on the surface of the components can be captured, and the detection accuracy of early faults can be improved. After the inspection task is completed, the multimodal data is uploaded to the operation platform, the data is analyzed and archived, and an inspection report containing temperature anomalies, fault locations and electromagnetic environment characteristics is generated, saving manual analysis time.

Description

Technical Field

[0001] The present invention relates to the technical field of power transmission line inspection, and in particular to a power transmission line unmanned aerial vehicle intelligent inspection system and method based on deep learning. Background Art

[0002] At present, UAV inspection technology for transmission lines is widely used in various complex terrains, and has significantly improved inspection efficiency in high-altitude and mountainous areas; however, in actual applications, this technology still faces some challenges that are difficult to avoid, mainly in complex electromagnetic environments and high-temperature fault point detection; the strong electromagnetic fields around transmission lines will interfere with the sensors and communication systems of drones, resulting in unstable data collection and affecting the continuity and safety of inspections; in addition, drones need to identify high-temperature fault points on the line during the inspection process to determine early fault warnings; however, traditional methods mostly rely on infrared imaging for temperature anomaly detection, but the accuracy of infrared images is greatly affected by distance and lighting conditions. It is difficult to focus at long distances or under high light interference, and is prone to missed detections and false detections.

[0003] Although the existing inspection system can optimize data collection and high-temperature detection in electromagnetic interference environments by increasing the resolution of infrared images and repeating shooting within a certain distance, it has not been able to completely overcome the impact of complex environments on image quality and detection accuracy, resulting in difficulty in ensuring the inspection quality of drones in strong electromagnetic areas, and reducing the effectiveness of identifying high-temperature fault points in different environments. Therefore, a deep learning-based transmission line drone intelligent inspection system and method is urgently needed to solve such problems. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] The present invention provides a power transmission line UAV intelligent inspection system and method based on deep learning to solve the problem of temperature anomaly detection relying on infrared imaging. However, the accuracy of infrared images is greatly affected by distance and lighting conditions. It is difficult to focus at long distances or under high light interference, and is prone to missed detections and false detections.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] On the one hand, the present invention provides a transmission line UAV intelligent inspection system based on deep learning, which includes:

[0008] The data acquisition module uses drone inspections to obtain environmental information and collects historical data from the power system. The environmental information and historical data include: electromagnetic interference sections and high-temperature fault point information; based on the GIS system and satellite remote sensing data, environmental maps are obtained, including line topography, direction and key node information;

[0009] The electromagnetic environment modeling module analyzes the electromagnetic environment based on the information provided by the data acquisition module and generates an electromagnetic environment model; it identifies high-interference sections based on the electromagnetic field strength in different areas;

[0010] The dynamic electromagnetic interference detection module monitors the electromagnetic environment around the drone in real time during the inspection task. It monitors the electromagnetic intensity based on sensors and identifies interference in a timely manner. When the interference intensity reaches the threshold, it triggers the adaptive filtering algorithm and dynamically adjusts the filtering parameters.

[0011] The multimodal data acquisition module collects multimodal data during the inspection process, uses a deep learning model to fuse different data sources, and adjusts image resolution and focus parameters for temperature abnormality areas;

[0012] High temperature fault point identification module, which identifies the abnormal temperature area, adjusts the shooting angle and distance, and obtains the fault point image;

[0013] After completing the inspection task, the operation platform analyzes and archives the collected data and generates a fault report.

[0014] On the other hand, the present invention provides a transmission line UAV intelligent inspection method based on deep learning, comprising:

[0015] Step S1, electromagnetic environment modeling and path planning,

[0016] Before the drone performs inspection tasks, it conducts a preliminary analysis of the electromagnetic environment of the target line based on historical inspection data and environmental maps to generate an electromagnetic environment model;

[0017] Step S2: Dynamic electromagnetic interference detection and path adjustment.

[0018] Perform inspection tasks based on the inspection route of step S1, and monitor the surrounding electromagnetic environment in real time during the inspection;

[0019] In step S2, when the monitored electromagnetic interference intensity reaches the threshold, the adaptive filtering algorithm is triggered to dynamically adjust the filtering parameters; if the interference continues to increase, the path is fine-tuned to guide the drone to a low-interference area;

[0020] Step S3, multimodal data fusion and image optimization,

[0021] During the inspection process, the drone collects visible light, infrared images and lidar point cloud data, and uses a deep learning model for multimodal data fusion;

[0022] Step S4, high temperature fault point identification,

[0023] In the temperature anomaly area detection in step S3, the drone automatically adjusts the shooting distance and angle to take close-up and multi-angle photos of key nodes.

[0024] Furthermore, in step S1, historical inspection data including records of electromagnetic interference and high temperature fault points are obtained based on previous drone inspections and the power system; environmental map data including terrain, line direction and key node information are obtained based on the GIS system and satellite remote sensing data, and the key nodes include insulators and connection points;

[0025] In step S1, the electromagnetic environment model identifies high-interference sections according to the electromagnetic field strength in different areas. Based on the identified high-interference sections, the drone adjusts the inspection path and flight altitude and pre-sets the inspection route.

[0026] Furthermore, in step S1, the electromagnetic environment model is constructed as follows:

[0027] Define historical inspection data as ,in, represents the spatial coordinate position, For location The electromagnetic field strength value, is the temperature value, is the total number of data points;

[0028] Define the environment map data as ,in, represents the spatial coordinate position, is the terrain height, is the distance from the node, is the total number of map data points;

[0029] Based on electromagnetic field strength and spatial coordinates Constructing electromagnetic field distribution model:

[0030] ,in, Indicates location The estimated value of the electromagnetic field strength is It is a parameter that controls the attenuation rate of the electromagnetic field intensity. It estimates the electromagnetic field intensity at different locations by smoothing the historical inspection data based on the Gaussian kernel function.

[0031] Setting the electromagnetic field strength threshold , define the high interference area as satisfying Calculate the interference risk weight of each area: ,in, Indicates location The interference risk weight is is the maximum electromagnetic field strength at all locations;

[0032] Adjust the inspection path of the drone according to the high interference section and define the inspection path optimization objective function for:

[0033] ,in, is the total number of path points, is the path point location, is the flight altitude of the waypoint, For the recommended safe flight altitude, is the trade-off coefficient between flight altitude and interference risk, and the goal is to minimize Find the optimal inspection path, avoid high interference areas and stay at a safe height.

[0034] Furthermore, in step S2, the drone performs the inspection task based on the inspection route generated in step S1 and monitors the electromagnetic environment in real time:

[0035] The inspection route is the optimal inspection route generated from step S1, and is set as the path point set During the inspection process, the drone flies point by point along the preset path and arrives at each location in turn. ,in is the total number of path points;

[0036] Set each path point The electromagnetic field strength is , when monitored Exceeding the set threshold When the adaptive filtering algorithm is triggered to reduce electromagnetic interference, the filtering parameters of the adaptive filter are Dynamic adjustment based on current electromagnetic field strength: ,in, For the waypoint The adaptive filter parameters are and are the minimum and maximum filter coefficients of the filter, respectively. Indicates the maximum electromagnetic field strength value measured in the inspection area;

[0037] like Continuously exceeding the threshold , then fine-tune the path to guide the drone to a low-interference area. The optimization objective function of path fine-tuning is Defined as: ,in, For waypoints The optimization target value of is the distance from the current path point to the low interference area, is the adjustment coefficient, which is used to balance the electromagnetic intensity and path distance. The minimum value of selects the next path point position and guides the UAV to gradually fly to the low-interference area.

[0038] Furthermore, in step S3, temperature abnormality area detection is performed, and a deep learning model is used to automatically optimize image resolution and focusing parameters for the temperature abnormality area.

[0039] Furthermore, in step S3, multimodal data fusion and deep learning models are used to identify temperature abnormality areas in real time, and dynamically adjust image resolution and focus parameters. Specifically:

[0040] At each waypoint The drone uses infrared sensors to collect temperature data , the temperature anomaly detection formula is defined as: ,in, For waypoints The temperature value at is the temperature abnormality threshold, if Exceed , then this point is regarded as a temperature anomaly point;

[0041] At the temperature anomaly At, collect multimodal data to represent visible light images , infrared images and LiDAR point cloud data , use the deep learning model to perform data fusion and calculate the fused image features: ,in, For waypoints The fusion characteristics of is the fusion function of the deep learning model;

[0042] For areas with abnormal temperature, image resolution Perform dynamic adjustments and control the inspection time at the same time. The resolution adjustment formula is: ,in, Temperature anomaly point Optimized resolution, is the base resolution, is the scaling factor of temperature anomaly, which is used to improve the resolution of high temperature areas;

[0043] Use a deep learning model to adjust image focus parameters: ,in, is the focusing parameter of the temperature anomaly point, is the basic focus value, is the adjustment coefficient.

[0044] Furthermore, in step S4, after the inspection is completed, the data is automatically uploaded to the operation platform. The platform uses image recognition and classification algorithms to analyze the image, archive and annotate the abnormal position and fault image; and finally generates an inspection report containing temperature anomalies, fault point location information and electromagnetic environment characteristics.

[0045] Furthermore, in step S4, after the abnormal temperature area is detected, the drone automatically adjusts the shooting distance and angle to perform close-range and multi-angle shooting:

[0046] Assume the position of the key node is , the current position of the drone is , the shooting distance is set to , the angle is set to and , respectively represent the pitch angle and azimuth angle, adjust the shooting distance To achieve the best shooting clarity: ,in, Indicates the distance from the drone to the key node, is the coordinate position of the key node, is the current position of the drone;

[0047] Optimize the angle and adjust the shooting pitch angle and azimuth :

[0048] ,

[0049] ,in, The pitch angle of the shooting is used to adjust the vertical angle of the drone. It is the shooting azimuth, which is used to adjust the horizontal angle of the drone.

[0050] Further, in step S4, the data uploading and analysis method is:

[0051] After the inspection is completed, the drone will collect the image data , Temperature data and electromagnetic intensity data Upload to the job platform and use the deep learning model Identify and classify image data to generate fault location annotations and type tags :

[0052] ,in, is the detected abnormal position, is the fault type label, Represents a deep learning model for image recognition and classification. Select a combination of convolutional neural network and recurrent neural network, YOLO target detection model, FasterRCNN target detection model and fault recognition model based on attention mechanism;

[0053] Generate an inspection report containing temperature anomalies, fault location and electromagnetic environment information. Defined as: ,in, It is an inspection report, including fault location, type, temperature and electromagnetic field strength information. Represents the total number of collected data points.

[0054] The beneficial effects of the present invention are:

[0055] The present invention, through electromagnetic environment modeling and path optimization, the UAV can identify high-interference sections and adjust the flight path and altitude accordingly, avoiding or reducing the impact of strong electromagnetic interference on sensors and communication systems. At the same time, by fusing data through infrared, visible light and lidar, the image resolution and focusing parameters are automatically optimized when an abnormal temperature area is detected to obtain clearer high-resolution images. The deep learning model is used to process multimodal data and focus on the abnormal temperature area, making the identification of high-temperature fault points more accurate.

[0056] According to the present invention, after detecting an abnormal temperature area, the drone automatically adjusts the shooting distance and angle, and shoots key components at close range and at multiple angles, so as to capture minute abnormalities on the surface of components and improve the detection accuracy of early faults. After the inspection task is completed, the multimodal data is uploaded to the operation platform, and the data is analyzed and archived through image recognition and classification algorithms, and an inspection report containing temperature anomalies, fault locations and electromagnetic environment characteristics is generated, saving manual analysis time. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0058] Figure 1 It is a structural schematic diagram of the power transmission line UAV intelligent inspection system based on deep learning of the present invention;

[0059] Figure 2 The figure is a flow chart of the deep learning-based intelligent inspection method for power transmission lines using drones of the present invention. DETAILED DESCRIPTION

[0060] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0061] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0062] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0063] Example 1, reference Figure 1 , this embodiment provides a transmission line drone intelligent inspection system based on deep learning, including:

[0064] The data acquisition module uses drone inspections to obtain environmental information and collects historical data from the power system. The environmental information and historical data include: electromagnetic interference sections and high-temperature fault point information; based on the GIS system and satellite remote sensing data, environmental maps are obtained, including line topography, direction and key node information;

[0065] The electromagnetic environment modeling module analyzes the electromagnetic environment based on the information provided by the data acquisition module and generates an electromagnetic environment model; it identifies high-interference sections based on the electromagnetic field strength in different areas;

[0066] The dynamic electromagnetic interference detection module monitors the electromagnetic environment around the drone in real time during the inspection task. It monitors the electromagnetic intensity based on sensors and identifies interference in a timely manner. When the interference intensity reaches the threshold, it triggers the adaptive filtering algorithm and dynamically adjusts the filtering parameters.

[0067] The multimodal data acquisition module collects multimodal data during the inspection process, uses a deep learning model to fuse different data sources, and adjusts image resolution and focus parameters for temperature abnormality areas;

[0068] High temperature fault point identification module, which identifies the abnormal temperature area, adjusts the shooting angle and distance, and obtains the fault point image;

[0069] After completing the inspection task, the operation platform analyzes and archives the collected data and generates a fault report.

[0070] Example 2, reference Figure 2, this embodiment provides a transmission line UAV intelligent inspection method based on deep learning, comprising the following steps:

[0071] Step S1, electromagnetic environment modeling and path planning,

[0072] Before the drone performs inspection tasks, it conducts a preliminary analysis of the electromagnetic environment of the target line based on historical inspection data and environmental maps to generate an electromagnetic environment model;

[0073] In step S1, historical inspection data including electromagnetic interference and high temperature fault point records are obtained based on previous drone inspections and the power system; environmental map data including terrain, line direction and key node information are obtained based on the GIS system and satellite remote sensing data. Key nodes include insulators and connection points.

[0074] In step S1, the electromagnetic environment model identifies high-interference sections according to the electromagnetic field strength in different areas. Based on the identified high-interference sections, the drone adjusts the inspection path and flight altitude and pre-sets the inspection route;

[0075] In step S1, the electromagnetic environment model is constructed as follows:

[0076] Define historical inspection data as ,in, represents the spatial coordinate position, For location The electromagnetic field strength value, is the temperature value, is the total number of data points;

[0077] Define the environment map data as ,in, represents the spatial coordinate position, is the terrain height, is the distance from the node, is the total number of map data points;

[0078] Based on electromagnetic field strength and spatial coordinates Constructing electromagnetic field distribution model:

[0079] ,in, Indicates location The estimated value of the electromagnetic field strength is It is a parameter that controls the attenuation rate of the electromagnetic field intensity. It estimates the electromagnetic field intensity at different locations by smoothing the historical inspection data based on the Gaussian kernel function.

[0080] Setting the electromagnetic field strength threshold , define the high interference area as satisfying Calculate the interference risk weight of each area: ,in, Indicates location The interference risk weight is is the maximum electromagnetic field strength at all locations;

[0081] Adjust the inspection path of the drone according to the high interference section and define the inspection path optimization objective function for:

[0082] ,in, is the total number of path points, is the path point location, is the flight altitude of the waypoint, For the recommended safe flight altitude, is the trade-off coefficient between flight altitude and interference risk, and the goal is to minimize Find the optimal inspection path, avoid high-interference areas and stay at a safe height;

[0083] Specifically, the introduction of data acquisition and electromagnetic environment modeling modules enables the UAV to model and analyze the electromagnetic environment of the mission area based on historical inspection data and environmental maps before performing inspection tasks, effectively plan inspection routes and flight altitudes, optimize task execution efficiency and reduce the impact of electromagnetic interference on inspections. Specifically, the electromagnetic environment model based on Gaussian kernel functions and the multi-level path optimization algorithm ensure that the UAV can avoid high-interference sections, thereby reducing the risk of data transmission fluctuations and signal interruptions, and improving the continuity of inspection tasks and data stability.

[0084] Step S2: Dynamic electromagnetic interference detection and path adjustment.

[0085] Perform inspection tasks based on the inspection route of step S1, and monitor the surrounding electromagnetic environment in real time during the inspection;

[0086] In step S2, when the monitored electromagnetic interference intensity reaches the threshold, the adaptive filtering algorithm is triggered to dynamically adjust the filtering parameters; if the interference continues to increase, the path is fine-tuned to guide the drone to a low-interference area;

[0087] In step S2, the drone performs the inspection task based on the inspection route generated in step S1 and monitors the electromagnetic environment in real time:

[0088] The inspection route is the optimal inspection route generated from step S1, and is set as the path point set During the inspection process, the drone flies point by point along the preset path and arrives at each location in turn. ,in is the total number of path points;

[0089] Set each path point The electromagnetic field strength is , when monitored Exceeding the set threshold When the adaptive filtering algorithm is triggered to reduce electromagnetic interference, the filtering parameters of the adaptive filter are Dynamic adjustment based on current electromagnetic field strength: ,in, For the waypoint The adaptive filter parameters are and are the minimum and maximum filter coefficients of the filter, respectively. Indicates the maximum electromagnetic field strength value measured in the inspection area;

[0090] like Continuously exceeding the threshold , then fine-tune the path to guide the drone to a low-interference area. The optimization objective function of path fine-tuning is Defined as: ,in, For waypoints The optimization target value of is the distance from the current path point to the low interference area, is the adjustment coefficient, which is used to balance the electromagnetic intensity and path distance. The minimum value of selects the next path point position and guides the UAV to gradually fly to the low-interference area.

[0091] Specifically, during the inspection process, the dynamic electromagnetic interference detection module is used to monitor and adjust the electromagnetic environment in real time, and an adaptive filtering algorithm is introduced to perform real-time filtering adjustments based on the current electromagnetic field strength. The path is fine-tuned to guide the drone to fly to low-interference areas, greatly improving the robustness of the inspection. In addition, the drone's multimodal data acquisition module further enhances the detection of temperature anomalies and image optimization functions, deeply integrates infrared, visible light and lidar data, and can accurately identify temperature anomalies under multiple interference conditions. Points, dynamically adjust image resolution and focusing parameters to obtain high-resolution images, thereby effectively improving the accuracy of fault identification.

[0092] Step S3, multimodal data fusion and image optimization,

[0093] During the inspection process, the drone collects visible light, infrared images and lidar point cloud data, and uses a deep learning model for multimodal data fusion;

[0094] In step S3, abnormal temperature areas are detected, and a deep learning model is used to automatically optimize image resolution and focus parameters for abnormal temperature areas, so as to obtain high-resolution images without increasing inspection time.

[0095] In step S3, multimodal data fusion and deep learning models are used to identify temperature abnormality areas in real time, and dynamically adjust image resolution and focus parameters. Specifically:

[0096] At each waypoint The drone uses infrared sensors to collect temperature data , the temperature anomaly detection formula is defined as: ,in, For waypoints The temperature value at is the temperature abnormality threshold, if Exceed , then this point is regarded as a temperature anomaly point;

[0097] At the temperature anomaly At, collect multimodal data to represent visible light images , infrared images and LiDAR point cloud data , use the deep learning model to perform data fusion and calculate the fused image features: ,in, For waypoints The fusion characteristics of is the fusion function of the deep learning model;

[0098] For areas with abnormal temperature, image resolution Perform dynamic adjustments and control the inspection time at the same time. The resolution adjustment formula is: ,in, Temperature anomaly point Optimized resolution, is the base resolution, is the scaling factor of temperature anomaly, which is used to improve the resolution of high temperature areas;

[0099] Use a deep learning model to adjust image focus parameters: ,in, is the focusing parameter of the temperature anomaly point, is the basic focus value, is the adjustment coefficient.

[0100] Specifically, in terms of fault detection, the high-temperature fault point identification module enables the drone to take clear close-up and multi-angle photos of key nodes, insulators and connection points, capture more detailed information, identify minor abnormal changes, and combine multimodal data analysis of deep learning models to achieve intelligent classification and labeling for different fault types, thereby improving the detection accuracy of key components and significantly reducing the probability of missed detection and false detection.

[0101] Step S4, high temperature fault point identification,

[0102] In the temperature abnormality area detection in step S3, the drone automatically adjusts the shooting distance and angle to take close-range and multi-angle shots of key nodes;

[0103] In step S4, after the inspection is completed, the data is automatically uploaded to the operation platform, which uses image recognition and classification algorithms to analyze the images, archive and annotate the abnormal locations and fault images, and finally generate an inspection report containing temperature anomalies, fault point location information, and electromagnetic environment characteristics;

[0104] In step S4, after the abnormal temperature area is detected, the drone automatically adjusts the shooting distance and angle to perform close-range and multi-angle shooting:

[0105] Assume the position of the key node is , the current position of the drone is , the shooting distance is set to , the angle is set to and , respectively represent the pitch angle and azimuth angle, adjust the shooting distance To achieve the best shooting clarity: ,in, Indicates the distance from the drone to the key node, is the coordinate position of the key node, is the current position of the drone;

[0106] Optimize the angle and adjust the shooting pitch angle and azimuth :

[0107] ,

[0108] ,in, The pitch angle of the shooting is used to adjust the vertical angle of the drone. It is the shooting azimuth, which is used to adjust the horizontal angle of the drone.

[0109] Step S4, data upload and analysis method is:

[0110] After the inspection is completed, the drone will collect the image data , Temperature data and electromagnetic intensity data Upload to the job platform and use the deep learning model Identify and classify image data to generate fault location annotations and type tags :

[0111] ,in, is the detected abnormal position, is the fault type label, Represents a deep learning model for image recognition and classification. Select a combination of convolutional neural network and recurrent neural network, YOLO target detection model, FasterRCNN target detection model and fault recognition model based on attention mechanism;

[0112] Generate an inspection report containing temperature anomalies, fault location and electromagnetic environment information. Defined as: ,in, It is an inspection report, including fault location, type, temperature and electromagnetic field strength information. Represents the total number of collected data points.

[0113] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A transmission line drone intelligent inspection system based on deep learning, characterized by: include, The data acquisition module uses drone inspections to obtain environmental information and collects historical data from the power system. The environmental information and historical data include: electromagnetic interference sections and high-temperature fault point information; based on the GIS system and satellite remote sensing data, environmental maps are obtained, including line topography, direction and key node information; The electromagnetic environment modeling module analyzes the electromagnetic environment based on the information provided by the data acquisition module and generates an electromagnetic environment model; it identifies high-interference sections based on the electromagnetic field strength in different areas; The dynamic electromagnetic interference detection module monitors the electromagnetic environment around the drone in real time during the inspection task. It monitors the electromagnetic intensity based on sensors and identifies interference in a timely manner. When the interference intensity reaches the threshold, it triggers the adaptive filtering algorithm and dynamically adjusts the filtering parameters. The multimodal data acquisition module collects multimodal data during the inspection process, uses a deep learning model to fuse different data sources, and adjusts image resolution and focus parameters for temperature abnormality areas; High temperature fault point identification module, which identifies the abnormal temperature area, adjusts the shooting angle and distance, and obtains the fault point image; The operation platform, after completing the inspection task, analyzes and archives the collected data and generates a fault report; The drone performs inspection tasks based on the inspection route and monitors the electromagnetic environment in real time: The inspection route is the optimal inspection route, set as the path point set During the inspection process, the drone flies point by point along the preset path and arrives at each location in turn. ,in is the total number of path points; Set each path point The electromagnetic field strength is , when monitored Exceeding the set threshold When the adaptive filtering algorithm is triggered to reduce electromagnetic interference, the filtering parameters of the adaptive filter are Dynamic adjustment based on current electromagnetic field strength: ,in, For the waypoint The adaptive filter parameters are and are the minimum and maximum filter coefficients of the filter, respectively. Indicates the maximum electromagnetic field strength value measured in the inspection area; like Continuously exceeding the threshold , then fine-tune the path to guide the drone to a low-interference area. The optimization objective function of path fine-tuning is Defined as: ,in, For waypoints The optimization target value of is the distance from the current path point to the low interference area, is the adjustment coefficient, which is used to balance the electromagnetic intensity and path distance. The minimum value of selects the next path point position and guides the UAV to gradually fly to the low-interference area.

2. A transmission line UAV intelligent inspection method based on deep learning, characterized in that: Based on the deep learning-based transmission line drone intelligent inspection system described in claim 1, The following steps are included: Step S1, electromagnetic environment modeling and path planning, Before the drone performs inspection tasks, it conducts a preliminary analysis of the electromagnetic environment of the target line based on historical inspection data and environmental maps to generate an electromagnetic environment model; Step S2: Dynamic electromagnetic interference detection and path adjustment. Perform inspection tasks based on the inspection route of step S1, and monitor the surrounding electromagnetic environment in real time during the inspection; In step S2, when the monitored electromagnetic interference intensity reaches the threshold, the adaptive filtering algorithm is triggered to dynamically adjust the filtering parameters; if the interference continues to increase, the path is fine-tuned to guide the drone to a low-interference area; In step S2, the drone performs the inspection task based on the inspection route generated in step S1 and monitors the electromagnetic environment in real time: The inspection route is the optimal inspection route generated from step S1, and is set as the path point set During the inspection process, the drone flies point by point along the preset path and arrives at each location in turn. ,in is the total number of path points; Set each path point The electromagnetic field strength is , when monitored Exceeding the set threshold When the adaptive filtering algorithm is triggered to reduce electromagnetic interference, the filtering parameters of the adaptive filter are Dynamic adjustment based on current electromagnetic field strength: ,in, For the waypoint The adaptive filter parameters are and are the minimum and maximum filter coefficients of the filter, respectively. Indicates the maximum electromagnetic field strength value measured in the inspection area; like Continuously exceeding the threshold , then fine-tune the path to guide the drone to a low-interference area. The optimization objective function of path fine-tuning is Defined as: ,in, For waypoints The optimization target value of is the distance from the current path point to the low interference area, is the adjustment coefficient, which is used to balance the electromagnetic intensity and path distance. The minimum value of selects the next path point position and guides the UAV to gradually fly to the low-interference area; Step S3, multimodal data fusion and image optimization, During the inspection process, the drone collects visible light, infrared images and lidar point cloud data, and uses a deep learning model for multimodal data fusion; Step S4, high temperature fault point identification, In the temperature anomaly area detection in step S3, the drone automatically adjusts the shooting distance and angle to take close-up and multi-angle photos of key nodes.

3. The method for intelligent inspection of power transmission lines by drones based on deep learning according to claim 2 is characterized in that: In step S1, historical inspection data including electromagnetic interference and high temperature fault point records are obtained based on previous drone inspections and the power system; environmental map data including terrain, line direction and key node information are obtained based on the GIS system and satellite remote sensing data. Key nodes include insulators and connection points. In step S1, the electromagnetic environment model identifies high-interference sections according to the electromagnetic field strength in different areas. Based on the identified high-interference sections, the drone adjusts the inspection path and flight altitude and pre-sets the inspection route.

4. The method for intelligent inspection of power transmission lines by drones based on deep learning according to claim 3 is characterized in that: In step S1, the electromagnetic environment model is constructed as follows: Define historical inspection data as ,in, represents the spatial coordinate position, For location The electromagnetic field strength value, is the temperature value, is the total number of data points; Define the environment map data as ,in, represents the spatial coordinate position, is the terrain height, is the distance from the node, is the total number of map data points; Based on electromagnetic field strength and spatial coordinates Constructing electromagnetic field distribution model: ,in, Indicates location The estimated value of the electromagnetic field strength is It is a parameter that controls the attenuation rate of the electromagnetic field intensity. It estimates the electromagnetic field intensity at different locations by smoothing the historical inspection data based on the Gaussian kernel function. Setting the electromagnetic field strength threshold , define the high interference area as satisfying Calculate the interference risk weight of each area: ,in, Indicates location The interference risk weight is is the maximum electromagnetic field strength at all locations; Adjust the inspection path of the drone according to the high interference section and define the inspection path optimization objective function for: ,in, is the total number of path points, is the path point location, is the flight altitude of the waypoint, For the recommended safe flight altitude, is the trade-off coefficient between flight altitude and interference risk, and the goal is to minimize Find the optimal inspection path, avoid high interference areas and stay at a safe height.

5. The method for intelligent inspection of power transmission lines by drones based on deep learning according to claim 4 is characterized in that: In step S3, temperature abnormality area detection is performed, and a deep learning model is used to automatically optimize image resolution and focusing parameters for the temperature abnormality area.

6. The method for intelligent inspection of power transmission lines by drones based on deep learning according to claim 5 is characterized in that: In step S3, multimodal data fusion and deep learning models are used to identify temperature abnormality areas in real time, and dynamically adjust image resolution and focus parameters. Specifically: At each waypoint The drone uses infrared sensors to collect temperature data , the temperature anomaly detection formula is defined as: ,in, For waypoints The temperature value at is the temperature abnormality threshold, if Exceed , then this point is regarded as a temperature anomaly point; At the temperature anomaly At, collect multimodal data to represent visible light images , infrared images and LiDAR point cloud data , use the deep learning model to perform data fusion and calculate the fused image features: ,in, For waypoints The fusion characteristics of is the fusion function of the deep learning model; For areas with abnormal temperature, image resolution Dynamic adjustment is performed while controlling the inspection time. The resolution adjustment formula is: ,in, Temperature anomaly point Optimized resolution, is the base resolution, is the scaling factor of temperature anomaly, which is used to improve the resolution of high temperature areas; Use a deep learning model to adjust image focus parameters: ,in, is the focusing parameter of the temperature anomaly point, is the basic focus value, is the adjustment coefficient.

7. The method for intelligent inspection of power transmission lines by drones based on deep learning according to claim 6 is characterized in that: In step S4, after the inspection is completed, the data is automatically uploaded to the operation platform. The platform uses image recognition and classification algorithms to analyze the image, archive and annotate the abnormal position and fault image; and finally generates an inspection report containing temperature anomalies, fault point location information and electromagnetic environment characteristics.

8. The method for intelligent inspection of power transmission lines by drones based on deep learning according to claim 7 is characterized in that: In step S4, after the abnormal temperature area is detected, the drone automatically adjusts the shooting distance and angle to perform close-range and multi-angle shooting: Assume the position of the key node is , the current position of the drone is , the shooting distance is set to , the angle is set to and , respectively represent the pitch angle and azimuth angle, adjust the shooting distance To achieve the best shooting clarity: ,in, Indicates the distance from the drone to the key node, is the coordinate position of the key node, is the current position of the drone; Optimize the angle and adjust the shooting pitch angle and azimuth : ， ,in, The pitch angle of the shooting is used to adjust the vertical angle of the drone. It is the shooting azimuth, which is used to adjust the horizontal angle of the drone.

9. The method for intelligent inspection of power transmission lines by drones based on deep learning according to claim 8, characterized in that: Step S4, data upload and analysis method is: After the inspection is completed, the drone will collect the image data , Temperature data and electromagnetic intensity data Upload to the job platform and use the deep learning model Identify and classify image data to generate fault location annotations and type tags : ,in, is the detected abnormal position, is the fault type label, Represents a deep learning model for image recognition and classification. Select a combination of convolutional neural network and recurrent neural network, YOLO target detection model, FasterRCNN target detection model and fault recognition model based on attention mechanism; Generate an inspection report containing temperature anomalies, fault location and electromagnetic environment information. Defined as: ,in, It is an inspection report, including fault location, type, temperature and electromagnetic field strength information. Represents the total number of collected data points.

Images