Power transmission line inspection system based on autonomous correction long-endurance unmanned airship

Abstract

The invention relates to the technical field of intelligent inspection of power equipment, and particularly discloses a power transmission line inspection system based on an autonomous correction long-endurance unmanned airship. The system aims to solve the problems of short endurance, small operation radius, single image angle, intermittent operation, leak detection, low precision, poor environmental adaptability and the like caused by dependence on manpower or GPS navigation in the conventional unmanned aerial vehicle inspection. The system comprises an energy supply module, a data acquisition module, an intelligent control module and a communication and data return module. Wherein the energy supply module is integrated with a solar energy collection, storage and intelligent scheduling unit to realize day-and-night continuous energy supply; the intelligent control module adopts an RT-DETR visual algorithm to identify a tower in real time, performs automatic deviation correction by fusing pose information, and accurately tracks a preset track; the data acquisition module supports multi-angle high-definition imaging, and the communication module guarantees real-time data return. And high-precision unmanned inspection with ultra-long endurance, full-automatic and wide-area continuous coverage is integrally realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection technology for power equipment, and more specifically, to a transmission line inspection system based on an autonomous correction long-endurance unmanned airship. Background Technology

[0002] With the continuous expansion of power systems and the increasing complexity of energy transmission networks, the operational status of transmission lines and their auxiliary equipment directly affects the safety and stability of the entire power grid. To ensure the long-term reliable operation of transmission equipment, regular inspections of transmission lines, towers, insulators, and hardware are necessary to detect defects such as corrosion, damage, electrical discharge, and foreign matter adhesion. Currently, transmission line inspection methods mainly include three types: manual inspection, ground-based vehicle-mounted inspection, and drone inspection.

[0003] Manual inspection: Traditional manual inspections typically require inspectors to walk along the line or climb towers to check equipment. This method is labor-intensive and dangerous, especially in complex terrains such as mountains, forests, and river crossings, where manual inspections are inefficient and unsafe. Furthermore, the results of manual inspections rely on experience-based judgment, leading to highly subjective data and hindering the creation of objective and traceable equipment records.

[0004] Ground-based vehicle-mounted inspection methods: In some areas, vehicle-mounted high-definition cameras or LiDAR equipment are used for ground inspections. While this method can improve work efficiency to some extent, it is limited by terrain conditions. Vehicles cannot approach the power lines themselves, and the camera's line of sight is limited, making it impossible to obtain clear images of the top of towers or conductor attachments. Therefore, vehicle-mounted methods are poorly adaptable to complex terrain.

[0005] Drone Inspection Method: In recent years, drone inspection technology has developed rapidly. Drones based on multi-rotor or fixed-wing platforms can carry high-definition cameras, infrared cameras, and other equipment to achieve automated aerial inspections, significantly improving efficiency and becoming the main method for detecting defects in intelligent power transmission lines. However, existing drones still have the following shortcomings:

[0006] Limited flight time: Multi-rotor drones rely on batteries for power and can usually only fly for 30 to 90 minutes; fixed-wing drones have a slightly longer flight time, but they are greatly affected by the environment and cannot hover stably for long periods of time to take pictures.

[0007] Limited operating radius: Battery capacity limitations result in a limited coverage area for a single flight, making it difficult to achieve continuous cross-regional inspections.

[0008] Insufficient image resolution: In order to reduce the load, most drones are equipped with cameras with limited focal lengths, which cannot clearly identify minute defects at high altitudes or long distances.

[0009] Data lacks multi-angle features: Traditional drones mostly use single-direction top-down or side-view shooting, which cannot form complete equipment spatial information, limiting the accuracy of subsequent 3D reconstruction and spatial point calibration.

[0010] Poor mission continuity: Due to limited endurance, inspection missions often require multiple takeoffs and landings, which can easily lead to data interruption and missed inspections.

[0011] Furthermore, existing long-endurance unmanned inspection platforms mostly rely on fixed-wing or vertical takeoff and landing (VTOL) designs. While these can extend endurance, they still cannot achieve true long-duration continuous flight and energy self-sufficiency. Especially in long-distance power transmission line scenarios, inspection missions often need to cover line areas of hundreds or even thousands of kilometers, which places higher demands on energy supply, communication stability, and data management capabilities.

[0012] On the other hand, with the development of artificial intelligence and deep learning, image recognition-based defect detection of power transmission equipment has gradually matured. However, this technology is highly sensitive to image quality, shooting angle, and resolution. If the acquisition equipment cannot provide high-quality multi-view images, it will seriously affect the detection accuracy and stability.

[0013] Therefore, there is an urgent need for an innovative inspection system that combines new energy supply, ultra-long line-of-sight imaging technology and intelligent three-dimensional perception algorithms to achieve long-term, wide-coverage, and high-precision unmanned intelligent inspection. Summary of the Invention

[0014] The purpose of this invention is to solve the problems of intermittent inspections, missed inspections, low accuracy, and poor environmental adaptability caused by the short endurance, small operating radius, single image acquisition angle, and reliance on manual or GPS navigation of existing UAV inspection systems. By constructing an intelligent power transmission line inspection system that integrates continuous solar power supply, RT-DETR visual autonomous correction navigation, multi-mode intelligent energy consumption scheduling, and a highly stable airship platform, this invention achieves unmanned and precise inspections with day and night operation, ultra-long flight time, fully automatic path tracking, multi-angle high-definition imaging, and wide-area continuous coverage.

[0015] To achieve the above objectives, the present invention employs the following technical means:

[0016] This invention provides a power transmission line inspection system based on an autonomous, long-endurance unmanned airship with self-correction capability, comprising:

[0017] The energy supply module is used to provide continuous energy support for the airship and its onboard equipment.

[0018] The data acquisition module is used to acquire multi-scale, high-precision images of power transmission lines and their auxiliary equipment;

[0019] The intelligent control module is used to fuse preset trajectories with real-time visual / pose information to achieve autonomous correction and target tracking;

[0020] The communication and data feedback module is used to realize the wireless transmission and remote interaction of inspection data;

[0021] The energy supply module includes an energy harvesting and conversion unit, an energy storage unit, a signal conversion unit, and a comprehensive scheduling unit installed on the surface of the airship, forming a closed-loop energy management system.

[0022] The intelligent control module uses a target detection algorithm based on the RT-DETR model to identify the position of the tower in real time, and calculates the flight offset by combining the camera imaging principle and sensor data, and automatically adjusts the airship attitude and heading to achieve accurate tracking of the preset trajectory.

[0023] In the above scheme, the energy harvesting and conversion unit is a high-efficiency solar panel laid on the top, wings and side surfaces of the airship, which is used to continuously harvest solar energy and convert it into electrical energy during flight.

[0024] In the above scheme, the integrated scheduling unit dynamically switches between working mode, cruise mode, and standby mode according to the flight status, wherein:

[0025] In operating mode, all modules are powered and the lateral motors are activated to assist movement;

[0026] In cruise mode, only the propulsion motor is powered, while the lateral motors are turned off, ensuring the lowest power consumption for autonomous correction, GPS positioning, and intelligent control modules.

[0027] In standby mode, power is turned off to all devices except the basic control module, and the hovering function is maintained through low power mode.

[0028] In the above scheme, the data acquisition module includes an auto-zoom camera assembly, a three-axis stabilized gimbal, an image acquisition and control unit, and an auxiliary sensor group. The auto-zoom camera assembly supports 20x continuous zoom and can automatically switch between long-distance observation and close-up fine imaging.

[0029] The implementation of the intelligent control module in the above scheme includes the following steps:

[0030] Step 1: Obtain the preset flight path based on the GPS information of the historical tower locations, control the airship to fly in the specified direction, and at the same time turn on the 2D high-definition camera to capture the scene;

[0031] Step 2: Detect the current scene lighting conditions. If it is daytime, enter cruise mode; if it is nighttime, enter standby mode.

[0032] Step 3: In cruise mode, start the cruise detection algorithm service based on the RT-DETR model to process 2D images in real time and detect transmission line tower targets;

[0033] Step 4: When a target tower is detected, initiate the autonomous correction service, which includes:

[0034] Step 4.1: At time t, keep the camera facing the same direction as the airship's movement;

[0035] Step 4.2: Obtain one or more tower target bounding boxes output by the target detection algorithm at time t;

[0036] Step 4.3: In the image coordinate system, calculate the Euclidean distance Ω from the image center point to the top center point of each predicted target, and decompose this distance into horizontal projections. and vertical projection ;

[0037] Step 4.4: Transform the horizontal projection to the world coordinate system, denoted as... By calculating the projection of the current position Horizontal projection to the next tower target on the predetermined flight path minimum difference value This yields the horizontal displacement. The calculation formula is:

[0038] in, This indicates the amount of inertial drift caused by engineering design flaws;

[0039] Step 4.5: After determining the target trajectory, based on the required horizontal distance between the unmanned airship and the tower along the flight path... Vertical distance is The position is determined to calculate the theoretical tilt angle of the airship entering the working mode. :

[0040] At time t, calculate the actual tilt angle. :

[0041]

[0042] in, Indicates the camera's focal length. and They represent Calculate the vertical height and horizontal distance of the airship from the tower at any given time; calculate the vertical movement of the airship. :

[0043] Calculate the distance to the target tower. :

[0044] Step 4.6: Based on the horizontal movement amount Vertical movement and distance Dynamically adjust the airship's flight direction and speed;

[0045] Step 5: When the cruise point is detected, shut down the forward propulsion system and allow the airship to hover;

[0046] Step 6: Activate the working mode and control the airship to move clockwise around the tower, sequentially executing the sequence of moving left, forward, right, and backward to acquire omnidirectional images of the tower and its auxiliary equipment;

[0047] Step 7: After completing the pole inspection, turn off the working mode, re-enter the cruise mode, and continue flying along the preset track to the next pole location.

[0048] In the above scheme, the RT-DETR model includes a feature extraction structure, a dynamic feature interaction structure, and a target query and decoding structure. The feature extraction structure uses the ConvNeXt structure to extract multi-scale feature maps. The dynamic feature interaction structure achieves multi-scale feature fusion through the Transformer mechanism. The target query and decoding structure uses a fixed number of query vectors to predict the tower target.

[0049] In the above scheme, the feature extraction structure uses a ConvNeXt structure to extract multi-scale feature maps. Let the input image be... After convolution and normalization operations, multi-scale features are obtained:

[0050]

[0051] in, Indicates the input image. and These represent the image height and width, respectively, and 3 represents the three RGB channels. This represents feature layers at different scales. Indicates the first Feature extraction functions at each scale;

[0052] In the dynamic feature interaction structure, let the first... The characteristics of each scale are ,in, This represents the number of spatial locations at this scale. For the channel dimension, first obtain the query, key, and value through linear transformation:

[0053]

[0054]

[0055]

[0056] in, The projection matrix is ​​learned, and it is normalized using the softmax function to obtain the attention distribution between positions. The attention weight matrix is ​​then calculated. :

[0057]

[0058] attention weight matrix Values ​​corresponding to the scale Multiplying these together yields the single-scale attention output features:

[0059]

[0060] Finally, the features from various scales are concatenated to obtain the final features:

[0061]

[0062] in, This represents the transpose of the key matrix. Indicates the dimension of the key / query vector; Concat indicates the feature concatenation operation. To output the projection matrix;

[0063] The target query and decoding structure uses a fixed number of query vectors. The system predicts targets, with each query corresponding to a possible target, and the final output is a set of predicted bounding boxes.

[0064]

[0065] in, The X-coordinate of the center point, the Y-coordinate of the center point, the width, and the height of the bounding box are given. Where N represents the target category and N is the preset maximum number of targets.

[0066] In the above scheme, the communication and data backhaul module includes a data encapsulation unit, a communication link management unit, a data compression and encryption unit, a link scheduling and caching unit, and a ground receiving and parsing unit. It adopts a multi-channel communication architecture and can select the best communication path in real time according to the signal strength and latency.

[0067] In the above scheme, the link scheduling and caching unit performs priority management on different types of data, ensures real-time transmission of high-priority data such as flight control commands and defect alarm images, and adopts a batch compression and delayed upload mechanism for ordinary inspection image data.

[0068] In the above scheme, the energy storage unit of the energy supply module adopts a combination of high-energy-density lithium battery and supercapacitor, which can store excess electrical energy during the day and release the stored energy in the absence of light or low light environment. It maintains the operation of propulsion system, flight control system and navigation system through low power consumption mode, and realizes uninterrupted inspection day and night.

[0069] Because the present invention employs the above-mentioned technical means, it has the following beneficial effects:

[0070] 1. This invention constructs a closed-loop energy supply system by configuring high-efficiency solar panels, high-energy-density lithium batteries, and supercapacitors. It solves the technical problems of traditional UAVs, which rely on limited-capacity batteries, resulting in short flight time, limited operating radius, and inability to achieve continuous day and night inspections. It achieves the effect of continuously collecting solar energy and converting it into electrical energy during the day, and releasing stored energy at night or in low-light environments to maintain basic flight functions, thereby significantly extending the system's operating time and supporting uninterrupted inspection of cross-regional, long-distance power transmission lines.

[0071] 2. This invention solves the problem of existing inspection platforms maintaining high energy consumption and wasting energy during non-operational phases by dynamically switching working mode, cruise mode and standby mode according to flight status through an integrated scheduling unit, and implementing differentiated power supply strategies for each module (such as only keeping the propulsion motor and the lowest power consumption intelligent control module powered in cruise mode, and shutting down unnecessary equipment in standby mode). It achieves the effect of maximizing energy utilization efficiency and extending the overall endurance while ensuring mission execution capability.

[0072] 3. This invention solves the technical problem that traditional UAVs, due to fixed focal length lenses or payload limitations, cannot simultaneously achieve both long-distance overview and close-range detailed imaging. By incorporating an automatic zoom camera assembly (supporting 20x continuous zoom), a three-axis stabilized gimbal, and an auxiliary sensor group, it achieves the effect of wide-area rapid scanning during the cruise phase and automatically switching to high-magnification zoom for multi-angle high-definition imaging of towers and their auxiliary equipment during the working mode, thereby improving the comprehensiveness and accuracy of defect identification.

[0073] 4. This invention solves the problems of poor detection stability and high false positive and false negative rates of traditional target detection methods under conditions such as large tower scale variations, complex backgrounds, and uneven illumination by using the ConvNeXt structure to extract multi-scale features, the Transformer mechanism to realize cross-scale feature interaction, and the end-to-end architecture of decoding target positions with fixed query vectors. It achieves the effect of efficiently and accurately locating transmission tower targets without anchor frames and non-maximum suppression, providing reliable visual input for subsequent autonomous correction.

[0074] 5. This invention uses an intelligent control module to identify the tower position in real time using a target detection algorithm based on the RT-DETR model, and calculates the flight offset (including horizontal projection) by combining camera imaging principles and sensor data. Vertical projection (and its conversion to the world coordinate system), thereby dynamically adjusting the airship's attitude and heading, solving the problem that relying on pure GPS navigation is susceptible to electromagnetic interference, terrain obstruction, or signal drift, which can lead to track deviation. It achieves highly robust and high-precision vision-driven autonomous correction and target tracking in complex power transmission line environments.

[0075] 6. This invention integrates the energy supply module, intelligent control module, data acquisition module, and communication feedback module at the system level, and achieves four-dimensional coordination of energy, perception, control, and communication based on flight mission status. For example, during the cruise phase, visual correction and propulsion are maintained with low power consumption; high-magnification zoom and omnidirectional surround shooting are automatically activated when approaching the tower; and day / night modes are switched according to lighting conditions, and the energy storage unit's power supply strategy is linked. This deep coupling of multiple technologies not only allows each module to play its independent advantages in its respective dimension, but also generates a positive closed-loop effect of "energy self-sufficiency supporting long-term perception - accurate perception driving efficient correction - correction accuracy ensuring high-quality imaging - high-quality imaging feeding back into the reliability of defect identification". This solves the problem of isolated operation of the energy, perception, control, and communication subsystems in existing inspection systems, making it difficult to form an overall performance improvement, and achieves a truly meaningful long-endurance, high-precision, fully automatic, and wide-coverage intelligent inspection and coordination effect for power transmission lines. Attached Figure Description

[0076] Figure 1 A flowchart of the airship inspection process;

[0077] Figure 2 This is a schematic diagram of cruise offset from the camera's perspective;

[0078] Figure 3 A schematic diagram of the working flight path for the airship. Detailed Implementation

[0079] The embodiments of the present invention will be described in detail below. Although the present invention will be described and illustrated in conjunction with some specific embodiments, it should be noted that the present invention is not limited to these embodiments. On the contrary, any modifications or equivalent substitutions made to the present invention should be covered within the scope of the claims of the present invention.

[0080] Furthermore, to better illustrate the present invention, numerous specific details are set forth in the following detailed embodiments. Those skilled in the art will understand that the present invention can be practiced without these specific details.

[0081] To address the current limitations of manual power transmission line inspection, such as high difficulty, low efficiency, and safety concerns, as well as the limitations of using drones to replace manual inspection in terms of endurance and operating radius, this invention proposes a method for power transmission line inspection using unmanned airships with autonomous correction and long endurance capabilities. This method aims to solve the problem of power transmission line inspection under the aforementioned limitations by providing a novel approach and filling a gap in the field.

[0082] refer to Figure 1 The flowchart shows the process of airship inspection. To achieve the above objectives, the technical solution of this invention is as follows: a power transmission line inspection system based on an autonomous correction long-endurance unmanned airship, characterized by the following four main modules: an energy supply module, a data acquisition module, an intelligent control module, and a communication and data feedback module.

[0083] I. Energy Supply Module

[0084] The energy supply module provides stable and continuous energy support for the airship and its onboard equipment, ensuring long-term operation and the normal functioning of its basic systems. This module consists of an energy harvesting and conversion unit, an energy storage unit, a signal conversion unit, and a comprehensive scheduling unit, forming a closed-loop energy management system that enables all-weather, dynamic energy supply and scheduling.

[0085] 1. Energy Harvesting and Conversion Unit: The energy harvesting and conversion unit is located on the top, wings, and sides of the airship, with high-efficiency solar panels evenly distributed on it to form a large-area photovoltaic harvesting layer. This unit is used for the efficient absorption and conversion of sunlight into electrical energy during the day, and mainly performs the following functions:

[0086] (1) Photovoltaic energy harvesting function: Taking advantage of the large surface area and long sunshine duration of the unmanned airship, the photovoltaic absorption area is maximized to increase the amount of solar energy that can be obtained per unit time.

[0087] (2) Real-time energy conversion function: Converts the collected solar energy into usable electrical energy to provide instant power for propulsion motors, image acquisition modules, flight control systems, etc.

[0088] (3) Energy surplus output function: When the lighting conditions are good and the power consumption is low, the surplus power is automatically delivered to the energy storage unit to achieve energy accumulation.

[0089] This unit enables the creation of a continuous clean energy source during flight, thereby increasing the airship's maximum flight time.

[0090] 2. Energy Storage Unit: The energy storage unit is used to centrally store the electrical energy output from the energy harvesting and conversion unit, and to provide a stable energy output to the system during periods of insufficient sunlight or at night. This unit includes high-energy-density lithium batteries, supercapacitors, or other energy storage devices, and has the following functions:

[0091] (1) High-density energy storage function: High-energy-density batteries are used to improve the energy storage capacity per unit weight and meet the needs of long-distance cruising;

[0092] (2) Daytime energy storage function: Stores excess electrical energy when the light is stable, providing sufficient reserves for nighttime cruising;

[0093] (3) Nighttime energy release and low power consumption support function: Release stored energy in the absence of light or low light environment, maintain the operation of propulsion system, flight control system and navigation system through low power consumption mode, and achieve continuous flight.

[0094] This unit ensures the unmanned airship's ability to conduct uninterrupted day and night inspections across regions.

[0095] 3. Signal Conversion Unit: The signal conversion unit is responsible for processing and converting the signals output from each module, providing precise energy control commands to the integrated dispatch unit. It mainly includes the following functional structures:

[0096] Control signal conversion structure: Receives commands from the flight control module and adjusts the power supply of the propulsion motor and lateral motor according to the flight status to achieve dynamic control of the airship's attitude and heading.

[0097] Scene-aware signal conversion structure: Based on the lighting environment and operational requirements, the signals collected from the outside are converted into commands for working, cruising, or standby modes, and fed back to the energy storage unit to realize the switching of operating modes.

[0098] Signal conversion structure: dynamically supplies power to the image acquisition module to ensure a balance between data acquisition and energy consumption, enabling continuous operation around the clock.

[0099] 4. Integrated Scheduling Unit: Based on instructions from the signal conversion unit, the integrated scheduling unit dynamically schedules the energy storage unit to achieve energy allocation under different operating modes. The main operating modes include:

[0100] Operating mode: Activated when the airship is performing operations at a designated location, maintaining power to all modules and simultaneously activating the lateral motors to assist movement, ensuring efficient execution of the task.

[0101] Cruise mode: Activated when moving between points, it only keeps the propulsion motor powered while shutting down the lateral motors, and ensures the lowest power consumption of the autonomous correction, GPS positioning and intelligent control modules to achieve energy optimization.

[0102] Standby mode: Enabled when not performing tasks and hovering in position, shutting off power to all devices except the basic control module, maintaining hovering function through low power mode, and maximizing battery life.

[0103] II. Data Acquisition Module

[0104] Figure 2 This is a schematic diagram of the cruise drift from the camera's perspective. The image acquisition module is used for multi-scale, high-precision image acquisition and dynamic recording of power transmission lines and their auxiliary equipment. This module integrates an automatic zoom camera assembly, a three-axis stabilized gimbal, and an intelligent image acquisition control unit. It can automatically adjust the focal length and acquisition strategy in different flight phases and inspection mission modes, achieving an organic combination of long-distance observation and close-range fine imaging.

[0105] Automatic zoom camera assembly: It adopts a programmable motorized zoom lens, which supports continuous zoom control from wide-angle to super telephoto, with a zoom ratio of up to 20x; through automatic focus and exposure control, it achieves high-definition imaging of targets at both near and far distances.

[0106] Three-axis stabilized gimbal: used to compensate for attitude disturbances of the airship during flight, maintaining image stability and directional accuracy.

[0107] Image Acquisition and Control Unit (ICC): Based on an embedded processing platform, it is responsible for automatically switching acquisition strategies according to flight mode, controlling lens parameters, and storing or transmitting image data in real time.

[0108] Auxiliary sensor group: including attitude sensor, GPS and distance measurement sensor, used to acquire spatial position and angle information, to realize geographic coordinate labeling and shooting angle optimization.

[0109] III. Intelligent Control Module

[0110] Figure 3 This is a schematic diagram of the airship's operational flight path. The intelligent control module is responsible for fusing the preset flight path (based on the GPS positions of the towers) with real-time visual / pose information to ensure the airship flies according to plan and achieves high-quality imaging of the tower targets. The preset flight path is obtained by combining GPS information from historical points to ensure the airship maintains the correct flight direction. For the airship to move from point A to point B, the intelligent control module performs the following processes, such as... Figure 1 As shown:

[0111] Based on historical location information, obtain the specified directions AB, activate the cruise mode, and simultaneously activate the 2D high-definition camera to capture the scene.

[0112] The system detects the current scene and determines the progress of the task based on the scene's lighting conditions. If it is daytime, it enters patrol mode; if it is nighttime, it enters standby mode.

[0113] Enable the cruise detection algorithm service to use 2D images for real-time detection of tower targets;

[0114] Upon detecting a target tower, the autonomous correction service is activated.

[0115] The cruise point has been reached; at this point, the forward propulsion system is shut down, and the airship hovers.

[0116] Turn on defect detection mode, turn on working mode, move around the tower, and photograph the equipment and defects at the location;

[0117] The airship moves in a clockwise direction, as shown in the flight diagram below. Figure 3 As shown. First, the right motor propels the airship to the left; upon reaching the left-hand point, the forward propulsion system is activated, moving the airship forward; upon reaching the forward point, the forward propulsion system is deactivated, and the left motor is activated, moving the airship to the right; upon reaching the right-hand point, the backward propulsion system is activated, moving the airship backward. Upon reaching the rear point, the right motor is activated, propelling the airship to the left to reach the cruising point.

[0118] Turn off working mode and re-enter cruise mode.

[0119] For the cruise detection algorithm service described in section 2, the target location of the tower points is determined based on the RT-DETR target detection algorithm model. The RT-DETR algorithm is a highly efficient improvement on the DETR series of algorithms. Its core idea is to complete the entire process from image input to target prediction through an end-to-end Transformer structure, avoiding the dependence on anchor boxes and non-maximum suppression in traditional target detection, thereby significantly improving detection accuracy and speed. This model consists of three main structures: a feature extraction structure, a dynamic feature interaction structure, and a target query and decoding structure, as follows:

[0120] Feature extraction structure: A ConvNeXt structure is used to extract multi-scale feature maps. Let the input image be... After convolution and normalization operations, multi-scale features are obtained:

[0121]

[0122] in These represent feature layers at different scales to accommodate changes in the size of the tower structure and its auxiliary equipment within the image.

[0123] Dynamic Feature Interaction Module: Let the feature at the i-th scale be... in, This represents the number of spatial locations at this scale. This is a channel-level transformation. First, a linear transformation is used to obtain the query, key, and value:

[0124]

[0125]

[0126]

[0127] in, Normalization is performed using the softmax function to obtain the attention distribution between positions, and the attention weight matrix is ​​then calculated. :

[0128]

[0129] attention weight matrix Values ​​corresponding to the scale Multiplying these together yields the single-scale attention output features:

[0130]

[0131] Finally, the features from various scales are concatenated to obtain the final features:

[0132]

[0133] Target query and decoding module: uses a fixed number of query vectors The system predicts the possible targets for each query. The final output is a set of predicted bounding boxes.

[0134]

[0135] in The X-coordinate of the center point, the Y-coordinate of the center point, the width, and the height of the bounding box are given. For the target category.

[0136] The self-correction service mentioned in point 3 mainly consists of the following three steps, as detailed below:

[0137] At any given moment t, the flight control system always keeps the camera facing the same direction as the airship's movement.

[0138] Obtain one or more target bounding boxes from the target detection algorithm at time t.

[0139] In the image coordinate system, calculate the Euclidean distance from the image center point to the top center point of each predicted target. And decompose the distance into horizontal projections and vertical projection For reference from the camera's perspective Figure 2 .

[0140] Horizontal projection transformation to the world coordinate system is denoted as: By calculating the projection of the current position Horizontal projection to the next tower target on the predetermined flight path minimum difference value This yields the horizontal displacement. The calculation is as follows:

[0141]

[0142] in, This represents the amount of inertial drift caused by engineering design, which is theoretically negligible. The goal is to find the minimum... Find the target track.

[0143] After determining the target trajectory, assume that the unmanned airship needs to be at a horizontal distance of [missing information] from the tower along the flight path. The vertical distance is If the airship enters the working mode at a certain position, then the theoretical tilt angle of the working mode position is... It should be:

[0144]

[0145] For time t, the actual tilt angle can be expressed as:

[0146]

[0147] in, Indicates the camera's focal length. , Let represent the vertical height and horizontal distance from the tower at time t, respectively. This can be directly obtained through sensors. The vertical movement of the airship is required to ensure it reaches the theoretical position. The distance can be expressed as:

[0148]

[0149] The distance from the target tower at this point can be expressed as:

[0150]

[0151] Repeat the above operation at each time point, by adjusting the horizontal movement amount. Vertical movement The distance d from the tower is used to continuously and dynamically adjust the direction and speed of the flight control system, shorten the cruise time, and quickly start data acquisition at the next location.

[0152] IV. Communication and Data Backhaul Module

[0153] This module enables wireless transmission and remote interaction of data collected by the unmanned airship during patrol and inspection, and is a crucial component of the system for completing closed-loop image information transmission. This module includes a data encapsulation unit, a communication link management unit, a data compression and encryption unit, a link scheduling and caching unit, and a ground receiving and parsing unit.

[0154] (1) Data Encapsulation Unit This unit is used to receive the raw image and detection result information from the image acquisition module and to format and encapsulate it in a structured manner. The encapsulated data packet includes the image frame sequence, detection result, timestamp, attitude angle parameters and position information.

[0155] (2) Communication Link Management Unit

[0156] This unit employs a multi-channel communication architecture, including a primary link and a backup link. The module has a built-in link monitoring mechanism that can select the optimal communication path in real time based on signal strength and latency.

[0157] (3) Data compression and encryption unit: In order to improve transmission efficiency and ensure data security, this unit uses an image encoding algorithm to adaptively compress image data and automatically adjusts the bit rate according to the network bandwidth.

[0158] (4) Link scheduling and caching unit This unit performs priority management and caching control for different types of data. For high-priority data such as flight control commands and defect alarm images, the module ensures real-time transmission; for ordinary inspection image data, a batch compression and delayed upload mechanism is adopted to reduce bandwidth usage.

[0159] (5) Ground Receiving and Parsing Unit: This unit corresponds to the airship communication module and is used to receive the transmitted inspection data and perform parsing, decoding, and storage. The parsed data is automatically imported into the inspection management system of the ground control center to realize the synchronous recording of tower equipment status identification, defect alarm, and inspection tasks.

[0160] I. The main technical challenges of this invention are as follows:

[0161] 1: Achieve stable, continuous, and low-error real-time target detection and trajectory comparison in complex power transmission line environments.

[0162] The inspection scenario of transmission lines has characteristics such as a wide variety of tower types, complex backgrounds, and large differences in illumination, which can easily lead to a decrease in the accuracy of visual detection. At the same time, during the flight of the airship, affected by wind force, airflow disturbance, attitude change, etc., the images will generate jitter and distortion, making the real-time detection stability of RT-DETR face challenges. In order to achieve autonomous correction, the system must compare the detection frame with the preset flight path in real time, and needs to process multi-source data such as 2D imaging, IMU data, and GPS information simultaneously, and complete matching and deviation calculation within an extremely short time. Insufficient accuracy in any link will cause flight path drift and make the correction control ineffective. Therefore, how to achieve high-precision, low-latency, stable and reliable visual detection and flight path comparison in a complex environment is one of the core technical difficulties of the system.

[0163] 2: The design of the autonomous correction control strategy based on visual input is highly difficult

[0164] This system uses the results of deep visual detection as the control input to adjust the speed, heading, and power of the airship in real time. This is different from the traditional method that relies on GPS or remote control operators, which requires the control system to have the ability to adapt to visual noise, delay, and misdetection. Due to the slow response speed and large inertia of the airship, if the control strategy is not designed reasonably, over-adjustment, jitter, and even instability are likely to occur. Therefore, it is necessary to deeply couple visual detection, flight path deviation calculation, and the airship dynamics model, and introduce filtering, dynamic weights, adaptive PID, or a continuous adjustment mechanism based on the DFL idea to achieve smooth and reliable correction control. How to ensure that the vision-driven control algorithm remains stable during long-term flight is an important technical problem faced by the system.

[0165] 3: The coordinated optimization of energy scheduling and task execution under long endurance conditions is complex

[0166] The system relies on solar energy to achieve long endurance, but the acquisition of solar energy is affected by factors such as weather, illumination angle, and flight attitude, making energy replenishment highly uncertain. At the same time, the inspection task itself requires the use of cameras, computing chips, and communication equipment, and the power consumption fluctuates greatly. To ensure continuous operation, the system must dynamically adjust the flight speed, power output, and detection frequency according to the remaining energy, illumination conditions, task intensity, and wind field environment. If the scheduling is improper, it may result in task interruption due to insufficient energy, or a decrease in inspection quality due to excessive energy conservation. Therefore, how to achieve the overall optimization among energy management, flight tasks, and correction control is the key challenge for realizing long-endurance autonomous inspection.

[0167] II. Technical advantages of the present invention

[0168] Technical advantage 1: Achieving a fully automatic flight path correction inspection mode driven by visual detection

[0169] Compared to traditional UAV inspections that require manual control or ground station intervention for flight path calibration, the unmanned airship system of this invention utilizes an RT-DETR model to identify tower structures in real time and automatically compares them with preset standard flight paths, forming a truly "intelligent visual navigation." The system can analyze the spatial offset of the target in real time and dynamically adjust the flight direction, speed, and power based on the deviation, thereby ensuring that the airship always follows the standard route for inspection. This approach overcomes the limitations of traditional GPS navigation in dealing with tower obstructions, electromagnetic interference, and the influence of mountainous terrain, while also reducing the safety risks of manual control, upgrading unmanned inspections from semi-automatic to truly fully automatic.

[0170] Technological Advantage 2: Solar-assisted power supply enables low-cost, long-endurance operation.

[0171] This invention achieves continuous energy replenishment during cruise by deploying high-efficiency solar panels on the top, wings, and sides of the airship, eliminating the need for frequent returns and significantly extending its endurance. The system supports maximizing energy harvesting during the day, storing excess energy in the energy storage unit; at night, it maintains flight and inspection based on an intelligent low-power strategy. Furthermore, it automatically adjusts flight speed, power output, and the computational intensity of the detection module based on day / night lighting conditions, task priorities, and changes in remaining energy, achieving optimal energy allocation. This system-level integration results in a long endurance advantage, enabling the unmanned airship to adapt to long-distance power transmission line inspection tasks across regions and time periods, overcoming the inherent bottlenecks of short endurance and frequent returns of traditional multi-rotor UAVs.

[0172] Technological Advantage 3: High degree of autonomy and robustness significantly improve inspection safety and data quality.

[0173] This invention not only utilizes deep learning models to improve pole identification accuracy at the perception level, but also achieves autonomous correction, adaptive control, and intelligent energy scheduling at the decision-making level, forming a highly intelligent and robust aerial inspection system. Its wind resistance, trajectory stability, and environmental adaptability are significantly higher than traditional UAVs. Furthermore, the airship platform itself has a large payload capacity, capable of carrying multimodal sensors (visible light, infrared, laser rangefinders, temperature sensors, etc.), enabling it to complete complex inspection tasks without human intervention.

Claims

1. A power transmission line inspection system based on an autonomous, long-endurance unmanned airship with self-correction capability, characterized in that, include: The energy supply module is used to provide continuous energy support for the airship and its onboard equipment. The data acquisition module is used to acquire multi-scale, high-precision images of power transmission lines and their auxiliary equipment; The intelligent control module is used to fuse preset trajectories with real-time visual / pose information to achieve autonomous correction and target tracking; The communication and data feedback module is used to realize the wireless transmission and remote interaction of inspection data; The energy supply module includes an energy harvesting and conversion unit, an energy storage unit, a signal conversion unit, and a comprehensive scheduling unit installed on the surface of the airship, forming a closed-loop energy management system. The intelligent control module uses a target detection algorithm based on the RT-DETR model to identify the position of the tower in real time, and calculates the flight offset by combining the camera imaging principle and sensor data, and automatically adjusts the airship attitude and heading to achieve accurate tracking of the preset trajectory.

2. The transmission line inspection system according to claim 1, characterized in that, The energy harvesting and conversion unit consists of high-efficiency solar panels laid on the top, wings, and sides of the airship, used to continuously harvest solar energy and convert it into electrical energy during flight.

3. The transmission line inspection system according to claim 1 or 2, characterized in that, The integrated scheduling unit dynamically switches between working mode, cruise mode, and standby mode based on flight status, wherein: In operating mode, all modules are powered and the lateral motors are activated to assist movement; In cruise mode, only the propulsion motor is powered, while the lateral motors are turned off, ensuring the lowest power consumption for autonomous correction, GPS positioning, and intelligent control modules. In standby mode, power is turned off to all devices except the basic control module, and the hovering function is maintained through low power mode.

4. The transmission line inspection system according to claim 1, characterized in that, The data acquisition module includes an auto-zoom camera assembly, a three-axis stabilized gimbal, an image acquisition and control unit, and an auxiliary sensor group. The auto-zoom camera assembly supports 20x continuous zoom and can automatically switch between long-distance observation and close-up fine imaging.

5. The transmission line inspection system according to claim 1, characterized in that, The implementation of the intelligent control module includes the following steps: Step 1: Obtain the preset flight path based on the GPS information of the historical tower locations, control the airship to fly in the specified direction, and at the same time turn on the 2D high-definition camera to capture the scene; Step 2: Detect the current scene lighting conditions. If it is daytime, enter cruise mode; if it is nighttime, enter standby mode. Step 3: In cruise mode, start the cruise detection algorithm service based on the RT-DETR model to process 2D images in real time and detect transmission line tower targets; Step 4: When a target tower is detected, initiate the autonomous correction service, which includes: Step 4.1: At time t, keep the camera facing the same direction as the airship's movement; Step 4.2: Obtain one or more tower target bounding boxes output by the target detection algorithm at time t; Step 4.3: In the image coordinate system, calculate the Euclidean distance Ω from the image center point to the top center point of each predicted target, and decompose this distance into horizontal projections. and vertical projection ; Step 4.4: Transform the horizontal projection to the world coordinate system, denoted as... By calculating the projection of the current position Horizontal projection to the next tower target on the predetermined flight path minimum difference value This gives the amount of movement in the horizontal direction. The calculation formula is: in, This indicates the amount of inertial drift caused by engineering design flaws; Step 4.5: After determining the target trajectory, based on the required horizontal distance between the unmanned airship and the tower along the flight path... Vertical distance is The position is determined to calculate the theoretical tilt angle of the airship entering the working mode. : At time t, calculate the actual tilt angle. : in, Indicates the camera's focal length. and They represent Calculate the vertical height and horizontal distance of the airship from the tower at any given time; calculate the vertical movement of the airship. : Calculate the distance to the target tower. : Step 4.6: Based on the horizontal movement amount Vertical movement and distance Dynamically adjust the airship's flight direction and speed; Step 5: When the cruise point is detected, shut down the forward propulsion system and allow the airship to hover; Step 6: Activate the working mode and control the airship to move clockwise around the tower, sequentially executing the sequence of moving left, forward, right, and backward to acquire omnidirectional images of the tower and its auxiliary equipment; Step 7: After completing the pole inspection, turn off the working mode, re-enter the cruise mode, and continue flying along the preset track to the next pole location.

6. The transmission line inspection system according to claim 1, characterized in that, The RT-DETR model includes a feature extraction structure, a dynamic feature interaction structure, and a target query and decoding structure. The feature extraction structure uses a ConvNeXt structure to extract multi-scale feature maps. The dynamic feature interaction structure achieves multi-scale feature fusion through a Transformer mechanism. The target query and decoding structure uses a fixed number of query vectors to predict tower targets.

7. The transmission line inspection system according to claim 6, characterized in that, The feature extraction structure uses a ConvNeXt structure to extract multi-scale feature maps. Let the input image be... After convolution and normalization operations, multi-scale features are obtained: in, Indicates the input image. and These represent the image height and width, respectively, and 3 represents the three RGB channels. This represents feature layers at different scales. Indicates the first Feature extraction functions at each scale; In the dynamic feature interaction structure, let the first... The characteristics of each scale are ,in, This represents the number of spatial locations at this scale. For the channel dimension, first obtain the query, key, and value through linear transformation: in, The projection matrix is ​​learned, and it is normalized using the softmax function to obtain the attention distribution between positions. The attention weight matrix is ​​then calculated. : attention weight matrix Values ​​corresponding to the scale Multiplying these together yields the single-scale attention output features: Finally, the features from various scales are concatenated to obtain the final features: in, This represents the transpose of the key matrix. Indicates the dimension of the key / query vector; Concat indicates the feature concatenation operation. To output the projection matrix; The target query and decoding structure uses a fixed number of query vectors. The system predicts targets, with each query corresponding to a possible target, and the final output is a set of predicted bounding boxes. in, The X-coordinate of the center point, the Y-coordinate of the center point, the width, and the height of the bounding box are given. Where N represents the target category and N is the preset maximum number of targets.

8. The transmission line inspection system according to claim 1, characterized in that, The communication and data backhaul module includes a data encapsulation unit, a communication link management unit, a data compression and encryption unit, a link scheduling and caching unit, and a ground receiving and parsing unit. It adopts a multi-channel communication architecture and can select the best communication path in real time according to signal strength and latency.

9. The transmission line inspection system according to claim 1 or 7, characterized in that, The link scheduling and caching unit manages the priority of different types of data, ensuring real-time transmission of high-priority data such as flight control commands and defect alarm images, and employing batch compression and delayed upload mechanisms for ordinary inspection image data.

10. The transmission line inspection system according to claim 1, characterized in that, The energy storage unit of the energy supply module uses a combination of high-energy-density lithium batteries and supercapacitors, which can store excess electrical energy during the day and release the stored energy in the absence of light or low light environment. It maintains the operation of the propulsion system, flight control system and navigation system through low power consumption mode, and realizes uninterrupted inspection day and night.

Images