Unmanned aerial vehicle intelligent inspection device for power transmission line

Abstract

The invention relates to the technical field of power transmission line inspection, in particular to a power transmission line unmanned aerial vehicle intelligent inspection device which comprises an unmanned aerial vehicle, an airborne computing system is carried on the unmanned aerial vehicle, an angle adjusting mechanism is mounted on the bottom surface of the unmanned aerial vehicle, a mounting shell is mounted at the bottom of the angle adjusting mechanism, and a laser radar and a visible light camera are mounted in the mounting shell; a feature extraction and matching module and an intelligent route generation module are arranged in the airborne computing system; the inspection device further comprises rear-end equipment, the airborne computing system communicates with the rear-end equipment, an intelligent report generation module is arranged in the rear-end equipment, and the rear-end equipment is used for processing images collected by the visible light camera. The tower type and size are quickly identified through the feature extraction and matching module, and the route is automatically planned through the intelligent route generation module, so that manual intervention can be reduced, the route planning efficiency is improved, and the method is particularly suitable for large-scale application of large power grid companies.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power transmission line inspection, in particular to a power transmission line unmanned aerial vehicle intelligent inspection device. BACKGROUND

[0002] In recent years, with the rapid development of unmanned aerial vehicle technology, it has been widely used in the field of power transmission line inspection. The power transmission line has the characteristics of long distance, high tower, wide distribution and complex terrain. The traditional manual inspection method is low in efficiency, high in cost and high in safety risk. The current mainstream "unmanned aerial vehicle fine autonomous inspection" mode adopts the technical path of "modeling first, planning second": first, the point cloud data is obtained by laser radar scanning to construct a fine three-dimensional model; then the technical personnel manually plan the flight route based on the model; finally, the unmanned aerial vehicle flies along the preset route for autonomous flight and data collection.

[0003] However, this mode has significant defects. First, the initial investment cost is extremely high. For large power grid companies with tens of thousands of towers, full-line laser point cloud collection and three-dimensional modeling require high cost and long cycle. Second, the flight route planning mainly relies on manual operation, which takes more than half an hour to plan complex towers, and the massive workload is difficult to bear. Finally, the mode lacks flexibility. When new tower types or line modifications occur, the whole process must be repeated, and the flexibility needs to be improved. Therefore, the industry urgently needs an intelligent inspection technology that can break away from the dependence on pre-fine modeling and achieve a low-cost, high-efficiency and self-adaptive inspection scheme.

[0004] The present application aims to improve the automation level and response speed of the inspection. SUMMARY

[0005] The present application provides a power transmission line unmanned aerial vehicle intelligent inspection device to solve the problems raised in the background art.

[0006] To solve the above technical problems, the technical solution adopted by the present application is as follows: a power transmission line unmanned aerial vehicle intelligent inspection device, comprising an unmanned aerial vehicle, an onboard computing system mounted on the unmanned aerial vehicle, an angle adjusting mechanism installed on the bottom surface of the unmanned aerial vehicle, an installation shell installed at the bottom of the angle adjusting mechanism, a laser radar and a visible light camera installed in the installation shell; the onboard computing system is built-in with a feature extraction and matching module for identifying towers and an intelligent flight path generation module for automatically planning flight paths; the inspection device further comprises a backend device, the onboard computing system communicates with the backend device, and the backend device is built-in with an intelligent report generation module for processing images collected by the visible light camera and generating reports.

[0007] Preferably, the unmanned aerial vehicle is provided with a millimeter wave radar on the front side, and the millimeter wave radar is connected to the onboard computing system, and the onboard computing system is provided with an obstacle avoidance decision module.

[0008] Preferably, the angle adjusting mechanism comprises a bracket connected to the bottom surface of the unmanned aerial vehicle through a rotating assembly, and the mounting shell is rotatably connected to the bracket through a rotating shaft, and the bracket is provided with a pitch motor on one side, and the output shaft of the pitch motor is in transmission connection with the rotating shaft.

[0009] An intelligent inspection method for power transmission lines by unmanned aerial vehicles, which is based on an inspection device and comprises the following steps: S1: controlling the unmanned aerial vehicle to fly to a preset initial position of a target tower; S2: scanning the tower by the onboard laser radar to obtain point cloud data, and a feature extraction and matching module identifying the type of the tower and calculating the structure size thereof based on the point cloud data; S3: an intelligent flight path generation module automatically generating an inspection flight path for the tower based on the identified type and structure size of the tower; S4: the unmanned aerial vehicle flying along the generated inspection flight path and collecting image data of the tower by the onboard visible light camera, while the millimeter wave radar continuously sensing the environment and the obstacle avoidance decision module controlling the unmanned aerial vehicle to avoid obstacles when detecting obstacles; S5: after the task is completed, an intelligent report generation module generating an inspection report based on the image data.

[0010] Preferably, in step S2, the feature extraction and matching module adopts a deep learning model based on a point cloud processing model of the PointNet++ architecture, which is trained by a large amount of standard tower point cloud data and can identify multiple tower types including cat head towers, wine cup towers and dry type towers, and accurately calculate the tower height, the number of cross arms, the layer spacing and the cross arm width.

[0011] Preferably, in step S3, the intelligent flight path generation module adopts an optimization strategy combining mixed integer programming and genetic algorithm, wherein the mixed integer programming is used to define the constraint conditions of the safety distance, the camera field of view overlap rate and the flight attitude, and the genetic algorithm is used to search for a refined inspection flight path that is optimal in terms of the total path length, the image acquisition quality and the estimated energy consumption in the solution space.

[0012] Preferably, the refined inspection flight path is a layered and surrounding path, the safety boundary thereof is a cylindrical space with the center of the tower as the axis and the radius being 1.5 times the identified tower height, and an independent sub-flight path loop is generated for each layer of cross arms, and the vertical height and the radius of the sub-flight path loop are adaptively adjusted according to the actual position and width of the corresponding cross arm.

[0013] Preferably, in step S4, the obstacle avoidance strategy of the obstacle avoidance decision module is: when the obstacle intrusion into the first level safety threshold is detected, the emergency hovering instruction is triggered; when the obstacle is static or the system judges that hovering is unsafe, the dynamic bypass algorithm is started, the local bypass path is planned and the original flight path is returned after obstacle avoidance.

[0014] Preferably, in step S5, the intelligent report generation module adopts a convolutional neural network-based image recognition model for batch analysis of high-definition images collected by the inspection, automatically identifies defect types, locates defect positions, evaluates risk levels, and generates inspection reports in accordance with specifications.

[0015] The present application has the following beneficial effects: (1) The present application quickly identifies the tower type and size through the feature extraction and matching module, and automatically plans the flight path through the intelligent flight path generation module, thereby reducing manual intervention, improving flight path planning efficiency, and being particularly suitable for large-scale application in large power grid companies; (2) By integrating the millimeter wave radar and the obstacle avoidance decision module, setting the safety threshold, and combining the dynamic bypass algorithm to realize real-time obstacle perception and obstacle avoidance, the stable operation of the unmanned aerial vehicle in complex environments is ensured, thereby improving the flight safety and reliability; (3) From data acquisition to report generation, the whole process is automated, the visible light camera automatically takes high-definition images, and the back-end intelligent report generation module automatically identifies defects, evaluates risks and generates standard reports, reducing the workload of manual analysis and improving the processing speed and accuracy of inspection data. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 Figure 1 is a structural schematic diagram of the inspection device of the present application; Figure 2 Figure 2 is a flowchart of the inspection method of the present application.

[0017] Reference signs: 1, unmanned aerial vehicle, 2, onboard computing system, 3, angle adjusting mechanism, 31, bracket, 32, rotating assembly, 33, rotating shaft, 34, pitch motor, 4, laser radar, 5, visible light camera, 6, millimeter wave radar, 7, mounting shell. DETAILED DESCRIPTION

[0018] The present application will be further described below in conjunction with the drawings.

[0019] As shown in Figure 1 The present application provides an unmanned aerial vehicle intelligent inspection device for power transmission lines, which comprises an unmanned aerial vehicle 1, an onboard computing system 2, an angle adjusting mechanism 3, a laser radar 4, a visible light camera 5, a millimeter wave radar 6 and a back-end device.

[0020] The unmanned aerial vehicle 1 is a multi-rotor unmanned aerial vehicle 1 platform, has GPS positioning, autonomous flight and hovering functions. The unmanned aerial vehicle 1 has a load capacity of not less than 5 kg, a continuous running time of more than 30 minutes, to adapt to long-distance inspection requirements, and a flight control system integrating a high-precision inertial measurement unit (IMU) and a barometer to ensure flight stability.

[0021] The airborne computing system 2 is mounted on the unmanned aerial vehicle 1, uses an embedded GPU computing platform, has a processing capacity of not less than 30 TOPS, is used for running a feature extraction and matching module, an intelligent route generation module and an obstacle avoidance decision module; a built-in storage unit (SSD, capacity 1 TB) in the system is used for temporarily storing point cloud and image data, and communicates with a rear-end device through a 5G or 4G module.

[0022] The angle adjusting mechanism 3 is installed on the bottom surface of the unmanned aerial vehicle 1 and is used for adjusting the viewing angle of the laser radar 4 and the visible light camera 5. The angle adjusting mechanism 3 comprises a support 31, the support 31 is connected with the bottom surface of the unmanned aerial vehicle 1 through a rotary assembly 32, the support 31 is horizontally rotated through the rotary assembly 32, a mounting shell 7 is rotationally connected with the support 31 through a rotating shaft 33, a pitch motor 34 is installed on one side of the support 31, an output shaft of the pitch motor 34 is in transmission connection with the rotating shaft 33, the pitch motor 34 controls the mounting shell 7 to swing in the vertical direction by ±90°, the laser radar 4 and the visible light camera 5 are fixed in the mounting shell 7, and the flexibility of scanning and shooting is ensured.

[0023] The laser radar 4 is a 32-line laser radar 4, has a detection distance of 200 m, and is used for quickly acquiring tower point cloud data and supporting real-time transmission to the airborne computing system 2.

[0024] The visible light camera 5 is a high-definition zoom camera and is used for collecting high-definition images of towers and lines. The camera supports automatic focusing and exposure control, and ensures image quality.

[0025] The millimeter wave radar 6 is installed on the front side of the unmanned aerial vehicle 1, adopts a 77GHz frequency band radar, has a detection distance of 150 m and an angle resolution of 5°, and is used for real-time sensing of front obstacles, and data input of the obstacle avoidance decision module of the airborne computing system 2.

[0026] The rear-end device is a ground server or a cloud platform, is equipped with a high-performance CPU and a GPU, runs an intelligent report generation module, and is connected with the airborne computing system 2 through a wireless network to receive image data and perform batch processing.

[0027] As shown in Figure 2 Another aspect of the present application provides a power transmission line unmanned aerial vehicle intelligent inspection method, which is based on the foregoing inspection device and comprises the following steps: S1: Control the unmanned aerial vehicle 1 to fly to the preset initial position of the target tower and hover, and during hovering, the unmanned aerial vehicle 1 keeps the position stable through GPS and IMU; S2: Start the laser radar 4 to quickly surround and scan the tower, the angle adjusting mechanism 3 controls the horizontal rotation of the laser radar 4, and the pitch motor 34 adjusts the vertical angle, so as to ensure that the point cloud covers the entire tower, obtain initial point cloud data, the feature extraction and matching module receives the point cloud data, and then uses the voxel grid filtering to denoise and filter the point cloud data, and then identifies the tower type and calculates the accurate size; S3: The intelligent flight path generation module automatically generates a patrol flight path for the tower based on the identified tower type and structure size; S4: The unmanned aerial vehicle 1 autonomously flies along the generated patrol flight path, the visible light camera 5 automatically zooms and shoots during flight, and at the same time, the millimeter wave radar 6 continuously scans the environment, when the obstacle intrusion safety threshold is detected, the obstacle avoidance decision module triggers emergency hovering, and a dynamic bypass algorithm is started to plan an arc-shaped path to bypass the obstacle, and then return to the original flight path; S5: After the task is completed, the unmanned aerial vehicle 1 returns to the base, the image data collected by the visible light camera 5 is transmitted to the backend equipment through wireless transmission, the intelligent report generation module processes the images in batches, automatically identifies the defect type, locates the defect position, evaluates the risk level, and generates a patrol report conforming to the specification.

[0028] In some embodiments, the feature extraction and matching module is implemented based on a deep learning model of PointNet++ architecture, the model training uses a large amount of standard tower point cloud data set, covering common types such as cat head tower, wine cup tower and dry type tower, the model input is the initial point cloud data obtained by the laser radar 4, the features are extracted through multi-layer perception and maximum pooling operation, and the tower type, tower height, cross arm layer number, layer spacing and cross arm width are output.

[0029] In some embodiments, the intelligent flight path generation module adopts an optimization strategy combining mixed integer programming (MIP) and genetic algorithm (GA), MIP is used to define the constraint conditions: safety distance (the minimum distance between the unmanned aerial vehicle 1 and the tower is 1.5 times the identified tower height), camera field of view overlap rate (not less than 80%), flight attitude (the pitch angle is limited within ±30°); GA is used to search the optimal path: the population size is set to 100, the iteration number is 500, the fitness function comprehensively considers the total path length, image acquisition quality and estimated energy consumption, and the output is a layered surrounding flight path, including multiple sub-flight path loops, each loop corresponds to a layer of cross arm, and the vertical height and radius are adaptively adjusted according to the cross arm position.

[0030] In some embodiments, the obstacle avoidance decision module implements an obstacle avoidance strategy based on the millimeter wave radar 6 data. The module sets two levels of safety thresholds: the first level is a warning threshold (obstacle distance 10 m), which triggers an emergency hover; the second level is a danger threshold (obstacle distance 5 m), which starts a dynamic detour algorithm. The dynamic detour algorithm uses the A* search algorithm to plan a local path in combination with a real-time environment map, and returns to the original flight path through PID control after avoiding obstacles.

[0031] In some embodiments, the intelligent report generation module is implemented based on a convolutional neural network image recognition model. The model uses a ResNet-50 architecture, is pre-trained on ImageNet, and is fine-tuned on a power defect dataset (including categories such as insulator damage and hardware corrosion). The model input is a high-definition image collected by the visible light camera 5, and the output is the defect type, location (mapped to the tower structure through image coordinates), and risk level (based on the severity of the defect). The report generation module automatically integrates data and outputs an inspection report, including a defect summary, image evidence, and recommended measures.

[0032] The above embodiments can be combined with each other.

[0033] The above embodiments do not limit the shape, material, structure, etc. of the present application in any form. Any simple modification, equivalent change, and modification of the above embodiments according to the technical essence of the present application are within the protection scope of the technical solution of the present application.

Claims

1. A transmission line unmanned aerial vehicle intelligent inspection device, comprising an unmanned aerial vehicle, characterized in that: The unmanned aerial vehicle is provided with an onboard computing system, an angle adjusting mechanism is installed on the bottom surface of the unmanned aerial vehicle, an installation shell is installed at the bottom of the angle adjusting mechanism, and a laser radar and a visible light camera are installed in the installation shell; The onboard computing system is provided with a feature extraction and matching module for identifying the tower and an intelligent route generation module for automatically planning a route; The inspection device further comprises a backend device, the onboard computing system communicates with the backend device, and the backend device is provided with an intelligent report generation module for processing images collected by the visible light camera and generating a report. 2.The power transmission line unmanned aerial vehicle intelligent inspection device according to claim 1, characterized in that: The unmanned aerial vehicle is provided with a millimeter wave radar on the front side, the millimeter wave radar is connected with the onboard computing system, and the onboard computing system is provided with an obstacle avoidance decision module. 3.The power transmission line unmanned aerial vehicle intelligent inspection device according to claim 1, characterized in that: The angle adjusting mechanism comprises a support, the support is connected with the bottom surface of the unmanned aerial vehicle through a rotating assembly, the installation shell is rotationally connected with the support through a rotating shaft, an elevation motor is installed on one side of the support, and an output shaft of the elevation motor is in transmission connection with the rotating shaft.

4. The method of claim 1, wherein: The inspection method is based on the inspection device of any one of claims 1-3, and comprises the following steps: S1: controlling the unmanned aerial vehicle to fly to a preset initial position of a target tower; S2: scanning the tower by the onboard laser radar to obtain point cloud data, and identifying the type of the tower and calculating the structure size thereof based on the point cloud data by the feature extraction and matching module; S3: automatically generating an inspection route for the tower based on the identified type and structure size of the tower by the intelligent route generation module; S4: flying the unmanned aerial vehicle along the generated inspection route, collecting image data of the tower by the onboard visible light camera, and continuously sensing the environment by the millimeter wave radar to detect obstacles and control the unmanned aerial vehicle to avoid obstacles by the obstacle avoidance decision module; S5: generating an inspection report based on the image data by the intelligent report generation module after the task is completed.

5. The transmission line unmanned aerial vehicle intelligent inspection method according to claim 4, characterized in that: In step S2, the feature extraction and matching module adopts a deep learning model based on a point cloud processing model of a PointNet++ architecture, which is trained by a large amount of standard tower point cloud data, can identify various tower types including cat head towers, wine cup towers and dry type towers, and accurately calculate the tower height, number of cross arms, layer spacing and cross arm width. 6.The power transmission line unmanned aerial vehicle intelligent inspection method according to claim 4, characterized in that: In step S3, the intelligent route generation module adopts an optimization strategy combining mixed integer programming and genetic algorithm, wherein the mixed integer programming is used to define the constraint conditions of safety distance, camera field of view overlap rate and flight attitude, and the genetic algorithm is used to search for a refined inspection route with optimal total length, image acquisition quality and estimated energy consumption in the solution space.

7. The transmission line drone intelligent inspection method of claim 6, wherein: The refined inspection route is a layered and surrounding path, the safety boundary thereof is a cylindrical space with the center of the tower as the axis and the radius being 1.5 times the identified tower height, and an independent sub-route loop is generated for each layer of cross arm, and the vertical height and radius of the sub-route loop are adaptively adjusted according to the actual position and width of the corresponding cross arm. 8.The power transmission line unmanned aerial vehicle intelligent inspection method of claim 4, wherein: In step S4, the obstacle avoidance strategy of the obstacle avoidance decision module is: when an obstacle intrudes into the first level safety threshold, an emergency hovering instruction is triggered; when the obstacle is static or the system judges that hovering is unsafe, a dynamic detour algorithm is started to plan a local detour path and return to the original route after avoiding the obstacle. 9.The power transmission line unmanned aerial vehicle intelligent inspection method of claim 4, wherein: In step S5, the intelligent report generation module adopts an image recognition model based on a convolutional neural network to perform batch analysis on the high-definition images collected during the inspection, automatically identify the defect type, locate the defect position, evaluate the risk level, and generate an inspection report that conforms to the specifications.

Images