A substation power drone automatic inspection method and device

Abstract

The present invention discloses a method and device for automatic inspection of power substations by UAVs, the method comprising collecting point cloud data of substations by using a laser radar; acquiring images in the substation by using an image collector; completing three-dimensional reconstruction of an unknown environment by the UAV based on the collected point cloud data and images in the substation, as well as the real-time position information of the UAV; and performing route planning based on the result of the three-dimensional reconstruction, which is used to control the power substation UAV to perform automatic inspections. The present invention realizes fully autonomous inspection of substations by UAVs without human participation, reduces the difficulty of operation for on-site personnel, frees operation and maintenance personnel from the heavy and boring routine inspection work, improves the overall operation efficiency, and enhances the automation and intelligent level of substation operation and maintenance.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent inspection of substations, and specifically relates to an automatic inspection method and device for substation power unmanned aerial vehicles. Background Art

[0002] In the power system, inspection, maintenance and management of substation equipment and timely discovery of equipment problems are one of the main tasks of substation operation and maintenance. Defects, hidden dangers, abnormal movement status of power equipment directly affect the safe operation of power equipment and the reliable power supply of the power grid, as well as the production and operation activities of industrial, agricultural and commercial users, and may even cause serious economic losses, which is not conducive to social harmony and stability. As substation operation and maintenance personnel, we must pay attention to the quality of equipment inspection. Timely discovery of equipment problems, improving power supply reliability and service level have become important means to improve the service level of power supply enterprises. Therefore, substations need to continuously strengthen the construction of inspection operation capabilities to create more economic and social benefits for power users. In the current widespread substation drone inspection applications, drones require manual operation by professionals, which greatly reduces the versatility of drone inspections; moreover, due to the physical limitations of the pilots themselves, it is difficult to concentrate their mental and physical strength for a long time. When operating drones in complex substation environments for flight operations, safety accidents such as explosions are prone to occur.

[0003] The existing daily inspection perspective of manual or robot inspections is generally from the bottom to the top of the equipment, and there are "blind spots" for inspections at high places and above the equipment. In the currently widely used substation drone inspection applications, drones need to be manually operated by professionals, which greatly reduces the versatility of drone inspections. Moreover, due to the physical limitations of the pilots themselves, it is difficult for them to concentrate their mental and physical strength for a long time. When operating drones to perform flight operations in a complex substation environment, safety accidents such as bombings are prone to occur. The existing drone inspection routes are taught by inspection personnel, and the debugging workload is very huge. In addition, the quality of waypoint settings and inspection collection pictures is related to the skill level of the pilots. There is no global verification of the safe flight path during the flight path design and flight process. Most of the existing inspection methods have the function of manually planning routes, but there are still problems such as irregular route planning and the lack of a joint complementary mechanism with other inspection systems (inspection robots, manual inspections), which weakens the inspection effect of drone inspections and fails to maximize the advantages of drone inspections. Summary of the invention

[0004] In response to the above problems, the present invention proposes a method and device for automatic inspection of substation power drones, which realizes fully autonomous inspection of substation drones without human participation, reduces the difficulty of operation for on-site personnel, frees operation and maintenance personnel from the heavy and boring routine inspection work, improves the overall operation efficiency, and enhances the automation and intelligence level of substation operation and maintenance.

[0005] In order to achieve the above technical objectives and the above technical effects, the present invention is implemented through the following technical solutions:

[0006] In a first aspect, the present invention provides a substation power drone automatic inspection method, comprising:

[0007] Use LiDAR to collect point cloud data of substations;

[0008] Use image collector to obtain images inside the substation;

[0009] Based on the collected point cloud data, images inside the substation, and the real-time location information of the drone, the drone can perform 3D reconstruction of the unknown environment.

[0010] Based on the results of 3D reconstruction, route planning is performed to control the substation power drone for automatic inspection.

[0011] Optionally, the point cloud data collection of the substation using a laser radar specifically includes the following steps:

[0012] According to the structure of the UAV, the ICP algorithm based on geometric features is used to calculate the installation position coordinates of RTK, GPS, image collector and lidar on the UAV, and obtain the installation position parameters at the coordinate position;

[0013] According to the installation location parameters obtained, single-point positioning is performed on RTK, GPS, image collector and lidar respectively to generate their own independent coordinate systems;

[0014] Select the origin of the drone as the origin of the coordinate system, calculate the spatial position coordinates of RTK, GPS, image collector and lidar according to the installation position coordinates, derive the transformation relationship between each independent coordinate system and the origin of the drone, establish an object-image relationship model, derive the calculation formula between the object-image point coordinates and the camera's external orientation elements, and the ground coordinate calculation formula of the laser point cloud;

[0015] During the normal flight of the UAV, GPS and RTK record the flight trajectory of the UAV in real time, while the image collector and lidar use spatial automatic scanning technology to obtain images and point cloud data within the substation.

[0016] Optionally, the three-dimensional reconstruction of the unknown environment comprises the following steps:

[0017] The positioning module receives point cloud data output by the laser radar in real time. The positioning module includes GPS and RTK. Combined with the flight trajectory of the drone, the real-time positioning of the drone in an unknown environment is completed through a specific algorithm. The output of the positioning module is a three-dimensional laser radar odometer, that is, the change in posture at the current moment relative to the original posture;

[0018] The point cloud data output by the laser radar is used for coloring and adding model texture, and the image in the substation is combined for image registration. All point cloud data are spliced ​​using a coordinate matrix through point cloud stitching to fully reflect the point cloud features of the scanned object;

[0019] By fusing the point cloud data output by the image collector and the lidar, radar point cloud information is generated. Combined with the location information obtained by the real-time positioning of the drone, the calculation formula between the object point coordinates and the camera's external orientation elements, as well as the ground coordinate calculation formula of the laser point cloud, the point cloud data fused with the radar point cloud information and the sensor is constructed to achieve three-dimensional reconstruction of the unknown environment by the drone.

[0020] Optionally, the step of using the positioning module to receive the point cloud data output by the laser radar in real time, combining the flight trajectory of the drone, and completing the real-time positioning of the drone in an unknown environment through a specific algorithm, further includes:

[0021] Identify and warn of possible dangers that the drone may encounter based on its current flight posture, adopt the drone control locking strategy, and execute predetermined strategies such as returning to the original route, prohibiting takeoff, and landing on the spot to ensure the safety of the drone.

[0022] Optionally, performing route planning based on the result of the three-dimensional reconstruction specifically includes the following steps:

[0023] Use the 3D reconstructed point cloud data to build substation domain knowledge and construct formulas based on first-order logic;

[0024] Through the automatic recognition algorithm of image objects, the identified objects are mapped to instances of the substation domain ontology, and the spatial position relationship between image objects corresponds to the object association attributes between instances in the domain ontology;

[0025] Using the identified target mapping, image preprocessing is performed to convert the image into a feature vector. The CNN convolutional neural network is selected to perform feature recognition and extraction on the feature vector of the identified target. Then, the optimal decision tree is trained by building a deep algorithm for classification.

[0026] The extracted target features are mapped to the object attribute association relationship between domain ontology instances, the image is tested using a decision binary tree, and logical reasoning is performed based on the constructed domain knowledge to perform retrieval;

[0027] A full-coverage flight strip design method is adopted, combined with the optimal decision tree and retrieval results, to realize the initial polygonal flight strip design. Then, the inspection area is divided based on the initial polygonal flight strip. While ensuring that the entire area can be covered, the route calculation is performed for each area to achieve precise route design for the substation scenario.

[0028] Optionally, the step of performing route planning based on the result of the three-dimensional reconstruction further includes:

[0029] Carry out automatic inspection of power substation by drone, and the results are displayed and presented in a one-stop data visualization development platform. After completion, check whether the results meet the substation inspection requirements.

[0030] In a second aspect, the present invention provides a substation power drone automatic inspection device, including: RTK, GPS, image collector and laser radar;

[0031] The laser radar is used to collect point cloud data for the substation;

[0032] The image collector is used to acquire images inside the substation;

[0033] Based on the collected point cloud data and images in the substation, as well as the real-time location information of the drone, the drone is used to complete the three-dimensional reconstruction of the unknown environment; the real-time location information of the drone is calculated based on the output information of RTK and GPS;

[0034] Based on the results of 3D reconstruction, route planning is performed to control the substation power drone for automatic inspection.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The present invention proposes a method and device for automatic inspection of substation power drones, which realizes fully autonomous inspection of substation drones without human participation, reduces the difficulty of operation for on-site personnel, frees operation and maintenance personnel from the heavy and boring routine inspection work, improves the overall operation efficiency, and enhances the automation and intelligence level of substation operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments and in conjunction with the accompanying drawings, wherein:

[0038] Figure 1 It is a block diagram of the automatic inspection method of power substation using drone. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0040] The application principle of the present invention is described in detail below in conjunction with the accompanying drawings.

[0041] Example 1

[0042] The embodiment of the present invention provides a substation power drone automatic inspection method, which realizes the substation drone fully autonomous inspection without human participation, such as Figure 1 As shown, the method comprises the following steps:

[0043] Step 1: According to the fuselage structure of the power inspection UAV, the ICP algorithm based on geometric features is used to calculate the installation position coordinates of the sensors (including RTK (real-time carrier phase difference), GPS (global positioning system), image collector (digital camera) and laser radar) on the UAV, and obtain the installation position parameters at the coordinate position;

[0044] Step 2: Based on the obtained installation position parameters, perform single-point positioning on each sensor to generate their own independent coordinate systems;

[0045] Step 3: Select the origin of the drone as the origin of the coordinate system, calculate the spatial position coordinates of each sensor according to the installation position coordinates, derive the transformation relationship between each independent coordinate system in step 2 and the origin of the drone, establish an object-image relationship model, derive the calculation formula between the object-image point coordinates and the camera's external orientation elements, and the ground coordinate calculation formula of the laser point cloud;

[0046] Step 4: During the normal flight of the drone, GPS and RTK record the flight trajectory of the drone in real time, while the image collector and lidar use spatial automatic scanning technology to obtain images and 3D lidar point cloud data in the substation;

[0047] Step 5: The positioning module relies on real-time reception of laser radar information. The positioning module includes GPS and RTK. Combined with the flight trajectory of the drone, the positioning module completes the real-time positioning of the drone in an unknown environment through a specific algorithm. Its input is the original data of the 3D laser radar point cloud obtained in step 4, and its output is the 3D laser radar odometer, that is, the change in posture at the current moment relative to the original posture;

[0048] Step 6: Use the 3D LiDAR point cloud data to add color and model texture, and combine it with the image inside the substation obtained in step 4 for image registration. Through point cloud stitching, all point cloud data are stitched together using the coordinate matrix to fully reflect the point cloud features of the scanned object, and remove noise and redundant point cloud data in the preprocessing stage;

[0049] Step 7: Generate radar point cloud information with rich texture and high spatial accuracy by fusing the visual sensor and the original point cloud data of the lidar. Combined with the location information obtained by the real-time positioning of the drone in step 5, the two calculation formulas derived in step 3 are used to construct the point cloud data of radar point cloud information and sensor fusion, so as to realize the three-dimensional reconstruction of the unknown environment by the drone and present it in the one-stop data visualization development platform.

[0050] Step 8: Use the point cloud data generated by the 3D reconstruction in step 7 to construct the substation domain knowledge, including instances, attribute associations, etc., and construct a formula based on first-order logic. Through the image target automatic recognition algorithm, the identified targets are mapped to instances of the substation domain ontology. The spatial position relationship between image targets corresponds to the object association attributes between instances in the domain ontology, and is used for automatic recognition of digital objects in power drone substations;

[0051] Step 9: Using the target mapping identified in step 8, first perform image preprocessing to convert the image into a feature vector, select the CNN convolutional neural network to perform feature recognition and extraction on the feature vector of the identified target, and then train the optimal decision tree for classification by building a deep algorithm;

[0052] Step 10: Map the target features extracted in step 9 to the object attribute association relationship between domain ontology instances, use the decision binary tree to test the image, and perform logical reasoning based on the constructed domain knowledge to perform retrieval;

[0053] Step 11: Using a full-coverage flight strip design method, combined with the optimal decision tree obtained in step 9 and the search results in step 10, an initial polygonal flight strip design is implemented. Then, the inspection area is divided based on the initial polygonal flight strip. While ensuring that the entire area can be covered, the flight routes are calculated one by one in each area to achieve precise flight route design for the substation scenario.

[0054] Step 12: Based on the precisely designed route, identify and warn the possible dangers that the drone may encounter according to the current flight posture obtained in step 5, adopt the drone control locking strategy, and execute the predetermined strategy such as returning to the original route, prohibiting takeoff, and landing on the spot to ensure the safety of the drone;

[0055] Step 13: Deploy drones at the substation and conduct automatic inspections of the substation power drones according to the route designed in step 11. The results are displayed and presented in the one-stop data visualization development platform. After completion, check whether the results meet the substation inspection requirements. If not, jump to step 4. If they meet the inspection requirements, the inspection is completed.

[0056] In a specific implementation of the embodiment of the present invention, the space automatic scanning technology refers to: using the surrounding environment to update the position of the drone, that is, by extracting the characteristics of the environment, combining the scene data collection technology that is resistant to airflow interference and the scene data collection technology that is resistant to strong light interference to obtain accurate odometer information, and when the drone moves around, it makes new observations and determines its own position information, and relies on accurate odometer information to accumulate and superimpose sensor information, so as to achieve original point cloud data acquisition and help subsequent map construction;

[0057] In a specific implementation of the embodiment of the present invention, the scene data acquisition technology with resistance to airflow interference performs edge contour detection and pixel information fusion processing on the collected laser radar scanning image, and uses a multi-level wavelet scale decomposition method to reduce image noise, thereby improving the clarity of the original point cloud data imaging;

[0058] In a specific implementation of the embodiment of the present invention, the scene data acquisition technology that resists strong light interference calculates the spatial three-dimensional coordinates of feature points in a specified area and their projections on different sub-eye images through spatial geometric association and stereoscopic vision of multiple sub-eyes (i.e., multiple sensors). When a sub-eye or part of the sub-eyes is strongly disturbed by light or background, the feature extraction results of other sub-eyes and the current viewing angle are used to focus on the specified area for feature extraction, thereby improving the clarity of the original point cloud data imaging;

[0059] In a specific implementation of the embodiment of the present invention, the laser point cloud data preprocessing method uses a 3D laser scanner to scan the point cloud of the object and add model texture, and performs image registration, and through point cloud stitching, all point cloud data are spliced ​​using a coordinate matrix to fully reflect the point cloud characteristics of the scanned object. Noise and redundant point cloud data are deleted in the preprocessing stage: relatively obvious external points can be manually selected and deleted, and inconspicuous external points can be screened and deleted using point cloud post-processing software;

[0060] In a specific implementation of the embodiment of the present invention, the automatic identification of digital objects in the substation first constructs the domain knowledge of the substation and the formula based on the first-order logic. Through the automatic recognition algorithm of the image target, the identified target is mapped to the instance of the domain ontology of the substation. The spatial position relationship between the image targets corresponds to the object association attribute between the instances in the domain ontology. Then, the specified specific target is identified from the visual data of the image, and the image is preprocessed to convert the image into a feature vector. The CNN convolutional neural network can be selected to extract the target to be identified as a feature. The image features obtained by the full link layer are classified by training the optimal (near) decision tree through the construction algorithm. After the features of the target are identified, they are mapped to the object attribute association relationship between the domain ontology instances, and the decision binary tree is obtained to test the image, and the logical reasoning based on the constructed domain knowledge is mined to obtain the retrieval result.

[0061] In a specific implementation of an embodiment of the present invention, the point cloud feature recognition and extraction processing adopts a CNN convolutional neural network method. First, a single layer of neurons is constructed layer by layer, and a single layer network is trained each time. The weights between layers except the top layer are changed into bidirectional weights. In this way, the top layer is still a single layer neural network, while other layers become graph models. Then, the wake-sleep algorithm is used to adjust all weights to make cognition and generation consistent, ensuring that the generated top layer can restore the underlying nodes as much as possible, which is conducive to feature extraction.

[0062] In a specific implementation of the embodiment of the present invention, the full-coverage flight path design method calculates the basic mathematical model of drone track planning through three-dimensionally reconstructed substation cloud data, drone fuselage structure and physical characteristics, inspection target characteristics, etc., uses the A* algorithm to solve the shortest path problem between two points in the substation space that meets the constraints of the mathematical model, and uses the genetic algorithm to select the one with the best fitness among all the shortest paths finally generated as the optimal solution of the algorithm;

[0063] In a specific implementation of the embodiment of the present invention, the drone control locking strategy adopted by the autonomous inspection control of the drone includes a drone airspace locking strategy, a ground proximity locking strategy, a dangerous weather locking strategy, and a flight conflict warning strategy. Based on the analysis of the drone track, the drone is identified and warned of the dangers that may be encountered according to the current flight situation;

[0064] In a specific implementation of an embodiment of the present invention, the one-stop data visualization development platform is adapted to a variety of data sources on and off the cloud, provides a rich variety of 2D and 3D visualization components, adopts drag-and-drop free layout, and integrates substation-related information into a highly versatile and extensible data visualization platform through visualization components, location information visualization, graphical editing interface and support for multiple data sources, so as to quickly customize and apply large data screens.

[0065] In summary, the present invention collects point cloud data of the substation through laser radar and establishes a high-precision three-dimensional model, so that the route planning function can be carried out in the established three-dimensional model of the substation in a more intuitive way, reducing the number of times the route planners go to the site, and the efficiency is improved; the route planning based on the three-dimensional model more accurately specifies the spatial position, shooting distance and shooting angle of the drone during the patrol, ensuring the standardization of each patrol operation. Under the premise of fully ensuring the flight safety of the drone substation, the patented method of the present invention focuses on improving the reliability and practicality of the patrol equipment, and through carrying auxiliary manual equipment (building) patrol, infrared temperature measurement, severe weather and dangerous environment patrol and other special patrols, it focuses on realizing the discovery of hidden dangers and defects such as rust, looseness, and falling of high-altitude equipment such as substation architecture, lightning rods, and busbars, and establishes an equipment patrol system that combines intelligent robot (high-definition video) patrol, drone patrol, manual patrol and professional live detection, reducing the workload of operation and maintenance personnel, reducing the risk of high-altitude patrol, and further improving the quality and efficiency of equipment patrol.

[0066] Example 2

[0067] The embodiment of the present invention provides a substation power drone automatic inspection device, including: RTK, GPS, image collector and laser radar;

[0068] The laser radar is used to collect point cloud data for the substation;

[0069] The image collector is used to acquire images inside the substation;

[0070] Based on the collected point cloud data and images in the substation, as well as the real-time location information of the drone, the drone is used to complete the three-dimensional reconstruction of the unknown environment; the real-time location information of the drone is calculated based on the output information of RTK and GPS;

[0071] Based on the results of 3D reconstruction, route planning is performed to control the substation power drone for automatic inspection.

[0072] The rest are the same as in Example 1.

[0073] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A substation power drone automatic inspection method, characterized in that: include: Use LiDAR to collect point cloud data of substations; Use image collector to obtain images inside the substation; Based on the collected point cloud data, images inside the substation, and the real-time location information of the drone, the drone can perform 3D reconstruction of the unknown environment. Based on the results of 3D reconstruction, route planning is performed to control the substation power drone for automatic inspection; The method of collecting point cloud data of the substation by using a laser radar and acquiring images in the substation by using an image collector specifically includes the following steps: According to the structure of the UAV, the ICP algorithm based on geometric features is used to calculate the installation position coordinates of RTK, GPS, image collector and lidar on the UAV, and obtain the installation position parameters at the coordinate position; According to the obtained installation position parameters, single-point positioning is performed on RTK, GPS, image collector and lidar respectively to generate their own independent coordinate systems; Select the origin of the drone as the origin of the coordinate system, calculate the spatial position coordinates of RTK, GPS, image collector and lidar according to the installation position coordinates, derive the transformation relationship between each independent coordinate system and the origin of the drone, establish an object-image relationship model, derive the calculation formula between the object-image point coordinates and the camera's external orientation elements, and the ground coordinate calculation formula of the laser point cloud; During the normal flight of the drone, GPS and RTK record the flight trajectory of the drone in real time, while the image collector and lidar use space automatic scanning technology to obtain images and point cloud data in the substation; The three-dimensional reconstruction of the unknown environment comprises the following steps: The positioning module receives point cloud data output by the laser radar in real time. The positioning module includes GPS and RTK. Combined with the flight trajectory of the drone, the real-time positioning of the drone in an unknown environment is completed through a specific algorithm. The output of the positioning module is a three-dimensional laser radar odometer, that is, the change in posture at the current moment relative to the original posture; The point cloud data output by the laser radar is used for coloring and adding model texture, and the image in the substation is combined for image registration. All point cloud data are spliced ​​using a coordinate matrix through point cloud stitching to fully reflect the point cloud features of the scanned object; By fusing the point cloud data output by the image collector and the lidar, radar point cloud information is generated. Combined with the location information obtained by the real-time positioning of the drone, the calculation formula between the object point coordinates and the camera's external orientation elements, as well as the ground coordinate calculation formula of the laser point cloud, the point cloud data fused with the radar point cloud information and the sensor is constructed to achieve three-dimensional reconstruction of the unknown environment by the drone.

2. The method for automatic inspection of power substation by drone according to claim 1 is characterized in that: The method further comprises: using the positioning module to receive the point cloud data output by the laser radar in real time, combining the flight trajectory of the drone, and completing the real-time positioning of the drone in an unknown environment through a specific algorithm, and further comprising: Identify and warn of possible dangers that the drone may encounter based on its current flight posture, adopt the drone control locking strategy, and execute predetermined strategies such as returning to the original route, prohibiting takeoff, and landing on the spot to ensure the safety of the drone.

3. The method for automatic inspection of power substation by drone according to claim 1 is characterized in that: The route planning based on the result of the three-dimensional reconstruction specifically includes the following steps: Use the 3D reconstructed point cloud data to build substation domain knowledge and construct formulas based on first-order logic; Through the automatic recognition algorithm of image objects, the identified objects are mapped to instances of the substation domain ontology, and the spatial position relationship between image objects corresponds to the object association attributes between instances in the domain ontology; Using the identified target mapping, image preprocessing is performed to convert the image into a feature vector. The CNN convolutional neural network is selected to perform feature recognition and extraction on the feature vector of the identified target. Then, the optimal decision tree is trained by building a deep algorithm for classification. The extracted target features are mapped to the object attribute association relationship between domain ontology instances, the image is tested using a decision binary tree, and logical reasoning is performed based on the constructed domain knowledge to perform retrieval; A full-coverage flight strip design method is adopted, combined with the optimal decision tree and retrieval results, to realize the initial polygonal flight strip design. Then, the inspection area is divided based on the initial polygonal flight strip. While ensuring that the entire area can be covered, the route calculation is performed for each area to achieve precise route design for the substation scenario.

4. The method for automatic inspection of power substation by drone according to claim 1 is characterized by: The step of performing route planning based on the result of the three-dimensional reconstruction further includes: Carry out automatic inspection of power substation by drone, and the results are displayed and presented in a one-stop data visualization development platform. After completion, check whether the results meet the substation inspection requirements.

5. A substation power drone automatic inspection device, characterized in that: include: RTK, GPS, image collector and LiDAR; The laser radar is used to collect point cloud data for the substation; The image collector is used to acquire images inside the substation; Based on the collected point cloud data and images in the substation, as well as the real-time location information of the drone, the drone is used to complete the three-dimensional reconstruction of the unknown environment; the real-time location information of the drone is calculated based on the output information of RTK and GPS; Based on the results of 3D reconstruction, route planning is performed to control the substation power drone for automatic inspection; The laser radar is used to collect point cloud data of the substation, and the image collector is used to obtain the image inside the substation. The specific steps include: According to the structure of the UAV, the ICP algorithm based on geometric features is used to calculate the installation position coordinates of RTK, GPS, image collector and lidar on the UAV, and obtain the installation position parameters at the coordinate position; According to the obtained installation position parameters, single-point positioning is performed on RTK, GPS, image collector and lidar respectively to generate their own independent coordinate systems; Select the origin of the drone as the origin of the coordinate system, calculate the spatial position coordinates of RTK, GPS, image collector and lidar according to the installation position coordinates, derive the transformation relationship between each independent coordinate system and the origin of the drone, establish an object-image relationship model, derive the calculation formula between the object-image point coordinates and the camera's external orientation elements, and the ground coordinate calculation formula of the laser point cloud; During the normal flight of the drone, GPS and RTK record the flight trajectory of the drone in real time, while the image collector and lidar use space automatic scanning technology to obtain images and point cloud data in the substation; The 3D reconstruction of an unknown environment includes the following steps: The positioning module receives point cloud data output by the laser radar in real time. The positioning module includes GPS and RTK. Combined with the flight trajectory of the drone, the real-time positioning of the drone in an unknown environment is completed through a specific algorithm. The output of the positioning module is a three-dimensional laser radar odometer, that is, the change in posture at the current moment relative to the original posture; The point cloud data output by the laser radar is used for coloring and adding model texture, and the image in the substation is combined for image registration. All point cloud data are spliced ​​using a coordinate matrix through point cloud stitching to fully reflect the point cloud features of the scanned object; By fusing the point cloud data output by the image collector and the lidar, radar point cloud information is generated. Combined with the location information obtained by the real-time positioning of the drone, the calculation formula between the object point coordinates and the camera's external orientation elements, as well as the ground coordinate calculation formula of the laser point cloud, the point cloud data fused with the radar point cloud information and the sensor is constructed to achieve three-dimensional reconstruction of the unknown environment by the drone.

Images