Laser radar SLAM and GNSS unmanned aerial vehicle global inspection system and method

Abstract

The invention provides a laser radar SLAM (Simultaneous Localization and Mapping) and GNSS (Global Navigation Satellite System) unmanned aerial vehicle global inspection system and method. The system comprises an unmanned aerial vehicle, an outdoor nest and a cloud server, the unmanned aerial vehicle comprises a laser radar SLAM module and a GNSS module; wherein communication connection is established between the unmanned aerial vehicle and the outdoor nest and between the outdoor nest and the cloud server through a 4G cellular network; the unmanned aerial vehicle comprises a plurality of navigation modes; the method comprises the following steps: step 1, performing laser radar SLAM unmanned aerial vehicle repetitive route execution calibration; step 2, deploying an outdoor nest; the cloud server deploys an inspection route and initializes inspection parameters; 3, judging the environment according to the navigation signal of the unmanned aerial vehicle; 4, in a laser radar SLAM navigation mode, a GNSS navigation mode and a fusion navigation mode, the unmanned aerial vehicle reads parameters and transmits the parameters to the cloud server side, and the cloud server side calculates and converts the parameters into trusted unmanned aerial vehicle coordinates. The flight adaptability of the unmanned aerial vehicle is improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent inspection technology for UAV engineering, and in particular to a laser radar SLAM and GNSS UAV full-area inspection system and method. Background Art

[0002] As infrastructure construction enters the stage of "equal emphasis on construction and maintenance", the traditional manual inspection model has the following problems: insufficient coverage and terrain restrictions; low efficiency of manual inspection; relying on working platforms such as aerial platforms or special inspection vehicles for detailed inspections, which requires large investment of manpower and equipment and high costs; high-altitude operations have high risks, and the risks of operations under severe weather conditions are particularly prominent.

[0003] Existing drone technology has been maturely applied in open scenarios such as power inspections, but the following technical bottlenecks still exist in inspections in GNSS-denied environments such as bridges, tunnels, and underground spaces: ordinary GNSS-navigated drones cannot effectively obtain positioning coordinates, resulting in drone loss of control or positioning drift. When the GNSS signal in the inspection environment is unstable, accidents such as flight loss and loss of connection are prone to occur; when ordinary GNSS drones need to carry out inspections in GNSS-denied environments, not only do they require human assistance, but it is also difficult to automate the entire inspection process; if the GNSS-denied operating space is beyond line of sight, ordinary GNSS drones are extremely difficult to operate, and equipment safety is difficult to ensure during the inspection process. Summary of the Invention

[0004] The present application provides a lidar SLAM and GNSS UAV global inspection system and method, which can be used to solve the technical problem of low navigation accuracy of UAVs in a single mode.

[0005] This application provides a LiDAR SLAM and GNSS UAV global inspection system, which includes:

[0006] It includes drones, outdoor nests and cloud servers; the drone contains a lidar SLAM module and a GNSS module; the lidar SLAM module is used to obtain three-dimensional point cloud perception data of the surrounding environment and calculate the relative coordinates of the drone and the surrounding environment in real time; the GNSS module is used to obtain GNSS coordinates.

[0007] Among them, communication connections are established between the drone and the outdoor nest, and between the outdoor nest and the cloud server through the 4G cellular network; when the drone is flying, flight parameters including altitude, speed, relative coordinates, battery power and three-dimensional point cloud perception data are sent to the nest through the 4G cellular network, and the nest forwards the data to the cloud server.

[0008] The drone includes multiple navigation modes; the navigation modes include: lidar SLAM navigation mode, fusion navigation mode, and GNSS navigation mode.

[0009] This application also provides a lidar SLAM and GNSS drone global inspection method, which is based on the system provided in this application and includes:

[0010] Step 1: perform repetitive route execution calibration of the LiDAR SLAM UAV;

[0011] Step 2: Deploy the outdoor machine nest; the cloud server deploys the inspection route and initializes the inspection parameters;

[0012] Step 3: Determine the environment based on the drone navigation signal;

[0013] Step 4: In the LiDAR SLAM navigation mode / GNSS navigation mode and fusion navigation mode, the UAV reads the parameters and transmits them to the cloud server, which calculates and converts them into trusted UAV coordinates. When the UAV is in fusion navigation mode, the cloud server calculates the weighted global coordinates under the two navigation modes using a real-time confidence dynamic weight allocation method.

[0014] The technical advantages of the present invention are mainly reflected in the following three aspects:

[0015] 1. Global navigation and autonomous inspection of bridges. By integrating LiDAR SLAM with GNSS navigation, drones can seamlessly switch between navigation modes in open areas and GNSS-denied areas such as bridge bottoms and piers, ensuring full coverage of the drone inspection range. Drones can autonomously locate and navigate even in GNSS-denied areas, overcoming the technical bottleneck of traditional drones relying on GNSS signals and enabling fully automated inspections.

[0016] 2. Positioning stability control method for navigation switching. During the navigation mode switching process, the system uses a weighted fusion algorithm and environmental prediction to achieve a smooth transition, avoiding positioning drift and loss of control, and ensuring the stability and safety of the UAV flight.

[0017] 3. Repeated Path Execution Stability Control Method: This method uses iterative correction and reverse compensation mechanisms to ensure the positioning accuracy of the drone when repeatedly executing routes in the LiDAR SLAM navigation mode. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The overall system architecture diagram provided for the embodiment of this application;

[0019] Figure 2 A flowchart for determining the drone navigation mode provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0021] The following first introduces the embodiments of the present application with reference to the accompanying drawings.

[0022] A laser radar SLAM and GNSS drone full-area inspection system includes a drone, an outdoor drone nest and a cloud server; the drone contains a laser radar SLAM module and a GNSS module; the laser radar SLAM module is used to obtain three-dimensional point cloud perception data of the surrounding environment and calculate the relative coordinates of the drone and the surrounding environment in real time; the GNSS module is used to obtain GNSS coordinates.

[0023] Among them, communication connections are established between the drone and the outdoor nest, and between the outdoor nest and the cloud server through the 4G cellular network; when the drone is flying, flight parameters including altitude, speed, relative coordinates, battery power and three-dimensional point cloud perception data are sent to the nest through the 4G cellular network, and the nest forwards the data to the cloud server.

[0024] The drone includes multiple navigation modes; navigation modes include: lidar SLAM navigation mode, fusion navigation mode, GNSS navigation mode

[0025] The hierarchical state logic is determined based on multi-dimensional parameter judgment; the navigation mode judgment criteria are: M uva When θ1≤θ1, the UAV is in the LiDAR SLAM navigation mode; θ1<M uva When θ<θ2, the UAV is in fusion navigation mode; M uva When ≥θ2, the UAV is in GNSS navigation mode; θ1 and θ2 are critical values, which are determined according to the environment. uva It is the drone navigation mode.

[0026] The global domain includes areas with strong GNSS signals, areas with weak GNSS signals, and areas with no GNSS signals. Global navigation is achieved using the drone's global coordinates as the positioning basis. The drone's global coordinates in areas with strong and weak GNSS signals can be directly obtained, while the drone's global coordinates in areas with no GNSS signals are obtained through conversion.

[0027] When in the critical state of GNSS navigation and lidar SLAM navigation, the weighted global coordinate P is calculated by using the real-time confidence dynamic weight allocation method. cal , use P cal It serves as the navigation reference coordinate and is transmitted back to the cloud server to ensure smooth transition of the drone's navigation mode switching.

[0028] The route offset after each iteration is calculated as the loss, and the spatial offset between waypoints is reversely compensated to the flight parameters of the UAV's route execution. The displacement deviation during the route execution process is calibrated, and the accuracy of the UAV's repeated route execution in the lidar SLAM navigation mode is guaranteed through an iterative correction mechanism.

[0029] The application method of UAV bridge full-area inspection is as follows: the UAV and the airport are deployed at the ideal location for power supply and network conditions in the bridge area to be inspected; the cloud service end demarcates the inspection area and plans the waypoint route, and the waypoint location is not constrained by the GNSS signal; the UAV takes off from the nest position, and when the UAV is unlocked and ready to fly, it reads the current GNSS coordinate P0 and the initial heading angle R0 of the UAV (the default is 0°, right is positive); when the UAV is in GNSS navigation mode, it sends back the GNSS coordinate P ai ; When the UAV is in navigation fusion mode, it will also send back GNSS coordinates P ai and the laser radar SLAM relative coordinate P ri , and return the global coordinate P when it is first in navigation fusion mode g , heading angle R g When the UAV enters the LiDAR SLAM navigation mode, it transmits the LiDAR SLAM relative coordinates P ri .

[0030] A laser radar SLAM and GNSS UAV global inspection method, the method comprising:

[0031] The first step is to calibrate the repetitive route execution of the LiDAR SLAM UAV.

[0032] In an environment without GNSS signals, the cloud server sets waypoints according to the laser radar SLAM navigation coordinate system to form a theoretical route. The coordinates of the i-th waypoint in the theoretical route are P(i)=(X i ,Y i ,Z i );

[0033] Subsequently, the cloud server sends a command to the machine nest, which forwards the command to the drone. The drone is in the lidar SLAM navigation mode and repeats the route k times. The coordinates of the actual waypoint i after the kth drone executes the route are The root mean square error between the theoretical path and the actual path after the UAV executes the route for the kth time is taken as the iterative loss after the kth execution of the route:

[0034]

[0035] in, is the offset between the actual waypoint and the theoretical waypoint, expressed as the modulus of the space vector from the ith actual waypoint to the theoretical waypoint when the route is executed for the kth time;

[0036] Secondly, the arithmetic mean of the k-iteration loss is used as the stage loss L total :

[0037]

[0038] Secondly, calculate the spatial offset compensation ΔS of the i-th waypoint i , Among them S i is the displacement of the i-th waypoint, and S is the total spatial displacement of the route;

[0039] Then, the UAV flight parameter matrix T is obtained when executing the kth route. k , T k is a second-order tensor of (i×6), then T k Correction ΔT k Expressed as:

[0040]

[0041] Then, ΔT is established through the relationship matrix C k With ΔS i The relationship between them is used to correct the flight parameters by using the spatial displacement compensation between waypoints:

[0042] ΔT k =ΔS i ·C.

[0043] The second step is to deploy the outdoor machine nest; the cloud server deploys the inspection route and initializes the inspection parameters.

[0044] Deploy outdoor drone nests: Select areas with good GNSS and 4G communication signals near the bridge to be inspected to ensure safe takeoff and landing of drones. Import the bridge BIM model or 3D point cloud model through the cloud server to define the inspection area boundaries.

[0045] The cloud server deploys inspection routes: The cloud server generates inspection routes based on the structural characteristics of the bridge. The routes can cover the bridge deck, bridge bottom, piers and supports. Users adjust the routes, flight altitude, set data collection points and navigation mode switching points on the cloud server.

[0046] The cloud server initializes the inspection parameters: the cloud server sends a command to control the opening of the machine nest. After the machine nest is opened, it forwards the command to control the drone to unlock and start executing the route mission; when taking off, the drone is in GNSS navigation mode and sends back the coordinates. The cloud server reads the GNSS coordinates at the time of takeoff and unlocking The initial heading angle R0 is used as a coordinate system parameter to establish a global coordinate system. P0 (0, 0, 0) is the origin of the coordinate system, R0 is the positive direction of the X axis of the coordinate system, and λ0 is the longitude. is the latitude.

[0047] The third step is to determine the environment based on the drone navigation signal.

[0048] First, the drone navigation mode M uva Determined by three parameters: the carrier-to-noise density ratio C / N0 of the GNSS signal, the number of detectable satellites n GNSS and HDOP (Horizontal Dilution of Precision); the UAV navigation mode parameters are determined by the following method:

[0049]

[0050] Among them, w a , w b and w c are the initial weights of the three parameters respectively; λ n and λ HDOP is the carrier-to-noise density ratio C / N0, the number of detectable satellites n GNSS and the normalized parameters of HDOP, the normalized denominator is determined by the best value within 15 seconds after the drone is unlocked.

[0051] Subsequently, according to M uva Numerical value for navigation mode determination: M uva When ≤θ1, the UAV is in SLAM navigation mode; θ1<M uva When θ<θ2, the UAV is in fusion navigation mode; M uva When ≥2, the UAV is in GNSS navigation mode; θ1 and θ2 are the critical maximum and minimum values, which are obtained from experiments.

[0052] In the fourth step, when the drone is in LiDAR SLAM navigation mode, GNSS navigation mode, or fusion navigation mode, the drone reads the parameters and transmits them to the cloud server, which calculates and converts them into trusted drone coordinates. When the drone is in fusion navigation mode, the cloud server calculates the weighted global coordinates of the two navigation modes using a method that dynamically assigns weights based on real-time confidence.

[0053] Step 41: When the UAV is in GNSS navigation mode, the GNSS coordinates sent back by the UAV at any time are Height H ai ; The cloud server receives GNSS coordinates and converts them into global coordinate system coordinates P GNSS (X, Y, Z); the coordinate conversion formula from GNSS coordinates to the global coordinate system is:

[0054] X=(N+h)·cosφ·cosλ;

[0055] Y=(N+h)·cosφ·sinλ;

[0056] Z=(N·(1-e 2 )+h)·sinφ;

[0057] Where N is the radius of curvature of the reference ellipsoid: a is the length of the Earth's semi-major axis; e is the eccentricity of the ellipsoid;

[0058] Step 42: When the drone is in fusion navigation mode, it will send back GNSS coordinates at the same time. and the laser radar SLAM relative coordinate P l (x ri ,y ri ,z ri ), the cloud server reads the GNSS global coordinates P when it is in navigation fusion mode for the first time gr (X r ,Y r ,Z r ), the heading angle θ is used as the coordinate system parameter to establish the local coordinate system, P gr The local coordinate system origin is located at the X-axis, and the Y-axis is located at 90° to the right of the aircraft head.

[0059] When the UAV is in fusion navigation mode, the relative coordinate P l (x ri ,y ri ,z ri ) is mapped to the global coordinate system: SLAM =R·P l +T;

[0060] Where, P SLAM is the global coordinate mapped from the relative coordinate of the laser radar SLAM; R represents the rotation matrix of the local coordinate system relative to the global coordinate system; T represents the base point drift of the local coordinate system relative to the global coordinate system; T g-l is the local coordinate system origin offset:

[0061] T g-l =P gr -P0=(X r ,Y r ,Z r )-(0,0,0)=[X r ,Y r ,Z r ] T ; R is the coordinate system rotation matrix, θ is the angular offset of the local coordinate system relative to the global coordinate system around the Z axis, that is, the yaw angle when establishing the local coordinate system;

[0062] Step 43: To ensure accurate and stable coordinate control when the UAV is in fusion navigation mode, a real-time confidence dynamic weight allocation method is used. The cloud server calculates the weighted global coordinates under the two navigation modes, uses the weighted global coordinates as the navigation reference coordinates, and transmits the corresponding instructions back to the UAV. The weighted coordinate calculation formula is:

[0063] P cal =W GNSS ·P GNSS +W SLAM ·P SLAM

[0064] Among them, P cal The coordinates are calculated by real-time weighting; W SLAM =1-W GNSS

[0065] When the UAV is in fusion navigation mode, the edge computing unit calculates P per second GNSS and P SLAM When the absolute error is greater than 0.1m, the drone is controlled to return to the previous waypoint, re-acquire the parameters to establish the local coordinate system, and then try to execute the mission again;

[0066] Step 44: When the UAV is in the LiDAR SLAM navigation mode, SLAM The coordinates are used as navigation reference coordinates and sent back to the cloud server; the flight control system uses P SLAM The coordinates continue to execute the route to ensure the continuity of the route; the drone hovers at the preset collection point, collects structural surface data through multiple sensors, and the edge computing module compresses and transmits it back to the cloud service end. The drone navigation mode determination logic is as shown in the attached manual. Figure 2 shown.

[0067] Finally, the drone returns to the nest and uploads the complete data set to the cloud server.

[0068] The above-described embodiments of the present application do not constitute a limitation on the scope of protection of the present application.

Claims

1. A laser radar SLAM and GNSS UAV global inspection system, characterized by: The system comprises: It includes drones, outdoor nests and cloud servers; the drone contains a lidar SLAM module and a GNSS module; the lidar SLAM module is used to obtain 3D point cloud perception data of the surrounding environment and calculate the relative coordinates of the drone and the surrounding environment in real time; the GNSS module is used to obtain GNSS coordinates; Among them, the drone and the outdoor nest, and the outdoor nest and the cloud server establish communication connections through the 4G cellular network. When the drone is flying, it sends flight parameters and 3D point cloud perception data to the nest through the 4G cellular network, and the nest forwards the data to the cloud server. The drone includes multiple navigation modes; the navigation modes include: lidar SLAM navigation mode, fusion navigation mode, and GNSS navigation mode.

2. The system according to claim 1, wherein: The navigation mode used by the drone is determined by the following critical values: Navigation mode criteria: M uva When θ1≤θ1, the UAV is in the LiDAR SLAM navigation mode; θ1<M uva When θ<θ2, the UAV is in fusion navigation mode; M uva When ≥θ2, the UAV is in GNSS navigation mode; θ1 and θ2 are critical values, which are determined according to the environment. uva It is the drone navigation mode.

3. The system according to claim 2, characterized in that When the UAV is in the critical state of GNSS navigation and lidar SLAM navigation, the weighted global coordinate P is calculated by using the real-time confidence dynamic weight allocation method. cal , use P cal Serves as navigation reference coordinates and transmits them back to the cloud server; The route offset after each iteration is calculated as the loss, and the spatial offset between waypoints is reversely compensated to the flight parameters of the UAV's route execution. The displacement deviation during the route execution process is calibrated, and the accuracy of the UAV's repeated route execution in the lidar SLAM navigation mode is guaranteed through an iterative correction mechanism.

4. The system according to claim 3, characterized in that The data transmitted by the drone to the cloud server includes: The drone takes off from the nest position. When the drone is unlocked and ready to fly, it reads the current GNSS coordinate P0 and the drone's initial heading angle R0 value of 0°, with rightward being positive; When the drone is in GNSS navigation mode, it sends back the GNSS coordinates P ai ; When the drone is in navigation fusion mode, it will also send back the GNSS coordinates P ai and the laser radar SLAM relative coordinate P ri , and return the global coordinate P when it is first in navigation fusion mode g , heading angle R g ; When the UAV enters the LiDAR SLAM navigation mode, it transmits the LiDAR SLAM relative coordinates P ri .

5. A lidar SLAM and GNSS drone global inspection method, the method being implemented based on the system of claims 1 to 4, the method comprising: Step 1: perform repetitive route execution calibration of the LiDAR SLAM UAV; Step 2: Deploy the outdoor machine nest; The cloud server deploys inspection routes and initializes inspection parameters; Step 3: Determine the environment based on the drone navigation signal; Step 4: When the drone is in the lidar SLAM navigation mode, GNSS navigation mode, and fusion navigation mode, the drone reads the parameters and transmits them to the cloud server, which calculates and converts them into trusted drone coordinates. When the drone is in fusion navigation mode, the cloud server calculates the weighted global coordinates under the two navigation modes using a real-time confidence dynamic weight allocation method.

6. The method according to claim 5, characterized in that Step 1: Calibrate the repetitive route execution of the LiDAR SLAM UAV, including: In an environment without GNSS signals, the cloud server sets waypoints according to the laser radar SLAM navigation coordinate system to form a theoretical route. The coordinates of the i-th waypoint in the theoretical route are P(i)=(X i ,Y i ,Z i ); Subsequently, the cloud server sends a command to the machine nest, which forwards the command to the drone. The drone is in the lidar SLAM navigation mode and repeats the route k times. The coordinates of the actual waypoint i after the kth drone executes the route are The root mean square error between the theoretical path and the actual path after the UAV executes the route for the kth time is taken as the iterative loss after the kth execution of the route: in, is the offset between the actual waypoint and the theoretical waypoint, expressed as the modulus of the space vector from the ith actual waypoint to the theoretical waypoint when the route is executed for the kth time; Secondly, the arithmetic mean of the k-iteration loss is used as the stage loss L total : Secondly, calculate the spatial offset compensation ΔS of the i-th waypoint i , Among them S i is the displacement of the i-th waypoint, and S is the total spatial displacement of the route; Then, the UAV flight parameter matrix T is obtained when executing the kth route. k , T k is a second-order tensor of (i×6), then T k Correction ΔT k Expressed as: Then, ΔT is established through the relationship matrix C k With ΔS i The relationship between them is used to correct the flight parameters by using the spatial displacement compensation between waypoints: ΔT k =ΔS i ·C。 7. The method according to claim 5, characterized in that Step 2: Deploy the outdoor machine nest; The cloud server deploys the inspection route and initializes the inspection parameters, including: Deploy outdoor inspection nests: Deploy outdoor inspection nests in areas with good GNSS and 4G communication signals near the bridge to be inspected. Import the bridge BIM model or 3D point cloud model through the cloud server to define the inspection area boundaries. The cloud server deploys inspection routes: The cloud server generates inspection routes based on the structural characteristics of the bridge, covering the bridge deck, bridge bottom, piers and supports; users adjust the routes, flight altitude, set data collection points and navigation mode switching points on the cloud server. The cloud server initializes the inspection parameters: the cloud server issues a command to control the opening of the machine nest. After the machine nest is opened, it forwards the command to control the drone to unlock and start executing the route mission; when taking off, the drone is in GNSS navigation mode and sends back the coordinates. The cloud server reads the GNSS coordinates at the time of takeoff and unlocking The initial heading angle R0 is used as a coordinate system parameter to establish a global coordinate system. P0 (0, 0, 0) is the origin of the coordinate system, R0 is the positive direction of the X axis of the coordinate system, and λ0 is the longitude. is the latitude.

8. The method according to claim 5, characterized in that Step 3: Determine the environment based on the drone navigation signal, including: First, the drone navigation mode M uva Determined by three parameters: the carrier-to-noise density ratio C / N0 of the GNSS signal, the number of detectable satellites n GNSS and HDOP (Horizontal Dilution of Precision); the UAV navigation mode parameters are determined by the following method: Among them, w a , w b and w c are the initial weights of the three parameters respectively; λ n and λ HDOP is the carrier-to-noise density ratio C / N0, the number of detectable satellites n GNSS and the normalized parameters of HDOP, the normalized denominator is determined by the best value within 15 seconds after the drone is unlocked. Subsequently, according to M uva Numerical value for navigation mode determination: M uva When ≤θ1, the UAV is in SLAM navigation mode; θ1<M uva When θ<θ2, the UAV is in fusion navigation mode; M uva When θ1 is greater than or equal to θ2, the UAV is in GNSS navigation mode; θ1 and θ2 are the critical maximum and minimum values, which are obtained from experiments.

9. The method according to claim 5, characterized in that Step 4: When the drone is in LiDAR SLAM navigation mode, GNSS navigation mode, or fusion navigation mode, the drone reads the parameters and transmits them to the cloud server, which calculates and converts them into trusted drone coordinates. This includes: Step 41: When the UAV is in GNSS navigation mode, the GNSS coordinates sent back by the UAV at any time are Height H ai ; The cloud server receives GNSS coordinates and converts them into global coordinate system coordinates P GNSS (X, Y, Z); the coordinate conversion formula from GNSS coordinates to the global coordinate system is: X=(N+h)·cosφ·cosλ; Y=(N+h)·cosφ·sinλ; Z=(N·(1-e 2 )+h)·sinφ; Where N is the radius of curvature of the reference ellipsoid: a is the length of the Earth's semi-major axis; e is the eccentricity of the ellipsoid; Step 42: When the drone is in fusion navigation mode, it will send back GNSS coordinates at the same time. and the laser radar SLAM relative coordinate P l (x ri ,y ri ,z ri ), the cloud server reads the GNSS global coordinates P when it is in navigation fusion mode for the first time gr (X r ,Y r ,Z r ), the heading angle θ is used as the coordinate system parameter to establish the local coordinate system, P gr The local coordinate system origin is located at the X-axis, and the Y-axis is located at 90° to the right of the aircraft head. When the UAV is in fusion navigation mode, the relative coordinate P l (x ri ,y ri ,z ri ) is mapped to the global coordinate system: SLAM =R·P l +T; Where, P SLAM is the global coordinate mapped from the relative coordinate of the laser radar SLAM; R represents the rotation matrix of the local coordinate system relative to the global coordinate system; T represents the base point drift of the local coordinate system relative to the global coordinate system; T g-l is the local coordinate system origin offset: T g-l =P gr -P0=(X r ,Y r ,Z r )-(0,0,0)=[X r ,Y r ,Z r ] T ; R is the coordinate system rotation matrix, θ is the angular offset of the local coordinate system relative to the global coordinate system around the Z axis, that is, the yaw angle when establishing the local coordinate system; Step 43: When the UAV is in the fusion navigation mode, the cloud server calculates the weighted global coordinates under the two navigation modes using a real-time confidence dynamic weight allocation method. The weighted global coordinates are used as navigation reference coordinates and the corresponding instructions are sent back to the UAV. The weighted coordinate calculation formula is: P cal =W GNSS ·P GNSS +W SLAM ·P SLAM Among them, P cal The coordinates are calculated by real-time weighting; W SLAM =1-W GNSS When the UAV is in fusion navigation mode, the edge computing unit calculates P per second GNSS and P SLAM When the absolute error is greater than 0.1m, the drone is controlled to return to the previous waypoint, re-acquire the parameters to establish the local coordinate system, and then try to execute the mission again; Step 44: When the UAV is in the LiDAR SLAM navigation mode, SLAM The coordinates are used as navigation reference coordinates and sent back to the cloud server; the flight control system uses P SLAM The coordinates continue to execute the route to ensure route continuity; the drone hovers at the preset collection point and collects structural surface data through multiple sensors. The edge computing module compresses and transmits it back to the cloud server; Finally, the drone returns to the nest and uploads the complete data set to the cloud server.

Images