Pressure pipeline axonometric mapping system based on unmanned aerial vehicle

Abstract

The invention relates to the technical field of unmanned aerial vehicle surveying and mapping, and discloses a pressure pipeline axonometric mapping system based on an unmanned aerial vehicle, and the system uses a pipeline anchoring relative navigation module to guide the unmanned aerial vehicle to fly along a line with a target pipeline as a reference; the outer diameter and three-dimensional center point coordinates of the pipeline are obtained in real time through an orthogonal contour scanning and parameter resolving module; a self-adaptive flight and measurement control module is used for adjusting the flight distance according to the measured outer diameter feedback, and adjusting sensor parameters according to the imaging quality feedback; meanwhile, the coordinates of the center point are incrementally constructed into a topological graph through an airborne real-time topological graph construction and node processing module, pipe fitting nodes are autonomously detected and recognized, and then a preset strategy is called to conduct refined scanning on the nodes. According to the invention, autonomous navigation independent of global positioning, self-adaptive closed-loop control of measurement and flight, and airborne integrated real-time modeling of pipe network topology and geometric information are realized, and the surveying and mapping precision, the automation level and the operation efficiency are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) mapping technology, specifically to a pressure pipeline isometric mapping system based on UAVs. Background Technology

[0002] As pressure-bearing special equipment, the periodic inspection, safety evaluation, and integrity management of pressure pipelines during their service life are crucial. Obtaining their precise geometric dimensions, three-dimensional morphology, and topological relationships is the data foundation for deformation analysis, corrosion assessment, stress calculation, and risk prediction. Currently, using drones for pipeline data acquisition has become an emerging technological approach. This typically involves generating a high-density 3D point cloud of the entire scene, followed by model reconstruction and analysis using post-processing software.

[0003] However, existing technologies still have several shortcomings when applied to the field of inspection and testing. First, these systems generally rely on Global Navigation Satellite Systems (GNSS) for positioning, making it difficult to effectively inspect many critical pipelines located in signal-obstructed environments such as indoor spaces and pipe corridors. Second, the traditional workflow of "collecting data first, then processing it offline" not only results in long inspection report cycles, but more importantly, the massive unstructured point cloud makes it difficult to quickly and accurately correlate the inspection data with specific pipe sections or fittings, hindering defect location and tracing. Furthermore, during data acquisition, non-orthogonal scanning postures introduce measurement errors, directly affecting the calculation accuracy of key safety evaluation parameters such as pipe outer diameter and ellipticity; fixed sensor parameters also struggle to handle conditions such as pipe surface corrosion and oil contamination, leading to unstable data quality. Simultaneously, conventional line-of-sight scanning methods cannot obtain the complete geometry of high-stress concentration areas such as valves and flanges, creating data blind spots in the risk assessment of these critical components. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a pressure pipeline isometric mapping system based on UAVs, so as to overcome the problems existing in the current pressure pipeline inspection and testing technology: it is difficult to enter indoor or obstructed environments due to the reliance on global positioning systems; the cumbersome data post-processing process leads to low efficiency in defect location and assessment; and the fixed acquisition method results in inaccurate measurement of key safety evaluation parameters and missing data of high-risk components.

[0005] To address the aforementioned technical problems, this invention provides a pressure pipeline isometric mapping system based on unmanned aerial vehicles (UAVs). The system includes: a pipeline anchoring relative navigation module, an orthogonal contour scanning and parameter calculation module, an adaptive flight and measurement control module, and an airborne real-time topology map construction and node processing module.

[0006] The pipeline anchoring relative navigation module is used to sense the relative pose between the UAV platform and the target pipeline in real time, and generate control commands accordingly to guide the UAV to fly stably along the pipeline axis.

[0007] The orthogonal contour scanning and parameter calculation module is used to actively project linear structured light onto the surface of the pipe during the flight of the UAV and capture the contour formed by it. Through geometric calculation, the outer diameter and three-dimensional center point coordinates of the pipe at the current cross section are calculated in real time.

[0008] The adaptive flight and measurement control module is used to receive the outer diameter output by the orthogonal contour scanning and parameter calculation module, and dynamically adjust the flight control parameters of the pipeline anchoring relative navigation module according to the outer diameter, and / or adjust the sensor operating parameters of the orthogonal contour scanning and parameter calculation module according to the imaging quality of the contour.

[0009] The airborne real-time topology map construction and node processing module is used to construct the topology of the pipeline network from the continuous three-dimensional center point coordinates output by the orthogonal contour scanning and parameter calculation module, and to detect and identify pipe fitting nodes on the pipeline path. After identifying the node, a preset strategy is invoked to control the UAV platform to perform a fine scan of the node.

[0010] In one embodiment, the pipeline anchoring relative navigation module includes a forward-looking binocular stereo vision system for generating a three-dimensional point cloud and segmenting the target pipeline therefrom to extract its central axis.

[0011] Furthermore, the module defines a local coordinate system for the pipeline that is dynamically associated with the UAV based on the extracted central axis; The lateral yaw angle of the UAV's body coordinate system relative to the local coordinate system of the pipeline is calculated in real time. Vertical pitch angle and radial distance ; And based on these three parameters and the target pose error vector The control commands are generated through closed-loop servo control.

[0012] One feasible control law is a proportional-integral-derivative (PID) controller, whose output is the UAV's yaw rate command. Pitch angular velocity command and lateral translation speed command They can be represented as: ; ; ; in, , , These are the PID controller gain coefficients corresponding to the error components. These gain coefficients can be adjusted through experimental tuning or adaptive algorithms.

[0013] In one embodiment, the orthogonal contour scanning and parameter calculation module includes a line laser emitter, a high-speed industrial camera, and a single-axis rotating gimbal for supporting the emitter and camera. To ensure measurement accuracy, the single-axis rotating gimbal is used to implement orthogonal constraint control. It receives the flight velocity vector of the UAV and drives the gimbal to rotate, such that the angle between the laser plane normal vector projected by the line laser emitter and the flight velocity vector approaches 90 degrees.

[0014] In one embodiment, the adaptive flight and measurement control module dynamically adjusts the flight control parameters in the following manner: Anchor the pipeline relative to the target radial distance of the navigation module Set to match the outer diameter measured in real time Related dynamic variables. A feasible computational model is as follows: ; in, At any moment The calculated radial distance the UAV should maintain from the target; At any moment The pipe's outer diameter obtained through real-time measurement; It is the preset field of view coefficient; It is the preset minimum safe offset.

[0015] In one embodiment, the adaptive flight and measurement control module adjusts the sensor operating parameters in the following manner: The imaging quality score of the contour is calculated online. ; Based on the error between the score and the preset target range, the power of the linear laser emitter and the exposure time of the industrial camera in the orthogonal contour scanning and parameter calculation module are adjusted in reverse.

[0016] A feasible model for calculating image quality score is as follows: ; in, The average gray value of the outline. The mean gradient along the contour normal direction. For overexposure rate, and These are the weighting coefficients.

[0017] In one embodiment, the airborne real-time topology graph construction and node processing module maintains a graph data structure in the airborne computer memory. To characterize the topology of the pipeline network. This structure includes: Vertex set Each vertex represents a pipe node and contains a node identifier, node type, and three-dimensional coordinate attributes; Edge set Each side represents a pipe segment connecting two vertices and includes identifiers of the starting and ending vertices of the connection, the average outer diameter, and the centerline trajectory attribute consisting of continuous three-dimensional center point coordinates.

[0018] In one embodiment, the airborne real-time topology map construction and node processing module includes a node classifier for identifying the type of pipe nodes perceived by the forward-looking sensors of the pipeline anchoring relative navigation module. Further, after the node classifier identifies the type of the pipe node, the module calls a refined scanning program corresponding to that node type from a preset scanning strategy library, and switches the UAV's flight mode to execute the program, thereby obtaining the complete three-dimensional geometry of the node.

[0019] This invention provides a pressure pipeline isometric mapping system based on unmanned aerial vehicles (UAVs). It has the following advantages: 1. This invention, by setting up a pipeline anchoring relative navigation module, achieves precise approach and stable tracking of pipelines in complex industrial environments without global positioning signals, solving the problem that traditional UAVs cannot easily enter indoor or pipe gallery areas for inspection and testing. Simultaneously, the adaptive flight and measurement control module dynamically adjusts the flight distance according to the pipeline's outer diameter, ensuring that the sensor is always within its optimal operating range, thus obtaining a high-quality, highly reliable data source for subsequent safety assessments. 2. This invention, through orthogonal contour scanning and parameter calculation modules and their orthogonal constraint control, can obtain precise cross-sectional geometric parameters of pipelines, providing a high-precision data foundation for the quantitative analysis of defects such as deformation, ellipticity, and corrosion in pressure pipelines. Furthermore, the ability to adaptively adjust sensor parameters ensures consistent and clear measurement data even when pipeline surface conditions (such as rust and oil stains) and lighting conditions change, significantly improving the reliability and consistency of inspection and testing results. 3. This invention utilizes an onboard real-time topology map construction and node processing module to generate a structured digital model from survey data in real time. This model not only contains precise geometric information of pipe sections but also clearly defines the type, location, and topological connection relationships of pipe fittings such as valves and elbows. This provides a clear global index for inspection and testing work, enabling precise location of discovered defects on specific pipe fittings or sections. Combined with the high-fidelity node model obtained from refined scanning, high-risk components can be given focused evaluation, greatly improving the efficiency and depth of safety assessment. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the functional modules of a pressure pipeline mapping system based on an unmanned aerial vehicle (UAV) according to an embodiment of the present invention;

[0021] Figure 2 This is a flowchart of a surveying method according to an embodiment of the present invention;

[0022] Figure 3 This is a schematic diagram of the coordinate system relationship according to an embodiment of the present invention.

[0023] Among them, 100 is the pipeline anchoring relative navigation module; 200 is the orthogonal contour scanning and parameter calculation module; 300 is the adaptive flight and measurement control module; and 400 is the airborne real-time topology map construction and node processing module. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the embodiments of this invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0025] See attached document Figure 1 , Figure 1 This is a schematic diagram of the functional modules of a UAV-based pressure pipeline mapping system according to an embodiment of the present invention. The present invention provides a UAV-based pressure pipeline isometric mapping system integrated onto a UAV platform. The system may include: a pipeline anchoring relative navigation module 100, an orthogonal contour scanning and parameter calculation module 200, an adaptive flight and measurement control module 300, and an airborne real-time topology map construction and node processing module 400.

[0026] The pipeline anchoring relative navigation module 100 is used to sense the relative pose between the UAV platform and the target pipeline in real time, and generate control commands accordingly to guide the UAV to fly stably along the pipeline axis.

[0027] The orthogonal contour scanning and parameter calculation module 200 is used to actively project linear structured light onto the surface of the pipe during the flight of the UAV and capture the contour formed by it. Through geometric calculation, the outer diameter and three-dimensional center point coordinates of the pipe at the current section are calculated in real time.

[0028] The adaptive flight and measurement control module 300 is used to receive real-time measurement data from the orthogonal profile scanning and parameter calculation module 200, and dynamically adjust the flight control parameters of the pipeline anchoring relative navigation module 100 and the sensor operating parameters of the orthogonal profile scanning and parameter calculation module 200 itself according to the data.

[0029] The airborne real-time topology map construction and node processing module 400 is used to construct the topology of the pipeline network from the continuous center point coordinates output by the orthogonal contour scanning and parameter calculation module 200, and to detect and identify pipe nodes on the pipeline path. After identifying the node, a preset strategy is called to control the UAV platform to perform a fine scan of the node.

[0030] See attached document Figure 2 , Figure 2 This is a flowchart of a surveying method according to an embodiment of the present invention. The overall workflow of the method provided by this embodiment of the present invention may include the following steps: S101, System Initialization and Target Pipeline Locking. The operator maneuvers the UAV platform to approach the initial pipe segment to be mapped. The pipeline anchoring relative navigation module 100 identifies the pipe segment and calculates the initial relative pose between the UAV platform and the pipe segment, completing the locking of the target pipeline and establishing the navigation reference for subsequent flight; S102, initiate axial cruise and real-time measurement. The pipeline anchoring relative navigation module 100 controls the UAV platform to fly autonomously along the locked pipeline axis. During flight, the orthogonal profile scanning and parameter calculation module 200 continuously operates to acquire the precise outer diameter and three-dimensional center point coordinates of the pipeline cross-section at high frequency; S103, execute adaptive control. The adaptive flight and measurement control module 300 receives the real-time calculated outer diameter of the pipe and feeds it back to the pipeline anchoring relative navigation module 100 to dynamically adjust the flight distance between the UAV and the pipe. At the same time, it adjusts the sensor parameters of the orthogonal contour scanning and parameter calculation module 200 according to the quality of the acquired contour image. S104, Dynamically constructing a topology map. The airborne real-time topology map construction and node processing module 400 constructs edges with topological relationships in the airborne computer from the continuously collected three-dimensional center point coordinates, forming pipe segment paths. At the same time, it uses the forward-looking sensor of the pipeline anchoring relative navigation module 100 to predict the pipeline structure ahead of the flight path. S105, Node Detection and Processing. When the airborne real-time topology map construction and node processing module 400 detects pipe fitting nodes such as elbows, tees, and valves ahead, the UAV platform decelerates and approaches the node. This module identifies the node type and, based on the identification result, calls the corresponding refined scanning program from the preset strategy library to control the UAV platform to perform specific scanning actions for that node; S106, Path Decision and Task Continuation. After processing a node, if the node is a T-junction or cross-junction with multiple branch paths, the system submits the information of the unexplored branches to the operator. After the operator selects the next mapping path, the system returns to the target locking in step S101 and repeats steps S102 to S105 along the new path until all target pipelines have been mapped. S107, Generate Structured Data. After all surveying tasks are completed, the system will output the final complete pipeline topology map formed in the airborne real-time topology map construction and node processing module 400 as a structured data file. This file contains the precise geometric, topological, and semantic information of all pipe segments and fitting nodes.

[0031] In one embodiment, the pipeline anchoring relative navigation module 100 provides the unmanned aerial vehicle platform with relative navigation capability using the target pipeline as a reference point, independent of a global positioning system (such as GPS).

[0032] The pipeline anchoring relative navigation module 100 includes a forward-looking binocular stereo vision system, which consists of two cameras with known relative positions and orientations, simultaneously acquiring left and right view images. Processing the acquired images to perceive the pipeline target and perform 3D localization may include the following steps: S201, Image Acquisition and Preprocessing. The binocular stereo vision system synchronously captures images from the left camera at a preset frequency. And right camera image To improve the robustness of subsequent processing, image preprocessing can be performed, including image distortion correction, grayscale conversion, and contrast enhancement.

[0033] S202, Stereo Matching and Depth Map Generation. A stereo matching algorithm is used to calculate the pixel disparity between left and right image pairs, thereby generating a depth map. For stereo matching algorithms, those skilled in the art can use known algorithms such as SGBM (Semi-Global Block Matching), the specific implementation of which is well-known in the field and will not be elaborated here. Each pixel value in the depth map corresponds to the depth information of that pixel in three-dimensional space.

[0034] S203, Point Cloud Generation and Pipeline Target Segmentation. Based on the camera intrinsic parameter matrix. and depth map It can convert two-dimensional image information into three-dimensional point clouds. For any pixel in the image and its depth value Its three-dimensional coordinates in the camera coordinate system It can be calculated using the following formula: ; ; ; in, , Let be the focal length of the camera along the x and y axes. , These are the coordinates of the camera's principal point; they are all part of the camera's intrinsic parameter matrix. The parameters are then used to convert the entire field of view into a 3D point cloud in the camera coordinate system. .

[0035] Subsequently, it is necessary to start from the original point cloud containing the background environment. In this process, the portion belonging to the target pipeline is segmented. One feasible implementation is to use a model-fitting segmentation algorithm, such as RANSAC (Random Sample Consensus). This algorithm iteratively samples the minimum set of points randomly from the point cloud, fits a cylindrical mathematical model, and calculates the number of interior points of other points in the point cloud relative to this model. Through multiple iterations, the cylindrical model that obtains the most interior point support is found, and the set of interior points of this model is considered the point cloud of the target pipeline. .

[0036] A cylindrical model can be defined by its axis (by a direction vector). and a point on the axis (definition) and radius To describe. A point in a point cloud. Distance to the axis of the cylinder for: ; If the distance is related to the radius The absolute value of the difference Less than the preset threshold Then point These are considered interior points of the model; S204, Pipeline centerline extraction and preliminary pose estimation. This is achieved after obtaining the point cloud data belonging to the target pipeline. Then, by performing principal component analysis (PCA) on the point cloud, the direction vector of its central axis can be obtained. Specifically, the point cloud is calculated... The covariance matrix of the pipeline is such that the eigenvector corresponding to the largest eigenvalue of this matrix is ​​the direction vector of the pipeline's central axis. Point clouds The geometric center can be taken as a point on the axis. .

[0037] At this point, the UAV has obtained a three-dimensional geometric description of the pipeline segment ahead, including its spatial orientation and position, in the camera coordinate system at the current moment. This information forms the basis for subsequent relative pose calculation and closed-loop tracking control.

[0038] See attached document Figure 3 , Figure 3 This is a schematic diagram of coordinate system relationships according to an embodiment of the present invention. After obtaining the three-dimensional information of the pipeline target through visual perception, the pipeline anchoring relative navigation module 100 establishes a dynamic local coordinate system based on the pipeline, calculates the relative pose of the UAV in the coordinate system in real time, and finally achieves stable tracking of the pipeline through closed-loop control.

[0039] One feasible implementation is to define a local coordinate system for the pipeline, denoted as . The origin of this coordinate system Defined as UAV body coordinate system The origin The projection point on the central axis of the pipeline. of An axis is defined as a direction vector relative to the central axis of the pipe. Consistent, pointing in the direction of the drone's flight. Axis is defined as arrive The vector direction is perpendicular to the pipe axis and points towards the drone. The shaft then passes through shaft and The cross product of the axes yields a coordinate system that satisfies the right-hand rule. This defines the local coordinate system of the pipeline. It is dynamically correlated with the current position of the drone body, providing a benchmark for the description of relative pose.

[0040] S205, Relative pose parameter calculation. This is done after defining the local coordinate system of the pipeline. Afterwards, the UAV body coordinate system Compared to The pose can be determined by a set of control parameters. Describe this set of parameters. Including a lateral yaw angle A vertical pitch angle and a radial distance .

[0041] Lateral yaw angle Defined as the UAV body coordinate system The direction of travel (i.e.) (axis) and local coordinate system of the pipeline of The angle between the axes (i.e., the tangential direction of the pipeline). This angle reflects the degree of alignment between the drone's heading and the pipeline's direction.

[0042] Vertical pitch angle Defined as the UAV body coordinate system of Axis in Projection on a plane, and The angle between the axes. This angle reflects the pitch attitude of the drone relative to the pipe in the vertical plane.

[0043] radial distance Defined as the UAV body coordinate system The origin The perpendicular distance to the central axis of the pipe, i.e., the vector. The distance reflects the spacing between the drone and the pipeline.

[0044] These three parameters can be obtained from the UAV's own attitude information (provided by the IMU) and the pipe axis vector calculated in the previous steps. and points on the axis It is obtained through vector operations and coordinate transformations.

[0045] S206, Target pose setting and error calculation. To achieve stable axial tracking flight, the system presets a target relative pose. In a typical embodiment, the goal is to align the UAV's heading parallel to the pipeline while maintaining a constant safe distance. The target's relative pose can then be set as follows: ; in, The target tracking distance is preset. The system will calculate the current relative pose in real time. Relative pose to the target By comparison, the error vector is obtained. : ; in, For yaw angle error, For pitch angle error, This represents the radial distance error.

[0046] S207 generates closed-loop servo control commands based on the calculated error vector. The system generates speed or attitude commands for the UAV platform through the controller. To drive the error closer to zero, a feasible implementation is to use a proportional-integral-derivative (PID) controller. A control law is designed independently for each error component. For example, the yaw rate command of a UAV. Pitch angular velocity command and lateral translation speed command They can be represented as follows: ; ; ; in, , , These are the PID controller gain coefficients corresponding to the error components. These gain coefficients can be adjusted through experimental tuning or adaptive algorithms. Meanwhile, the UAV maintains a constant forward speed. Flying along the axis of the pipeline.

[0047] Through the above steps, the pipeline anchoring relative navigation module 100 can lock the UAV's flight status in a relative coordinate system based on the pipeline in real time, thereby achieving high-precision and robust axial tracking flight of the pipeline without relying on external positioning sources.

[0048] In one embodiment, the orthogonal contour scanning and parameter calculation module 200 functions to actively acquire the contours of each cross-section of the pipeline along the flight path at a high frequency when the UAV platform flies along the pipeline axis, and to provide three-dimensional data input for subsequent parameter calculation.

[0049] The core components of this module include a line laser emitter and a high-speed industrial camera, both fixedly mounted on a single-axis rotating gimbal with a defined relative pose. The contour 3D reconstruction process based on active structured light may include the following steps: S301, Coordinate System Definition and Calibration. To achieve 3D reconstruction, it is necessary to define and calibrate the coordinate system within the module. The camera coordinate system is defined as follows: Its origin is at the optical center of the camera. The axis is along the optical axis. The coordinate system of the scanning module is defined as follows: This coordinate system is fixed to the rotating gimbal. Camera coordinate system. Relative to the coordinate system of the scanning module homogeneous transformation matrix and line laser emitters in The equations of the projected light plane can all be accurately obtained through a pre-calibration process. For calibration methods, those skilled in the art can use known techniques such as the Zhang Zhengyou calibration method, which will not be elaborated upon here.

[0050] Line laser in the scanning module coordinate system A fixed light plane is formed in the middle, and its plane equation can be expressed as: ; in, for The coordinates of the points in the system, , , , These are the coefficients of the plane equation obtained after calibration.

[0051] S302, Image Acquisition and Contour Pixel Extraction. During operation, a line laser emitter projects a fan-shaped laser beam onto the pipe surface, forming a laser line. A high-speed industrial camera simultaneously captures an image containing this laser line. To extract the laser line contour from the image, an image segmentation algorithm can be used. One feasible implementation is to threshold the image to highlight the laser line, and then use a centerline extraction method such as the Steger algorithm to obtain a sub-pixel-level two-dimensional coordinate point set of the laser line on the image. .

[0052] S303, 3D coordinate calculation of contour points. For the contour pixel set... any point in The goal is to calculate the coordinates of this point in three-dimensional space. Based on the pinhole camera imaging model, the pixel... Corresponding to the camera coordinate system A three-dimensional ray in the equation, where any point on the ray... The coordinates can be represented as: ; in, This is the intrinsic parameter matrix of the camera. Let be the depth value to be determined, representing Distance along the ray direction.

[0053] because It is formed by laser projection, and therefore it must lie on the laser beam plane. Therefore, From the camera coordinate system Transform to the scanning module coordinate system After that, the points obtained The equations of the laser plane must be satisfied. The coordinate transformation relationship is as follows: ; Expanding the above equation and substituting it into the laser plane equation, we can obtain the value for the unknown depth. The linear equation is given. By solving this equation, the linear equation can be uniquely determined. The value of .

[0054] After determining each pixel corresponding depth Then, the coordinate system of that point in the camera coordinate system can be obtained. Precise three-dimensional coordinates By iterating through all the contour pixels, a three-dimensional point set representing the pipe cross-section contour in the camera coordinate system is finally obtained. This set of points forms the data basis for subsequent fitting of pipe cross-sectional parameters.

[0055] After obtaining the three-dimensional spatial point set representing the pipe profile, the orthogonal profile scanning and parameter calculation module 200 needs to process the point set to extract the pipe's cross-sectional parameters, namely the outer diameter and the coordinates of the three-dimensional center point, in real time and accurately. To ensure measurement accuracy, this embodiment of the invention introduces an orthogonal constraint mechanism.

[0056] S304, Implement orthogonal constraint control. To eliminate measurement errors caused by the tilt of the projection angle and ensure that the calculated profile accurately reflects the cross-section of the pipe, the single-axis rotating gimbal in this embodiment is used to actively adjust the attitude of the scanning module. Specifically, the pipeline anchoring relative navigation module 100 outputs the UAV's flight velocity vector in real time, which can be approximated as the tangent direction of the pipe at the current position. The control system of the single-axis rotating gimbal receives this flight velocity vector and drives the gimbal to rotate, so that the angle between the normal vector of the line laser projection plane and the velocity vector always approaches 90 degrees. This active attitude compensation ensures that the laser scanning plane is always physically orthogonal to the central axis of the pipe, thereby ensuring that the collected profile point set is located on a true cross-section of the pipe.

[0057] S305, the mathematical model of the cross-sectional parameters. The three-dimensional point set obtained after orthogonal constraint scanning. Theoretically, these two points can coexist on the same plane and form an arc. The goal of parameter solving is to find the center of the circle containing this arc. and radius This is a spatial circle fitting problem, where the center of the circle... That is, the coordinates of the center point of the pipe at the current cross-section in the camera coordinate system, while the radius... The corresponding pipe radius is then doubled to obtain the pipe's outer diameter. .

[0058] S306, Execute the robust fitting algorithm. Due to measurement noise and environmental interference, the collected point set may contain outliers, and directly using the standard least squares method for fitting may lead to biased results. Therefore, a feasible implementation is to use a robust estimation algorithm.

[0059] First, the three-dimensional point set Projecting onto the best-fit plane determined by these points reduces the 3D fitting problem to 2D. This best-fit plane can be obtained by applying a set of points... Principal component analysis (PCA) is performed to find that the eigenvector corresponding to the smallest eigenvalue of the covariance matrix of the point set is the normal vector of the plane.

[0060] Subsequently, in the two-dimensional plane, a circle fitting algorithm such as Random Sampling Consensus (RANSAC) is used. The iterative process of this algorithm includes:

[0061] 1. Randomly select 3 points from the set of 2D points (the minimum number of points required to determine a circle).

[0062] 2. Calculate the center of a candidate circle based on these three points. and radius .

[0063] 3. Iterate through all other points in the point set and calculate their distances to the candidate circle. If the distance is less than a preset interior point threshold... If the point is an interior point, then that point is recorded as an interior point.

[0064] 4. Repeat steps 1 to 3 several times, and select the candidate circle that receives the most in-point support as the final fitting result.

[0065] After determining the optimal set of interior points, the circle parameters can be finally optimized using the least squares method with all interior points to improve accuracy. The objective function of this optimization problem is: ; in, and The final optimized coordinates and radius of the pipeline center point.

[0066] S307, Parametric Coordinate Transformation and Output. The center of the circle obtained from the fitting. In the camera coordinate system The following representation is needed. For use in global topology graph construction, it must be transformed to the global world coordinate system. This transformation is achieved by using the transformation matrix from the world coordinate system to the body coordinate system obtained through the UAV's own positioning and attitude determination system (such as RTK-INS). And the pre-calibrated transformation matrix from the body coordinate system to the camera coordinate system. To achieve: ; in, This refers to the final coordinates of the pipeline's center point in the global world coordinate system. The orthogonal profile scanning and parameter calculation module 200 will ultimately calculate the global center point coordinates in real time. and pipe outer diameter The output is then sent to subsequent functional modules.

[0067] In one embodiment, the adaptive flight and measurement control module 300 establishes a feedback loop from real-time measurement results to flight control behavior, enabling the UAV platform's flight attitude to proactively adapt to changes in the geometric dimensions of the measured object pipeline.

[0068] In traditional UAV inspection or mapping missions, flight control parameters are typically set to fixed values ​​before the mission begins and remain unchanged throughout the mission. This approach cannot adapt to actual working conditions such as changes in pipe diameter, potentially leading to poor measurement field of view or safety risks. This invention proposes a measurement-navigation closed-loop control method, the specific implementation of which may include the following steps: S401 receives real-time measurement data. The adaptive flight and measurement control module 300 receives the pipe outer diameter measurement value output from the orthogonal profile scanning and parameter calculation module 200 in real time via its internal data bus. This data is updated immediately after each successful calculation of the cross-sectional parameters.

[0069] S402, dynamically calculate the target flight distance. The adaptive flight and measurement control module 300 anchors the aforementioned pipeline relative to the target radial distance in the navigation module 100. It transforms from a static constant into a dynamic variable. This dynamic variable is the real-time measured pipe diameter. The function. One feasible implementation is to use a linear function model to calculate the dynamic target radial distance. : ; in, At any moment The calculated radial distance the UAV should maintain from the target; At any moment The pipe's outer diameter obtained through real-time measurement; It is a preset field of view coefficient, which is set according to the field of view angle of the camera and the subtended angle of the line laser in the orthogonal contour scanning and parameter calculation module 200. Its physical meaning is the ratio of the minimum distance required to ensure that the camera can capture the complete laser contour arc under any pipe diameter to the pipe diameter. It is a preset minimum safety offset, used to ensure that even with a very small pipe diameter, the drone can maintain an absolutely safe distance from the pipe to prevent collisions.

[0070] For example, when the pipe is measured to be thicker, Increase, calculated The pressure will increase accordingly, and the system will instruct the drone to move appropriately away from the pipe. Conversely, when the pipe is detected to be narrowing, Decrease The pressure will also decrease accordingly, and the system will instruct the drone to move appropriately closer to the pipeline.

[0071] S403, Update navigation target and execute control. Calculate the dynamic target distance. The data is transmitted in real time to the pipeline anchoring relative navigation module 100, which updates the target relative pose within it. The radial distance component in the target distance is then calculated. The PID controller of the pipeline anchoring relative navigation module 100 then recalculates the radial distance error based on the new target distance. And generate corresponding control commands (such as lateral translation speed commands). This drives the drone platform to adjust its lateral position until the actual radial distance matches the new target distance.

[0072] Through the above steps, the system establishes a closed loop that directly feeds back real-time measurement results (pipe diameter) to the flight control system (navigation target). This adaptive adjustment mechanism ensures that the measurement module always operates under optimal field-of-view conditions, regardless of changes in pipe diameter. This guarantees the integrity and quality of the measurement data and enhances the autonomy and safety of the UAV in complex pipeline environments.

[0073] To further enhance the system's adaptability and measurement reliability in varying field environments, the adaptive flight and measurement control module 300 in this embodiment of the invention also constructs a measurement-sensor closed loop, from imaging quality to sensor parameters. The function of this closed loop is to enable the system to adjust the hardware parameters of the measurement sensor in reverse based on the real-time quality of the acquired laser contour image, thereby proactively counteracting signal quality degradation caused by factors such as differences in pipe surface material and changes in ambient lighting.

[0074] One feasible implementation may include the following steps: S404, Online Image Quality Assessment. After each frame of laser contour image is captured by the orthogonal contour scanning and parameter calculation module 200, the system does not immediately perform 3D reconstruction, but first performs a quality assessment on the image. To achieve quantitative assessment, the system calculates one or more imaging quality indices.

[0075] One implementation method is to define a comprehensive quality score. The calculation of this score can take into account the following aspects:

[0076] 1. Contour Brightness: After extracting the contour pixel set, calculate the average gray value of these pixels. .

[0077] 2. Overexposure level: The number of pixels whose grayscale values ​​reach saturation (e.g., 255 for an 8-bit image) within a statistically significant area. Percentage of total outline pixels The ratio, i.e., overexposure rate .

[0078] 3. Contour sharpness: Calculate the mean image gradient along the normal direction of the contour pixels. A larger gradient value usually indicates a sharper contour edge and a clearer image.

[0079] Taking all the above factors into account, the image quality score It can be defined as a function, for example: ; in, and These are preset weighting coefficients used to balance the importance of different indicators. This score... It can quantitatively reflect whether the current image is moderately bright, not overexposed, and clear enough.

[0080] S405, calculates the quality error. The system has a preset ideal imaging quality target range. The quality score calculated in real time. Compare with the target interval. If Below This indicates that the image is too dark or blurry; if it is higher than... If the image quality is too high, there may be a risk of overexposure or unnecessary energy consumption; if it is within the acceptable range, the image quality is considered ideal. Based on this, a quality error can be defined. .

[0081] S406 generates parameter adjustment instructions. Based on quality error... The adaptive flight and measurement control module 300 calculates the adjustment amount of the hardware parameters in the orthogonal contour scanning and parameter calculation module 200 through a preset adjustment strategy. The hardware parameters to be adjusted mainly include the power of the line laser emitter. Exposure time of industrial cameras .

[0082] The adjustment strategy can be a set of rules:

[0083] like If the image quality is low, proceed according to priority: first, increase the laser power within a safe range. If the laser power has reached its maximum, then appropriately increase the camera exposure time. .

[0084] like Or overexposure rate detected If the threshold is exceeded (image quality is too high or overexposed), the following steps are taken in order of priority: first, reduce the camera exposure time. If the exposure time has reached the lower limit, reduce the laser power. .

[0085] The adjustment step size can be fixed or adjusted according to the error. The size is proportional to the convergence to achieve faster convergence.

[0086] S407, applying new sensor parameters. The newly calculated parameters... and These parameters are sent and set to the hardware controllers of the line laser emitter and industrial camera via the driver interface. These new parameters will take effect in the next frame acquisition cycle.

[0087] Through the above steps, the system forms a continuous closed loop of "acquisition-evaluation-decision-adjustment". This mechanism enables the measurement system to perceive and adapt to environmental changes, autonomously maintain high-quality data source input, and thus significantly improve the robustness and final measurement accuracy of the entire surveying system when facing pipe surfaces of different colors, reflective properties, and cleanliness levels, as well as operating under different lighting conditions.

[0088] In one embodiment, the airborne real-time topology map construction and node processing module 400 abandons the traditional surveying workflow of "first collecting complete point clouds and then performing offline 3D reconstruction and modeling", and instead directly and incrementally organizes the real-time measurement data acquired during UAV flight into a structured pipeline model with topological logic.

[0089] To achieve this functionality, the system maintains a graph data structure in the onboard computer's memory. This data structure is used to represent the pipeline network. The definition of this data structure is as follows: Vertex set This represents a topological node in a pipeline network. These nodes are critical locations where pipelines connect, turn, or their properties change, such as elbows, tees, crosses, valves, flanges, instruments, and the start and end points of the pipeline. Each vertex in the set... It contains a set of attributes, which may include at least: a unique node identifier. A node type (Its value can be "elbow", "valve", etc.); the three-dimensional coordinates of this node in the global world coordinate system. ; and a list to store the identifiers of all edges connected to the node.

[0090] Edge set This represents a continuous, unbranched straight or curved pipe segment connecting two vertices. Each edge in the set... It contains a set of attributes, which may include at least: a unique edge identifier. The identifiers of the starting and ending vertices of the connection. and ; and data describing the geometric and physical characteristics of the pipe section, such as an ordered list of continuous three-dimensional center point coordinates. This list defines the centerline trajectory of the pipe segment; and the average outer diameter of the pipe segment. .

[0091] The airborne incremental construction process of this topology map may include the following steps: S501, Topology Initialization. At the start of the surveying mission, after the UAV locks onto the initial target pipe segment, the onboard real-time topology construction and node processing module 400 creates an empty graph structure in memory. Then, based on the initial locked position, the system creates the first vertex in the graph. This vertex represents the starting point of the mapping task. Simultaneously, a new edge is created. and set its starting vertex to .at this time, It is marked as "currently active edge".

[0092] S502, dynamic extension of the edge. During the UAV's flight along the pipeline axis, the orthogonal contour scanning and parameter calculation module 200 continuously outputs the global center point coordinates at a high frequency. and pipe outer diameter The airborne real-time topology map construction and node processing module 400 receives these data streams in real time. For each newly received center point coordinate... The system adds it to the currently active edge. Centerline trajectory list The end of.

[0093] At the same time, the system utilizes the newly received outer diameter value Dynamically update the average outer diameter attribute of the currently active edge. One feasible implementation is to use a moving average or weighted average algorithm to smooth out measurement noise. The length attribute of the edge can also be updated cumulatively based on the Euclidean distance between the newly added point and the previous point.

[0094] S503, Vertex Generation and Edge Closure. During flight, the subsequent node detection mechanism (detailed later) continuously analyzes the path ahead. When a new topological node is detected (e.g., a bend), the system creates (or "instantiates") a new vertex at the global location corresponding to that node. .

[0095] New peak After creation, the system will display the "currently active edge". Update the state and identify the terminating vertex. Set as a newly created vertex identifier At this point, the edge representing the first pipe segment... It is fully defined and closed. Then, the system creates a new edge. Set its starting vertex to and will This is marked as the new "currently active edge". The drone then continues flying, and the data points collected will be used to extend this new edge. .

[0096] By repeatedly executing steps S502 and S503, the system organizes the measurement data into a topological chain or network consisting of alternating vertices and edges as the UAV flies over each section of the pipeline, thus obtaining a complete pipeline model without post-processing at the end of the mission.

[0097] In the airborne real-time topology map construction and node processing module 400, in order to achieve incremental construction of the topology map, the system needs to have the ability to predict the path ahead and detect and identify the topology nodes that will be encountered in advance while flying along the pipeline.

[0098] This forward-looking perception function is executed in parallel with the UAV's axial flight mission, utilizing real-time image data and 3D information acquired by the forward-looking binocular stereo vision system anchored relative to the navigation module 100. One feasible implementation may include the following steps:

[0099] S504, Remote Path Structure Analysis and Node Detection. The system continuously analyzes the 3D point cloud reconstructed by the forward-looking vision system at the far end of its perception range. By performing cylindrical fitting on this remote point cloud, the centerline direction of the preceding pipe segment can be extracted. The system monitors the rate of change of this centerline direction vector in real time. When a drastic change exceeding a preset angle threshold is detected in this direction vector over a short distance, the system determines that a curved node (such as a bend) has been detected.

[0100] Another detection method is to detect abrupt changes in pipeline geometry. For example, by continuously monitoring the fitted radius of the distal pipeline, if a step change in radius is detected, a reducing joint node is identified. If the intersection of another cylindrical structure is detected on the sidewall of the main pipeline, a branch-type node (such as a tee) is identified. The output of this step is a preliminary node existence determination and the approximate location of the potential node in the camera coordinate system.

[0101] S505, Node Region Segmentation and Type Recognition. After detecting potential topological nodes, the system defines the 3D spatial region or 2D image region where the node is located as a Region of Interest (ROI). The airborne real-time topology map construction and node processing module 400 then calls a node classifier to analyze the ROI to determine its specific type.

[0102] One feasible implementation is to employ a machine learning-based classification method. The system pre-trains a convolutional neural network (CNN) offline using a dataset containing a large number of labeled pipe fitting images (such as elbows, tees, flanges, and various valves). During actual flight, the ROIs (Regions of Interest) extracted from real-time images are input into this trained CNN classifier. The network extracts and analyzes image features, ultimately outputting the probability distribution of the node's category. The system selects the category with the highest probability as the node's identification result, for example, outputting "gate valve" or "90-degree elbow." For the training and implementation of this classifier, those skilled in the art can use mature deep learning frameworks; the specific details are well-known in the field and will not be elaborated upon here.

[0103] Another alternative implementation is to use a method based on 3D geometric template matching. The system pre-configures a 3D model library of standard pipe fittings. The local point cloud reconstructed by the forward-looking system within the node's ROI is then 3D registered with the models of each standard pipe fitting in the library. By calculating the similarity metric after registration (such as chamfer distance or Hausdorff distance), the standard pipe fitting with the highest similarity is selected as the node's recognition result.

[0104] S506, Trigger vertex instantiation. When the node type is identified and the UAV platform flies to a distance less than a preset trigger distance threshold from the node, the onboard real-time topology graph construction and node processing module 400 will generate an instruction. This instruction contains the identified type of the node. And the global three-dimensional coordinates of the node estimated by the navigation system. The instruction is sent to the topology management unit to trigger the operation in step S503, namely closing the current edge and creating a new vertex using the node's information as an attribute. This mechanism ensures that the construction of the topology graph is synchronized with the physical progress of the UAV.

[0105] Conventional axial scanning along pipes can only acquire the centerline and diameter information of pipe segments, failing to capture the complete three-dimensional geometry of complex pipe fittings such as elbows, tees, and valves. Therefore, this invention proposes a cognitive-action closed-loop autonomous refined scanning method.

[0106] S507, Scanning Mode Switching and Task Scheduling. In the aforementioned step S506, once the type of a topology node is successfully identified and the UAV has reached a safe hovering point near that node, the onboard real-time topology map construction and node processing module 400 will issue a high-priority command to the flight control system. This command will temporarily suspend the conventional axial tracking task of the pipeline anchoring relative to the navigation module 100, causing the UAV to switch from the "flying along the line" mode to the "detailed inspection around the point" mode.

[0107] After the entire surveying task is completed and all target pipelines have been traversed and modeled, the system executes the final data integration and output process. This process aims to transform the pipeline network diagram data structure, which is dynamically constructed in the onboard computer memory and contains complete geometric and topological information, into a persistent, portable, and directly compatible structured data file for downstream applications.

[0108] S601, Data Generation Triggered. The triggering conditions for this process can be: the onboard real-time topology map construction and node processing module 400 confirms that there are no more unexplored branch paths; or, the remote operator issues a task completion command. Upon triggering, the system stops all data acquisition and real-time construction activities and locks the final version of the network topology map in memory. .

[0109] S602, data serialization. The system serializes graph data structures in memory. Perform serialization. Serialization is the process of converting the state information of objects in memory into a form that can be stored or transmitted. One feasible implementation is to convert the entire graph structure into a common data exchange format, such as JSON or XML. These formats have good self-descriptive and hierarchical structures, can form a natural mapping relationship with the vertex-edge structure of the graph, and are also easy to be parsed by various computer programs.

[0110] S603 defines the output file structure. The generated structured data file has a well-defined, predefined hierarchical structure. A typical file structure may include the following main parts:

[0111] A top-level metadata object is used to store global information for this surveying task, such as: task unique identifier, surveying date and time, and the global coordinate system used.

[0112] A list of vertices containing all vertex objects in the graph. Each vertex object represents a topological node, and its data structure includes at least:

[0113] A node_id field stores the unique identifier of the node.

[0114] A node_type field stores the semantic type of the node as identified by the classifier, such as "90_degree_elbow" or "gate_valve".

[0115] A `position` field stores the node's three-dimensional coordinates in the global coordinate system. .

[0116] For nodes that have obtained detailed models through fine scanning, there will also be a detailed_model field, which stores a relative path to a separate 3D model file (e.g., PLY or OBJ format), or directly embeds the mesh data of the model.

[0117] A list of edges containing all edge objects in the graph. Each edge object represents a pipe segment, and its data structure includes at least:

[0118] An edge_id field stores the unique identifier of this pipe segment.

[0119] The start_node_id and end_node_id fields store the identifiers of the starting and ending vertices connected to the pipe segment, respectively, defining the connection relationship of the pipe network.

[0120] A diameter field stores the average outer diameter value of this pipe section.

[0121] A centerline field whose content is a 3D coordinate point An ordered array is formed, which describes the spatial trajectory of the centerline of this pipe segment.

[0122] S604, File Generation and Output. Following the structure described above, the system converts the graph data in memory into corresponding fields and values, ultimately generating a complete structured data file. This file can be stored on the UAV's onboard storage device or transmitted wirelessly to a ground station or cloud server after the mission.

[0123] This type of output file is not the raw point cloud data, but a highly structured digital twin model containing rich geometric and semantic information. This model requires no additional and time-consuming post-processing and can be easily converted to standard exchange formats such as IFC and DXF, thus achieving compatibility with various CAD (Computer-Aided Design), GIS (Geographic Information System), or BIM (Building Information Modeling) software. This provides a directly usable data foundation for subsequent pressure pipeline inspection, integrity management, and safety assessment, constituting the final data output of the entire surveying and mapping system.

[0124] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A pressure pipeline isometric mapping system based on unmanned aerial vehicles (UAVs), characterized in that, include: The pipeline anchoring relative navigation module is used to sense the relative pose between the UAV platform and the target pipeline in real time, and generate control commands accordingly to guide the UAV to fly stably along the pipeline axis. The orthogonal contour scanning and parameter calculation module is used to actively project linear structured light onto the pipe surface during the flight of the UAV and capture the contour formed by it. Through geometric calculation, the outer diameter and three-dimensional center point coordinates of the pipe at the current section are calculated in real time. An adaptive flight and measurement control module is used to receive the outer diameter output by the orthogonal contour scanning and parameter calculation module, and dynamically adjust the flight control parameters of the pipeline anchoring relative navigation module according to the outer diameter, and / or adjust the sensor operating parameters of the orthogonal contour scanning and parameter calculation module according to the imaging quality of the contour. The airborne real-time topology map construction and node processing module is used to construct the topology of the pipeline network from the continuous three-dimensional center point coordinates output by the orthogonal contour scanning and parameter calculation module, and to detect and identify pipe fitting nodes on the pipeline path. After identifying the node, a preset strategy is invoked to control the UAV platform to perform a fine scan of the node.

2. The pressure pipeline isometric mapping system based on UAV according to claim 1, characterized in that, The pipeline anchoring relative navigation module includes a forward-looking binocular stereo vision system for generating a three-dimensional point cloud and segmenting the target pipeline from it to extract its central axis.

3. The pressure pipeline isometric mapping system based on UAV according to claim 2, characterized in that, The pipeline anchoring relative navigation module is also used for: Based on the extracted central axis, a local coordinate system for the pipeline is defined and dynamically associated with the UAV. The lateral yaw angle, vertical pitch angle, and radial distance of the UAV's body coordinate system relative to the local coordinate system of the pipeline are calculated in real time. Based on the error between these three parameters and the target pose, the control command is generated through closed-loop servo control.

4. The pressure pipeline isometric mapping system based on UAV according to claim 1, characterized in that, The orthogonal contour scanning and parameter calculation module includes: a line laser emitter, a high-speed industrial camera, and a single-axis rotating gimbal for supporting the emitter and camera.

5. The pressure pipeline isometric mapping system based on UAV according to claim 4, characterized in that, The single-axis rotating gimbal is used to implement orthogonal constraint control. It receives the flight velocity vector of the UAV and drives the gimbal to rotate, so that the angle between the laser plane normal vector projected by the line laser emitter and the flight velocity vector approaches 90 degrees.

6. The pressure pipeline isometric mapping system based on UAV according to claim 1, characterized in that, The adaptive flight and measurement control module uses the following method to dynamically adjust flight control parameters: The target radial distance between the pipeline anchoring and the navigation module is set as a dynamic variable related to the real-time measured outer diameter, so that the distance between the UAV and the pipeline can be adaptively adjusted as the outer diameter of the pipeline changes.

7. The pressure pipeline isometric mapping system based on UAV according to claim 1, characterized in that, The adaptive flight and measurement control module adjusts the sensor operating parameters in the following specific way: Calculate the imaging quality score of the contour online; Based on the error between the score and the preset target range, the power of the linear laser emitter and the exposure time of the industrial camera in the orthogonal contour scanning and parameter calculation module are adjusted in reverse.

8. The pressure pipeline isometric mapping system based on UAV according to claim 1, characterized in that, The airborne real-time topology graph construction and node processing module maintains a graph data structure in the airborne computer memory to represent the topology of the pipeline network. This graph data structure includes: A set of vertices, where each vertex represents a pipe node and contains a node identifier, node type, and 3D coordinate attributes; An edge set, where each edge represents a pipe segment connecting two vertices, and includes identifiers of the starting and ending vertices of the connection, the average outer diameter, and the centerline trajectory attribute consisting of the coordinates of continuous three-dimensional center points.

9. The pressure pipeline isometric mapping system based on UAV according to claim 1, characterized in that, The airborne real-time topology map construction and node processing module includes a node classifier, which is used to identify the type of pipe nodes perceived by the forward-looking sensors of the pipeline anchoring relative navigation module.

10. The pressure pipeline isometric mapping system based on UAV according to claim 9, characterized in that, The airborne real-time topology map construction and node processing module is also used to: after the node classifier identifies the type of the pipe node, call the refined scanning program corresponding to the node type from the preset scanning strategy library, and switch the UAV's flight mode to execute the program, thereby obtaining the complete three-dimensional geometric shape of the node.

Images