Multi-unmanned aerial vehicle cooperative surveying and mapping method and system for areas along railway

Abstract

The embodiment of the invention provides a multi-unmanned aerial vehicle cooperative surveying and mapping method and system for areas along a railway, and belongs to the technical field of railway surveying and mapping. The method comprises the steps that the unmanned aerial vehicle collects monitoring data in the flight process; the relative position and posture between the aircrafts are dynamically adjusted, and an observation configuration with variable baseline parameters is constructed; laser ranging modules carried by the unmanned aerial vehicles are controlled to execute synchronous ranging operation, and multiple sets of laser point cloud data covering the railway structure area are generated; performing coordinate registration and fusion on the collected laser point cloud data, and combining position information and attitude information of each unmanned aerial vehicle; and extracting geometric dimension parameters of the railway structure from the compensated three-dimensional model, and outputting the geometric dimension parameters as measurement results. According to the scheme, high-precision non-contact measurement of geometric dimension parameters such as the track gauge, the height difference and the roll angle of a railway track structure is achieved.

Description

Technical Field

[0001] The present invention relates to the field of railway surveying and mapping technology, and in particular to a multi-UAV collaborative surveying and mapping method for areas along railways and a multi-UAV collaborative surveying and mapping system for areas along railways. Background Art

[0002] The geometric structural parameters of railway lines, including track gauge, height difference, and roll angle, are core indicators for ensuring train safety and assessing line condition. Currently, widely used non-contact measurement methods are mostly based on image acquisition and visual reconstruction technologies. Using camera-equipped track inspection vehicles, track robots, or low-altitude flying platforms, they collect image data along the railway line. Then, using algorithms such as stereo vision, structured light, or SfM (Structure from Motion), they reconstruct a 3D model and extract structural parameters.

[0003] However, in practical applications, such methods still face two key technical challenges. First, image reconstruction methods are extremely sensitive to shooting angle, lighting conditions, and feature contrast. In complex environments (such as tunnels, bridges, and curved sections), problems such as image blur, insufficient parallax, or feature extraction failures are prone to occur, resulting in incomplete reconstructed model structures or large coordinate errors. Second, the observation configuration of multi-view data lacks effective control, and the baseline parameters between different measurement nodes fluctuate greatly, resulting in unstable quality of the acquired parallax information. This makes it difficult to standardize spatial registration and error compensation during the 3D reconstruction process, affecting the final geometric measurement accuracy.

[0004] Therefore, the urgent technical problem to be solved is: how to achieve stable control of measurement perspective, observation configuration and measurement data quality in the complex structure area along the railway, so as to improve the reliability and accuracy of geometric measurement. Summary of the Invention

[0005] The purpose of the embodiments of the present invention is to provide a multi-UAV collaborative surveying method and system for areas along railways, so as to at least solve the problems of uncontrollable multi-view measurement configuration and unstable three-dimensional reconstruction accuracy in the prior art.

[0006] To achieve the above objectives, the present invention provides, in a first aspect, a multi-UAV collaborative mapping method for areas along a railway. The method comprises: deploying multiple UAVs equipped with laser ranging modules in the area along the railway, and controlling each UAV to collect monitoring data including laser ranging data, UAV position information, and UAV attitude information during flight; dynamically adjusting the relative positions and attitudes of the UAVs according to the structural characteristics of the target area during flight, and constructing an observation configuration with variable baseline parameters to achieve optimized collection of multi-angle laser ranging data. After reaching the target observation configuration, the laser ranging modules carried by each UAV are controlled to perform synchronous ranging operations to generate multiple sets of laser point cloud data covering the railway structure area; the collected laser point cloud data are coordinate-aligned and fused, and combined with the position and attitude information of each UAV to compensate for the 3D reconstruction errors caused by changes in the observation configuration; the geometric dimension parameters of the railway structure are extracted from the compensated 3D model and output as the measurement results.

[0007] Optionally, when controlling each UAV to collect monitoring data during flight, the method also includes: based on a unified time synchronization strategy, synchronously collecting laser ranging information, UAV position information and UAV attitude information, and adding time tags in a unified format to various types of data; organizing the laser ranging information, UAV position information and UAV attitude information with unified time tags into independent data frames; organizing the data frames in the order of time tags to construct a measurement data set in the form of a time series.

[0008] Optionally, the relative position and attitude between the aircraft are dynamically adjusted according to the structural characteristics of the target area, including: identifying the structural type of the target area, and selecting a configuration parameter group according to the structural type; wherein the structural characteristics include the geometric characteristics of the track and the environmental structural characteristics; the geometric characteristics include straight segments and curve segments, and the environmental structural characteristics include bridge areas and tunnel areas; the configuration parameter group includes multiple preset baseline length and attitude angle combinations; according to the current flight state, a configuration parameter group that meets the target area adaptation conditions is selected, and the changes in the flight path and attitude angle are controlled to obtain measurement data under this configuration.

[0009] Optionally, an observation configuration with variable baseline parameters is constructed, including: adjusting the spatial spacing between aircraft according to the change in point cloud density in the current area; when the point cloud density in a certain area is lower than a set threshold, adjusting the relative position or attitude direction between adjacent aircraft.

[0010] Optionally, coordinate registration and fusion are performed on the collected laser point cloud data, including: calculating initial spatial transformation parameters based on the drone position information and drone attitude information associated with the ranging information; uniformly transforming the point cloud coordinates corresponding to each laser ranging information into the same spatial reference system; performing iterative error optimization based on the preliminary registration to adjust the spatial transformation parameters and update the point cloud coordinate values ​​to complete data fusion.

[0011] Optionally, the position information and attitude information of each UAV are combined to compensate for the three-dimensional reconstruction error caused by the change of the observation configuration, including: for each set of laser ranging information, obtaining the corresponding UAV position information and UAV attitude information; calculating the change of the ranging direction in three-dimensional space based on the UAV attitude information; inferring the elastic deformation of the frame according to the force state of the aircraft body during flight; correcting the actual spatial position and direction of the sensor based on the elastic deformation; performing attitude vector inverse transformation and body deformation compensation on the point cloud coordinate value, updating the three-dimensional coordinate value corresponding to the ranging information, and forming error-compensated point cloud data.

[0012] Optionally, geometric dimension parameters of the railway structure are extracted from the compensated three-dimensional model, including: extracting a set of spatial points representing the track area from the point cloud data in the compensated three-dimensional model; determining multiple cross-sectional profiles in the set of spatial points; performing a curve fitting operation on the profile point set, and calculating the distance, height difference and inclination value between the left and right boundary points based on the fitting results.

[0013] Optionally, the method further includes: performing regional block processing on multiple measurement data frames corresponding to the laser ranging information, dividing each measurement data frame into multiple spatial sub-regions; performing density estimation on the corresponding laser points in each spatial sub-region, and marking the positions of low-density or void areas after comparison with various preset standards as areas where sparse data exists; based on the areas where sparse data exists, adjusting the relative flight paths between drones so that subsequent laser ranging information collection spatially covers the areas where sparse data exists in the previous round.

[0014] A second aspect of the present invention provides a multi-UAV collaborative mapping system for areas along railways, the system comprising: an acquisition unit for deploying multiple UAVs equipped with laser ranging modules in the area along the railway, and controlling each UAV to collect monitoring data including laser ranging data, UAV position information, and UAV attitude information during flight; an observation configuration construction unit for dynamically adjusting the relative positions and attitudes between the UAVs according to the structural characteristics of the target area during the flight of each UAV, and constructing an observation configuration with varying baseline parameters to achieve optimized acquisition of multi-angle laser ranging data; a processing unit for controlling the laser ranging modules carried by each UAV to perform synchronous ranging operations after reaching the target observation configuration, and generating multiple sets of laser point cloud data covering the railway structure area; a three-dimensional reconstruction unit for performing coordinate registration and fusion on the collected laser point cloud data, and combining the position information and attitude information of each UAV to compensate for the three-dimensional reconstruction error caused by the change in the observation configuration; and a reading unit for extracting the geometric dimension parameters of the railway structure from the compensated three-dimensional model, and outputting the geometric dimension parameters as measurement results.

[0015] On the other hand, the present invention provides a computer-readable storage medium having instructions stored thereon, which, when executed on a computer, enables the computer to execute the above-mentioned multi-UAV collaborative mapping method for the area along the railway.

[0016] Through the above-mentioned technical solution, the present invention achieves multi-angle, high-density laser ranging data collection of structures along the railway through the coordinated deployment and flight control of multiple drones. By dynamically adjusting the relative position and attitude between the aircraft, an observation configuration with variable baseline parameters is constructed, effectively improving measurement parallax and spatial coverage. Based on the multiple sets of laser point cloud data generated by synchronous ranging operations, coordinate alignment and error compensation are combined with the drone position and attitude information to reduce reconstruction deviations caused by configuration changes. Ultimately, structural parameters such as track gauge and height difference can be extracted from the high-precision point cloud model, improving the stability and accuracy of railway geometric measurements.

[0017] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following detailed description, they are used to explain the embodiments of the present invention, but do not constitute a limitation of the embodiments of the present invention. In the accompanying drawings: Figure 1 This is a flowchart of the steps of a multi-UAV collaborative mapping method for areas along a railway provided by one embodiment of the present invention; Figure 2 This is a system structure diagram of a multi-UAV collaborative mapping system for areas along railways provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following describes the specific embodiments of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention and are not intended to limit the present invention.

[0020] Figure 1 This is a flowchart of a method for collaborative mapping of multiple UAVs along railway lines provided by one embodiment of the present invention. Figure 1 As shown, an embodiment of the present invention provides a multi-UAV collaborative mapping method for an area along a railway, the method comprising: Step S10: deploy multiple drones equipped with laser ranging modules in the area along the railway, and control each drone to collect monitoring data including laser ranging data, drone position information and drone attitude information during flight.

[0021] Specifically, when controlling each drone to collect monitoring data during flight, the method also includes: based on a unified time synchronization strategy, synchronously collecting laser ranging information, drone position information, and drone attitude information, and adding time tags in a unified format to each type of data; organizing the laser ranging information, drone position information, and drone attitude information with unified time tags into independent data frames; and organizing the data frames in the order of time tags to construct a measurement data set in the form of a time series.

[0022] In the implementation of the present invention, multiple drones equipped with laser ranging capabilities are deployed along the railway line, enabling non-contact spatial measurement of large-scale track structures. In this solution, multiple drones are controlled to collect monitoring data, including laser ranging information, drone position information, and drone attitude information, during flight. This provides raw data support for the subsequent construction of three-dimensional models and extraction of geometric dimensional parameters. To ensure the temporal and spatial consistency of multi-source monitoring data, the solution introduces a unified time synchronization strategy to coordinate the collection and unified identification of various types of measurement data.

[0023] Specifically, during drone flight, laser ranging information records the distance returned by the current ranging device after emitting laser pulses toward the ground or structure. This information is highly time-sensitive. To ensure that this information corresponds to the drone's position and attitude at the moment of measurement, it is necessary to simultaneously record spatial position information (such as 3D coordinates or latitude and longitude data) and attitude information (including pitch, roll, and yaw angles) at the moment of flight. The solution of the present invention uses a unified time reference source, such as GNSS timing or a high-stability internal clock, to uniformly time-calibrate all data types involved in the measurement.

[0024] During the synchronous acquisition process, a unified time tag, or timestamp, is attached to each set of laser ranging data, drone position information, and drone attitude information. This unique and globally consistent time tag ensures that data from different sources are correctly associated with the same sampling moment during subsequent processing. To ensure efficient data processing and clear structural logic, all collected data is encapsulated into independent data frames. Each frame contains a set of laser ranging values, the drone's position coordinates at the corresponding moment, and the attitude angle value.

[0025] Furthermore, all time-tagged data frames are organized in chronological order to construct a measurement dataset in the form of a time series. This time series not only preserves the spatial continuity of the laser point cloud data but also provides a foundation for subsequent point cloud stitching, attitude compensation, and coordinate alignment. This approach not only ensures that each laser point fully matches its flight state at the time of acquisition, but also allows the generation process of each measurement point in space to be accurately tracked when constructing the 3D point cloud model, reducing geometric errors caused by attitude drift or measurement delays.

[0026] This solution effectively addresses common issues in multi-UAV collaborative measurement, such as data misalignment, time drift, and attitude-range decoupling. It ensures the consistency of collected data in both the temporal and spatial domains, thereby improving the accuracy and stability of subsequent 3D model construction. By unifying time control and data frame construction, it supports data fusion across different flight paths and attitudes, maintaining the geometric continuity and spatial integrity of measurement data even in complex railway structural areas (such as curves, tunnels, and bridges). This provides a reliable data foundation for the high-precision extraction of key geometric parameters such as track gauge, elevation difference, and roll angle.

[0027] Step S20: During the flight of each UAV, the relative position and attitude between the aircraft are dynamically adjusted according to the structural characteristics of the target area, and an observation configuration with variable baseline parameters is constructed to achieve optimized collection of multi-angle laser ranging data.

[0028] Specifically, the relative positions and attitudes between aircraft are dynamically adjusted according to the structural characteristics of the target area, including: identifying the structural type of the target area and selecting a configuration parameter group according to the structural type; wherein the structural characteristics include the geometric characteristics of the track and the environmental structural characteristics; the geometric characteristics include straight segments and curve segments, and the environmental structural characteristics include bridge areas and tunnel areas; the configuration parameter group includes multiple preset baseline length and attitude angle combinations; according to the current flight state, a configuration parameter group that meets the target area adaptation conditions is selected, and the changes in the flight path and attitude angle are controlled to obtain measurement data under the configuration.

[0029] Furthermore, an observation configuration with variable baseline parameters is constructed, including: adjusting the spatial spacing between aircraft according to the changes in the point cloud density in the current area; when the point cloud density in a certain area is lower than the set threshold, adjusting the relative position or attitude direction between adjacent aircraft.

[0030] In an embodiment of the present invention, to obtain laser ranging data with high spatial coverage and measurement accuracy during surveying and mapping missions along railway lines, it is necessary to dynamically adjust the flight configurations of multiple drones based on the structural characteristics of different terrain environments. By constructing an observation configuration with variable baseline parameters, the present invention allows each drone to flexibly adjust its relative position and posture throughout the flight process based on changes in the actual mission area. This enables multi-angle, multi-parallax laser ranging data acquisition, thereby improving the spatial resolution capabilities of the survey area and the integrity of the point cloud data.

[0031] The proposed solution identifies the structural type of the target area during flight. This identification is based on the spatial morphological characteristics of the railway line, primarily encompassing two types of information: track geometry, such as straight segments, horizontal and vertical curves; and environmental structural characteristics, such as open areas, tunnel sections, bridge structures, and station areas. By combining navigation data, pre-stored track geographic information, or a quick pre-scan performed before the mission begins, the type of area to be surveyed can be classified.

[0032] For different area types, the method presets multiple configuration parameter groups. The configuration parameter groups include multiple control variables used to describe the spatial arrangement between drones, mainly including the following categories: 1) Baseline length range, that is, the relative distance between the two UAVs on the horizontal plane, is used to control observation parallax.

[0033] 2) Observation angle range, that is, the angle formed between the laser ranging directions of each drone, which is used to adjust the ranging angle coverage range.

[0034] 3) Attitude angle setting, namely the pitch angle, roll angle and yaw angle parameters of each drone, which are used to construct different ranging fields of view.

[0035] 4) Relative height difference: adjust the elevation when there is a risk of occlusion or when you want to increase the vertical parallax distribution.

[0036] Each configuration parameter group corresponds to a typical structural area, such as using a linear array configuration in a straight section, a fan-shaped arrangement in a curved section, and a near-parallel flight mode in a tunnel to avoid reflection interference.

[0037] During flight, the method selects a matching configuration parameter set based on the current flight state (including position, attitude, speed, etc.) and the predicted type of the area ahead. Once the target configuration is confirmed, the control logic initiates the corresponding adjustment strategy, instructing the drone to adjust its relative position or attitude angle. For example, if entering a curved section, the original parallel formation is adjusted to an observation formation with a certain angle. If entering a tunnel, the baseline length is reduced and the laser ranging direction is made as parallel as possible to avoid ranging interference caused by the curved structure.

[0038] After selecting and initially adjusting the configuration parameter set, the present invention dynamically modifies the configuration based on the density distribution of real-time point cloud data to further enhance the effectiveness of ranging data. Point cloud density can be assessed by counting the number of valid laser reflection points per unit voxel (e.g., 0.5 m³). If the point cloud density in a certain area continuously falls below a set threshold (e.g., fewer than N points per cubic meter), the area is considered to have a ranging blind spot or insufficient spatial coverage.

[0039] At this point, the method is modified in two ways: 1) Adjust the relative distance between adjacent drones, such as shortening the baseline appropriately to increase the overlap rate of parallax on the same structural details.

[0040] 2) Adjust attitude and direction, such as changing the ranging angle to avoid obstructions or enhance oblique measurement capabilities. This adaptive configuration adjustment based on point cloud data feedback improves the scene adaptability of the observation configuration, enabling drone swarms to maintain high-quality coverage of target structures even in changing terrain.

[0041] In addition, to ensure data continuity during the configuration transformation process, the solution of the present invention adopts a frame-by-frame configuration transition mechanism. That is, the target configuration parameters are used as the target state of the next frame configuration, and the flight path and attitude angle are gradually adjusted through a certain step size to avoid acquisition interruption or reconstruction error caused by large configuration jumps. This mechanism can enable the UAV to continue to maintain stable laser ranging data acquisition capabilities during configuration adjustment. The solution of the present invention also includes a configuration rationality verification step in its implementation. By calculating the global spatial coverage, laser ranging direction intersection angle distribution and parallax distribution index of the current observation configuration, it is determined whether the configuration meets the three-dimensional reconstruction input conditions, such as the minimum baseline length, maximum parallax coverage, etc. If the verification fails, it automatically falls back to the last stable configuration or selects other backup solutions from the configuration template library for replacement.

[0042] This solution, based on the present invention, introduces structural adaptability during the laser ranging data acquisition phase through dynamic configuration control, significantly improving point cloud coverage and measurement perspective diversity in complex areas (such as bridge junctions, curved ramps, and tunnel exits). Compared to fixed flight paths or uniform configuration approaches, this configuration adaptation strategy optimizes the matching of ranging direction and configuration based on target structural characteristics, effectively alleviating measurement obstacles such as ranging blind spots, insufficient reconstruction parallax, and severe occlusion. This enhanced configuration controllability also enhances 3D model stability and geometric measurement reliability during subsequent coordinate alignment, point cloud splicing, and error compensation.

[0043] Step S30: After reaching the target observation configuration, the laser ranging modules carried by each UAV are controlled to perform synchronous ranging operations to generate multiple sets of laser point cloud data covering the railway structure area.

[0044] Specifically, after reaching the target observation configuration, synchronized ranging operations are performed across multiple aircraft to ensure consistency in both spatial and temporal laser ranging data. During the preparation phase for synchronized ranging, a unified ranging trigger sequence is preset for each aircraft. This sequence is based on a shared time base and initialized before flight commences via a high-precision timing mechanism. During flight, each aircraft continuously adjusts its clock to the unified time base to ensure synchronized responses to ranging commands at the moment of triggering.

[0045] During synchronized ranging, all aircraft complete laser beam emission and echo reception within a predetermined time window, recording the corresponding distance information, acquisition time, 3D position, and attitude angle in the same coordinate reference system. To improve spatial resolution, multiple sets of laser transmissions can be executed within each synchronized acquisition cycle, with each set of laser data corresponding to a point cloud subset. By simultaneously measuring the same structural area with multiple aircraft in different attitudes and relative positions, laser point cloud data can be obtained from multiple angles, enhancing the spatial sampling density and feature representation capabilities of the structural surface.

[0046] Based on the solution of the present invention, the use of this synchronous ranging method not only improves the temporal consistency between different point cloud data, but also ensures the spatial correlation and reconstruction consistency between point clouds, effectively improves the geometric restoration accuracy of railway structures under multi-angle measurement, and establishes a reliable data foundation for subsequent point cloud fusion and error compensation.

[0047] Step S40: coordinate registration and fusion of the collected laser point cloud data, and combining the position information and attitude information of each UAV to compensate for the 3D reconstruction error caused by the change of the observation configuration.

[0048] Specifically, the collected laser point cloud data are subjected to coordinate registration and fusion, including: calculating the initial spatial transformation parameters based on the drone position information and drone attitude information associated with the ranging information; uniformly transforming the point cloud coordinates corresponding to each laser ranging information into the same spatial reference system; performing iterative error optimization based on the preliminary registration to adjust the spatial transformation parameters and update the point cloud coordinate values ​​to complete data fusion.

[0049] Furthermore, the position information and attitude information of each UAV are combined to compensate for the three-dimensional reconstruction error caused by the change of the observation configuration, including: for each set of laser ranging information, the corresponding UAV position information and UAV attitude information are obtained; based on the UAV attitude information, the change of the ranging direction in three-dimensional space is calculated; according to the force state of the aircraft body during flight, the elastic deformation of the frame is calculated; based on the elastic deformation, the actual spatial position and direction of the sensor are corrected; the attitude vector inverse transformation and body deformation compensation are performed on the point cloud coordinate values, and the three-dimensional coordinate values ​​corresponding to the ranging information are updated to form error-compensated point cloud data.

[0050] In an embodiment of the present invention, coordinate registration and fusion of collected laser point cloud data is a key step in achieving high-precision 3D reconstruction of railway structures. Because multiple aircraft collect laser ranging information at different positions and attitudes, the original coordinates of the generated point cloud data are distributed in their respective local coordinate systems, and there are certain spatial deviations caused by attitude errors, position drift, and configuration changes. Therefore, it is necessary to unify the various local point clouds into the same global coordinate reference system and improve the consistency and geometric accuracy of the fused point cloud through a multi-level registration and error compensation mechanism.

[0051] The initial registration phase calculates the spatial transformation parameters at the time of acquisition based on the drone position and attitude information associated with each set of laser ranging information. Position information provides a three-dimensional translation vector, while attitude information includes three-dimensional rotations of pitch, roll, and yaw. By converting the attitude angles into a rotation matrix, a homogeneous transformation matrix from the local coordinate system to the target reference system can be constructed for each data frame. The three-dimensional coordinate vector in each point cloud is multiplied by this transformation matrix to obtain point cloud data unified to the world coordinate system. This step decouples the multi-source point clouds from their respective drone reference frames, forming an initial alignment of point clouds with global spatial significance.

[0052] After the initial transformation, an iterative error optimization is performed in a unified coordinate system to improve fusion quality. This optimization process is based on point pairs in the spatial overlap region and calculates the residual vectors between adjacent point clouds using an iterative closest point (ICP) algorithm or its improved variants (such as weighted ICP, color ICP, and normal ICP). By minimizing these residuals, the rotation and translation parameters in the spatial transformation matrix are optimized to correct the accumulated errors in the initial transformation. This process can be performed locally to correct the point cloud overlap errors of adjacent aircraft, or it can be optimized globally to improve the overall model consistency. During the registration optimization process, multi-level registration accuracy thresholds can be set to gradually refine the error convergence range to achieve fine alignment of the fused point clouds in terms of spatial continuity and local geometric structure.

[0053] After completing coordinate registration and point cloud fusion, residual 3D reconstruction errors may still exist due to changes in the observed configuration. To further improve the spatial accuracy of the point cloud, it is necessary to combine the drone position information and attitude information of each set of laser ranging information to compensate for configuration errors. For each set of laser ranging information, the corresponding drone position information and attitude information must be obtained simultaneously. Position information is generally represented by GNSS coordinates or spatial three-dimensional position calculated by an inertial navigation system (INS), while attitude information usually includes pitch, roll, and yaw angles to describe the orientation of the drone in space. At this time, it is necessary to ensure that the pose data and laser ranging data have a unified time tag to ensure the temporal and spatial correspondence of the data.

[0054] Based on the acquired drone attitude information, the change in the laser ranging direction in three-dimensional space can be calculated. Specifically, the attitude angles are converted into a rotation matrix. The fixed emission direction vector of the laser in the drone's coordinate system is then transformed to the ground reference system using this rotation matrix. This results in the true emission direction of each ranging beam in three-dimensional space. This step is fundamental processing, used to correctly project the local ranging data into the global coordinate system.

[0055] Furthermore, the effects of elastic deformation of the aircraft body during flight due to aerodynamic forces, load changes, or ambient temperature fluctuations can be considered. To estimate the elastic deformation of the frame, a simplified frame elastic model can be established by referring to flight acceleration data, attitude change rate, wind speed information, and external temperature data. For example, the frame can be treated as a simplified beam structure model, and the small displacements of each node can be calculated using conventional mechanical formulas. Using a small-angle approximation, the deflection changes of the frame can be calculated into elastic deformations on the order of micrometers to millimeters based on the applied load and material stiffness parameters (such as Young's modulus).

[0056] After determining the elastic deformation of the airframe, the actual spatial position and orientation of the sensor must be corrected based on this deformation. Specifically, the sensor mounting matrix obtained through initial static calibration is superimposed with subtle pose changes (such as small translations and rotations) calculated in real time to produce a dynamically updated sensor pose description. The sensor's launch origin position and ranging direction vector are then finely adjusted to match the spatial offset caused by airframe deformation during dynamic flight.

[0057] Subsequently, the original coordinate values ​​of the laser point cloud are subjected to dual processing of attitude vector inverse transformation and body deformation compensation. First, the attitude inverse transformation is applied to convert the ranging points from the local aircraft reference system to the ground absolute reference system; then, based on the sensor posture corrected by elastic deformation, the three-dimensional coordinates of each ranging point are further adjusted. The adjustment method can adopt the vector superposition method, that is, the spatial displacement of the sensor due to deformation is added to the point cloud coordinates after the preliminary inverse transformation. Through the above process, the final error-compensated point cloud dataset can be formed. This dataset not only corrects the measurement error caused by the change of the aircraft's attitude, but also further compensates for the measurement offset caused by the elastic deformation of the aircraft body, which can effectively improve the spatial accuracy and consistency of the three-dimensional reconstructed model under multi-UAV collaborative mapping.

[0058] In one possible implementation, the dual compensation mechanism described above can reduce the local spatial error of point cloud data by over 30% on average under flight conditions with wind speeds exceeding 5 m / s and drone payload variations exceeding 10%. This significantly improves the stability and reliability of track geometry parameter extraction (such as gauge, elevation difference, and roll angle). Especially in long-distance, multi-batch surveying and mapping missions, elastic deformation compensation can effectively suppress cumulative point cloud drift, improving the consistency of overall measurement results and possessing significant engineering value.

[0059] Based on the solution of the present invention, the coordinate registration and error compensation method described above can effectively resolve point cloud geometric inconsistencies caused by observation configuration changes, attitude disturbances, and position information errors between multiple drones. Compared to directly splicing point clouds without processing, this method significantly improves the overall structural alignment, detail fidelity, and spatial geometric stability of the fused point cloud. In particular, in curved sections, tunnels, or areas with occluding structures, it ensures that point clouds from all directions are effectively aligned, improving the density and integrity of the overall model. This ensures the accuracy of subsequent geometric parameter extraction (such as track gauge, height difference, and roll angle), meeting the core requirement for high-reliability 3D data in railway non-contact measurement.

[0060] Step S50: extracting geometrical dimension parameters of the railway structure from the compensated three-dimensional model, and outputting the geometrical dimension parameters as measurement results.

[0061] Specifically, the geometric dimension parameters of the railway structure are extracted from the compensated three-dimensional model, including: extracting a set of spatial points representing the track area from the point cloud data in the compensated three-dimensional model; determining multiple cross-sectional profiles in the set of spatial points; performing a curve fitting operation on the profile point set, and calculating the distance, height difference and inclination value between the left and right boundary points based on the fitting results.

[0062] In this embodiment of the present invention, extracting the geometric dimensions of railway structures from the compensated 3D model is a key step in the entire surveying and mapping process, directly related to the quantitative analysis and assessment of the railway line's structural condition. After completing point cloud coordinate registration and attitude error compensation, a set of laser point cloud data with a unified coordinate reference, high geometric accuracy, and spatial integrity is obtained. This dataset represents the spatial morphology of the railway structures along the line in three dimensions. Based on this, geometric analysis methods are used to extract and quantify key structural areas, resulting in multiple geometric indicators such as track gauge, height difference, and roll angle, which serve as the final output of the railway survey.

[0063] Specifically, the first step is to identify a set of spatial points representing the track structure in the three-dimensional point cloud data. This step can be accomplished by combining methods such as setting a ground height threshold, filtering point density features, and geometric morphology recognition. Since the track structure appears in the point cloud as a high-density, parallel, strip-like region extending longitudinally, the consistency of the surface normal distribution and the spacing between structural lines can be used to extract the point cloud segments of the two rails and construct the spatial boundary of the track area. To eliminate the influence of non-target structures (such as trackside facilities and ground interference objects) on subsequent calculations, the extracted results must be spatially filtered and outliers removed.

[0064] After extracting the track area point cloud, multiple cross-sectional point sets are captured at regular intervals along the track's longitudinal direction. Each cross-sectional point set contains the left and right rail structural points at the current cross-sectional location. By projecting the point cloud into the cross-sectional coordinate system, an approximate two-dimensional cross-sectional structure of the track can be obtained. Curve fitting is performed on each set of cross-sectional point clouds, using methods such as quadratic polynomials, spline interpolation, and circular arc fitting to describe the rail contour boundary shape. During the fitting process, the rail head or tread edge points are preferentially identified as structural references to prevent stray points on the outside of the curve from interfering with the fitting results.

[0065] After curve fitting is complete, structural parameter calculations are performed based on the geometric relationships between the fitted curves. Track gauge is calculated as the shortest horizontal distance between the rail curves on either side of the profile, typically taking the Euclidean distance between representative points (such as the inner edge of the rail top). Height difference is determined by comparing the vertical elevation difference between the fitted curves on both sides of the rail surface within the same section, reflecting the unevenness of the track cross section. The roll angle is calculated based on the angle between the normal to the rail surface tangent and the horizontal plane, revealing the lateral tilt of the track structure.

[0066] Preferably, the method further includes: performing regional block processing on multiple measurement data frames corresponding to the laser ranging information, dividing each measurement data frame into multiple spatial sub-regions; performing density estimation on the corresponding laser points in each spatial sub-region, and marking the positions of low-density or void areas after comparison with various preset standards as areas with sparse data; based on the areas with sparse data, adjusting the relative flight paths between drones so that subsequent laser ranging information collection spatially covers the areas with sparse data in the previous round.

[0067] In this embodiment of the present invention, after completing the initial laser ranging information collection, to improve the spatial integrity and structural coverage of the laser point cloud data, the collected data can be further processed by performing regional segmentation and spatial sparsity detection. This processing step is primarily used to identify areas of insufficient coverage or low density in the point cloud data. By optimizing the flight path, subsequent ranging tasks can focus on compensating for these sparse data areas, thereby improving the spatial continuity and geometric accuracy of the overall 3D model.

[0068] In specific operations, multiple measurement data frames corresponding to the laser ranging information are first divided into a 3D grid at a uniform spatial resolution. The point cloud data within each measurement data frame is divided into several spatial subregions with a fixed voxel size, for example, a voxel side length of 0.5 or 1 meter. The number of laser points within each subregion is counted and used as a density value to reflect the spatial sampling status of that subregion. To improve statistical accuracy, the data from multiple frames can be superimposed and calculated to form a temporally fused density map.

[0069] The point cloud density of each spatial subregion is then compared to a preset threshold. This threshold is determined based on the measurement scenario and structural complexity. For example, a track structure area must have at least 100 valid points per cubic meter. If a region has insufficient points, it is marked as a "low-density area" or "data hole area." Furthermore, spatial segments with multiple consecutive subregions showing low density can be further identified as areas with severe structural obstruction or ranging blind spots, requiring compensation through configuration or path adjustments.

[0070] After the marking is completed, the path optimization adjustment for the next stage of the ranging task is carried out according to the spatial distribution results of the sparse area. The adjustment strategy may include: appropriately reducing the lateral spacing between UAVs to improve the coverage of the observation overlapping area; adjusting the attitude angle or the downward angle of some aircraft to change the direction of the laser beam illumination to compensate for the area that was originally blocked or the measurement angle was insufficient; or inserting an additional scanning path above the sparse area to form a local coverage enhancement segment. This data acquisition adjustment mechanism based on sparse area feedback has high adaptability and local compensation capabilities, and is particularly suitable for measurement scenarios with complex structures and frequent occlusions (such as bridge connection sections, track intersection sections or side wall occlusion sections). Compared with the traditional uniform scanning path, this strategy can significantly improve the density and completeness of point cloud data in key parts of the structure and reduce reconstruction errors caused by data loss.

[0071] The proposed approach, through regional segmentation and sparsity analysis of laser ranging data, not only automatically identifies measurement blind spots but also establishes a closed-loop mechanism that drives flight path adjustments from the measurement results. This method improves the data closure of ranging tasks and the boundary integrity of point cloud models, thereby enhancing the stability and reliability of subsequent geometric parameter extraction, providing a solid data foundation for high-precision modeling and structural analysis of railway structures.

[0072] Figure 2 This is a system structure diagram of a multi-UAV collaborative mapping system for railway areas provided by an embodiment of the present invention. Figure 2 As shown, an embodiment of the present invention provides a multi-UAV collaborative mapping system for areas along railways, the system comprising: an acquisition unit for deploying multiple UAVs equipped with laser ranging modules in the area along the railway, and controlling each UAV to acquire monitoring data including laser ranging data, UAV position information, and UAV attitude information during flight; an observation configuration construction unit for dynamically adjusting the relative positions and attitudes between the UAVs according to the structural characteristics of the target area during flight, and constructing an observation configuration with variable baseline parameters to achieve optimized acquisition of multi-angle laser ranging data; a processing unit for controlling the laser ranging modules carried by each UAV to perform synchronous ranging operations after reaching the target observation configuration, and generating multiple sets of laser point cloud data covering the railway structure area; a three-dimensional reconstruction unit for performing coordinate registration and fusion on the acquired laser point cloud data, and combining the position information and attitude information of each UAV to compensate for the three-dimensional reconstruction error caused by the change in the observation configuration; and a reading unit for extracting geometric dimension parameters of the railway structure from the compensated three-dimensional model and outputting the geometric dimension parameters as measurement results.

[0073] An embodiment of the present invention also provides a computer-readable storage medium, which stores instructions that, when executed on a computer, enable the computer to execute the above-mentioned multi-UAV collaborative mapping method for areas along the railway.

[0074] Those skilled in the art will appreciate that all or part of the steps in the methods described in the aforementioned embodiments can be performed by instructing the relevant hardware through a program. The program, stored in a storage medium, includes instructions for causing a microcontroller, chip, or processor to execute all or part of the steps in the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0075] The above describes in detail the optional embodiments of the present invention in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the technical concept of the embodiments of the present invention, a variety of simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the scope of protection of the embodiments of the present invention. It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner unless there is any contradiction. In order to avoid unnecessary repetition, the embodiments of the present invention will no longer describe the various possible combinations separately.

[0076] In addition, the various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the embodiments of the present invention, they should also be regarded as the contents disclosed in the embodiments of the present invention.

Claims

1. A multi-UAV collaborative mapping method for areas along railways, characterized in that: The method comprises: Deploy multiple drones equipped with laser ranging modules along the railway line and control each drone to collect monitoring data including laser ranging data, drone position information, and drone attitude information during flight; During the flight of each UAV, the relative position and attitude of the aircraft are dynamically adjusted according to the structural characteristics of the target area, and an observation configuration with variable baseline parameters is constructed to achieve optimized collection of multi-angle laser ranging data; After reaching the target observation configuration, the laser ranging modules carried by each UAV are controlled to perform synchronous ranging operations to generate multiple sets of laser point cloud data covering the railway structure area; The collected laser point cloud data is aligned and fused, and the position and attitude information of each drone is combined to compensate for the 3D reconstruction error caused by the change of the observation configuration; The geometrical dimension parameters of the railway structure are extracted from the compensated three-dimensional model and output as measurement results.

2. The multi-UAV collaborative mapping method for railway areas according to claim 1 is characterized in that: When controlling each UAV to collect monitoring data during flight, the method further includes: Based on a unified time synchronization strategy, laser ranging information, drone position information, and drone attitude information are collected synchronously, and time tags in a unified format are added to each type of data; The laser ranging information, UAV position information and UAV attitude information with unified time tags are organized into independent data frames; The data frames are organized in the order of time labels to construct a measurement dataset in the form of a time series.

3. The multi-UAV collaborative mapping method for railway areas according to claim 1 is characterized in that: Dynamically adjust the relative position and attitude of aircraft based on the structural characteristics of the target area, including: Identify the structural type of the target area and select a configuration parameter group according to the structural type; The structural features include the geometric features of the track and the environmental structural features; the geometric features include straight segments and curved segments, and the environmental structural features include bridge areas and tunnel areas; The configuration parameter group includes a plurality of preset baseline length and posture angle combinations; According to the current flight state, a configuration parameter group that meets the adaptation conditions of the target area is selected, and the changes in the flight path and attitude angle are controlled to obtain the measurement data under this configuration.

4. The multi-UAV collaborative mapping method for railway areas according to claim 1 is characterized in that: Construct observation configurations with varying baseline parameters, including: Adjust the spatial distance between aircraft based on the changes in point cloud density in the current area; When the point cloud density in a certain area is lower than the set threshold, the relative position or attitude direction between adjacent aircraft is adjusted.

5. The multi-UAV collaborative mapping method for railway areas according to claim 1 is characterized in that: Coordinate registration and fusion of collected laser point cloud data, including: Calculating initial spatial transformation parameters based on the UAV position information and the UAV attitude information associated with the ranging information; The point cloud coordinates corresponding to each laser ranging information are uniformly transformed into the same spatial reference system; Based on the preliminary registration, iterative error optimization is performed to adjust the spatial transformation parameters and update the point cloud coordinate values ​​to complete data fusion.

6. The multi-UAV collaborative mapping method for railway areas according to claim 1, characterized in that: Combining the position and attitude information of each drone to compensate for 3D reconstruction errors caused by changes in the observation configuration, including: For each set of laser ranging information, obtain the corresponding drone position information and drone attitude information; Calculate the change of ranging direction in three-dimensional space based on the UAV attitude information; Calculate the elastic deformation of the frame according to the stress state of the aircraft body during flight; Correct the actual spatial position and orientation of the sensor based on the elastic deformation; Perform inverse posture vector transformation and body deformation compensation on the point cloud coordinate values, update the three-dimensional coordinate values ​​corresponding to the ranging information, and form error-compensated point cloud data.

7. The multi-UAV collaborative mapping method for railway areas according to claim 1, characterized in that: Extract the geometric parameters of the railway structure from the compensated 3D model, including: Extracting a set of spatial points representing the track area from the point cloud data of the compensated three-dimensional model; determining a plurality of cross-sectional profiles in a set of spatial points; A curve fitting operation is performed on the profile point set, and the distance, height difference, and inclination angle between the left and right boundary points are calculated based on the fitting results.

8. The multi-UAV collaborative mapping method for railway areas according to claim 1, characterized in that: The method further comprises: Performing regional block processing on multiple measurement data frames corresponding to the laser ranging information, dividing each measurement data frame into multiple spatial sub-regions; Density estimation is performed on the corresponding laser points in each spatial sub-region, and after comparison with various preset standards, the locations of low-density or empty areas are marked as areas with sparse data; Based on the areas with sparse data, the relative flight paths between drones are adjusted so that the subsequent laser ranging information collection spatially covers the areas with sparse data in the previous round.

9. A multi-UAV collaborative mapping system for areas along railways, characterized by: The system comprises: The acquisition unit is used to deploy multiple drones equipped with laser ranging modules along the railway line and control each drone to collect monitoring data including laser ranging data, drone position information, and drone attitude information during flight; The observation configuration construction unit is used to dynamically adjust the relative position and attitude of each UAV according to the structural characteristics of the target area during the flight of each UAV, and to construct an observation configuration with variable baseline parameters to achieve optimized collection of multi-angle laser ranging data; The processing unit is used to control the laser ranging modules carried by each UAV to perform synchronous ranging operations after reaching the target observation configuration, thereby generating multiple sets of laser point cloud data covering the railway structure area; The 3D reconstruction unit is used to coordinate registration and fusion of the collected laser point cloud data, and to compensate for the 3D reconstruction errors caused by changes in the observation configuration by combining the position and attitude information of each UAV; The reading unit is used to extract geometric dimension parameters of the railway structure from the compensated three-dimensional model and output the geometric dimension parameters as measurement results.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the multi-UAV collaborative mapping method for an area along a railway as described in any one of claims 1 to 8.

Images