A terrain mapping method and device based on uncertainty analysis, electronic equipment and storage medium

Abstract

The present application relates to a kind of topographic mapping methods based on uncertainty analysis, including generating terrain candidate point set;Obtain effective voxel set;Obtain plane feature set;Obtain octree hash map structure;Update octree hash map structure, complete topographic mapping.The present application also relates to a kind of topographic mapping device based on uncertainty analysis, including acquisition, first execution, second execution, processing and update module.The present application also relates to an electronic device and storage medium.Using the present application designs a kind of topographic mapping method, device, electronic equipment and storage medium based on uncertainty analysis, can solve point cloud map, plane representation map or fixed resolution voxel mapping technology in existence measurement noise and pose error, construction and query overhead, poor stability, difficult to take into account sparse and dense area, difficult to achieve balance in mapping, cause the problem of registration accuracy, real-time and map compactness limit of unmanned aerial vehicle three-dimensional environment mapping.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicles, and in particular to a terrain mapping method and device based on uncertainty analysis, an electronic device and a storage medium. BACKGROUND

[0002] To meet the needs of autonomous navigation and mapping for real-time and high-precision three-dimensional mapping, in the prior art, point cloud maps, plane representation maps or fixed resolution voxel mapping are usually used. However, point cloud maps are difficult to explicitly model map uncertainty, often treat local planes as deterministic quantities, and ignore measurement noise and pose errors; as the map size increases, the construction and query overheads based on kd-tree increase; plane representation maps are sensitive to sparse / uneven point clouds, and their stability is affected; fixed voxel mapping is difficult to balance the expression ability of sparse and dense areas, and it is difficult to balance between coarse and fine when facing various structures such as ground, wall and debris; the above problems limit the registration accuracy, real-time performance and map compactness of unmanned aerial vehicle three-dimensional environment mapping.

[0003] Therefore, the prior art has defects and needs to be improved and developed. SUMMARY

[0004] The present application provides a terrain mapping method and device based on uncertainty analysis, an electronic device and a storage medium, which are used to solve the problems that in the prior art, point cloud maps, plane representation maps or fixed resolution voxel mapping are usually used, but point cloud maps are difficult to explicitly model map uncertainty, often treat local planes as deterministic quantities, and ignore measurement noise and pose errors; as the map size increases, the construction and query overheads based on kd-tree increase; plane representation maps are sensitive to sparse / uneven point clouds, and their stability is affected; fixed voxel mapping is difficult to balance the expression ability of sparse and dense areas, and it is difficult to balance between coarse and fine when facing various structures such as ground, wall and debris; the above problems also limit the registration accuracy, real-time performance and map compactness of unmanned aerial vehicle three-dimensional environment mapping.

[0005] In a first aspect of the present application, a terrain mapping method based on uncertainty analysis is provided, comprising:

[0006] Collecting original point cloud data and initial pose estimation obtained by an unmanned aerial vehicle to generate a terrain candidate point set;

[0007] Performing voxel division on the terrain candidate point set to obtain an effective voxel set;

[0008] Performing plane fitting on the effective voxel set to obtain a plane feature set;

[0009] Processing the plane feature set to obtain an octree hash map structure;

[0010] Collecting newly added original point cloud data and initial pose estimation obtained by the unmanned aerial vehicle to update the octree hash map structure and complete the terrain mapping.

[0011] Further, the original point cloud data and initial pose estimation obtained by the unmanned aerial vehicle are collected to generate a terrain candidate point set, including:

[0012] Collecting the original point cloud data obtained by the radar carried on the unmanned aerial vehicle and the initial pose estimation output by the state estimation module installed in the unmanned aerial vehicle;

[0013] Converting the original point cloud data to a current global reference coordinate system according to the initial pose estimation to obtain converted point cloud data;

[0014] Sorting the converted point cloud data according to the height channel to filter out terrain candidate points according to a preset minimum height percentage;

[0015] Collecting all the filtered terrain candidate points to generate the terrain candidate point set.

[0016] Further, the terrain candidate point set is subjected to voxel division to obtain an effective voxel set, including:

[0017] Dividing the terrain candidate point set into voxel blocks at a coarse resolution;

[0018] Performing principal component analysis on each voxel block to obtain covariance matrix eigenvalues;

[0019] Judging planarity. If the ratio of the smallest eigenvalue to the second largest eigenvalue in the covariance matrix eigenvalues is not significantly reduced, or the angle between the fitted plane obtained through the covariance matrix eigenvalues and the horizontal plane is not less than a preset angle threshold, the planarity is insufficient, and the voxel block is subdivided into smaller sub-voxel blocks, and the principal component analysis and the planarity judgment are repeatedly performed on each sub-voxel block until the planarity is sufficient or the minimum voxel size is reached. If the ratio of the smallest eigenvalue to the second largest eigenvalue in the covariance matrix eigenvalues is significantly reduced, and the angle between the fitted plane obtained through the covariance matrix eigenvalues and the horizontal plane is less than the preset angle threshold, the planarity is sufficient;

[0020] Retaining all voxel blocks with sufficient planarity to generate an effective voxel set.

[0021] Further, the effective voxel set is subjected to plane fitting to obtain a plane feature set, including:

[0022] Performing plane fitting on each point set in the effective voxel set to obtain a plane normal vector n, a plane center point c, and a plane covariance matrix Σ;

[0023] constructing a triple (n, c,∑) according to the plane normal vector n, the plane center point c and the plane covariance matrix∑;

[0024] collecting the triple of each point set in the effective voxel set to obtain the plane feature set.

[0025] Further, the plane feature set is processed to obtain an octree hash map structure, comprising:

[0026] Each plane feature in the plane feature set is converted into an integer grid index according to the voxel space coordinates to which it belongs, and a hash table Hash[(x, y, z)]=(n, q,∑) is constructed, where x, y and z are the horizontal coordinate, vertical coordinate and vertical coordinate of the voxel space coordinates, respectively.

[0027] The octree structure is used to manage the spatial hierarchy, and the octree hash map structure is obtained.

[0028] Further, the newly added original point cloud data obtained by the unmanned aerial vehicle and the newly added initial pose estimation are collected to update the octree hash map structure, and the terrain mapping is completed, comprising:

[0029] The newly added original point cloud data is searched in the effective voxel set, if the newly added original point cloud data exists in the effective voxel set, the cumulative observation number is accumulated and the newly added point cloud data is added to the effective voxel set, when the cumulative observation number reaches the cumulative observation number threshold, the original point cloud data existing in the effective voxel set is deleted; if the newly added original point cloud data does not exist in the effective voxel set, the plane fitting is performed on the newly added original point cloud data and is added to the plane feature set.

[0030] Further, after the terrain mapping is completed, it further comprises:

[0031] Collecting the current point cloud data obtained by the unmanned aerial vehicle;

[0032] According to the current point cloud data, the to-be-processed voxel containing the current point cloud data is located in the octree hash map structure;

[0033] The plane features contained in the to-be-processed voxel are traversed to match the current point cloud data and calculate the point-to-plane distance residual error;

[0034] The point-to-plane distance residual error is weighted according to the covariance to construct an optimization equation;

[0035] The optimization equation is iteratively solved to obtain the current pose estimation of the unmanned aerial vehicle under the current point cloud data;

[0036] feedback the current pose estimation to the UAV to optimize the UAV pose.

[0037] In a second aspect, the present application provides a terrain mapping device based on uncertainty analysis, comprising:

[0038] a collection module configured to collect raw point cloud data and initial pose estimation obtained by the UAV to generate a terrain candidate point set;

[0039] a first execution module configured to perform voxel division on the terrain candidate point set to obtain an effective voxel set;

[0040] a second execution module configured to perform plane fitting on the effective voxel set to obtain a plane feature set;

[0041] a processing module configured to process the plane feature set to obtain an octree hash map structure;

[0042] an updating module configured to collect new raw point cloud data and new initial pose estimation obtained by the UAV to update the octree hash map structure and complete terrain mapping.

[0043] In a third aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is executed by the processor to implement the terrain mapping method based on uncertainty analysis.

[0044] In a fourth aspect, the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the terrain mapping method based on uncertainty analysis.

[0045] Advantages:

[0046] According to the above technical solution, the present application provides a terrain mapping method based on uncertainty analysis, which combines multi-source data processing of a UAV carrying a radar, voxelized space modeling, probabilistic plane feature expression, and efficient indexing and online updating mechanism, and has the following advantages:

[0047] 1. Covariance modeling is introduced to realize quantitative expression of map uncertainty: in each effective voxel, the normal vector and the center point are calculated simultaneously by plane fitting, and the plane parameter covariance matrix is explicitly estimated. The covariance matrix considers the sources of sensor ranging noise, attitude estimation error, etc., so that each plane unit in the map has a confidence expression in a statistical sense. This representation not only can dynamically adjust the residual weight in registration optimization and reduce the interference of low confidence constraints, but also can provide a reliability reference index for subsequent tasks (such as path planning and obstacle avoidance decision).

[0048] 2. Adaptive voxel resolution for efficient mapping and detail preservation: Principal Component Analysis (PCA) planarity detection combined with an angle threshold criterion is used to maintain larger voxel sizes in flat and low-information areas, reducing redundant computations. In complex or detailed areas (e.g., building edges, obstacle perimeters), automatic subdivision into smaller voxels captures key geometric information. This mechanism avoids the common problems of "over-refinement in sparse areas" and "information loss in dense areas" in fixed-resolution voxel mapping, ensuring a balance between spatial resolution and computational overhead.

[0049] 3. "Octree + Hash Table" combined structure for efficient retrieval and management: The octree structure is responsible for spatial hierarchical management and local subdivision control, while the hash table uses voxel coordinates as keys to achieve approximate constant-time neighborhood queries and insertion operations. This combination of dual structures significantly improves retrieval efficiency when performing point-to-plane matching, map updating, and other high-frequency operations, reducing traversal and comparison computational overhead compared to traditional kd-tree or pure octree schemes.

[0050] 4. Online incremental updating and redundancy elimination for compactness and timeliness of the map: Each new frame only updates newly covered areas or voxels with significantly changed observation angles, avoiding the time consumption of full-map reconstruction. For voxels with a threshold number of observations and convergent plane parameters, the original point data is deleted, and only the plane statistics are retained, enabling long-term compressed storage of map data. When significant changes in plane parameters are detected, local reconstruction is automatically triggered, effectively addressing dynamic environmental changes (e.g., new obstacles, structural changes).

[0051] 5. Loop closure optimization for improved positioning and mapping accuracy: In the pose optimization phase, the plane covariance is used to weight the point-to-plane residuals, reducing the influence of high-uncertainty constraints and making the optimization results more stable. The optimized current pose serves as the initial pose input for the next frame of data processing, forming a "mapping-positioning" loop that continuously improves the overall accuracy and robustness of the system. This loop mechanism maintains stable pose estimation and map updating under adverse conditions such as multi-path reflections, local occlusions, and sparse point clouds.

[0052] 6. Adaptability to multiple scenarios and usability under resource constraints: The method can adaptively adjust voxel resolution and map update frequency in different scenarios such as low-altitude urban inspection, mountain mapping, and forest exploration. The map storage format only retains plane triplet parameters and covariance, significantly reducing storage space requirements compared to directly storing original point clouds, making it suitable for UAV platforms with limited onboard computing resources. Efficient retrieval and updating strategies enable real-time operation in embedded computing environments without relying on high-performance ground computing stations.

[0053] 7. Provide directly usable uncertainty information for subsequent tasks: In addition to mapping and localization, the planar covariance matrix can be used as input to provide additional decision basis for path planning, obstacle avoidance strategy, robustness evaluation, etc. The introduction of uncertainty information makes the map not only contain geometric information, but also have a quantifiable reliability description, thereby improving the autonomy and safety of the entire unmanned aerial vehicle system in complex environments.

[0054] It should be understood that all combinations of the foregoing concepts and additional concepts described in greater detail below (provided such concepts are not mutually inconsistent) are contemplated as being part of the inventive subject matter.

[0055] The foregoing and other aspects, embodiments and features of the present teachings can be better understood from the following detailed description taken in conjunction with the accompanying drawings. Additional features of the present teachings will be described or will become apparent in the course of the following detailed description. BRIEF DESCRIPTION OF DRAWINGS

[0056] The accompanying drawings are not intended to be drawn to scale. In the drawings, each identical, or nearly identical, component that is illustrated in various figures is represented with a like numeral. For purposes of clarity, not every component is called out in every drawing. There is now being described by way of example various aspects of the present teachings. These aspects will be described with reference to the accompanying figures in which:

[0057] Figure 1 Flow chart of a terrain mapping method based on uncertainty analysis in an embodiment of the present application.

[0058] Figure 2 Flow chart of step S102 of a terrain mapping method based on uncertainty analysis in an embodiment of the present application.

[0059] Figure 3 Flow chart of step S104 of a terrain mapping method based on uncertainty analysis in an embodiment of the present application.

[0060] Figure 4 Flow chart of step S106 of a terrain mapping method based on uncertainty analysis in an embodiment of the present application.

[0061] Figure 5 Flow chart of step S108 of a terrain mapping method based on uncertainty analysis in an embodiment of the present application.

[0062] Figure 6 Schematic diagram of terrain mapping of a flat ground using a terrain mapping method based on uncertainty analysis in an embodiment of the present application.

[0063] Figure 7 Fig. 1 is a schematic diagram of a topographic mapping method based on uncertainty analysis for a slope land in an embodiment of the present application. DETAILED DESCRIPTION

[0064] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the embodiments of the present application will be described clearly and completely below with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the described embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort belong to the scope of protection of the present application. Unless otherwise defined, the technical terms or scientific terms used herein should have the usual meanings understood by those of ordinary skill in the art to which the present application belongs.

[0065] The terms "first", "second", and similar terms used in the patent application specification and claims of the present application do not denote any order, quantity, or importance, but are only used to distinguish different components. Similarly, the singular forms "a", "an", and "the" and the like do not denote a quantity restriction, but denote the existence of at least one, unless the context clearly indicates otherwise. The terms "include" or "contain" and the like mean that the elements or objects appearing before "include" or "contain" cover the features, integers, steps, operations, elements, and / or components listed after "include" or "contain", and do not exclude the existence or addition of one or more other features, integers, steps, operations, elements, components, and / or sets thereof. "Up", "down", "left", "right", and the like are only used to represent relative positional relationships, and when the absolute positions of the described objects change, the relative positional relationships may also change accordingly.

[0066] In the prior art, point cloud maps, planar representation maps, or fixed resolution voxel maps are usually used. However, point cloud maps are difficult to explicitly model map uncertainty, often treat local planes as deterministic quantities, and ignore measurement noise and pose errors; as the map size increases, the construction and query overhead based on kd-tree increases; planar representation maps are more sensitive to sparse / uneven point clouds, and their stability is affected; fixed voxel mapping is difficult to balance the expression ability of sparse and dense areas, and it is difficult to balance between coarse and fine when facing various structures such as ground, wall, and debris; the above problems limit the registration accuracy, real-time performance, and map compactness of unmanned aerial vehicle three-dimensional environment mapping.

[0067] In view of this, with reference to Figure 1 The embodiment of the present application provides a topographic mapping method based on uncertainty analysis, comprising:

[0068] Step S102: Collect original point cloud data acquired by the UAV and initial pose estimation to generate a terrain candidate point set.

[0069] Step S104: Perform voxel division on the terrain candidate point set to obtain an effective voxel set.

[0070] Step S106: Perform plane fitting on the effective voxel set to obtain a plane feature set.

[0071] Step S108: Process the plane feature set to obtain an octree hash map structure.

[0072] Step S110: Collect new original point cloud data acquired by the UAV and new initial pose estimation to update the octree hash map structure and complete terrain mapping.

[0073] In some embodiments, referring to Figure 2 , collecting original point cloud data acquired by the UAV and initial pose estimation to generate a terrain candidate point set includes:

[0074] Step S1021: Collect original point cloud data acquired by the radar mounted on the UAV and initial pose estimation output by the state estimation module installed in the UAV.

[0075] When the UAV carrier acquires radar point cloud, the initial pose estimation is provided synchronously by the state estimation module, which can output the initial pose based on IMU / GNSS / visual odometry.

[0076] Step S1022: Convert the original point cloud data to the current global reference coordinate system according to the initial pose estimation, for example, convert the original point cloud data from the radar coordinate system to the current global system by homogeneous transformation, and then obtain converted point cloud data.

[0077] Step S1023: Sort the converted point cloud data according to the height channel to filter out terrain candidate points according to a preset minimum height percentage, for example, set the preset minimum height percentage to 5%, and filter out terrain candidate points with a height percentage not less than 5%.

[0078] In addition to filtering out possible ground point set by the minimum height percentage, the possible ground point set can also be filtered out according to the height threshold to filter out non-terrain points obviously higher than the ground. Filtering according to the height threshold or the minimum height percentage reduces the calculation amount and provides a positioning reference for subsequent plane extraction.

[0079] Step S1024: Collect all filtered terrain candidate points to generate a terrain candidate point set.

[0080] Through steps S1021 to S1024, height percentile screening is completed, non-terrain points are eliminated as early as possible, reducing the burden on the voxelization and fitting stages; coordinate unification is completed in advance, eliminating subsequent cumulative errors caused by inconsistencies in multi-source coordinates; and a more uniform set of candidate points is generated, which is beneficial to the stability of subsequent planarity criteria.

[0081] In some embodiments, reference is made to Figure 3 Voxel partitioning is performed on the candidate terrain points to obtain an effective voxel set, including:

[0082] Step S1041: Divide the terrain candidate point set into voxel blocks according to coarse resolution.

[0083] To adapt to the diversity of environmental structure and point cloud distribution, the area to be mapped is first divided into grids at a coarse resolution. For example, the three-dimensional space of the terrain is divided into voxel blocks at a coarse resolution, with each voxel block having a volume of 2 cubic meters.

[0084] Step S1042: Perform principal component analysis on each voxel block to obtain the eigenvalues ​​of the covariance matrix.

[0085] Within each voxel, the covariance matrix of the point set is calculated using principal component analysis (PCA), and its eigenvalues ​​are taken.

[0086] Step S1043: Determine the flatness. If the ratio of the smallest eigenvalue to the second largest eigenvalue in the covariance matrix does not decrease significantly, or the angle between the fitted plane obtained through the covariance matrix eigenvalues ​​and the horizontal plane is not less than a preset angle threshold (e.g., the preset angle threshold is 15°), then the flatness is insufficient. The voxel block is then subdivided into smaller sub-voxel blocks, and principal component analysis and flatness determination are repeated for each sub-voxel block until the flatness is sufficient or the minimum voxel size is reached. If the ratio of the smallest eigenvalue to the second largest eigenvalue in the covariance matrix decreases significantly, and the angle between the fitted plane obtained through the covariance matrix eigenvalues ​​and the horizontal plane is less than a preset angle threshold, then the flatness is sufficient. The method of judging flatness by whether it is significantly reduced or not means comparing the ratio of the smallest eigenvalue to the second largest eigenvalue with a predetermined threshold for judging flatness. If the ratio is less than the threshold, the ratio is not significantly reduced, and the flatness is significant. If the ratio is not less than the threshold, the ratio is significantly reduced, and the flatness is insufficient. Furthermore, the threshold for judging flatness depends on the sensor noise and the scene, and is generally set between 0.01 and 0.1. The smaller the value, the higher the required flatness. For example, a value of 0.01 indicates an extremely smooth plane, while a value of 0.1 allows for a certain degree of roughness.

[0087] If the ratio of the smallest eigenvalue to the second largest eigenvalue does not decrease significantly, or the angle between the fitted plane and the horizontal plane exceeds a preset threshold (e.g., within 15° is considered as near-horizontal terrain), it is considered that the point cloud in the voxel does not meet the requirement of large plane structure. At this time, the voxel is split by 1 / 8 octree, recursively divided into 8 sub-voxels, and the above checking and subdivision operations are repeated for each sub-voxel until the point cloud planarity in the voxel is sufficient or the minimum voxel size (e.g., 0.2 m³) is reached.

[0088] Step S1044: Retain all voxel blocks with sufficient planarity to generate an effective voxel set. Finally, only voxel units that meet the planar feature are retained for terrain map modeling, thereby eliminating non-terrain data that fails to be identified in step S102, where non-terrain data refers to a plane that is not less than a preset angle threshold (e.g., 15°).

[0089] Finally, only voxel units that meet the planar feature are retained for map modeling, thereby avoiding over-refinement of low-information areas and improving mapping efficiency. Based on point set planarity, it is determined whether to perform octree recursive division, preventing low-information areas from being excessively divided and improving mapping efficiency. Combined with hash fast positioning and spatial recursive division of octree, the query and update speed is improved. Point cloud data is no longer retained for convergent planes, and only statistics are maintained, realizing compressed mapping. Point-to-plane distance and covariance weighting are used for tight coupling pose optimization.

[0090] Steps S1041-S1044 complete the maintenance of large voxels in information-poor areas, reduce calculation and storage; automatically subdivide in detail areas to ensure local geometric expression; the criterion for "whether to subdivide" is established on the basis of statistical and geometric dual indicators, reducing misjudgment caused by uneven sampling; and the output "effective voxel set" provides a quality-controllable input set for subsequent probability plane fitting.

[0091] In some embodiments, with reference to Figure 4 Plane fitting is performed on the effective voxel set to obtain a plane feature set, including:

[0092] Step S1061: Perform plane fitting on each point set in the effective voxel set to obtain a plane normal vector n, a plane center point c, and a plane covariance matrix Σ.

[0093] Step S1062: Construct a triple (n, c, Σ) according to the plane normal vector n, the plane center point c, and the plane covariance matrix Σ.

[0094] Step S1063: Collect the triples of each point set in the effective voxel set to obtain a plane feature set.

[0095] A plane fitting is performed on the point set in each valid voxel to obtain a plane normal vector n, a plane center point c, and a plane covariance matrix Σ, which quantifies the mapping uncertainty considering ranging error, pose error, etc. Each plane is described by a three-tuple (n, c, Σ) without saving the original points.

[0096] For each valid voxel point set passing the planarity test, a plane fitting is performed, such as using least squares fitting or RANSAC to suppress outliers, to obtain a normal vector and a point cloud centroid, and a plane covariance matrix Σ is combined to describe the plane. Unlike traditional deterministic representation, the present application calculates the uncertainty of the plane parameters, i.e. the plane covariance matrix Σ, by measuring the noise model and the current pose estimation error. For example, according to the laser ranging error and pose error propagation formula, the plane normal vector n, the plane center point c, and the plane covariance matrix Σ are derived. In this way, each plane feature is explicitly represented by a three-tuple, which can quantitatively depict the uncertainty factors in the mapping process.

[0097] Steps S1061-S1063 provide a weight basis for subsequent registration with the statistically quantized plane covariance matrix Σ, which is more reliable than the "deterministic plane"; by discarding the original points and retaining only the three-tuple, a compressed map storage is possible; the plane-level uncertainty provides a threshold reference for subsequent change detection.

[0098] In some embodiments, with reference to Figure 5 , the set of plane features is processed to obtain an octree hash map structure, including:

[0099] Step S1081: Each plane feature in the set of plane features is converted into an integer grid index according to the voxel space coordinates to which it belongs and a hash table Hash[(x, y, z)] = (n, q, Σ) is constructed, where x, y, and z are the horizontal, vertical, and vertical coordinates of the voxel space coordinates, respectively.

[0100] Step S1082: The octree structure is used to manage the spatial hierarchy and obtain the octree hash map structure.

[0101] All plane structures are organized into a hash table with voxel coordinates as the key and plane three-tuple parameter description as the value; the octree structure is used for local lookup and split control.

[0102] All the built plane features are organized into a hash table according to their voxel space coordinates, while the spatial hierarchy is managed by octree structure. Specifically, the grid coordinates of voxels or the octree node numbers are used as hash keys, and the corresponding plane parameters, plane normal vector n, plane center point c and plane covariance matrix Σ are stored as values. This way of combining hash indexing with octree can complete the neighborhood query and insertion update in approximately constant time. Compared with the traditional kd-tree, the octree hash structure has higher efficiency for continuous queries and dynamic insertion.

[0103] Steps S1081-S1082 realize the positioning and neighborhood query in approximately constant time by using the hash table with the octree structure, reduce the search cost in the registration and update stages, and hierarchical management helps local reconstruction and visualization, and can provide a unified index to be called by the upper algorithm.

[0104] In some embodiments, the newly added original point cloud data acquired by the unmanned aerial vehicle and the newly added initial pose estimation are collected to update the octree hash map structure, and the terrain mapping is completed, including:

[0105] Step S1101: searching for the newly added original point cloud data in the effective voxel set, if the newly added original point cloud data exists in the effective voxel set, accumulating the observation times and adding the newly added point cloud data in the effective voxel set, and when the accumulated observation times reach the accumulated observation times threshold, deleting the original point cloud data existing in the effective voxel set; if the newly added original point cloud data does not exist in the effective voxel set, performing plane fitting on the newly added original point cloud data and adding it in the plane feature set.

[0106] The radar scanning frequency is 10 times per second, and after each frame of radar scanning, only the voxels covering new areas or having significantly different observation angles are incrementally updated. When the cumulative observation times of a certain plane feature reach a threshold (such as 50 times), the original points are no longer retained to prevent the map from expanding.

[0107] For each new point cloud scan, first, the point cloud is registered into the map coordinate system using the current estimated drone pose. For points falling into a new region without a corresponding voxel, a new voxel is established and an initial plane fitting is performed; for points falling into an existing voxel, they are added to the voxel point set and the plane parameters and covariance are updated in real time. As the number of observations increases, the uncertainty of the plane parameters converges rapidly, i.e., it tends to be stable after the cumulative number of observations reaches a threshold. Once the plane uncertainty converges, the historical raw points in the voxel are deleted, and only the converged plane normal vector n, plane center point c and plane covariance matrix Σ are retained, which can significantly reduce the storage overhead and avoid the expansion of the map over time. If the subsequent scan adds new points that cause a large deviation of the plane parameters, it is considered that the structure of the region has changed, triggering the re-division of the voxel and the reconstruction of the plane feature. Through incremental updating, the map can maintain the latest and compact, while eliminating redundant data, and realize real-time online maintenance.

[0108] Step S1101 controls the map volume and memory occupation, reduces the storage pressure of long-term operation; event-oriented updating driven by statistical convergence and change detection reduces invalid recalculation and maintains the freshness and availability of the map in a dynamic environment.

[0109] In some embodiments, after completing the terrain mapping, further comprising:

[0110] Collecting current point cloud data obtained by the drone.

[0111] Locating the to-be-processed voxel containing the current point cloud data in the octree hash map structure according to the current point cloud data.

[0112] Traversing the plane features contained in the to-be-processed voxel to match the current point cloud data and calculate the point-to-plane distance residual.

[0113] According to the covariance weighting of the point-to-plane distance residual, an optimization equation is constructed.

[0114] Iteratively solving the optimization equation to obtain the current pose estimation of the drone under the current point cloud data.

[0115] Feeding back the current pose estimation to the drone to optimize the drone pose.

[0116] The map information is beneficial to the registration and optimization of the pose estimation of the current frame. The point-to-plane residual of the current frame point cloud and the plane in the map is calculated, a residual weight matrix is constructed based on the plane covariance, the mapping error is explicitly considered, and the state estimation module is given for attitude optimization update to realize high-precision positioning.

[0117] The current frame point cloud is finely registered with the constructed map plane features after rough position transformation. For each new point from the radar, the voxel containing the point is quickly located in the hash octree, and then the plane features contained in the voxel are traversed for matching. The point-to-plane distance residual is calculated during matching. Since the plane normal vector n and the plane center point c both have estimation uncertainties, the variance of the distance residual can be calculated. In the optimization process, each effective matched point-plane residual is weighted according to its covariance, for example, in the form of weighted least squares or Kalman filtering, to construct an optimization equation. By solving the optimization equation, the UAV pose and local plane parameters can be updated simultaneously, the measurement noise and mapping error are explicitly considered, and a high-precision registration result is achieved.

[0118] The above steps directly map the uncertainty of the map into the optimization weight, reduce the influence of unreliable constraints, form a closed loop of “mapping-locating”, improve the consistency of attitude estimation and reduce the cumulative drift, and provide a more stable pose input for subsequent planning and control modules.

[0119] Figure 6 and Figure 7 respectively show a flat land and a slope land terrain mapping schematic diagram using an uncertainty analysis based terrain mapping method according to an embodiment of the present application, wherein: Figure 6 shown is the mapping effect in a standardized farmland, the voxel color represents the z-axis height information at the position, the lower part is the farmland, and the upper part is the field power line; Figure 7 shown is the mapping effect in a slope land, the voxel color represents the z-axis height information at the position, and the figure shows the bow-shaped flight route of the UAV, which is an example of scanning and mapping the slope.

[0120] Another embodiment of the present application also provides an uncertainty analysis based terrain mapping device, comprising:

[0121] The acquisition module is configured to acquire raw point cloud data and initial pose estimation obtained by the UAV to generate a terrain candidate point set.

[0122] The first execution module is configured to perform voxel division on the terrain candidate point set to obtain an effective voxel set.

[0123] The second execution module is configured to perform plane fitting on the effective voxel set to obtain a plane feature set.

[0124] The processing module is configured to process the plane feature set to obtain an octree hash map structure.

[0125] The update module is configured to acquire new raw point cloud data and new initial pose estimation obtained by the UAV to update the octree hash map structure and complete terrain mapping.

[0126] Another embodiment of the present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program, when executed by the processor, implements the method for topographic mapping based on uncertainty analysis.

[0127] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The processor is a control center of a data processing apparatus of the gateway, and connects each part of the data processing apparatus of the gateway through various interfaces and lines.

[0128] Another embodiment of the present application also provides a computer readable storage medium, which stores a computer program, the computer program, when executed by a processor, implements the method for topographic mapping based on uncertainty analysis.

[0129] The memory can include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required by a function; and the data storage area can store data created by the processor, etc. In addition, the memory is preferably but not limited to a high-speed random access memory, for example, can also be a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can also optionally include a memory remotely arranged relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0130] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by a computer program instructing the relevant hardware. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiment methods. The storage medium can be a magnetic disc, an optical disc, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), etc. The storage medium can also include a combination of the above-mentioned types of memories.

[0131] In summary, the present application provides a terrain mapping method based on uncertainty analysis, device, electronic equipment and storage medium, combined with multi-source data processing of unmanned aerial vehicle carrying radar, voxel space modeling, probability plane feature expression and efficient index and online updating mechanism, can be applied to real-time positioning and navigation in unmanned aerial vehicle low-altitude three-dimensional environment mapping, path planning, obstacle avoidance, logistics distribution, plant protection flight, city perception and other tasks, therefore the present application provides a terrain mapping method based on uncertainty analysis, device, electronic equipment and storage medium have the following beneficial effects: introduce covariance modeling, realize map uncertainty quantitative expression: in each effective voxel, through the plane fitting calculation normal vector and center point, estimate the plane parameter covariance matrix at the same time. Covariance matrix considers the source of sensor ranging noise, attitude estimation error, etc., so that each plane unit in the map has a confidence expression in a statistical sense. This representation not only can dynamically adjust the residual weight in registration optimization, reduce the interference of low confidence constraints, but also can provide reliability reference index for subsequent tasks (such as path planning, obstacle avoidance decision). Adaptive voxel resolution, improve mapping efficiency and detail compatibility: adopt principal component analysis (PCA) planarity detection combined with angle threshold criterion, maintain larger voxel size in flat and low information area, reduce redundant calculation. In the complex structure or rich detail area (such as building edge, obstacle around), automatically subdivide into smaller voxel, capture key geometric information. This mechanism avoids the common problems of "over refinement in sparse areas" and "information loss in dense areas" in fixed resolution voxel mapping, ensuring the balance between spatial resolution and computational overhead. "Octree + hash table" combined structure realizes efficient retrieval and management: octree structure is responsible for spatial hierarchical management and local subdivision control, and hash table takes voxel coordinates as key to realize approximate constant time neighborhood query and insertion operation. This combination of double structures significantly improves retrieval efficiency when performing point-to-plane matching, map updating and other high-frequency operations, compared with traditional kd-tree or pure octree scheme, reducing the computational overhead of traversal and comparison. Online incremental update and redundancy elimination, maintain the compactness and timeliness of the map: only update the new covered area or the voxel with significant observation angle change in each new data, avoid the time consumption caused by full map reconstruction. For the voxel whose observation number reaches the threshold and the plane parameter converges, delete the original point data, only keep the plane statistics, realize long-term compressed storage of map data. When significant changes in plane parameters are detected, automatically trigger local reconstruction, effectively deal with environmental dynamic changes (such as new obstacles, structure changes). Closed-loop registration optimization, improve the accuracy of positioning and mapping: in the pose optimization stage, the plane covariance is used to weight the point-to-plane residual, reduce the influence of high uncertainty constraints, make the optimization result more stable. The optimized current pose is used as the initial pose input for the next frame data processing, forming a "mapping-positioning" closed loop, continuously improving the overall accuracy and robustness of the system.The closed-loop mechanism can still maintain stable pose estimation and map updating under adverse conditions such as multi-path reflection, local occlusion, and sparse point cloud. Adaptability to multiple scenes and availability under resource constraints: In different scenes such as low-altitude city inspection, mountain mapping, and forest detection, the voxel resolution and map update frequency can be adaptively adjusted. The map storage form only retains the plane triple parameters and covariance, which can significantly reduce the storage space requirement compared with directly storing the original point cloud, and is suitable for unmanned aerial vehicle platforms with limited on-board computing resources. Efficient search and update strategy enables the method to run in real time in an embedded computing environment without relying on high-performance ground computing stations. Provides directly usable uncertainty information for subsequent tasks: In addition to mapping and positioning, the plane covariance matrix can be used as input to provide additional decision basis for path planning, obstacle avoidance strategy, and robustness evaluation. The introduction of uncertainty information makes the map not only contain geometric information, but also have a quantifiable reliability description, thereby improving the autonomy and safety of the entire unmanned aerial vehicle system in complex environments.

[0132] Although the present application has been described in connection with the preferred embodiment thereof with reference to the drawings, it is not intended to be limited to the embodiment but it will be apparent to those skilled in the art that various modifications and improvements can be made within the scope of the present application. Accordingly, it is intended to cover all such modifications and improvements that fall within the scope of the appended claims.

Claims

1. A terrain mapping method based on uncertainty analysis, characterized in that, include: The process involves collecting raw point cloud data and initial pose estimation from a UAV, converting the raw point cloud data to the current global reference coordinate system based on the initial pose estimation to obtain converted point cloud data, sorting the converted point cloud data by height channel, and filtering terrain candidate points according to a preset minimum height percentage, and aggregating all the filtered terrain candidate points to generate a terrain candidate point set. Voxel partitioning is performed on the terrain candidate point set, dividing the terrain candidate point set into voxel blocks at a coarse resolution; principal component analysis is performed on each voxel block to obtain the eigenvalues ​​of the covariance matrix; Planarity is determined based on a threshold used for planeness assessment. If the ratio of the smallest eigenvalue to the second largest eigenvalue in the covariance matrix is ​​less than the planeness threshold, then the planeness is considered insufficient if the ratio does not decrease significantly, or if the angle between the fitted plane obtained from the covariance matrix eigenvalues ​​and the horizontal plane is not less than a preset angle threshold. The voxel block is then subdivided into smaller sub-voxel blocks, and principal component analysis and planeness assessment are repeated for each sub-voxel block until planeness is sufficient or the minimum voxel size is reached. If the ratio of the smallest eigenvalue to the second largest eigenvalue in the covariance matrix is ​​not less than the planeness threshold, then the planeness is considered sufficient if the ratio decreases significantly, and the angle between the fitted plane obtained from the covariance matrix eigenvalues ​​and the horizontal plane is less than a preset angle threshold. All voxel blocks with sufficient planeness are retained to generate an effective voxel set. Perform plane fitting on the effective voxel set to obtain a planar feature set; Process the planar feature set to obtain an octagonal hash map structure; New raw point cloud data acquired by UAVs and new initial pose estimates are collected to update the octagonal hash map structure and complete terrain mapping.

2. The terrain mapping method based on uncertainty analysis according to claim 1, characterized in that, The process of collecting raw point cloud data and initial pose estimation from the UAV to generate a terrain candidate point set includes: The raw point cloud data acquired by the radar mounted on the UAV and the initial pose estimate output by the state estimation module installed in the UAV are collected.

3. The terrain mapping method based on uncertainty analysis according to claim 1, characterized in that, The step of performing plane fitting on the effective voxel set to obtain a planar feature set includes: Perform plane fitting on each point set in the effective voxel set to obtain the plane normal vector n, the plane center point c, and the plane covariance matrix Σ. Based on the plane normal vector n, the plane center point c, and the plane covariance matrix Σ, construct a triple (n, c, Σ). The planar feature set is obtained by combining the triples of each point set in the effective voxel set.

4. The terrain mapping method based on uncertainty analysis according to claim 3, characterized in that, The process of processing the planar feature set to obtain the octagonal hash map structure includes: Each planar feature in the planar feature set is converted into an integer grid index according to its voxel space coordinates and a hash table Hash[(x, y, z)]=(n, q, Σ) is constructed, where x, y, and z are the x-coordinate, y-coordinate, and z-coordinate of the voxel space coordinates, respectively. An octree structure is used to manage the spatial hierarchy and obtain the octree hash map structure.

5. The terrain mapping method based on uncertainty analysis according to claim 1, characterized in that, The process of collecting newly acquired raw point cloud data and initial pose estimates from UAVs to update the octagonal hash map structure and complete terrain mapping includes: The newly added original point cloud data is searched in the effective voxel set. If the newly added original point cloud data exists in the effective voxel set, the number of observations is accumulated and the newly added point cloud data is added to the effective voxel set. When the number of accumulated observations reaches the accumulated observation threshold, the original point cloud data existing in the effective voxel set is deleted. If the newly added original point cloud data does not exist in the effective voxel set, the newly added original point cloud data is subjected to plane fitting and then added to the plane feature set.

6. The terrain mapping method based on uncertainty analysis according to claim 1, characterized in that, After completing the topographic mapping, it also includes: Collect current point cloud data acquired by the drone; Based on the current point cloud data, locate the voxel containing the current point cloud data in the octagonal hash map structure; Traverse the planar features contained within the voxel to be processed to match the current point cloud data and calculate the point-to-plane distance residual; An optimization equation is constructed by weighting the point-to-plane distance residuals according to their covariance. The optimization equation is solved iteratively to obtain the current pose estimate of the UAV under the current point cloud data; The current pose estimate is fed back to the UAV to optimize the UAV pose.

7. A terrain mapping device based on uncertainty analysis, characterized in that, include: The acquisition module is used to acquire raw point cloud data and initial pose estimation obtained by the UAV; convert the raw point cloud data to the current global reference coordinate system based on the initial pose estimation to obtain converted point cloud data; sort the converted point cloud data according to the height channel to filter terrain candidate points according to a preset minimum height percentage; and collect all the filtered terrain candidate points to generate a terrain candidate point set. The first execution module is used to perform voxel partitioning on the terrain candidate point set, dividing the terrain candidate point set into voxel blocks according to coarse resolution; and to perform principal component analysis on each voxel block to obtain the eigenvalues ​​of the covariance matrix. Planarity is determined based on a threshold used for planeness assessment. If the ratio of the smallest eigenvalue to the second largest eigenvalue in the covariance matrix is ​​less than the planeness threshold, then the planeness is considered insufficient if the ratio does not decrease significantly, or if the angle between the fitted plane obtained from the covariance matrix eigenvalues ​​and the horizontal plane is not less than a preset angle threshold. The voxel block is then subdivided into smaller sub-voxel blocks, and principal component analysis and planeness assessment are repeated for each sub-voxel block until planeness is sufficient or the minimum voxel size is reached. If the ratio of the smallest eigenvalue to the second largest eigenvalue in the covariance matrix is ​​not less than the planeness threshold, then the planeness is considered sufficient if the ratio decreases significantly, and the angle between the fitted plane obtained from the covariance matrix eigenvalues ​​and the horizontal plane is less than a preset angle threshold. All voxel blocks with sufficient planeness are retained to generate an effective voxel set. The second execution module is used to perform plane fitting on the effective voxel set to obtain a planar feature set; The processing module is used to process the planar feature set to obtain an octagonal hash map structure; The update module is used to collect new raw point cloud data acquired by the UAV and add initial pose estimation to update the octagonal hash map structure and complete terrain mapping.

8. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the terrain mapping method based on uncertainty analysis as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the terrain mapping method based on uncertainty analysis as described in any one of claims 1 to 6.

Images