Point cloud topographic feature intelligent extraction method based on regional quasi-geoid constraint

Abstract

The invention relates to the technical field of point cloud extraction, in particular to a point cloud topographic feature intelligent extraction method based on regional quasi-geoid constraint. The method comprises the following steps: acquiring airborne laser radar point cloud and global navigation satellite observation data, constructing a regional quasi-geoid refinement model, and realizing point cloud elevation datum unification; constructing a ground point cloud triangulation network model based on the constrained point cloud elevation, calculating a terrain inclination angle and a linear expression error of a triangular surface, and adaptively determining a feature extraction parameter; and traversing the point cloud data in the triangular surface projection range, extracting the significant topographic feature points meeting the relative height condition, optimizing the triangulation network structure through iterative subdivision until a termination condition is met, and outputting a topographic feature extraction result. According to the method, through the synergistic effect of the regional quasi-geoid constraint and the adaptive feature discrimination mechanism, the unification of the elevation reference of the complex terrain point cloud, the effective suppression of the pseudo features and the synchronous improvement of the terrain feature extraction precision and stability are realized.

Description

Technical Field

[0001] This invention relates to the field of point cloud extraction technology, and in particular to an intelligent method for extracting point cloud terrain features based on regional geoid constraints. Background Technology

[0002] In complex water networks and linear watersheds, the terrain is fragmented and undulating, and influenced by hydrological and geological conditions. Traditional cross-section and topographic surveys often rely on benchmarks or existing control points as elevation datums, which are prone to settlement or instability, making it difficult to maintain consistent elevation datums over the long term. This results in insufficient connectivity and reliability between data from different survey periods. Furthermore, these areas have dense vegetation and widespread water bodies, making measurement methods such as those using global navigation satellites susceptible to obstruction, leading to significant cumulative measurement errors and failing to meet the requirements for high-precision topographic representation.

[0003] On the other hand, although airborne lidar can quickly acquire high-density point cloud data, existing point cloud processing technologies are mostly based on ellipsoidal height or simple correction results for terrain analysis, lacking a unified and high-precision regional elevation benchmark constraint. Moreover, existing point cloud terrain feature extraction methods generally use fixed thresholds or single-scale rules, failing to fully consider the error differences under different terrain conditions, and are prone to generating false features or feature omissions in complex terrain, making it difficult to balance extraction accuracy and automation efficiency. Summary of the Invention

[0004] Therefore, it is necessary to provide an intelligent extraction method for point cloud terrain features based on regional geoid constraints to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, a method for intelligent extraction of point cloud topographic features based on regional quasi-geoid constraints is proposed, the method comprising the following steps: Step S1: Acquire airborne lidar point cloud data and corresponding global navigation satellite observation data for the area to be measured; construct a refined geoid model of the region based on the global navigation satellite observation data; Step S2: Based on the regional geoid refinement model, perform elevation datum transformation on the ellipsoidal height in the airborne lidar point cloud data to obtain point cloud elevation data constrained by the regional geoid. Step S3: Construct a ground point cloud triangulation model based on point cloud elevation data and use triangular faces as the basic unit for terrain representation, calculate the terrain dip angle and linear representation error of each triangular face; Step S4: Adjust the slope threshold and scale factor based on the terrain tilt angle and linear expression error to confirm the feature extraction parameter system; Step S5: Based on the feature extraction parameter system, traverse the airborne lidar point cloud data within the projection range of each triangular face, extract significant terrain feature points that meet the relative height threshold, and add the significant terrain feature points to the ground point cloud triangulation model for iterative subdivision. Step S6: Repeat step S5 until the preset iteration termination condition is met, and output the point cloud terrain feature extraction results.

[0006] The beneficial effects of this invention are as follows: by introducing a regional quasi-geoid refinement model to uniformly constrain and transform the elevation of airborne lidar point clouds, it effectively overcomes the problems of elevation inconsistency and error accumulation caused by directly using ellipsoidal heights or simplified corrections in traditional methods, significantly improving the stability, comparability, and reliability of point cloud elevation data in complex regions. Simultaneously, based on a ground point cloud triangulation model as the basic terrain representation framework, it comprehensively considers the triangular facet terrain tilt angle, linear representation error, and elevation accuracy, adaptively adjusting the slope threshold and scale factor to avoid the problem of false features or missing features easily generated under complex terrain conditions by fixed thresholds and single-scale rules. During feature extraction, the point with the largest vertical distance within the triangular facet projection range is selected as a candidate feature point, and the validity of the candidate points is judged by combining a relative height threshold and error control mechanism. Only significant terrain feature points that meet the conditions are subjected to splitting operations, thereby suppressing noise interference while ensuring feature extraction accuracy. Furthermore, for invalid feature points that do not meet the conditions, the topological structure of their respective triangular faces is maintained to avoid unnecessary triangulation reconstruction, improving overall computational efficiency and model stability. In summary, this project has achieved unified elevation benchmarks, adaptive feature extraction parameters, and controllable evolution of triangular mesh structures in complex terrain environments, significantly improving the accuracy, robustness, and automation level of point cloud terrain feature extraction, and has good engineering application value. Attached Figure Description

[0007] Figure 1 This is a flowchart illustrating the steps of an intelligent extraction method for point cloud terrain features based on regional geoid constraints. Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S4. Figure 3 This is a schematic diagram of the ellipsoid, geoid, and terrain surface of the intelligent extraction method for point cloud terrain features based on regional geoid constraints proposed in this application. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0008] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0009] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0010] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0011] To achieve the above objectives, please refer to Figures 1 to 3 A method for intelligent extraction of point cloud terrain features based on regional geoid constraints, the method comprising the following steps: Step S1: Acquire airborne lidar point cloud data and corresponding global navigation satellite observation data for the area to be measured; construct a refined geoid model of the region based on the global navigation satellite observation data; In one embodiment, an aerial survey is first performed on the area to be measured. An airborne lidar system mounted on a flight platform scans the area to be measured, acquiring airborne lidar point cloud data covering the area. The airborne lidar point cloud data includes three-dimensional spatial coordinate information of surface targets, and the three-dimensional spatial coordinates are expressed using ellipsoidal height as the elevation reference.

[0012] Simultaneously, during the airborne lidar scanning process, corresponding Global Navigation Satellite System (GNSS) observation data is acquired. This GNSS observation data includes the positioning information of the flight platform during the scanning process and observation data from ground-based GNSS reference stations deployed within the area. This observation data reflects the elevation benchmark variation characteristics of the area under test.

[0013] After acquiring global navigation satellite (GNSS) observation data, the data undergoes preprocessing, including outlier removal, time synchronization correction, and observation accuracy assessment, to obtain a valid GNSS observation dataset that meets the modeling accuracy requirements. The preprocessed GNSS observation data is used to characterize the relationship between ellipsoidal height and normal height within the measured region.

[0014] Based on the aforementioned effective global navigation satellite observation dataset, a refined model of the regional quasi-geoid is constructed. Specifically, by analyzing the elevation differences at global navigation satellite observation points, the spatial variation relationship of the quasi-geoid within the area to be measured is established, and an refined model of the quasi-geoid covering the entire area to be measured is generated using interpolation or fitting methods.

[0015] Step S2: Based on the regional geoid refinement model, perform elevation datum transformation on the ellipsoidal height in the airborne lidar point cloud data to obtain point cloud elevation data constrained by the regional geoid. In one embodiment, after acquiring airborne lidar point cloud data and global navigation satellite observation data as described in step S1, and constructing a regional quasi-geoid refinement model, the elevation information in the airborne lidar point cloud data undergoes a unified elevation datum transformation. The original elevation in the airborne lidar point cloud data is an ellipsoidal height relative to a reference ellipsoid, which cannot directly reflect the true elevation characteristics of the ground. Therefore, a regional quasi-geoid refinement model is needed for constraint correction.

[0016] Specifically, each point in the airborne lidar point cloud data is treated as a processing object. Its corresponding planar coordinate information is read, and its position is matched in a regional quasi-geoid refinement model based on the planar coordinates to obtain the regional quasi-geoid information corresponding to the spatial position of the point. The regional quasi-geoid refinement model can reflect the influence of gravity anomalies, topographic relief, and changes in the shape of the earth on the elevation datum within the region, thereby providing a more refined elevation reference.

[0017] In one embodiment, the ellipsoidal heights of the point cloud points are transformed point by point to map the ellipsoidal heights to elevation values ​​constrained by a regional geoid. This transformation process can employ a point-by-point correction method to ensure that the elevation correction result of each point cloud point accurately corresponds to its spatial location, thereby avoiding the impact of overall translation or local distortion on the accuracy of the point cloud elevation.

[0018] To improve the stability and consistency of elevation transformation, in one embodiment, the refined model of the regional quasi-geoid can be interpolated to form a continuous elevation reference surface within the point cloud coverage area. When point cloud points are located between model nodes, the quasi-geoid information at the corresponding location is obtained through interpolation, thereby ensuring the continuity and smoothness of the elevation transformation process.

[0019] After completing the ellipsoidal elevation transformation of all point cloud points, point cloud elevation data constrained to a regional geoid are obtained. This point cloud elevation data is expressed under the same elevation datum, which can accurately reflect the relative elevation differences and absolute elevation characteristics of ground features, providing a reliable data foundation for subsequent terrain modeling, feature extraction, and elevation accuracy analysis.

[0020] Step S3: Construct a ground point cloud triangulation model based on point cloud elevation data and use triangular faces as the basic unit for terrain representation, calculate the terrain dip angle and linear representation error of each triangular face; In one embodiment, after completing the point cloud elevation benchmark conversion described in step S2, a ground point cloud triangulation model is constructed based on the point cloud elevation data constrained by the regional geoid. First, the point cloud elevation data is filtered to identify ground points, distinguishing those belonging to the ground surface from vegetation points, building points, and noise points, to ensure that the triangulation model can accurately reflect the terrain undulation characteristics.

[0021] After obtaining ground point cloud data, based on the planar spatial distribution of the ground point cloud, spatial neighborhoods are divided into the ground point cloud, and adjacent ground points are connected according to preset triangulation rules to gradually generate a ground point cloud triangular network model composed of multiple triangular faces. The triangular network model covers the entire area to be measured, and uses triangular faces as the basic unit for terrain representation to describe the spatial morphological changes of local terrain.

[0022] In one embodiment, for each triangular facet in the ground point cloud triangulation model, based on the spatial coordinate information of the three vertices constituting the triangular facet, the spatial attitude change of the triangular facet relative to the horizontal plane is analyzed, thereby obtaining the corresponding terrain tilt features, which are used to characterize the terrain slope change of the area where the triangular facet is located. By calculating the terrain tilt features of all triangular facests, a terrain tilt angle dataset reflecting the overall terrain undulation characteristics of the region can be formed.

[0023] Meanwhile, in one embodiment, the linear approximation degree of the terrain represented by the triangular facet is analyzed based on the spatial distribution of ground points within the triangular facet. By evaluating the deviation between the actual point cloud elevation distribution within the area covered by the triangular facet and the linear terrain represented by the triangular facet, the linear representation error of the triangular facet is determined, thereby reflecting the accuracy level of the triangular facet in describing local complex terrain.

[0024] In one embodiment, the terrain slope angle and linear representation error can be used as joint terrain feature parameters to distinguish between flat areas, gentle slope areas, and complex terrain areas with large undulations. When the linear representation error of the triangular facet is large, it can serve as an important reference for subsequent detailed modeling or feature point extraction.

[0025] Step S4: Adjust the slope threshold and scale factor based on the terrain tilt angle and linear expression error to confirm the feature extraction parameter system; In one embodiment, after completing the construction of the ground point cloud triangulation model and the calculation of the terrain slope angle and linear representation error of each triangular face as described in step S3, the parameter system used for terrain feature extraction is adaptively adjusted based on the acquired terrain feature parameters. The parameter system includes at least a slope threshold for distinguishing the degree of terrain undulation and a scale factor for controlling the scale of feature extraction.

[0026] In one embodiment, the terrain slope distribution of all triangular faces in the ground point cloud triangulation model is first statistically analyzed to obtain the overall range of terrain slope variation and local distribution characteristics within the region. Based on the concentration range and dispersion of terrain slope, the initially set slope threshold is segmented and corrected to match the actual terrain undulation characteristics, thereby avoiding over-feature extraction in flat areas or feature omission in steep areas.

[0027] Meanwhile, in one embodiment, the scale factor is dynamically adjusted based on the linear representation error distribution of each triangular facet. When the linear representation error is small, indicating that the triangular facet can reflect local terrain changes well, the scale factor is appropriately increased to improve the overall stability of feature extraction; when the linear representation error is large, indicating that the local terrain undulations are complex or there are nonlinear changes, the scale factor is decreased to enhance the ability to perceive detailed terrain changes.

[0028] In one embodiment, the terrain slope angle and linear expression error can be jointly analyzed to collaboratively correct the slope threshold and scale factor. For example, when a region exhibits both large terrain slope angle changes and high linear expression error, the feature extraction parameters for that region can be adjusted to a more sensitive state to adapt to complex terrain structures; while in regions with gentle terrain changes and low linear expression error, a relatively conservative parameter configuration is adopted to improve the robustness of the feature extraction results.

[0029] In one embodiment, after adjusting the slope threshold and scale factor, the adjustment results are integrated to form a feature extraction parameter system that matches the terrain features of the current area to be measured. The parameter system is then used as the basic control parameters for subsequent terrain feature point screening, geomorphic structure identification, and fine modeling.

[0030] Step S5: Based on the feature extraction parameter system, traverse the airborne lidar point cloud data within the projection range of each triangular face, extract significant terrain feature points that meet the relative height threshold, and add the significant terrain feature points to the ground point cloud triangulation model for iterative subdivision. In one embodiment, after confirming the feature extraction parameter system described in step S4, the spatial coverage of each triangular face in the ground point cloud triangulation model is traversed one by one based on the feature extraction parameter system. Specifically, the projection range of each triangular face on the horizontal plane is used as the search area, and candidate point cloud data falling within the projection range are selected from the airborne lidar point cloud data to form a candidate point cloud set for the corresponding triangular face.

[0031] In one embodiment, for each candidate point cloud set, the vertical distance change of the point cloud data relative to the corresponding triangular facet fitting plane is calculated, and the candidate point cloud is judged point by point in conjunction with the relative height threshold in the feature extraction parameter system. When a point cloud shows a significant height deviation relative to the triangular facet that is higher or lower than the fitting plane, and the degree of deviation exceeds the set relative height threshold, the point cloud is judged as a significant terrain feature point with significant terrain representation significance.

[0032] In one embodiment, to avoid interference from local noise points or abnormal echoes on the feature point extraction results, the determination of significant terrain feature points can be supplemented by combining the height continuity and spatial distribution stability of the point cloud within its local neighborhood. Only when a candidate point cloud maintains a consistent height deviation trend among its neighboring point clouds is it finally confirmed as a valid significant terrain feature point, thereby improving the reliability of the feature point extraction results.

[0033] In one embodiment, the confirmed significant terrain feature points are added one by one to the ground point cloud triangulation model, and the newly added feature points are used as constraint nodes to locally reconstruct the original triangular faces. Specifically, an iterative subdivision operation is performed on the triangular faces containing significant terrain feature points, decomposing the original triangular faces into multiple new sub-triangular faces, so that the triangulation model can more accurately fit the actual terrain undulations.

[0034] In one embodiment, the above-described traversal, judgment, and iterative subdivision process can be executed in multiple rounds in a coarse-to-fine manner. In each iteration, the projection range of the triangular face is retraced based on the updated triangular mesh model, and new significant terrain feature points are continuously introduced in conjunction with the current feature extraction parameter system until no new feature points that meet the conditions are detected within the preset height variation range.

[0035] Step S6: Repeat step S5 until the preset iteration termination condition is met, and output the point cloud terrain feature extraction results.

[0036] In one embodiment, after completing the extraction of significant terrain feature points and iterative subdivision of the ground point cloud triangulation model as described in step S5, the iteration termination judgment stage is entered. Based on the ground point cloud triangulation model updated in the current round, a comprehensive statistical analysis is performed on the number of newly added significant terrain feature points, their spatial distribution, and the corresponding changes in the triangular facet structure during this iteration.

[0037] In one embodiment, the preset iteration termination condition includes at least one or more of the following: the number of newly added significant terrain feature points is lower than a preset minimum threshold in two adjacent iterations; or the terrain dip angle variation of each triangle face in the triangular mesh model tends to stabilize overall; or the terrain linear representation error remains within an allowable range in multiple consecutive iterations. By judging the above conditions, it is used to identify whether the current terrain representation has reached the expected accuracy requirement.

[0038] In one embodiment, when it is determined that the current iteration result has not met any iteration termination condition, the process automatically returns to step S5. Based on the updated ground point cloud triangulation model and feature extraction parameter system, the process continues to perform the next round of significant terrain feature point extraction and iterative subdivision processing on the point cloud data within the projection range of each triangular face, thereby gradually refining the terrain representation structure.

[0039] In one embodiment, when at least one preset iteration termination condition is met, the repeated execution of step S5 is terminated, and the current ground point cloud triangulation model is determined as the final stable model. At this time, the set of nodes contained in the model constitutes the point cloud terrain feature extraction result, wherein the significant terrain feature points corresponding to each node can accurately reflect the key undulation features and structural change locations of the actual terrain.

[0040] In one embodiment, before outputting the point cloud terrain feature extraction results, a consistency check and integrity check can be performed on the final model to confirm the absence of isolated feature points, abnormally subdivided triangular faces, or topological breaks. After passing the check, the point cloud terrain feature extraction results are output in structured data form for subsequent terrain analysis, digital elevation model construction, or terrain change monitoring applications.

[0041] Preferably, the calculation of the terrain dip angle and linear expression error of each triangular facet in step S3 is as follows: For each triangular facet, the corresponding original terrain slope angle is calculated based on the spatial angle between the normal vector of the triangular facet and the horizontal plane. The distribution of terrain types in the area to be measured is determined based on the original terrain dip angle, and the linear expression error corresponding to different terrain type distributions is statistically analyzed. By applying probability weights to the linear representation errors corresponding to different terrain types, the linear representation errors corresponding to triangular faces are obtained. Extract the geometric dimensions of the triangular faces and correct the original terrain dip angle by combining the linear expression error to obtain the corrected terrain dip angle of each triangular face.

[0042] In one embodiment, after constructing the ground point cloud triangulation model, each triangular face is selected sequentially according to the connection relationship of the triangulation, and the planar coordinates and point cloud elevation data corresponding to the three vertices of the triangular face are read. Based on the elevation difference between the three vertices and their distribution relationship in the plane, the tilt state of the triangular face relative to the local horizontal plane is determined, and the original terrain tilt angle of the triangular face is obtained.

[0043] In this embodiment, the triangular facet is used as a local terrain representation unit. The point cloud elevation data within the area covered by the triangular facet is projected onto the triangular facet plane, and the deviation between the point cloud elevation and the triangular facet plane is statistically analyzed. This characterizes the degree of deviation generated when the triangular facet is used to represent the terrain using a linear plane, and the linear representation error of the triangular facet is obtained.

[0044] In this embodiment, to avoid the influence of different terrain conditions on the tilt angle calculation results, the triangular facet is divided into terrain undulation degrees according to the original terrain tilt angle, and the linear expression error under the corresponding undulation degree is used as a reference to correct the original terrain tilt angle, thereby obtaining a corrected terrain tilt angle that matches the actual terrain undulation characteristics.

[0045] It should be noted that in a specific application scenario, when the point cloud area covered by a certain triangular facet is located in an area with relatively gentle terrain undulations such as a slope or plateau, the deviation of the elevation of each point cloud within the triangular facet relative to the triangular facet plane is relatively small overall, and the linear expression error is concentrated in the lower range. At this time, the correction range of the original terrain tilt angle is relatively small, thereby ensuring the continuity and stability of the slope information.

[0046] For example, when the elevation of the point cloud within a triangular area exhibits a slow, unidirectional change without any significant abrupt change, the triangular area is identified as a terrain unit with high linear representation adaptability. Only the original terrain tilt angle is slightly modified to avoid excessively amplifying the impact of local noise on the slope calculation results.

[0047] Conversely, when there are local undulations or significant elevation fluctuations within the area covered by a triangular facet, the deviation between the point cloud elevation and the triangular facet plane increases significantly, and the linear representation error distribution becomes discrete. In this case, the triangular facet is identified as a terrain unit with low linear representation adaptability, and the original terrain tilt angle is corrected accordingly. This corrected terrain tilt angle can more realistically reflect the actual terrain change trend, thus providing a reliable basis for subsequent slope threshold adjustment and extraction of significant terrain feature points.

[0048] As an example of the present invention, reference is made to Figure 2As shown, step S4 in this example includes: Step S41: Obtain the slope threshold; calculate the elevation accuracy of the area where the triangular face is located based on the terrain inclination angle of each triangular face and the preset grid digital elevation accuracy model, and confirm the slope deviation based on the elevation accuracy. Step S42: Adjust the slope threshold by the slope deviation to obtain the corrected slope threshold; Step S43: Divide the triangular facets into flat, gentle, or steep slope types according to the corrected slope threshold, and set the corresponding scale factor; Step S44: Create a feature extraction parameter system using the corrected slope threshold and slope factor.

[0049] In one embodiment, an initial slope threshold is first preset based on project requirements or industry experience, such as a reference slope used to distinguish between gentle and undulating terrain. Subsequently, for each ground point cloud triangle, its corresponding grid cell in the grid digital elevation model is read, and the elevation accuracy level of that grid cell is obtained, such as low-precision area, medium-precision area, or high-precision area.

[0050] Based on this, the terrain slope angle of the triangular facet is compared with the elevation accuracy level: when the triangular facet is located in an area with low elevation accuracy, it is considered that its slope calculation result is greatly affected by noise, and a larger slope deviation is set accordingly; when the triangular facet is located in an area with high elevation accuracy, a smaller slope deviation is set.

[0051] For example, if a triangular facet is located in an area with dense vegetation cover and sparse point clouds, the corresponding grid elevation accuracy is low, and its slope deviation is set to "large level"; if another triangular facet is located in an area with bare ground and dense point clouds, the slope deviation is set to "small level".

[0052] Based on the slope deviation obtained in step S41, the initial slope threshold is adaptively adjusted. Specifically, when the slope deviation is large, the slope threshold is relaxed to avoid flat areas being misclassified as steep slopes due to local elevation errors; when the slope deviation is small, the slope threshold is maintained or tightened to enhance the terrain classification ability.

[0053] For example, for triangular faces in low-precision regions, the corresponding slope threshold is appropriately increased compared to the initial threshold; while for triangular faces in high-precision regions, the original slope threshold remains unchanged.

[0054] The terrain type of each triangular facet is classified using a modified slope threshold. When the terrain slope angle of the triangular facet is lower than the first modified threshold, it is classified as flat land; when the slope angle is between the first and second modified thresholds, it is classified as gentle slope; and when the slope angle is higher than the second modified threshold, it is classified as steep slope.

[0055] Meanwhile, scale factors are set for different terrain types to control subsequent feature point extraction. For example, a larger scale factor is set for flat terrain to reduce redundant feature points; a medium scale factor is set for gentle slope terrain; and a smaller scale factor is set for steep slope terrain to enhance the response to detailed terrain changes.

[0056] The corrected slope threshold and its corresponding scale factor are uniformly encapsulated to form a feature extraction parameter system oriented towards the triangular facet. This parameter system serves as the direct input for subsequent steps traversing the point cloud data within the projection range of the triangular facet, controlling the strength of the relative height threshold determination and the selection density of significant terrain feature points.

[0057] For example, in the parameter system corresponding to the steep slope triangle, the relative height determination condition is more sensitive, so that terrain details such as ditches and steep transition points can be identified first; while in the parameter system corresponding to the flat land triangle, the repeated extraction of noise points is suppressed by the more lenient determination condition.

[0058] Preferably, step S5 includes the following steps: Step S51: Determine the extraction range of the triangular face according to the feature extraction parameter system; Step S52: Within the extraction range of the triangular face, traverse the airborne lidar point cloud data of each triangular face within the horizontal projection range, and calculate the vertical distance between the point cloud and the corresponding triangular face. Step S53: Within each triangular face, select the point with the largest vertical distance as a candidate terrain feature point; Step S54: Based on the relative height of the candidate terrain feature points and the linear expression error of their corresponding terrain types, determine whether the candidate terrain feature points meet the preset relative height threshold condition; Step S55: When a candidate terrain feature point meets the preset relative height threshold condition, it is determined as an effective terrain feature point, and the effective terrain feature point is added to the ground point cloud triangular mesh model to perform splitting of the triangular face.

[0059] In one embodiment, the feature extraction parameter system generated in step S44 is first read, and the corresponding point cloud extraction range is determined for each triangular face to be processed.

[0060] The extraction range is based on the projection area of ​​the triangular facet onto the horizontal plane, and is expanded or contracted according to the terrain type to which the triangular facet belongs. For example, for a triangular facet identified as flat land, only its own projection area is used as the extraction range; while for a triangular facet of gentle slope or steep slope, a certain buffer area is appropriately extended outside its projection boundary to ensure that slope break or abrupt change points are not missed.

[0061] In one specific embodiment, within a defined extraction range, the airborne lidar point cloud data falling within that range are traversed point by point.

[0062] For each point in the point cloud, its three-dimensional coordinates are obtained, and the vertical distance from that point to the plane containing the current triangular face is calculated to characterize the degree of undulation of the point relative to the triangular face. For example, within a gentle slope triangular face, some points in the traversed point cloud are closely distributed along the triangular face, with relatively small vertical distances; while points located on the slope shoulder or at micro-topographic protrusions have significantly larger vertical distances than the surrounding points.

[0063] In one specific embodiment, the vertical distances of all point clouds calculated within the same triangular facet are compared, and the point with the largest vertical distance is selected as a candidate terrain feature point for that triangular facet. This method ensures that the selected point reflects the most significant terrain undulation within the triangular facet. For example, in a locally raised, gently sloping area, although most point cloud points are close to the triangular facet, the point cloud point corresponding to the top of the raised area has the largest vertical distance and is therefore selected as a candidate terrain feature point.

[0064] In one specific embodiment, the relative height of the candidate terrain feature point with respect to the average elevation of the triangular face is obtained, and the validity of the candidate terrain feature point is judged by combining the linear expression error corresponding to the terrain type to which the triangular face belongs.

[0065] When the relative height of a candidate terrain feature point is significantly higher than the allowable linear expression error range for that terrain type, the point is considered to reflect the actual terrain change; otherwise, the point is considered to be likely to originate from noise or local measurement error.

[0066] For example, in a flat triangular facet, even if a point with the largest vertical distance appears, its relative height may still fall within the allowable error range for flat terrain, thus being deemed invalid; while in a steep triangular facet, the same change in height may be considered a valid feature.

[0067] In one specific embodiment, when a candidate terrain feature point meets a preset relative height threshold condition, it is confirmed as an effective terrain feature point and added to the existing ground point cloud triangulation model.

[0068] Subsequently, using the effective shape feature points as new vertices, the original triangular face is split and divided, so that the original single triangular face is decomposed into multiple new sub-triangular faces, thereby improving the accuracy of the triangular mesh in representing the real terrain shape.

[0069] For example, in a triangular facet that originally covered the location of a slope break, by introducing effective shape feature points, the triangular facet is split into multiple smaller triangular faces, so that the location of the slope break is clearly depicted in the model.

[0070] Preferably, step S51 includes: The terrain tilt angle, slope threshold and scale factor corresponding to each triangle face are extracted by the ground point cloud triangulation model. Based on the correspondence between terrain dip angle and slope threshold, the terrain type to which the current triangular face belongs is determined, and a scale factor matching the terrain type is selected as the extraction control parameter for the triangular face. The horizontal projection area of ​​the triangular face is used as the basic extraction area, and the basic extraction area is expanded or shrunk using extraction control parameters to obtain the candidate extraction range. The accuracy and effectiveness of the candidate extraction range are verified, and regions that do not have the significance of feature extraction are eliminated, thereby obtaining the extraction range of the triangular face.

[0071] In one embodiment, firstly, by accessing the attribute information associated with the current triangular facet in the ground point cloud triangulation model, the terrain slope angle corresponding to the triangular facet is directly read. Simultaneously, the slope threshold determined in steps S41-S44 and the matching scale factor are obtained. This information serves as the source of basic parameters for subsequent feature point extraction control of the triangular facet.

[0072] Subsequently, the terrain slope angle is compared with a slope threshold to determine the terrain type of the current triangular facet. For example, when the terrain slope angle of the triangular facet is lower than the slope threshold, the triangular facet is classified as flat land; when the terrain slope angle is close to the slope threshold, it is classified as gentle slope; and when the terrain slope angle is significantly higher than the slope threshold, it is classified as steep slope. Based on the determination result, a scale factor pre-set to correspond to the terrain type is selected as the extraction control parameter for the triangular facet.

[0073] Based on this, the projection area of ​​the triangular facet onto the horizontal plane is used as the basic extraction area, and the range of the basic extraction area is adjusted using the extraction control parameters. Specifically, for flat triangular facests, the basic extraction area remains unchanged; for gently sloping triangular facests, the basic extraction area is appropriately expanded; for steeply sloping triangular facests, the boundary of the extraction area is further expanded to cover possible slope breaks or abrupt changes, thereby forming a candidate extraction range.

[0074] Finally, the accuracy and effectiveness of the candidate extraction range are verified. By checking the point cloud density, spatial continuity and the discernibility of elevation changes within the candidate extraction range, areas with sparse point clouds, broken boundaries or no terrain feature expression are eliminated, and only the effective areas that meet the feature extraction requirements are retained. Finally, the actual extraction range of the current triangular face is determined.

[0075] For example, in a hilly transition area, a triangular facet is identified as a gentle slope. After its basic projection area expands outward under the scale factor, the initial candidate range covers the slope shoulder. After accuracy and validity verification, the edge areas with overly sparse point clouds are eliminated, and the continuous area that can reflect the slope change is retained as the final extraction range of the triangular facet.

[0076] Preferably, the preset relative height threshold condition in step S54 is as follows: The vertical distance of a candidate terrain feature point relative to its triangular face is greater than the allowable elevation error of the digital elevation model under the terrain inclination angle corresponding to the triangular face, and is a threshold after being corrected by a safety factor. The allowable elevation error of the digital elevation model is used to characterize the reasonable error range of point cloud elevation data under the current terrain conditions, and the safety factor is used to suppress the interference of pseudo-feature points caused by measurement noise or data errors.

[0077] In one embodiment, the vertical distance of a candidate terrain feature point relative to its corresponding triangular face is obtained. This vertical distance reflects the degree of deviation of the point from the actual elevation expressed linearly by the triangular face. Simultaneously, based on the terrain slope angle corresponding to the triangular face, the allowable elevation error range of the digital elevation model used under the given terrain conditions is determined. This allowable elevation error characterizes the error range within which the point cloud elevation data can still be considered a reasonable undulation under the current slope conditions.

[0078] Subsequently, a safety factor is introduced to correct the allowable elevation error, forming a relative height threshold for judgment. This safety factor is used to suppress false elevation abrupt changes caused by laser ranging noise, attitude errors, or local anomalies in the point cloud, thereby avoiding misclassification of non-realistic terrain undulations as terrain feature points.

[0079] When the vertical distance between a candidate terrain feature point and its triangular face is greater than the allowable elevation error threshold after safety factor correction, the candidate terrain feature point is determined to meet the preset relative height threshold condition; otherwise, the candidate terrain feature point is determined not to meet the condition.

[0080] For example, in a gently sloping area, the digital elevation model corresponding to a certain triangular facet allows for a certain degree of elevation fluctuation. If a candidate terrain feature point only exhibits slight undulations consistent with the surrounding point cloud, its vertical distance does not exceed the corrected allowable error range, and the point is judged as a pseudo feature point caused by measurement error. However, when the vertical distance of the candidate terrain feature point significantly exceeds the error range, it can be confirmed that it reflects the actual terrain abrupt change, thus satisfying the relative height threshold condition.

[0081] Preferably, step S55 includes: When a candidate terrain feature point meets the preset relative height threshold condition, the candidate terrain feature point is determined to be an effective terrain feature point. For effective shape feature points, record their spatial coordinate information and establish their association with the corresponding triangular face; Effective shape feature points are added as new nodes to the ground point cloud triangulation model, and the effective shape feature points are used as dividing nodes to perform a fission partitioning operation on their respective triangular faces, so that the triangular face is divided into multiple new sub-triangular faces.

[0082] In one embodiment, the three-dimensional spatial coordinate information of the terrain feature points that are identified as valid is recorded, and the correspondence between the valid terrain feature points and their respective triangular faces is established in the data structure to ensure that the source triangular face of the feature point can be accurately traced during the subsequent triangulation update process.

[0083] Subsequently, the effective shape feature points are introduced as new nodes into the ground point cloud triangulation model, and the effective shape feature points are used as dividing nodes to perform a fission partitioning operation on their respective original triangular faces. The fission partitioning operation divides the original triangular face into multiple new sub-triangular faces by connecting the effective shape feature points with the vertices of the original triangular face, thereby improving the local fineness of terrain representation while maintaining the overall topological continuity.

[0084] For example, in a certain slope area, the original triangular facet is too large to accurately represent the local convex terrain. When a candidate terrain feature point that is significantly higher than the linear representation of the triangular facet is detected and confirmed as an effective terrain feature point, the point is introduced into the triangular mesh model. This allows the original triangular facet to be subdivided into multiple sub-triangular facets, thereby providing a more accurate description of the terrain undulations in the slope area.

[0085] Preferably, step S55 further includes: When a candidate terrain feature point does not meet the preset relative height threshold condition, the candidate terrain feature point is judged and marked as an invalid feature point, and the triangular face splitting process of the invalid feature point is terminated. Preserve the topology of the triangular face to which invalid feature points belong and perform new node or split operations on the interrupted ground point cloud triangulation model.

[0086] In one embodiment, the candidate terrain feature point is marked as invalid, and its corresponding relative height value, the number of the triangular face to which it belongs, and the judgment result are recorded as invalid feature point information, so as to avoid repeatedly triggering feature extraction judgment for the same location in subsequent iterations.

[0087] After invalidation is completed, the triangular face splitting process for the invalid feature point is immediately terminated. The invalid feature point is not introduced into the ground point cloud triangulation model as a new node, thereby avoiding unnecessary model refinement caused by measurement noise, local point cloud density fluctuations or linear expression errors.

[0088] At the same time, the original topology of the triangle to which the invalid feature point belongs remains unchanged, and the update process of the current ground point cloud triangulation model is interrupted, so that the model only continues to process other triangles that have not yet been judged, without performing new node or fission operations on the current triangle.

[0089] For example, in a flat area with minimal topographic relief, a candidate topographic feature point may appear as the highest point in the local point cloud, but its relative height is still within the allowable error range of the digital elevation model. In this case, classifying the candidate topographic feature point as an invalid feature point and terminating the splitting process can effectively avoid introducing too many sub-triangles in the flat area, thereby ensuring a balance between overall accuracy and computational efficiency in the triangulation model.

[0090] Preferably, the method for obtaining the topological structure of the triangular facet to which the invalid feature point belongs includes: Based on the spatial location of invalid feature points in the ground point cloud triangulation model, determine the target triangular face in which they are located; Extract the node index information of the target triangle in the ground point cloud triangular network model to obtain the vertex coordinates that constitute the target triangle; Based on vertex coordinates, extract the identification information of adjacent triangles that are adjacent to the target triangle to establish the adjacency relationship between the target triangle and the surrounding triangles. The topological structure of the triangle to which invalid feature points belong is obtained based on vertex information and adjacency relationships.

[0091] In one embodiment, based on the spatial coordinates of invalid feature points in the ground point cloud triangulation model, a point-to-surface attribution determination process is performed. The invalid feature points are projected into the current triangulation model, and the target triangular face to which they fall or are closest is determined, thereby confirming the target triangular face to which the invalid feature points belong.

[0092] After determining the target triangle, the node index information of the target triangle is read from the ground point cloud triangulation model, and the coordinate information of each vertex constituting the target triangle is extracted based on the node index to characterize the geometric boundary and spatial shape of the target triangle.

[0093] Subsequently, using each edge of the target triangle as the retrieval reference, the adjacent triangles sharing the boundary with it are queried in the ground point cloud triangulation model, and the corresponding adjacent triangle identification information is extracted, thereby establishing a description of the adjacency relationship between the target triangle and the surrounding triangles.

[0094] After completing the extraction of vertex information and the construction of adjacency relationships, the vertex coordinate information of the target triangle face and the identification information of its adjacent triangle faces are combined to form topological structure data that describes the local connection state of the target triangle face, that is, to obtain the complete topological structure of the triangle face to which the invalid feature point belongs.

[0095] For example, in a relatively flat area, an invalid feature point may be located inside a target triangle connected to multiple surrounding triangles by three edges. By obtaining the vertex distribution of this target triangle and its connection relationships with adjacent triangles in the above manner, the overall connectivity and stability of the triangular mesh model in this area can be accurately maintained without introducing new nodes.

[0096] Preferably, determining the target triangular face containing invalid feature points based on their spatial location in the ground point cloud triangulation model includes: Based on the three-dimensional spatial coordinates of the invalid feature points, confirm their corresponding positions in the triangulation plane projection; Based on the planar projection position, retrieve the set of triangular faces in the ground point cloud triangulation model that intersect or cover the projection position to form a candidate triangular face set; In the candidate triangle face set, based on the geometric inclusion relationship between invalid feature points and each candidate triangle face, the triangle face containing the projection position of the invalid feature points is determined as the target triangle face. When the projected position of an invalid feature point is located in the common boundary or vertex region of multiple adjacent triangular faces, the corresponding target triangular face is determined according to the principle of minimizing the vertical distance between the invalid feature point and each adjacent triangular face.

[0097] In one embodiment, to accurately obtain the attribution relationship of invalid feature points in the ground point cloud triangulation model, the three-dimensional spatial coordinates of the invalid feature points are first processed. Specifically, the three-dimensional coordinate information of the invalid feature points is read, and their elevation components are ignored. The coordinates are then mapped to the horizontal reference plane corresponding to the ground point cloud triangulation model to obtain the corresponding position of the invalid feature point in the triangulation plane projection.

[0098] After obtaining the planar projection position, a spatial retrieval operation is performed in the ground point cloud triangulation model with the projection position as the center to find triangular faces within the planar projection range that intersect or cover the projection position, and the retrieved triangular faces are summarized to form a candidate triangular face set.

[0099] Subsequently, a geometric inclusion relationship judgment is performed on each triangle in the candidate triangle facet set, specifically determining whether the planar projection position of the invalid feature point is located within the internal region of the corresponding triangle facet. If a triangle facet can completely contain the projection position, then that triangle facet is directly determined as the target triangle facet to which the invalid feature point belongs.

[0100] In another implementation, if the planar projection position of the invalid feature point happens to be located in the common boundary or common vertex region of multiple adjacent triangular faces, and the target triangular face cannot be uniquely determined by geometric inclusion relationships, a vertical distance determination rule is further introduced. Specifically, the vertical distance from the invalid feature point to the plane containing each adjacent triangular face is calculated, and the triangular face with the smallest vertical distance is selected as the target triangular face to which the invalid feature point belongs.

[0101] For example, in an area with relatively small topographic relief but dense triangulation, the projection of an invalid feature point may fall on the common boundary of two adjacent triangular faces. In this case, by comparing the perpendicular distances of the invalid feature point to the two triangular faces, it can be accurately determined that it is closer to one of the triangular faces, thus identifying that triangular face as the target triangular face and avoiding the instability of topological determination caused by boundary ambiguity.

[0102] Of particular importance, step S2 includes: Read the ellipsoidal height information corresponding to the airborne lidar point cloud data, and use the ellipsoidal height as the original elevation data to be converted; Based on the spatial distribution range of airborne lidar point cloud data, the regional geoid refinement model corresponding to the regional spatial range is invoked, and the elevation benchmark constraint relationship of the regional geoid refinement model at the location of the point cloud is determined. Based on the elevation datum constraint relationship, the ellipsoidal height of each laser echo point in the original elevation data is transformed by elevation datum conversion, so that the ellipsoidal height is converted from the ellipsoidal elevation datum to the elevation value with the regional quasi-geoid as the reference, so as to obtain point cloud elevation data constrained by the regional quasi-geoid.

[0103] In one embodiment, the ellipsoidal height information corresponding to each laser echo point is read point by point from the airborne lidar point cloud data, and the ellipsoidal height is stored uniformly as the raw elevation data to be processed. The ellipsoidal height is usually calculated by the airborne positioning and attitude determination system and corresponds one-to-one with the planar coordinates of the point cloud.

[0104] Subsequently, based on the overall spatial distribution range of the airborne lidar point cloud data, the geographical area covered by the point cloud data is determined, and a regional quasi-geoid refinement model matching this spatial range is called from a pre-built regional quasi-geoid refinement model library. This model can reflect the local undulation characteristics of the quasi-geoid within the target area and is used to establish the elevation benchmark constraint relationship between the ellipsoidal height and the normal height.

[0105] Based on this, the elevation reference constraint information of the corresponding location is obtained in the regional geoid refinement model using the planar position of each laser echo point as an index. According to the constraint relationship, the ellipsoidal height in the original elevation data is transformed point by point, so that the elevation value of each laser echo point is transformed from the ellipsoidal elevation reference to the elevation value with the regional geoid as a reference.

[0106] For example, in hilly terrain regions, due to local variations in the quasi-geoid, directly using the ellipsoidal height as the elevation benchmark can easily introduce systematic biases. Through the aforementioned transformation process, the point cloud elevations within the same terrain unit can better reflect the actual surface undulations, thus obtaining point cloud elevation data constrained by the regional quasi-geoid. This provides a unified and reliable elevation benchmark for subsequent ground point cloud modeling and terrain feature extraction.

[0107] For details, please refer to Figure 3 The picture As potential energy, To define the potential energy expression of the geoid in space. For normal potential energy, For normal potential energy representation, the surface height (i.e., topographic elevation) relative to the height of the ellipsoid is denoted as . , This reflects the undulation of the geoid relative to the ellipsoid; the height relative to the geoid is denoted as . This geometric ellipsoid and its normal ellipsoid potential are called the geodetic reference system.

[0108] Of particular importance is the reading of the ellipsoidal height information corresponding to the airborne lidar point cloud data, and the use of this ellipsoidal height as the raw elevation data to be converted, including: Data structure analysis is performed on the airborne lidar point cloud data to identify the location information field corresponding to each laser echo point; Extract the three-dimensional spatial coordinate data corresponding to each laser echo point from the location information field, and determine the ellipsoidal height information used to characterize the elevation datum in the three-dimensional spatial coordinate data; A one-to-one correspondence is established between the ellipsoidal height information and the planar position information of the corresponding laser echo point, thereby obtaining the original elevation data.

[0109] In one embodiment, the acquired airborne lidar point cloud data is parsed. The point cloud data is stored in a preset data format, where each laser echo point contains a location information field. By parsing the record structure of the point cloud data, the location information field corresponding to each laser echo point is identified and read line by line.

[0110] Subsequently, the three-dimensional spatial coordinate data of each laser echo point is extracted from the location information field. The three-dimensional spatial coordinate data includes at least planar coordinate information and a height component representing an elevation datum, wherein the height component is the ellipsoidal height information based on a reference ellipsoid. This height component is distinguished from other coordinate components by field identifiers or data order determination, thereby clarifying the corresponding position of the ellipsoidal height in the three-dimensional spatial coordinate system.

[0111] Based on this, the extracted ellipsoidal height information is associated and stored with the corresponding planar position information of the laser echo points, so that each ellipsoidal height has a one-to-one correspondence with its corresponding planar coordinates. Through the above processing, the ellipsoidal heights of each laser echo point in the airborne lidar point cloud are uniformly organized into a raw elevation dataset to be converted, providing standardized input data for subsequent elevation datum conversion processing based on the regional quasi-geoid refinement model.

[0112] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0113] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for intelligent extraction of point cloud topographic features based on regional quasi-geoid constraints, characterized in that, Includes the following steps: Step S1: Acquire airborne lidar point cloud data and corresponding global navigation satellite observation data for the area to be measured; construct a refined geoid model of the region based on the global navigation satellite observation data; Step S2: Based on the regional geoid refinement model, perform elevation datum transformation on the ellipsoidal height in the airborne lidar point cloud data to obtain point cloud elevation data constrained by the regional geoid. Step S3: Construct a ground point cloud triangulation model based on point cloud elevation data and use triangular faces as the basic unit for terrain representation, calculate the terrain dip angle and linear representation error of each triangular face; Step S4: Adjust the slope threshold and scale factor based on the terrain tilt angle and linear expression error to confirm the feature extraction parameter system; Step S5: Based on the feature extraction parameter system, traverse the airborne lidar point cloud data within the projection range of each triangular face, extract significant terrain feature points that meet the relative height threshold, and add the significant terrain feature points to the ground point cloud triangulation model for iterative subdivision. Step S6: Repeat step S5 until the preset iteration termination condition is met, and output the point cloud terrain feature extraction results.

2. The intelligent extraction method for point cloud terrain features based on regional quasi-geoid constraints according to claim 1, characterized in that, Step S3 involves calculating the topographic dip angle and linear expression error of each triangular facet as follows: For each triangular facet, the corresponding original terrain slope angle is calculated based on the spatial angle between the normal vector of the triangular facet and the horizontal plane. The distribution of terrain types in the area to be measured is determined based on the original terrain dip angle, and the linear expression error corresponding to different terrain type distributions is statistically analyzed. By applying probability weights to the linear representation errors corresponding to different terrain types, the linear representation errors corresponding to triangular faces are obtained. Extract the geometric dimensions of the triangular faces and correct the original terrain dip angle by combining the linear expression error to obtain the corrected terrain dip angle of each triangular face.

3. The intelligent extraction method for point cloud terrain features based on regional quasi-geoid constraints according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Obtain the slope threshold; calculate the elevation accuracy of the area where the triangular face is located based on the terrain inclination angle of each triangular face and the preset grid digital elevation accuracy model, and confirm the slope deviation based on the elevation accuracy. Step S42: Adjust the slope threshold by the slope deviation to obtain the corrected slope threshold; Step S43: Divide the triangular facets into flat, gentle, or steep slope types according to the corrected slope threshold, and set the corresponding scale factor; Step S44: Create a feature extraction parameter system using the corrected slope threshold and slope factor.

4. The intelligent extraction method for point cloud terrain features based on regional quasi-geoid constraints according to claim 1, characterized in that, Step S5 includes the following steps: Step S51: Determine the extraction range of the triangular face according to the feature extraction parameter system; Step S52: Within the extraction range of the triangular face, traverse the airborne lidar point cloud data of each triangular face within the horizontal projection range, and calculate the vertical distance between the point cloud and the corresponding triangular face. Step S53: Within each triangular face, select the point with the largest vertical distance as a candidate terrain feature point; Step S54: Based on the relative height of the candidate terrain feature points and the linear expression error of their corresponding terrain types, determine whether the candidate terrain feature points meet the preset relative height threshold condition; Step S55: When a candidate terrain feature point meets the preset relative height threshold condition, it is determined as an effective terrain feature point, and the effective terrain feature point is added to the ground point cloud triangular mesh model to perform splitting of the triangular face.

5. The intelligent extraction method for point cloud terrain features based on regional quasi-geoid constraints according to claim 4, characterized in that, Step S51 includes: The terrain tilt angle, slope threshold and scale factor corresponding to each triangle face are extracted by the ground point cloud triangulation model. Based on the correspondence between terrain dip angle and slope threshold, the terrain type to which the current triangular face belongs is determined, and a scale factor matching the terrain type is selected as the extraction control parameter for the triangular face. The horizontal projection area of ​​the triangular face is used as the basic extraction area, and the basic extraction area is expanded or shrunk using extraction control parameters to obtain the candidate extraction range. The accuracy and effectiveness of the candidate extraction range are verified, and regions that do not have the significance of feature extraction are eliminated, thereby obtaining the extraction range of the triangular face.

6. The intelligent extraction method for point cloud terrain features based on regional quasi-geoid constraints according to claim 4, characterized in that, The preset relative height threshold condition in step S54 is as follows: The vertical distance of a candidate terrain feature point relative to its triangular face is greater than the allowable elevation error of the digital elevation model under the terrain inclination angle corresponding to the triangular face, and is a threshold after being corrected by a safety factor. The allowable elevation error of the digital elevation model is used to characterize the reasonable error range of point cloud elevation data under the current terrain conditions, and the safety factor is used to suppress the interference of pseudo-feature points caused by measurement noise or data errors.

7. The intelligent extraction method for point cloud terrain features based on regional quasi-geoid constraints according to claim 4, characterized in that, Step S55 includes: When a candidate terrain feature point meets the preset relative height threshold condition, the candidate terrain feature point is determined to be an effective terrain feature point. For effective shape feature points, record their spatial coordinate information and establish their association with the corresponding triangular face; Effective shape feature points are added as new nodes to the ground point cloud triangulation model, and the effective shape feature points are used as dividing nodes to perform a fission partitioning operation on their respective triangular faces, so that the triangular face is divided into multiple new sub-triangular faces.

8. The intelligent extraction method for point cloud terrain features based on regional quasi-geoid constraints according to claim 4, characterized in that, Step S55 also includes: When a candidate terrain feature point does not meet the preset relative height threshold condition, the candidate terrain feature point is judged and marked as an invalid feature point, and the triangular face splitting process of the invalid feature point is terminated. Preserve the topology of the triangular face to which invalid feature points belong and perform new node or split operations on the interrupted ground point cloud triangulation model.

9. The intelligent extraction method for point cloud terrain features based on regional quasi-geoid constraints according to claim 8, characterized in that, Methods for obtaining the topological structure of the triangular facet to which invalid feature points belong include: Based on the spatial location of invalid feature points in the ground point cloud triangulation model, determine the target triangular face in which they are located; Extract the node index information of the target triangle in the ground point cloud triangular network model to obtain the vertex coordinates that constitute the target triangle; Based on vertex coordinates, extract the identification information of adjacent triangles that are adjacent to the target triangle to establish the adjacency relationship between the target triangle and the surrounding triangles. The topological structure of the triangle to which invalid feature points belong is obtained based on vertex information and adjacency relationships.

10. The intelligent extraction method for point cloud terrain features based on regional quasi-geoid constraints according to claim 9, characterized in that, Based on the spatial location of invalid feature points in the ground point cloud triangulation model, the target triangular face containing them is determined as follows: Based on the three-dimensional spatial coordinates of the invalid feature points, confirm their corresponding positions in the triangulation plane projection; Based on the planar projection position, retrieve the set of triangular faces in the ground point cloud triangulation model that intersect or cover the projection position to form a candidate triangular face set; In the candidate triangle face set, based on the geometric inclusion relationship between invalid feature points and each candidate triangle face, the triangle face containing the projection position of the invalid feature points is determined as the target triangle face. When the projected position of an invalid feature point is located in the common boundary or vertex region of multiple adjacent triangular faces, the corresponding target triangular face is determined according to the principle of minimizing the vertical distance between the invalid feature point and each adjacent triangular face.

Images