A line marking robot working area three-dimensional environment reconstruction system

Abstract

The application relates to the technical field of three-dimensional environment reconstruction, in particular to a three-dimensional environment reconstruction system for a marking robot working area, which comprises a point cloud attribute calculation module, which performs ground scanning on the marking robot working area, obtains original point clouds, performs coordinate system alignment on the original point clouds, traverses all data points and calculates the principal curvature value and the principal direction value of each data point. In the application, after the ground scanning on the marking robot working area and the acquisition of the original point clouds, the coordinate system alignment and the principal curvature and the principal direction calculation of each data point are performed, on the basis of the establishment of the curvature attribute point cloud with local geometric attributes, the data points with higher principal curvature values are screened by adopting a set threshold value, and the sensitive response capability to the fine ground marking traces is improved; the feature point cluster established in combination with the spatial proximity and the geometric direction consistency constructs the discrete marking feature expression with stronger spatial continuity and more accurate shape recognition.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional environment reconstruction technology, and in particular to a three-dimensional environment reconstruction system for the working area of ​​a line-marking robot. Background Technology

[0002] The field of 3D environment reconstruction technology involves data acquisition, digital processing, and 3D model reconstruction of real physical environments to obtain high-precision spatial geometric structure and texture information.

[0003] In practice, 3D environment reconstruction technology often relies on static processing and geometric mapping of point cloud data, which can easily lead to problems such as blurred representation and unclear feature boundaries in areas with complex structures or poor texture continuity. Furthermore, curvature attribute calculations are often limited to global fitting or mean filtering, resulting in a delayed response to locally highly variable areas and difficulty in identifying fine-grained line features. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a three-dimensional environment reconstruction system for the working area of ​​a line-marking robot.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a three-dimensional environment reconstruction system for the working area of ​​a line-marking robot includes:

[0006] The point cloud attribute calculation module performs ground scanning on the working area of ​​the line drawing robot to obtain the original point cloud, performs coordinate system alignment on the original point cloud, traverses all data points and calculates the principal curvature value and principal direction value of each data point to establish a curvature attribute point cloud.

[0007] The line drawing feature recognition module, based on the curvature attribute point cloud, sets a curvature filtering threshold and traverses all data points, filters out points whose main curvature value is greater than the curvature filtering threshold, and obtains a trace candidate point set. Based on the trace candidate point set, a discrete line drawing feature cluster is established.

[0008] The trace network construction module calculates the center position and main direction of each cluster based on the discrete line feature clusters, connects clusters whose spatial distance is less than a threshold and whose main directions are consistent, generates a line connection path graph, and fits and establishes a line vectorization structure based on the line connection path graph.

[0009] The multi-resolution mesh generation module divides the working area of ​​the line-drawing robot into initial cells based on the curvature attribute point cloud, establishes an initial complexity mesh, and obtains an adaptive terrain mesh model based on the initial complexity mesh.

[0010] Preferably, the steps for obtaining the curvature attribute point cloud are as follows:

[0011] The work area is scanned by ground laser using a line-marking robot, and the spatial three-dimensional position coordinates of the area are collected and recorded to construct a set of spatial coordinate points and generate an original point cloud.

[0012] Based on the original point cloud, and with a predefined reference coordinate system as a reference, the position of each spatial coordinate point is adjusted to the corresponding position under the reference coordinate system by performing coordinate transformation on the set of spatial coordinate points one by one, thereby generating an aligned original point cloud.

[0013] Based on the aligned original point cloud, the spatial geometric distribution within the neighborhood of each spatial coordinate point is calculated point by point. The principal curvature value and the corresponding principal direction vector of each spatial coordinate point are extracted, and the calculated principal curvature value and principal direction vector are assigned to the corresponding spatial coordinate points in sequence to generate a curvature attribute point cloud.

[0014] Preferably, the step of obtaining the candidate trace point set is as follows:

[0015] Based on the curvature attribute point cloud, the principal curvature values ​​of all spatial coordinate points in the curvature attribute point cloud are statistically analyzed, and the mean value of the principal curvature values ​​of all spatial coordinate points is calculated. The mean value is taken as the curvature filtering threshold to generate the curvature filtering threshold.

[0016] Based on the curvature filtering threshold, the principal curvature values ​​of each spatial coordinate point in the curvature attribute point cloud are traversed point by point. It is determined whether the principal curvature values ​​of each spatial coordinate point exceed the curvature filtering threshold. Spatial coordinate points whose principal curvature values ​​exceed the curvature filtering threshold are marked one by one to generate a trace candidate point set.

[0017] Preferably, the step of obtaining the discrete line feature cluster is as follows:

[0018] Based on the set of candidate trace points, the three-dimensional coordinates and local direction vector of each candidate trace point are extracted, and the spatial position and direction relationship between each candidate trace point is constructed by traversing point by point to form a set of candidate trace point pairs.

[0019] Based on the set of candidate trace points, calculate the comprehensive morphological similarity between each pair of candidate trace points;

[0020] Based on the comprehensive morphological similarity between each pair of trace candidate points, a threshold is used to filter the similarity between all pairs of trace candidate points. The trace candidate point pairs with a comprehensive morphological similarity higher than the set threshold are used as the connecting edges of the same clustering unit. All trace candidate points that form a connection relationship are aggregated into independent subsets, and discrete line-drawing feature clusters are obtained according to the aggregation relationship.

[0021] Preferably, the step of obtaining the line-connected path diagram is as follows:

[0022] Based on the discrete scribing feature cluster, the three-dimensional spatial coordinates of all trace candidate points in each discrete scribing feature cluster are traversed one by one. The average value of the spatial coordinates of all trace candidate points is taken and the average value is defined as the center position of the discrete scribing feature cluster, thereby generating a set of center positions of discrete scribing feature clusters.

[0023] Based on the set of center positions of the discrete line-drawing feature clusters, all candidate trace points of the discrete line-drawing feature clusters are traversed one by one. Principal component analysis is used to extract the principal component feature vector of each discrete line-drawing feature cluster. The principal component feature vector is defined as the principal direction vector of the discrete line-drawing feature cluster, and a set of principal direction vectors of discrete line-drawing feature clusters is generated.

[0024] Based on the set of center positions of the discrete line feature clusters and the set of main direction vectors of the discrete line feature clusters, the spatial distance and the angle between the main directions of any two discrete line feature clusters are compared one by one. Discrete line feature clusters with spatial distance values ​​less than a preset spatial threshold and main direction angles less than a preset angle threshold are paired and connected to form a connection relationship and defined as a connection path, generating a line connection path map.

[0025] Preferably, the step of obtaining the line vectorization structure is as follows:

[0026] Based on the line connection path map, all discrete line feature clusters in each connection path in the line connection path map are traversed one by one. The three-dimensional spatial coordinates of all trace candidate points contained in each discrete line feature cluster are extracted. All three-dimensional spatial coordinates of each discrete line feature cluster are summarized to generate a set of three-dimensional spatial coordinates of the connection path.

[0027] Based on the set of three-dimensional spatial coordinates of the connection path, linear fitting of the spatial point cloud is performed for each connection path. The direction vector and center position of the fitted line segment are calculated by the least squares method. The fitted line segment parameters of the connection path are obtained one by one, and a set of fitted line segment parameters of the connection path is generated.

[0028] Based on the set of line segment parameters fitted by the connection path, the parameters of the fitted line segments are converted into a standard vectorized representation format one by one, and the three-dimensional coordinates and vector directions of the start and end points of the fitted line segments are recorded in sequence to construct a line vectorization structure.

[0029] Preferably, the step of obtaining the initial complexity grid is as follows:

[0030] Based on the curvature attribute point cloud, the boundary range and spatial size parameters of the working area of ​​the line-marking robot are set. The entire working area of ​​the line-marking robot is divided into regular cubic cells according to a uniform side length. Each curvature attribute point in the curvature attribute point cloud is mapped and assigned to the cubic cell according to its three-dimensional position coordinates, thereby generating the initial set of cells for the working area of ​​the line-marking robot.

[0031] Based on the initial set of cells in the working area of ​​the line-marking robot, calculate the coupling morphological complexity of each initial cell;

[0032] Based on the coupled morphological complexity of each initial cell, the entire set of initial cells in the working area of ​​the line-marking robot is traversed. The spatial position index and coupled morphological complexity of all initial cells are paired and encoded one by one, and a unified organization is formed to form a complexity expression grid of spatial structure changes within the region, thus generating the initial complexity grid.

[0033] Preferably, the steps for obtaining the adaptive terrain mesh model are as follows:

[0034] Based on the initial complexity grid, the coupling morphological complexity of each initial cell in the initial complexity grid is traversed one by one, and the values ​​are compared with the pre-set coupling morphological complexity threshold. At the same time, it is detected whether the spatial position of each initial cell has a spatial geometric intersection relationship with the line vectorization structure. Initial cells that meet the conditions of coupling morphological complexity exceeding the threshold or having an intersection relationship are selected to generate a set of cells to be subdivided.

[0035] Based on the set of cells to be subdivided, each cell to be subdivided is evenly divided along the midpoint of the three-dimensional spatial coordinate axis to form eight smaller sub-cells, and the three-dimensional spatial position and boundary information of all sub-cells are recorded one by one to generate the first recursive subdivided cell set.

[0036] Based on the first recursive subdivision cell set, the steps of obtaining the cell set to be subdivided and obtaining the first recursive subdivision cell set are repeated until the coupling morphological complexity of all cells meets the set threshold and no longer intersects with the line vectorized structure. Then, all cells are connected by grid topology to generate an adaptive terrain grid model.

[0037] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0038] In this invention, after scanning the ground and acquiring the original point cloud of the working area of ​​the marking robot, coordinate system alignment and principal curvature and principal direction calculations are performed for each data point. Based on the establishment of a curvature attribute point cloud with local geometric properties, a threshold is set to filter data points with higher principal curvature values, improving the sensitivity to subtle ground marking marks. By combining feature point clusters established based on spatial proximity and geometric direction consistency, a discrete marking feature expression with stronger spatial continuity and more accurate shape recognition is constructed. The center position and principal direction between point clusters are determined, and point clusters with geometric coherence are connected to complete path fitting, so that the marking information forms a structured vector expression in space, which facilitates subsequent path extraction and task planning. By dividing the space into grids and introducing coupled morphological complexity calculation based on the determinant of the structural tensor and the average curvature of each cell, a complexity grid that can dynamically respond to terrain changes is constructed. The grid is further subdivided based on whether it intersects with the marking structure, forming an adaptive terrain grid model with spatial adaptability and multi-scale expression capabilities, which enhances the completeness of the marking area feature expression and the accuracy of environmental modeling. Attached Figure Description

[0039] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0041] Please see Figure 1 This invention provides a technical solution: a three-dimensional environment reconstruction system for the working area of ​​a line-marking robot, comprising:

[0042] The point cloud attribute calculation module performs ground scanning on the working area of ​​the line drawing robot to obtain the original point cloud, performs coordinate system alignment on the original point cloud, traverses all data points and calculates the principal curvature value and principal direction value of each data point to establish a curvature attribute point cloud.

[0043] The line drawing feature recognition module, based on the curvature attribute point cloud, sets a curvature filtering threshold and traverses all data points, filtering out points whose main curvature value is greater than the curvature filtering threshold to obtain a trace candidate point set, and establishes a discrete line drawing feature cluster based on the trace candidate point set.

[0044] The trace network construction module calculates the center position and main direction of each cluster based on discrete line feature clusters, connects clusters whose spatial distance is less than the threshold and whose main directions are consistent, generates a line connection path graph, and fits and establishes a line vectorization structure based on the line connection path graph.

[0045] The multi-resolution mesh generation module divides the working area of ​​the line-drawing robot into initial cells based on the curvature attribute point cloud, establishes an initial complexity mesh, and obtains an adaptive terrain mesh model based on the initial complexity mesh.

[0046] The steps to obtain a point cloud with curvature attributes are as follows:

[0047] The work area is scanned by ground laser using a line-marking robot, and the spatial three-dimensional position coordinates of the area are collected and recorded to construct a set of spatial coordinate points and generate an original point cloud.

[0048] Based on the original point cloud, and with a predefined reference coordinate system as a reference, the position of each spatial coordinate point is adjusted to the corresponding position under the reference coordinate system by performing coordinate transformation on the set of spatial coordinate points one by one, thereby generating an aligned original point cloud.

[0049] Based on the aligned original point cloud, the spatial geometric distribution within the neighborhood of each spatial coordinate point is calculated point by point. The principal curvature value and the corresponding principal direction vector of each spatial coordinate point are extracted, and the calculated principal curvature value and principal direction vector are assigned to the corresponding spatial coordinate points in sequence to generate a curvature attribute point cloud.

[0050] Specifically, a marking robot performs a ground laser scan of the work area. This involves controlling a 3D lidar sensor (e.g., a lidar) mounted on the robot to scan the entire work area without omission, following a preset scanning path, such as an "arch" or "S" shaped path. During the scan, the lidar continuously emits laser beams and receives signals reflected from the ground. Based on the signal's time of flight or phase difference, the 3D spatial coordinates of each reflection point relative to the lidar's own coordinate system are calculated. ,in (Representing the robot coordinate system), and at the same time, these three-dimensional coordinate data and corresponding high-precision timestamp information are recorded in real time. All discrete spatial three-dimensional position coordinate points acquired during the entire scanning process are gathered together to form a spatial coordinate point set containing a large number of data points. This set is the original point cloud.

[0051] Based on the original point cloud, and using a pre-calibrated reference coordinate system within the working area as a reference, this reference coordinate system can be established by selecting at least three non-collinear fixed control points with clear physical identifiers within the working area, and using a total station or differential GPS to determine the three-dimensional coordinates of these control points within a globally unified coordinate framework. For each spatial coordinate point in the original point cloud, its initial coordinates are relative to the robot's own sensor coordinate system during scanning. To unify all points to the reference coordinate system, a coordinate transformation is required. First, based on the output data of the attitude and position sensors (such as inertial measurement units (IMU) and GPS-RTK) on the robot, combined with the sensor calibration parameters, the pose transformation matrix of the robot's sensor coordinate system relative to the predefined reference coordinate system at each scanning moment can be calculated. This matrix contains rotation and translation information, and then this transformation matrix is... Applied to each spatial coordinate point in the original point cloud Through matrix multiplication (For homogeneous coordinates, and (As a four-dimensional vector), the position of each spatial coordinate point is adjusted and transformed from the robot sensor coordinate system to the corresponding position in the predefined reference coordinate system, thereby generating the aligned original point cloud.

[0052] Based on the aligned original point cloud, a neighborhood analysis is performed point-by-point for each spatial coordinate point to calculate its local spatial geometric distribution characteristics. First, the spatial coordinate points being processed are... Define a neighborhood, which can be based on a spherical neighborhood search with a fixed radius. For example, if the average point spacing of the point cloud is 0.01 meters, the search radius can be set. The distance is 0.05 meters, which is 5 times the average point spacing, to ensure that the neighborhood contains a sufficient number of points (e.g., at least 15 to 20 points are expected) for robust geometric property estimation, or a K-nearest neighbor (KNN) search method can be used to select points at a distance from the target point. Recent There are points, among which The value can be chosen by trade-off between point cloud density and computational efficiency, for example, setting... The value is 25. After collecting the point set in the neighborhood, the covariance matrix of the 3D coordinates of these neighborhood points is calculated. Specifically, the centroid of the neighborhood points is first calculated, and then the deviation vector of each neighborhood point relative to the centroid is calculated, thereby constructing... The covariance matrix is ​​obtained by performing eigenvalue decomposition on it, yielding three eigenvalues. and their corresponding eigenvectors The principal curvature value at this point can be derived from these eigenvalues; for example, a commonly used curvature measure is... ,in A larger value indicates a higher degree of curvature or unevenness in the local area at that point, and vice versa. The principal direction value is usually selected from the eigenvectors corresponding to the direction of maximum curvature change or the direction of the most significant geometric feature, such as the eigenvector corresponding to the largest eigenvalue. Associated feature vectors As the principal direction, the principal curvature value and the corresponding principal direction vector of each spatial coordinate point calculated in this way are used as new attribute information and assigned to the corresponding spatial coordinate points in sequence, thereby generating a curvature attribute point cloud.

[0053] The steps for obtaining the candidate trace point set are as follows:

[0054] Based on the curvature attribute point cloud, the principal curvature values ​​of all spatial coordinate points in the curvature attribute point cloud are statistically analyzed, and the mean value of the principal curvature values ​​of all spatial coordinate points is calculated. The mean value is taken as the curvature filtering threshold to generate the curvature filtering threshold.

[0055] Based on the curvature filtering threshold, the principal curvature values ​​of each spatial coordinate point in the curvature attribute point cloud are traversed point by point. It is determined whether the principal curvature value of each spatial coordinate point exceeds the curvature filtering threshold. Spatial coordinate points whose principal curvature values ​​exceed the curvature filtering threshold are marked one by one to generate a set of trace candidate points.

[0056] Specifically, based on curvature attribute point clouds, the first step is to comprehensively statistically analyze the principal curvature values ​​associated with each spatial coordinate point recorded in the point cloud. This process involves reading the principal curvature values ​​of all spatial coordinate points within the curvature attribute point cloud one by one, summing all the read principal curvature values ​​to obtain a total, and simultaneously counting the total number of spatial coordinate points involved in the summation. Then, the sum of all calculated principal curvature values ​​is divided by the total number of spatial coordinate points to obtain the average value of these principal curvature values. This average value is directly adopted and set as the curvature filtering threshold to identify points with significant geometric features in subsequent steps. For example, if a curvature attribute point cloud contains 500,000 spatial coordinate points, and the principal curvature values ​​of these 500,000 points are extracted, for example, if these principal curvature values ​​mainly range from 0.05 to 0.7, and the sum of all 500,000 principal curvature values ​​is 150,000, then the average value of these principal curvature values ​​is... Therefore, in this case, the curvature screening threshold is determined to be 0.30, thus generating the curvature screening threshold.

[0057] Based on the curvature filtering threshold, the system then performs point-by-point checks and filtering on each spatial coordinate point in the curvature attribute point cloud. Specifically, the system sequentially accesses each spatial coordinate point in the curvature attribute point cloud, extracts the principal curvature value recorded at that point, and compares this principal curvature value with the curvature filtering threshold previously calculated based on the average of the principal curvature values ​​of all spatial coordinate points in the curvature attribute point cloud. It determines whether the principal curvature value of the current spatial coordinate point is strictly greater than this curvature filtering threshold. For example, if the previously calculated curvature filtering threshold is 0.30, for a spatial coordinate point with a principal curvature value of 0.45, since 0.45 is strictly greater than 0.30, this spatial coordinate point is considered... Coordinate points will be judged to meet the filtering conditions. If the principal curvature value of another spatial coordinate point is 0.25, since 0.25 is not greater than 0.30, the point does not meet the filtering conditions. For spatial coordinate points whose principal curvature value is exactly equal to 0.30, since they do not meet the strict greater than condition, they will also not be selected. All spatial coordinate points whose principal curvature value is strictly greater than the curvature filtering threshold will be marked by the system one by one. For example, a specific Boolean flag will be set to true for these points in the data structure, or their index will be added to a new list structure. All spatial coordinate points that have passed this filtering and been successfully marked are summarized together to generate a trace candidate point set.

[0058] The steps for obtaining discrete line feature clusters are as follows:

[0059] Based on the candidate trace point set, the three-dimensional coordinate value and local direction vector of each candidate trace point are extracted. The spatial position and direction relationship between each candidate trace point is constructed by traversing point by point to form a set of candidate trace point pairs.

[0060] Based on the set of candidate trace points, the comprehensive morphological similarity between each pair of candidate trace points is calculated using the following formula:

[0061] ;

[0062] in, Candidate points for traces With trace candidate points The overall morphological similarity between them Candidate points for traces With trace candidate points The Euclidean distance between them For the feature scale length, and These are candidate trace points. With trace candidate points The normalized local direction vector, This is the result of the dot product of two local direction vectors. The adjustment parameter is affected by spatial distance. Adjustment parameters are affected by directional consistency;

[0063] Based on the comprehensive morphological similarity between each pair of trace candidate points, a threshold is used to filter the similarity between all pairs of trace candidate points. The trace candidate point pairs with a comprehensive morphological similarity higher than the set threshold are used as the connecting edges of the same clustering unit. All trace candidate points that form a connection relationship are aggregated into independent subsets, and discrete line-drawing feature clusters are obtained according to the aggregation relationship.

[0064] Specifically, based on the trace candidate point set, where each trace candidate point was previously selected through principal curvature filtering and carries its original three-dimensional spatial coordinate information as well as the principal direction vector obtained when calculating the curvature properties, the stored three-dimensional coordinate values ​​are directly extracted for each trace candidate point in the trace candidate point set. And the associated local direction vector, which is the principal direction vector calculated for that point in its neighborhood, representing the main extension direction of the local geometry of that point. Next, to systematically analyze the relationships between these candidate trace points, all points in the candidate trace point set are paired, that is, for each pair of points containing... A set of candidate trace points is used to systematically generate all possible traces. An unordered pair of points, for example, if the candidate point set of traces contains points Then it will be built sequentially. , ,..., ,Then ,..., until For each such candidate pair of trace points (e.g., point...) and points Record their respective three-dimensional coordinate values. and local direction vector This combination process is essentially to prepare basic data for subsequent calculation of the similarity between point pairs. By traversing all possible point pairs and recording their spatial location and orientation information, a set of candidate point pairs for traces is formed.

[0065] formula: The advantage of this formula lies in its ability to effectively measure both the spatial proximity and directional consistency of candidate trace point pairs. This is crucial for accurately identifying lines with linear extension characteristics from point clouds. The exponential decay term penalizes distance, ensuring that only spatially close points are considered similar. The directional consistency term further filters out points with approximately parallel local directions. The combination of these two methods effectively eliminates irrelevant noise points and enhances the ability to identify genuine trace segments. (Parameters...) and The introduction of this feature allows for flexible adjustment of the relative importance of distance and orientation factors in similarity calculation based on the characteristics of the actual line drawing and the quality of the point cloud data, thereby improving the robustness and accuracy of line drawing feature extraction.

[0066] parameter The steps to obtain it are as follows:

[0067] Indicate trace candidate points With trace candidate points The Euclidean distance between the two points is calculated directly from their coordinates in three-dimensional space. Points are extracted from the candidate trace point pair set constructed in the previous step. 3D coordinates and points 3D coordinates Then, the standard Euclidean distance formula is applied to calculate it. For example, if point... Coordinates are Rice, dot Coordinates are Rice, then rice.

[0068] parameter The steps to obtain it are as follows:

[0069] This parameter, representing the feature scale length, defines the characteristic distance at which points are considered "nearest neighbors" within a local area. Its setting is based on the typical physical dimensions of the lines to be identified within the working area, such as the average width of the lines or the typical point spacing of the laser-scanned point cloud on such targets. Geometric feature parameters are obtained through preliminary research and measurement of the expected line type within the working area of ​​the line-marking robot. For example, if the target line is a common white solid lane line on urban roads, its standard width is typically 0.10 to 0.15 meters. Considering the density of the scanned point cloud and the need to capture the continuity of line segments, this parameter is applied. The feature scale length can be set to a value similar to or slightly larger than this width to ensure that neighboring points forming the same line segment can be effectively associated. In this embodiment, by measuring multiple line samples, the average feature width of the line is approximately 0.12 meters. Considering the point cloud density, to ensure sufficient points participate in the calculation without excessively smoothing details, the feature scale length is... It was set to 0.10 meters.

[0070] parameter and The steps to obtain it are as follows:

[0071] and These represent candidate trace points. With trace candidate points The normalized local direction vectors are derived from the principal direction vectors calculated for each original point cloud point in the preceding steps. When a point is selected as a trace candidate point, its principal direction vector information is retained. This principal direction vector describes the point's... or point Before substituting these directions into the formula, it is necessary to ensure that the local direction vectors are normalized, i.e., their magnitude is 1. If the original principal direction vector is... Then its normalized form Calculated as For example, point The local direction vector is Its module length is After normalization ,point The local direction vector is Its module length is After normalization .

[0072] parameter The steps to obtain it are as follows:

[0073] This is a parameter for adjusting the spatial distance, used to control the spatial distance. Similarity The decay rate of the influence; this parameter is a positive real number, and its value is set to ensure that the distance exceeds a certain multiple of the feature scale length. The similarity between point pairs is significantly reduced to enhance local connectivity. The value is typically determined through experimental tuning on sample point cloud data containing known line features. The goal is to find a balance that connects neighboring points on the same line while effectively distinguishing between different lines or noise points. For example, if we want the distance between two points to reach twice the feature scale length (i.e., ... When ), the contribution of the spatial distance term decay to Then at this time The approximation is equal to Thus, it is deduced that In this embodiment, after multiple experiments on the test dataset, it was found that when When set to 1.0, it can better distinguish between points inside and outside the reticle, therefore it is set... .

[0074] parameter The steps to obtain it are as follows:

[0075] The parameter representing the influence of orientation consistency is used to adjust the contribution of the consistency of local orientation vectors between two trace candidate points to the overall similarity. This parameter is a positive real number. The larger the value, the heavier the penalty for deviation in direction, and the higher the requirement for directional consistency in similarity. Its setting also depends on the analysis and experimental testing of actual line-drawing data features. If the line-drawing features have high directional continuity, a larger value should be selected. The value is positive; conversely, if the line allows for a certain degree of curvature or directional change, then... The value can be appropriately reduced; for example, when it is necessary to evaluate the angle between two direction vectors. At that time, the direction consistency item The attenuation situation, if it is required that the angle between the directions is... ( When ), the value of the direction consistency term is not less than 0.85, that is... ,like The result is 0.933; if The result is If the requirements are met, The result is However, this approach would be too strict. In this embodiment, considering that the actual lines may have slight bends, to ensure the robustness of the connection and to penalize obvious directional inconsistencies, a [specific method] is chosen. .

[0076] Calculation process:

[0077] With a pair of trace candidate points and Let's take an example to calculate:

[0078] point 3D coordinates Meters, their normalized local direction vectors .

[0079] point 3D coordinates Meters, their normalized local direction vectors .

[0080] The parameter values ​​are set as follows:

[0081] Feature Scale Length rice.

[0082] Spatial distance affects adjustment parameters .

[0083] Directional consistency affects adjustment parameters .

[0084] First, calculate the Euclidean distance between the two points. :

[0085] ;

[0086] ;

[0087] rice.

[0088] Then, calculate the dot product of the two local direction vectors. :

[0089] ;

[0090] .

[0091] Next, calculate the spatial distance term:

[0092] ;

[0093] .

[0094] Then, calculate the direction consistency term:

[0095] ;

[0096] .

[0097] Finally, calculate the overall morphological similarity. :

[0098] .

[0099] This result indicates candidate trace points. With trace candidate points Comprehensive morphological similarity between The similarity score is approximately 0.73338. This value falls between 0 and 1. Higher values ​​(closer to 1) indicate that the two points are very similar in both spatial location and local direction, while lower values ​​(closer to 0) indicate poor similarity. The calculated similarity score of 0.73338 is relatively high, indicating that the points... and points They are very likely to belong to the same segment of line features.

[0100] Based on the comprehensive morphological similarity between each pair of trace candidate points calculated from the set of trace candidate point pairs. All these calculated similarity values ​​are then subjected to a unified threshold screening. This "threshold setting," or comprehensive morphological similarity threshold, is based on the known pairs of crisscrossing feature points and non-feature point pairs in a large dataset. The values ​​were obtained through analysis of their distribution characteristics; specifically, a series of typical line-drawing scenarios were collected and calculated. Values ​​are plotted and histograms are drawn to observe whether they can effectively distinguish between the internal connections of the lines and noise or the connections between different lines. Value range, for example, in testing, it was found that pairs of points belonging to the same clearly defined line, their The values ​​are typically distributed between 0.6 and 0.9, while random noise pairs or pairs of points belonging to different lines have different values. Since most values ​​are below 0.4, to maximize precision while maintaining recall, the overall morphological similarity threshold can be set in the valley region of this distribution or fine-tuned based on the expected clustering effect, for example, set to 0.55. Any value with an overall morphological similarity of 0.4 or higher... Candidate trace point pairs (points) with values ​​higher than this set threshold (e.g., 0.55) and points If they are strongly correlated, they are marked as belonging to the same potential line segment. Conceptually, this is equivalent to establishing a connecting edge in a graph with trace candidate points as nodes and similarity between them above a threshold as edges. Subsequently, all trace candidate points that are connected in this way are aggregated, that is, all connected components in the graph are searched. All trace candidate points in each connected component together form an independent subset. According to this aggregation relationship, a series of discrete sets of trace candidate points with high morphological similarity are finally obtained, namely discrete line feature clusters.

[0101] The steps to obtain the path diagram by drawing lines are as follows:

[0102] Based on discrete scribing feature clusters, the three-dimensional spatial coordinates of all trace candidate points within each discrete scribing feature cluster are traversed one by one. The average value of the spatial coordinates of all trace candidate points is taken and the average value is defined as the center position of the discrete scribing feature cluster, thereby generating a set of center positions of discrete scribing feature clusters.

[0103] Based on the set of center positions of discrete line-drawing feature clusters, all candidate trace points of discrete line-drawing feature clusters are traversed one by one. Principal component analysis is used to extract the principal component feature vector of each discrete line-drawing feature cluster. The principal component feature vector is defined as the principal direction vector of the discrete line-drawing feature cluster, and the set of principal direction vectors of discrete line-drawing feature clusters is generated.

[0104] Based on the set of center positions of discrete line feature clusters and the set of main direction vectors of discrete line feature clusters, the spatial distance and the angle between the main directions of any two discrete line feature clusters are compared one by one. Discrete line feature clusters whose spatial distance is less than a preset spatial threshold and whose main direction angle is less than a preset angle threshold are paired and connected to form a connection relationship and defined as a connection path, thus generating a line connection path map.

[0105] Specifically, based on discrete scribing feature clusters, the following operations are performed for each discrete scribing feature cluster: First, the system accesses all candidate trace points contained within the discrete scribing feature cluster and extracts the three-dimensional spatial coordinates of these candidate trace points one by one. If a discrete line feature cluster contains There are 10 candidate trace points, with coordinates as follows: The system will then use these spatial coordinates , , Each component is summed separately, and then each sum is divided by the total number of candidate trace points. In this way, the discrete line feature cluster is calculated. , , The average coordinates across the three dimensions, i.e., the center position of the cluster. , , For example, a discrete line feature cluster contains three candidate trace points with coordinates P1(1.0, 2.0, 0.5), P2(1.1, 2.1, 0.6), and P3(1.2, 2.0, 0.4). Then the x-coordinate of its center position is... The y-coordinate is The z-coordinate is Therefore, the center position of the discrete line feature cluster is (1.1, 2.033, 0.5). The average three-dimensional spatial coordinates obtained by this calculation are defined as the center position of the discrete line feature cluster. The center positions of all discrete line feature clusters are organized to generate a set of discrete line feature cluster center positions.

[0106] Based on the discrete scribing feature clusters generated in the previous steps, the system processes each discrete scribing feature cluster one by one to determine its principal direction. For the currently processed discrete scribing feature cluster, the system first collects the three-dimensional spatial coordinates of all trace candidate points contained within it. Using the coordinate data of these points, the system employs principal component analysis (PCA) to extract the principal direction of the point set. Specifically, when implementing PCA, the system first calculates the covariance matrix of the three-dimensional coordinates of all trace candidate points within the discrete scribing feature cluster. This covariance matrix describes the degree of dispersion of the point set in various directions. Then, the system performs eigenvalue decomposition on the covariance matrix to obtain three eigenvalues ​​and their corresponding eigenvectors. These eigenvectors are mutually orthogonal and point in the direction of the largest data variance. The eigenvector associated with the largest eigenvalue, i.e., the first principal component, represents the most important direction of the data point set distribution, which is the main trend direction of the discrete scribing feature cluster extending in space. This first principal component eigenvector (usually a three-dimensional unit vector) is defined as the principal direction vector of the discrete scribing feature cluster. This process is repeated for all discrete scribing feature clusters, and the calculated principal direction vectors are collected to generate a set of principal direction vectors for discrete scribing feature clusters.

[0107] Based on the set of discrete line-marking feature cluster center positions and the set of discrete line-marking feature cluster principal direction vectors obtained in previous steps, the system begins to perform pairwise comparisons on any two different discrete line-marking feature clusters to determine whether there is a connection between them. For any two selected discrete line-marking feature clusters, such as cluster A and cluster B, the system first extracts their respective center position coordinates from the set of discrete line-marking feature cluster center positions. and And calculate the Euclidean distance between the two center locations. Simultaneously, the principal direction vectors of each discrete line feature cluster are extracted from the set of principal direction vectors. and And calculate the angle between the two principal direction vectors. Next, the calculated spatial distance will be... Compare with a preset spatial threshold and set the angle of the principal direction. Compared to a preset angle threshold, the "preset spatial threshold" is set based on the continuity characteristics and point cloud density of the actual scribing. For example, if the maximum allowable break length between scribing segments is 0.5 meters, the spatial threshold can be set to 0.5 meters, or slightly larger, such as 0.6 meters, to tolerate some measurement error and data sparsity. The "preset angle threshold," on the other hand, is set based on the allowable curvature of the scribing. For example, for relatively straight scribing lines, the angle threshold can be set smaller, such as 15 degrees (approximately 0.26 radians), to ensure good directional consistency among connected clusters. If the spatial distance... Less than a preset space threshold (e.g.) (meters) and the included angle of the main direction Less than a preset angle threshold (e.g.) If the two discrete line-drawing feature clusters are considered to have a potential connection, the system will establish a connection edge between the nodes representing the two clusters and define this connection relationship as a connection path. By repeating this comparison and connection process for all possible discrete line-drawing feature cluster pairs, a network structure containing all connected discrete line-drawing feature clusters and their connection paths will be finally constructed, generating a line-drawing connection path graph.

[0108] The steps to obtain the line vectorized structure are as follows:

[0109] Based on the line-connection path graph, all discrete line feature clusters in each connection path of the line-connection path graph are traversed one by one. The three-dimensional spatial coordinates of all trace candidate points contained in each discrete line feature cluster are extracted. All three-dimensional spatial coordinates of each discrete line feature cluster are summarized to generate a set of three-dimensional spatial coordinates of the connection path.

[0110] Based on the set of three-dimensional spatial coordinates of the connection path, linear fitting of the spatial point cloud is performed for each connection path. The direction vector and center position of the fitted line segment are calculated by the least squares method. The fitted line segment parameters of the connection path are obtained one by one, and a set of fitted line segment parameters of the connection path is generated.

[0111] Based on the set of parameters of the fitted line segments along the connection path, the parameters of the fitted line segments are converted into a standard vectorized representation format one by one. The three-dimensional coordinates and vector directions of the start and end points of the fitted line segments are recorded in sequence to construct a line vectorization structure.

[0112] Specifically, based on the scribbled connection path graph, each connection path in the graph is processed independently. First, the system traverses the currently processed connection path and identifies all discrete scribbled feature clusters that constitute the path. For each discrete scribbled feature cluster within the path, the system further extracts the three-dimensional spatial coordinates of all trace candidate points contained in that cluster. For example, a connection path may be formed by sequentially connecting discrete scribing feature clusters A, B, and C. The system will first collect the coordinates of all trace candidate points in discrete scribing feature cluster A, then collect the coordinates of all trace candidate points in discrete scribing feature cluster B, and finally collect the coordinates of all trace candidate points in discrete scribing feature cluster C. All three-dimensional spatial coordinates extracted from all discrete scribing feature clusters in the connection path will be merged to form a large point set containing the coordinates of all relevant trace candidate points for the current connection path. This point set is the three-dimensional spatial coordinate set of the connection path.

[0113] Based on the set of 3D spatial coordinates of the connecting paths, a linear fitting operation of the spatial point cloud is performed for each set of 3D spatial coordinates corresponding to the connecting path. Specifically, for a specific connecting path and its corresponding set of 3D spatial coordinates, the least squares method is used to determine a 3D straight line that best fits these spatial points. The goal of the least squares method is to find a straight line that minimizes the sum of the squares of the perpendicular distances from all points to the line. In 3D space, a straight line can be parameterized by a point (e.g., the center of the line or any point on the line) and the direction vector of the line. By applying the least squares algorithm (e.g., ...) to the 3D coordinate data of the point set... The covariance matrix of the decentralized point set can be processed by Singular Value Decomposition (SVD). The eigenvector corresponding to the largest eigenvalue is the direction vector of the fitted line, and the average coordinates of all points is the center position of the fitted line. The direction vector and center position of the best-fit line segment can be calculated. This direction vector describes the overall direction of the connection path, and the center position marks the approximate location of the path in space. This linear fitting process is repeated for each connection path in the line connection path graph. The fitted line segment parameters of each connection path are obtained one by one, including its direction vector and center position. These parameters are then organized to generate a set of fitted line segment parameters for the connection path.

[0114] Based on the set of fitted line segment parameters for each connecting path, which includes the direction vector and center position obtained by fitting each connecting path using the least squares method, each fitted line segment parameter in this set is processed individually to transform it into a standardized vectorized representation format. This standard format typically requires explicitly recording the three-dimensional coordinates of the start and end points of the line segment, as well as the possible line segment directions (although sometimes these can be calculated from the start and end points). For each fitted line segment parameter (i.e., direction vector) of the connecting path... and central position First, it is necessary to determine the actual coverage area of ​​the line segment on the original point cloud data. This can be achieved by projecting the candidate points of the original traces contained in the connecting path onto the fitted line, and then finding the farthest two ends of the projection point distribution on the fitted line. The three-dimensional coordinates of these two farthest endpoints are defined as the three-dimensional coordinates of the start and end points of the fitted line segment. For example, if the direction vector of the fitted line of a connecting path is... The central position is By analyzing the projection of points in the three-dimensional spatial coordinate set of the corresponding connection path onto the fitted straight line, the coordinates of its starting point can be determined. and endpoint coordinates Meanwhile, the vector direction of this line segment can be directly obtained by fitting the direction vector. , or by The calculated and normalized information (three-dimensional coordinates of the start and end points, vector direction) is recorded sequentially according to a predetermined data structure or file format (e.g., commonly used GIS vector data format or custom format). This conversion and recording process is repeated for all connected paths, and finally, the overall vectorized representation of the line is constructed, that is, the line vectorization structure.

[0115] The steps to obtain the initial complexity grid are as follows:

[0116] Based on the curvature attribute point cloud, the boundary range and spatial size parameters of the working area of ​​the line-marking robot are set. The entire working area of ​​the line-marking robot is divided into regular cubic cells according to a uniform side length. Each curvature attribute point in the curvature attribute point cloud is mapped and assigned to the cubic cell according to its three-dimensional position coordinates, thus generating the initial set of cells for the working area of ​​the line-marking robot.

[0117] Based on the initial set of cells in the work area of ​​the marking robot, the coupling morphological complexity of each initial cell is calculated using the following formula:

[0118] ;

[0119] in, For the first The coupling morphological complexity of an initial cell, For the first Determinant of the orientation structure tensor of the initial cell. For the first The maximum eigenvalue of the initial cell orientation structure tensor For the first The average curvature value of all curvature attribute points within an initial cell. The characteristic curvature scale constant, The normal distribution response exponent, The curvature response index, , For the first The orientation structure tensor of the initial cell For the first The number of curvature attribute points contained in each initial cell. For the first The first initial cell The unit normal direction vector of each curvature attribute point This is the transpose of the unit normal direction vector. The matrix product of the unit normal direction vector and its transpose is used to represent its directional contribution tensor.

[0120] Based on the coupled morphological complexity of each initial cell, the entire set of initial cells in the working area of ​​the line-marking robot is traversed. The spatial position index and coupled morphological complexity of all initial cells are paired and encoded one by one, and a unified organization is formed to form a complexity expression grid of spatial structure changes within the region, thus generating the initial complexity grid.

[0121] Specifically, based on curvature attribute point clouds, the first step is to accurately define the boundary of the entire 3D working area where the marking robot performs its tasks. This is typically achieved by using minimum and maximum coordinates in a global coordinate system. Coordinates are used to define areas; for example, a work area is defined as... rice, rice, The unit length is set to 0.25 meters, which refers to the uniform side length of the regular cubic cells used for subsequent division. The selection of this side length requires a balance between computational efficiency and detail capture capability. For example, if the average density of the point cloud is high and fine terrain changes need to be identified, a smaller side length, such as 0.2 meters, can be chosen. If the point cloud is sparse or the focus is mainly on macroscopic terrain, a larger side length, such as 0.5 meters, can be chosen. In this embodiment, based on the size of the typical marking area and the density of the laser-scanned point cloud, the uniform side length is set to 0.25 meters. Next, according to this uniform side length, the entire defined marking robot working area is divided into a series of closely arranged, non-overlapping regular cubic cells in three-dimensional space. For each curvature attribute point in the curvature attribute point cloud, its three-dimensional position coordinates in the reference coordinate system are extracted. The coordinates are compared with the spatial range of each cube cell, and then precisely mapped and assigned to the unique cube cell to which they belong. This mapping process can be achieved through simple integer division operations; for example, if the cell side length is... The starting point of the work area is Then point Index of the cell It can be done , , Calculations show that after all curvature attribute points are assigned, all cube cells containing at least one curvature attribute point together constitute the initial cell set of the work area of ​​the marking robot.

[0122] formula: The advantage of the formula lies in its low coupling complexity. By comprehensively considering the anisotropy of the point cloud normal distribution and the magnitude of the average curvature within a cell, it can more comprehensively characterize the geometric complexity of local regions. The first term... This reflects the consistency of the distribution of normal directions: when the normal directions are uniformly distributed (e.g., on a plane) or linearly distributed, this term approaches 0; when the normal directions are isotropically distributed (e.g., at corners or on surfaces with irregular heights), this term approaches 1. The second term... This is directly related to the degree of surface curvature; the greater the average curvature, the larger this value. This information is then combined with adjustable parameters. and By combining these methods, areas containing significant topographical changes or complex geometric features can be identified;

[0123] parameter The steps to obtain it are as follows:

[0124] For the first The initial cell's orientation structure tensor is calculated using the following formula: ,in It is the first The number of curvature attribute points contained in each initial cell is determined by iterating through the initial cell set of the work area of ​​the marking robot and checking each cell. The curvature attribute points within the cell are counted. For example, if the cell... If it contains 20 curvature attribute points, then ,parameter It is the first The first initial cell The unit normal direction vector of a curvature attribute point, for a cell Each curvature attribute point within Its unit normal vector Specifically, it is estimated by analyzing the local neighborhood geometry of the curvature attribute point cloud, for example, the point... Select its three-dimensional space nearest neighbors (e.g.) (Searching from the entire curvature attribute point cloud) for this Points (including) Principal component analysis (PCA) is performed on the three-dimensional coordinates of the local point set. The eigenvector corresponding to the smallest eigenvalue is the normal direction of that local point set. After normalization, we obtain... , yes The product of the matrix with its own transpose (outer product) represents its directional contribution tensor, which is a A symmetric matrix, which will contain cells All Summing the directional contribution tensors of each point and then averaging them yields the result. .

[0125] parameter The steps to obtain it are as follows:

[0126] For the first The orientation structure tensor of the initial cell The determinant, after calculation, is obtained matrix Then, its determinant can be directly obtained using the standard determinant calculation formula. For example, if ,but .

[0127] parameter The steps to obtain it are as follows:

[0128] For the first Initial cell orientation structure tensor The largest eigenvalue, obtained after calculation Then, by solving its characteristic equation (in It is the identity matrix. (These are eigenvalues) to obtain three eigenvalues. The largest of them is .

[0129] parameter The steps to obtain it are as follows:

[0130] For the first The average curvature value of all curvature attribute points within an initial cell, for each cell Each curvature attribute point within (common (points), extract the calculated principal curvature values ​​from the curvature attribute point cloud. Then calculate the average of these curvature values, for example, if the cell There are 3 points with principal curvature values ​​of 0.2, 0.3, and 0.4. Then... .

[0131] parameter The steps to obtain it are as follows:

[0132] This is a characteristic curvature scale constant; this parameter is used to measure the mean curvature. Normalization is performed, and the value should be set according to the curvature range of typical features or demarcation characteristics within the working area. By analyzing known scene data, the distribution of curvature values ​​of features of interest such as demarcation edges and road surface undulations is statistically analyzed, and a representative curvature value is selected as... For example, if analysis shows that the curvature values ​​of significant topographic changes are typically between 0.1 and 0.8, then... The value is set to the median or mean of this range, such as 0.4. In this embodiment, through statistical analysis of sample point clouds containing various road surface features, it was found that the average principal curvature value of the marked edges and areas of minor road surface damage is usually around 0.35. Therefore, the value is set to... .

[0133] parameter The steps to obtain it are as follows:

[0134] The normal distribution response exponent is a positive real number used to adjust the sensitivity of the coupled morphological complexity to the degree of anisotropy of the normal distribution. Its value is determined experimentally on multiple sample cells containing different normal distribution features (such as planes, edges, and corners). Items, and adjustments The value is used to observe its contribution to distinguishing these features. The goal is to make this value effectively amplify the influence of complex normal distributions. Generally, the value ranges from 0.5 to 2.0. In this embodiment, through testing, when At that time, it can better reflect the complexity of the normal distribution, therefore, it is set .

[0135] parameter The steps to obtain it are as follows:

[0136] The curvature response exponent is a positive real number used to adjust the sensitivity of the coupling morphological complexity to the average curvature magnitude. Its setting method is similar to... Similarly, experiments were conducted on sample cells with different mean curvatures, and adjustments were made. Value, such that the curvature term To appropriately reflect the complexity of high curvature regions, the value is generally between 0.5 and 2.0. In this embodiment, to ensure that the complexity responds appropriately to changes in the average curvature, a value is set... .

[0137] Calculation process:

[0138] Starting with an initial cell Taking this as an example, the cell contains... One curvature attribute point.

[0139] Their unit normal direction vectors are as follows: , , .

[0140] The curvature values ​​of these points are as follows: , , .

[0141] The default parameters are: , , .

[0142] Calculate the directional structure tensor :

[0143] , , .

[0144] .

[0145] calculate and :

[0146] .

[0147] because It is a diagonal matrix whose eigenvalues ​​are the diagonal elements: .

[0148] therefore, .

[0149] Calculate the normal distribution term:

[0150] .

[0151] Calculate the mean curvature :

[0152] .

[0153] Calculate the curvature term:

[0154] .

[0155] Calculate the morphological complexity of coupling :

[0156] .

[0157] The result indicates that the first The coupling morphological complexity of the initial cells Approximately 5.196. A higher value indicates a more complex local region represented by the cell in terms of normal distribution and average curvature. For example, a cell containing sharp edges or corners and with a sharply curved surface... The value will be relatively high, while a cell representing a flat, uniform surface will have a lower value. The value will be lower (for example, if the normal distribution term is 0, the curvature term is 0). ,but ).

[0158] Based on the previously calculated coupling morphological complexity of each initial cell The system will perform a complete traversal of the initial set of cells in the entire work area of ​​the marking robot. During the traversal, for each initial cell, the system will record its unique spatial position index in the 3D mesh. For example, if the entire work area is divided into one... In a three-dimensional raster, each cell can be accessed through its integer coordinates within the raster. (in , , The system uses this information as a unique identifier and also reads the calculated coupling morphological complexity value for that cell. Then, index this spatial location. Its corresponding coupling morphological complexity value One-to-one pairing is performed, and this pairing information is stored. For example, a data structure can be constructed in which each entry contains the cell index and its complexity value. In this way, all initial cells and their respective coupled morphological complexities are uniformly organized to form a complexity expression grid that can reflect the degree of spatial structure change within the region. This grid is actually based on the initial regular cube cells, with each cell assigned an attribute value that quantifies its geometric complexity, ultimately generating the initial complexity grid.

[0159] The steps for obtaining an adaptive terrain mesh model are as follows:

[0160] Based on the initial complexity grid, the coupling morphological complexity of each initial cell in the initial complexity grid is traversed one by one, and the values ​​are compared with the pre-set coupling morphological complexity threshold. At the same time, it is checked whether the spatial position of each initial cell has a spatial geometric intersection relationship with the line vectorized structure. Initial cells that meet the conditions of coupling morphological complexity exceeding the threshold or having an intersection relationship are selected to generate a set of cells to be subdivided.

[0161] Based on the set of cells to be subdivided, each cell is evenly divided along the midpoint of the three-dimensional coordinate axis to form eight smaller sub-cells. The three-dimensional spatial position and boundary information of all sub-cells are recorded one by one to generate the first recursive subdivision set of cells.

[0162] Based on the first recursive subdivision of the cell set, the steps of obtaining the cell set to be subdivided and obtaining the first recursive subdivision of the cell set are repeated until the coupling morphological complexity of all cells meets the set threshold and no longer intersects with the line vectorized structure. Then, all cells are connected by grid topology to generate an adaptive terrain grid model.

[0163] Specifically, based on the initial complexity grid, the system will check and evaluate each initial cell in the grid one by one. For the currently traversed initial cell, the previously calculated coupling morphological complexity value will be extracted first. Then, this value is compared with a "pre-set coupling morphological complexity threshold." This threshold is set based on statistical analysis of the distribution of initial cell coupling morphological complexity under a large number of different scenarios, as well as a trade-off between the accuracy and computational efficiency of the final mesh model. Specifically, by analyzing the expression effect of the mesh model generated under different complexity thresholds on known terrain features, a critical value that can effectively distinguish between areas that need refinement (i.e., complex terrain or containing important features) and areas that do not need refinement (i.e., relatively flat and simple terrain) can be selected. For example, if statistics show that most flat areas... A value below 2.0, but containing areas with significant topographical changes or demarcated structures. If the value is typically higher than 3.5, the coupling morphological complexity threshold can be set to 3.0. Simultaneously with this threshold, the system also checks whether the initial cell's spatial position has a spatial geometric intersection with the previously constructed line vectorization structure. This detection process includes checking whether the cell's bounding box intersects with any line segment (defined by the start and end points) in the line vectorization structure. If the coupling morphological complexity value of an initial cell is... Strictly greater than a pre-set threshold for coupling morphological complexity (e.g.) If the initial cell's spatial range intersects or overlaps geometrically with any line segment in the line vectorization structure, then the initial cell is determined to need further refinement. All initial cells that meet one of these two conditions (i.e., complexity exceeds the threshold or intersects with the line) will be filtered out and summarized to form a set of cells to be subdivided.

[0164] Based on the set of cells to be subdivided, each initial cell marked as to be subdivided in the set is processed. Specifically, the cell to be subdivided is moved along its coordinates in three-dimensional space. , , The midpoints of the three coordinate axes are evenly divided, for example, a side with a length of... The cube cell, which is in The range of the axial direction is (in ), then its The position of the midpoint of the axis is Similarly, calculate shaft and The midpoints of the axes, through which three mutually orthogonal planes are formed, precisely divide the original cell to be subdivided into eight identical cubic cells. The side length of each of these eight sub-cells is half the side length of the original cell. For each newly generated subcell, the system records its position in three-dimensional space (e.g., by its minimum corner coordinates or center coordinates) and its new boundary range (i.e., the minimum and maximum values ​​on each coordinate axis). All subcells generated after such an octree partitioning of all cells to be subdivided are collected to form the first recursive subdivided cell set.

[0165] Based on the first recursive subdivision of the cell set, the system will initiate an iterative refinement process. The core of this process is to repeatedly execute the same steps as generating the cell set to be subdivided and generating the first recursive subdivision cell set. Specifically, for each sub-cell in the first recursive subdivision cell set, the coupled morphological complexity of its internal curvature attribute points will be recalculated (if the sub-cell still contains curvature attribute points), and its coupled morphological complexity will be checked again to see if it exceeds a pre-set coupled morphological complexity threshold (this threshold remains unchanged during the recursive process, for example, still 3.0). Simultaneously, the spatial position of the sub-cell will be re-checked to see if it corresponds to the line vectorization structure. In spatial geometric intersection relationships, if a sub-cell satisfies one of these two conditions, it will be added to a new "set of cells to be subdivided". Then, each cell in this new set of cells to be subdivided is again octree-partitioned to generate a smaller set of sub-cells. This iterative process of "computational complexity, checking intersections, filtering cells to be subdivided, and subdividing cells to be subdivided" continues. Each iteration further refines the areas that need to be subdivided until the following two termination conditions are met: First, the coupling morphological complexity of all cells (regardless of their size) no longer exceeds a pre-set coupling morphological complexity threshold (e.g., all cells' coupling morphological complexity is less than or equal to the pre-set threshold). Second, when the spatial positions of all cells no longer geometrically intersect with the vectorized line structure, the recursive subdivision process stops. At this point, all cells in the system (including the initial cells that have not been subdivided and the sub-cells that have reached the final size after multiple levels of subdivision) together constitute the final cell set. Finally, mesh topology connections are made between all cells in this final cell set. For example, the MarchingCubes algorithm can be used to generate triangular patches connecting adjacent cells based on whether each cell corner is inside the object (by checking whether there are curvature attribute points or their relationship with the vectorized line structure). This constructs a three-dimensional mesh model that can adaptively reflect the terrain complexity and accurately express the line position, i.e., an adaptive terrain mesh model.

Claims

1. A three-dimensional environment reconstruction system for the working area of ​​a line-marking robot, characterized in that, The system includes: The point cloud attribute calculation module performs ground scanning on the working area of ​​the line drawing robot to obtain the original point cloud, performs coordinate system alignment on the original point cloud, traverses all data points and calculates the principal curvature value and principal direction value of each data point to establish a curvature attribute point cloud. The line drawing feature recognition module, based on the curvature attribute point cloud, sets a curvature filtering threshold and traverses all data points, filters out points whose main curvature value is greater than the curvature filtering threshold, and obtains a trace candidate point set. Based on the trace candidate point set, a discrete line drawing feature cluster is established. The trace network construction module calculates the center position and main direction of each cluster based on the discrete line feature clusters, connects clusters whose spatial distance is less than a threshold and whose main directions are consistent, generates a line connection path graph, and fits and establishes a line vectorization structure based on the line connection path graph. The multi-resolution mesh generation module divides the working area of ​​the line-drawing robot into initial cells based on the curvature attribute point cloud, establishes an initial complexity mesh, and obtains an adaptive terrain mesh model based on the initial complexity mesh. The steps for obtaining the adaptive terrain mesh model are as follows: Based on the initial complexity grid, the coupling morphological complexity of each initial cell in the initial complexity grid is traversed one by one, and the values ​​are compared with the pre-set coupling morphological complexity threshold. At the same time, it is detected whether the spatial position of each initial cell has a spatial geometric intersection relationship with the line vectorization structure. Initial cells that meet the conditions of coupling morphological complexity exceeding the threshold or having an intersection relationship are selected to generate a set of cells to be subdivided. Based on the set of cells to be subdivided, each cell to be subdivided is evenly divided along the midpoint of the three-dimensional spatial coordinate axis to form eight smaller sub-cells, and the three-dimensional spatial position and boundary information of all sub-cells are recorded one by one to generate the first recursive subdivided cell set. Based on the first recursive subdivision cell set, the steps of obtaining the cell set to be subdivided and obtaining the first recursive subdivision cell set are repeated until the coupling morphological complexity of all cells meets the set threshold and no longer intersects with the line vectorized structure. Then, all cells are connected by grid topology to generate an adaptive terrain grid model.

2. The three-dimensional environment reconstruction system for the working area of ​​the marking robot according to claim 1, characterized in that, The steps for obtaining the curvature attribute point cloud are as follows: The work area is scanned by ground laser using a line-marking robot, and the spatial three-dimensional position coordinates of the area are collected and recorded to construct a set of spatial coordinate points and generate an original point cloud. Based on the original point cloud, and with a predefined reference coordinate system as a reference, the position of each spatial coordinate point is adjusted to the corresponding position under the reference coordinate system by performing coordinate transformation on the set of spatial coordinate points one by one, thereby generating an aligned original point cloud. Based on the aligned original point cloud, the spatial geometric distribution within the neighborhood of each spatial coordinate point is calculated point by point. The principal curvature value and the corresponding principal direction vector of each spatial coordinate point are extracted, and the calculated principal curvature value and principal direction vector are assigned to the corresponding spatial coordinate points in sequence to generate a curvature attribute point cloud.

3. The three-dimensional environment reconstruction system for the working area of ​​the marking robot according to claim 1, characterized in that, The steps for obtaining the candidate trace point set are as follows: Based on the curvature attribute point cloud, the principal curvature values ​​of all spatial coordinate points in the curvature attribute point cloud are statistically analyzed, and the mean value of the principal curvature values ​​of all spatial coordinate points is calculated. The mean value is taken as the curvature filtering threshold to generate the curvature filtering threshold. Based on the curvature filtering threshold, the principal curvature values ​​of each spatial coordinate point in the curvature attribute point cloud are traversed point by point. It is determined whether the principal curvature values ​​of each spatial coordinate point exceed the curvature filtering threshold. Spatial coordinate points whose principal curvature values ​​exceed the curvature filtering threshold are marked one by one to generate a trace candidate point set.

4. The three-dimensional environment reconstruction system for the working area of ​​the marking robot according to claim 1, characterized in that, The steps for obtaining the discrete line-drawing feature cluster are as follows: Based on the set of candidate trace points, the three-dimensional coordinates and local direction vector of each candidate trace point are extracted, and the spatial position and direction relationship between each candidate trace point is constructed by traversing point by point to form a set of candidate trace point pairs. Based on the set of candidate trace points, calculate the comprehensive morphological similarity between each pair of candidate trace points; Based on the comprehensive morphological similarity between each pair of trace candidate points, a threshold is used to filter the similarity between all pairs of trace candidate points. The trace candidate point pairs with a comprehensive morphological similarity higher than the set threshold are used as the connecting edges of the same clustering unit. All trace candidate points that form a connection relationship are aggregated into independent subsets, and discrete line-drawing feature clusters are obtained according to the aggregation relationship.

5. The three-dimensional environment reconstruction system for the working area of ​​the marking robot according to claim 1, characterized in that, The steps for obtaining the line-connected path diagram are as follows: Based on the discrete scribing feature cluster, the three-dimensional spatial coordinates of all trace candidate points in each discrete scribing feature cluster are traversed one by one. The average value of the spatial coordinates of all trace candidate points is taken and the average value is defined as the center position of the discrete scribing feature cluster, thereby generating a set of center positions of discrete scribing feature clusters. Based on the set of center positions of the discrete line-drawing feature clusters, all candidate trace points of the discrete line-drawing feature clusters are traversed one by one. Principal component analysis is used to extract the principal component feature vector of each discrete line-drawing feature cluster. The principal component feature vector is defined as the principal direction vector of the discrete line-drawing feature cluster, and a set of principal direction vectors of discrete line-drawing feature clusters is generated. Based on the set of center positions of the discrete line feature clusters and the set of main direction vectors of the discrete line feature clusters, the spatial distance and the angle between the main directions of any two discrete line feature clusters are compared one by one. Discrete line feature clusters with spatial distance values ​​less than a preset spatial threshold and main direction angles less than a preset angle threshold are paired and connected to form a connection relationship and defined as a connection path, generating a line connection path map.

6. The three-dimensional environment reconstruction system for the working area of ​​the marking robot according to claim 1, characterized in that, The steps for obtaining the line vectorization structure are as follows: Based on the line connection path map, all discrete line feature clusters in each connection path in the line connection path map are traversed one by one. The three-dimensional spatial coordinates of all trace candidate points contained in each discrete line feature cluster are extracted. All three-dimensional spatial coordinates of each discrete line feature cluster are summarized to generate a set of three-dimensional spatial coordinates of the connection path. Based on the set of three-dimensional spatial coordinates of the connection path, linear fitting of the spatial point cloud is performed for each connection path. The direction vector and center position of the fitted line segment are calculated by the least squares method. The fitted line segment parameters of the connection path are obtained one by one, and a set of fitted line segment parameters of the connection path is generated. Based on the set of line segment parameters fitted by the connection path, the parameters of the fitted line segments are converted into a standard vectorized representation format one by one, and the three-dimensional coordinates and vector directions of the start and end points of the fitted line segments are recorded in sequence to construct a line vectorization structure.

7. The three-dimensional environment reconstruction system for the working area of ​​the marking robot according to claim 1, characterized in that, The steps for obtaining the initial complexity grid are as follows: Based on the curvature attribute point cloud, the boundary range and spatial size parameters of the working area of ​​the line-marking robot are set. The entire working area of ​​the line-marking robot is divided into regular cubic cells according to a uniform side length. Each curvature attribute point in the curvature attribute point cloud is mapped and assigned to the cubic cell according to its three-dimensional position coordinates, thereby generating the initial set of cells for the working area of ​​the line-marking robot. Based on the initial set of cells in the working area of ​​the line-marking robot, calculate the coupling morphological complexity of each initial cell; Based on the coupled morphological complexity of each initial cell, the entire set of initial cells in the working area of ​​the line-marking robot is traversed. The spatial position index and coupled morphological complexity of all initial cells are paired and encoded one by one, and a unified organization is formed to form a complexity expression grid of spatial structure changes within the region, thus generating the initial complexity grid.

Images