Track-driven vector lane line topology automatic mapping method and system

Abstract

The invention discloses a vector lane line topology automatic mapping method based on track driving. The method comprises the following steps: receiving driving track data and vector lane line data as input; generating separation lines according to a preset interval along the driving track direction, and dividing the driving track data into a plurality of track fragments; for each track segment, forming a marking data set in the track segment according to a preset space correlation condition; in each track segment, analyzing a logic relation between the intercepted vector lane marking lines, combining the associated vector lane marking lines into a lane group, generating a lane segment based on the lane group, and establishing a left-right adjacent topological relation between lanes in the track segment; performing connection relation judgment on the currently generated lane segment and the front-drive lane segment thereof to construct a front-and-back-drive lane topology network; and outputting a vector lane line map based on the front and rear drive lane topology network.

Description

Technical Field

[0001] This invention relates to the field of map production technology, specifically to a trajectory-driven automated mapping method and system for vector lane topology. Background Technology

[0002] In the field of crowdsourced map production, traditional lane-level road network construction methods typically rely on manual processing. This process first generates independent, unrelated vector lane lines and road boundaries based on algorithms, and then manually interprets and edits these elements to convert them into a lane-level road network with topological connections. This approach has the following significant drawbacks:

[0003] Insufficient automation capabilities mean that the mapping process relies heavily on the experience and subjective judgment of operators, resulting in slow mapping speeds and high labor costs.

[0004] Faced with complex situations such as broken, overlapping, intersecting, and merging lane markings commonly found in input data, there is a lack of unified and robust automated processing models. Manual identification, judgment, and repair are required for each issue, making it difficult to guarantee processing consistency.

[0005] Especially at complex traffic structures such as intersections, U-turn areas, and merging points of main and auxiliary roads, existing methods struggle to automatically infer and generate complete and reasonable lane connection relationships (including necessary virtual connecting lanes), often causing topology network interruptions and affecting the integrity of downstream applications.

[0006] The lack of an effective mechanism for fusing and processing multiple driving trajectory data makes it difficult to achieve automatic deduplication, interruption, and connection at the topology level, which can easily lead to the generation of redundant or contradictory lane data, thus restricting the quality and reliability of data results. Summary of the Invention

[0007] To overcome the above-mentioned technical defects, the present invention provides a method and system for automated mapping of vector lane topology based on trajectory driving.

[0008] To solve the above problems, the present invention is implemented according to the following technical solution:

[0009] In a first aspect, the present invention provides a trajectory-driven method for automated mapping of vector lane line topology, comprising the following steps: receiving driving trajectory data and vector lane marking data as input; generating dividing lines at preset intervals along the driving trajectory direction to divide the driving trajectory data into multiple trajectory segments; for each trajectory segment, extracting vector lane markings spatially associated with the trajectory segment from the vector lane marking data according to preset spatial association conditions to form a marking dataset within the trajectory segment; analyzing the logical relationships between the extracted vector lane markings within each trajectory segment, combining the associated vector lane markings into lane groups, and generating lane segments based on the lane groups, while establishing left-right adjacency topology relationships between lanes within the trajectory segment; determining the connection relationship between the currently generated lane segment and its preceding lane segment to construct a front-to-rear drive lane topology network; and outputting a vector lane line map based on the front-to-rear drive lane topology network.

[0010] In conjunction with the first aspect, the present invention provides a first specific implementation of the first aspect. Specifically, for multiple driving trajectory data, different lane groups are split and divided into regions by road boundary elements to achieve deduplication and fusion of the topology; in intersection areas lacking lane lines, connecting lanes are automatically generated to complete the road network based on the lanes associated with the trajectory segments before and after, and the vector lane line map is updated.

[0011] In conjunction with the first aspect, the present invention provides a second specific implementation of the first aspect. Specifically, for each trajectory segment, according to preset spatial association conditions, vector lane markings spatially associated with the trajectory segment are extracted from the vector lane marking data to form a marking dataset within the trajectory segment. This includes the following steps: defining the spatial scope of each trajectory segment and establishing a spatial index for the vector lane marking data; based on the spatial index, obtaining all vector lane markings that intersect with the spatial scope of the current trajectory segment or are located within a preset buffer distance, forming a candidate marking set; traversing each vector lane marking in the candidate marking set and filtering based on the preset spatial association conditions; and aggregating all vector lane markings filtered through the spatial association conditions to generate a marking dataset corresponding to the current trajectory segment.

[0012] In conjunction with the first aspect, the present invention provides a third specific implementation of the first aspect. Specifically, the step of traversing each vector lane marking in the candidate lane marking set and filtering based on the preset spatial association conditions includes the following steps: calculating the shortest spatial distance between the vector lane marking and all trajectory points in the current trajectory segment, and determining whether the shortest spatial distance is less than a first preset threshold; calculating the elevation values ​​of the vector lane marking and the nearest point pair in the current trajectory segment, and determining whether the elevation difference between the two is less than a second preset threshold; calculating the angle between the average direction of the vector lane marking and the overall travel direction of the current trajectory segment, and determining whether the angle is less than a third preset threshold; determining whether there is a road boundary element blocking the vector lane marking and the current trajectory segment through spatial relationship analysis; and determining that the vector lane marking passes the spatial association condition filtering if and only if the shortest distance is less than the first preset threshold, the elevation difference is less than the second preset threshold, the angle is less than the third preset threshold, and there is no road boundary element blocking.

[0013] In conjunction with the first aspect, the present invention provides a fourth specific implementation of the first aspect. Specifically, the step of combining associated vector lane markings into a lane group includes the following steps: performing a parallelism judgment on the vector lane markings in the marking dataset, and dividing the vector lane markings that are parallel to each other and whose spacing is within a preset lane width range into the same candidate group; calculating the direction vector of each vector lane marking in the candidate group, comparing it with the average direction of the current trajectory segment, and retaining the markings with the same direction as valid lane markings; sorting the valid lane markings according to the lateral position of the valid lane markings relative to the driving trajectory, according to the distance perpendicular to the driving direction of the trajectory, to generate an ordered lane marking sequence; and combining the ordered lane marking sequence into a lane group.

[0014] In conjunction with the first aspect, the present invention provides a fifth specific implementation of the first aspect. Specifically, the step of generating lane segments based on lane groups and simultaneously establishing left-right adjacency topology relationships between lanes within the trajectory segment includes the following steps: In the lane group, for every two sequentially adjacent lane markings, a lane centerline is generated between them, each lane centerline representing a drivable lane; based on the ordered lane marking sequence, the left and right boundary markings of each lane centerline are determined; for any two adjacent lane centerlines, they share the same lane marking as a common boundary, and a left-right adjacency relationship is established based on the common boundary; all lane centerlines generated within the current trajectory segment and their left-right adjacency topology relationships are combined to form lane segment data.

[0015] In conjunction with the first aspect, the present invention provides a sixth specific implementation of the first aspect. Specifically, the step of determining the connection relationship between the currently generated lane segment and its preceding lane segment to construct a front-to-back lane topology network includes the following steps: spatially superimposing the starting end of the current lane segment and the ending end of the preceding lane segment, and calculating the spatial overlap of the corresponding lane markings at the two endpoints; comparing the number of lanes, lane type, and lane width attributes of the current lane segment and the preceding lane segment at the connection endpoints; if the number of lanes and the type of lanes of the current lane segment and the preceding lane segment are the same, and the spatial overlap of the corresponding lane markings at the endpoints is higher than a preset splicing threshold, then it is determined to be a smooth splicing scenario; if the number of lanes at the connection endpoints changes, then it is identified as a merging / diverging scenario; if the current lane segment and the preceding lane segment neither meet the smooth splicing conditions nor constitute a merging / diverging relationship, then topology breaking processing is performed at the connection point, and a preceding-to-follower topology link relationship is established between the spatially nearest lanes.

[0016] In conjunction with the first aspect, the present invention provides a seventh specific implementation of the first aspect. Specifically, the step of identifying a merging or splitting scenario when the number of lanes at the connection endpoint changes includes: when it is detected that one lane marking of a front-drive lane segment is connected to two lane markings of a subsequent lane segment, it is identified as a merging scenario, and a virtual right lane marking is generated for the merging lane; when it is detected that two lane markings of a front-drive lane segment are connected to one lane marking of a subsequent lane segment, it is identified as a merging scenario, and a virtual left lane marking is generated for the lane before merging.

[0017] Secondly, the present invention also provides a trajectory-driven vector lane topology automated mapping system, comprising: a data receiving module for receiving driving trajectory data and vector lane marking data; a trajectory segmentation module for generating dividing lines at preset intervals along the driving trajectory direction to segment the driving trajectory data into multiple trajectory segments; an association processing module for extracting vector lane markings spatially associated with the trajectory segment from the vector lane marking data for each trajectory segment according to preset spatial association conditions, forming a marking dataset within the trajectory segment; a topology construction module for analyzing the logical relationships between the extracted vector lane markings within each trajectory segment, combining the associated vector lane markings into lane groups, generating lane segments based on the lane groups, and simultaneously establishing left-right adjacency topology relationships between lanes within the trajectory segment; a network generation module for determining the connection relationship between the currently generated lane segment and its preceding lane segment, constructing a front-to-rear-drive lane topology network; and an output module for outputting a vector lane map based on the front-to-rear-drive lane topology network.

[0018] In conjunction with the second aspect, the present invention provides a first specific implementation of the second aspect, which specifically includes: a trajectory fusion module, used to segment and divide different lane groups into regions using road boundary elements for multiple driving trajectory data, so as to achieve deduplication and fusion of the topology; and a road network completion module, used to automatically generate connecting lanes to complete the road network in intersection areas lacking lane lines, based on the lanes associated with the trajectory segments before and after, and update the vector lane line map.

[0019] Compared with the prior art, the beneficial effects of the present invention are:

[0020] By employing trajectory-driven and automated segmentation processing, end-to-end generation from raw data to the topological road network is achieved, significantly reducing the need for manual intervention and operational costs, and improving the overall efficiency of map production. Relying on the analysis of lane marking logical relationships within trajectory segments and the automatic lane group generation mechanism, complex situations such as missing, overlapping, intersecting, and merging / diverging lane markings can be effectively handled, reducing reliance on human experience and judgment. Through the construction of lane adjacency relationships within segments and the judgment of connections between preceding and following segments, continuous and reasonable lane-level topologies, including virtual connecting lanes, can be automatically generated in key areas such as intersections and ramp merging points, effectively avoiding network breaks and supporting application requirements such as high-precision navigation. Attached Figure Description

[0021] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, wherein:

[0022] Figure 1 This is a flowchart of a trajectory-driven vector lane topology automated mapping method according to the present invention. Detailed Implementation

[0023] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0024] like Figure 1 As shown, this invention relates to a trajectory-driven vector lane topology automated mapping method and system.

[0025] Example 1

[0026] like Figure 1As shown, a trajectory-driven automated vector lane topology mapping method includes the following steps: receiving driving trajectory data and vector lane marking data as input; generating dividing lines at preset intervals along the driving trajectory direction to divide the driving trajectory data into multiple trajectory segments; for each trajectory segment, extracting vector lane markings spatially associated with the trajectory segment from the vector lane marking data according to preset spatial association conditions to form a marking dataset within the trajectory segment; analyzing the logical relationships between the extracted vector lane markings within each trajectory segment, combining associated vector lane markings into lane groups, and generating lane segments based on the lane groups, while establishing left-right adjacency topology relationships between lanes within the trajectory segment; determining the connection relationship between the currently generated lane segment and its preceding lane segment to construct a front-to-rear drive lane topology network; and outputting a vector lane map based on the front-to-rear drive lane topology network.

[0027] In a preferred embodiment, driving trajectory data and vector lane marking data are received as input.

[0028] Specifically, it receives driving trajectory data and vector lane marking data as input. Driving trajectory data typically comes from vehicle sensors such as GPS and IMU, and includes information such as timestamps, location coordinates (latitude and longitude or planar coordinates), and speed. Vector lane marking data comes from high-precision maps or remote sensing images, representing lane lines in vector form (such as polylines or curves), and includes geometric and attribute information (such as lane line type, color, etc.).

[0029] In a preferred embodiment, dividing lines are generated at preset intervals along the driving trajectory direction to divide the driving trajectory data into multiple continuous trajectory segments.

[0030] Specifically, dividing lines are generated at preset intervals along the vehicle trajectory direction, dividing the trajectory data into multiple trajectory segments. The preset interval can be a fixed distance (e.g., 50 meters) or a dynamic distance (adjusted according to the trajectory curvature). The dividing lines are perpendicular to the trajectory tangent direction, cutting the trajectory into multiple segments. Each trajectory segment contains a continuous range of trajectory points for subsequent local processing.

[0031] In a preferred embodiment, for each trajectory segment, according to preset spatial association conditions, vector lane markings spatially associated with the trajectory segment are extracted from the vector lane marking data to form a marking dataset within the trajectory segment.

[0032] Specifically, for each trajectory segment, based on preset spatial association conditions, vector lane markings spatially associated with that trajectory segment are extracted from the vector lane marking data to form a marking dataset within that trajectory segment. The spatial association conditions include:

[0033] First preset threshold: The distance between the vector lane marking and the current trajectory segment is less than a set value (e.g., 30 meters).

[0034] Second preset threshold: The elevation value of the vector lane marking and the nearest point pair in the current trajectory segment is less than the set value (e.g., 2 meters).

[0035] The third preset threshold is: the angle between the vector lane marking and the current driving trajectory direction is less than the set value (e.g., 30 degrees).

[0036] There are no road boundary features obstructing the vector lane markings and the current trajectory segment.

[0037] In a preferred embodiment, within each trajectory segment, the logical relationship between the captured vector lane markings is analyzed, the associated vector lane markings are combined into lane groups, and lane segments are generated based on the lane groups. At the same time, the left and right adjacency topology relationship between each lane within the trajectory segment is established.

[0038] Specifically, associated lane markings are identified based on their geometric characteristics (such as parallelism and continuity) and attribute characteristics (such as consistency in type). For example, multiple parallel lane markings of the same type may belong to the same lane group. Associated lane markings are grouped into lane groups, each representing one or more lanes. The lane markings within a lane group are used to generate lane centerlines or lane boundaries. Based on the lane groups, lane segments are generated through interpolation or fitting, represented as vector geometry (such as polylines). The left-right adjacency relationships between lane segments are established based on their relative positions (such as left-to-right order) and stored as topological data.

[0039] In a preferred embodiment, the connection relationship between the currently generated lane segment and its front-drive lane segment is determined to construct a front-drive lane topology network.

[0040] Specifically, based on the geometric endpoints, directional consistency, and overlap of lane segments, it is determined whether the current lane segment is connected to the preceding lane segment. For example, if the distance between the starting point of the current lane segment and the ending point of the preceding lane segment is less than a threshold, and their directions are continuous, then it is considered a connection. Connected lane segments are linked together to form a preceding-follower relationship chain, constructing a global topology network. This network represents the travel direction of lanes and the connections between multiple lanes.

[0041] In a preferred embodiment, a vector lane line map is output based on the front and rear drive lane topology network.

[0042] Specifically, based on the front and rear drive lane topology network, a vector lane line map is output. The output format can be a common geographic data format (such as Shapefile, GeoJSON, or a custom format), containing lane line geometry data, attribute data (such as lane type and width), and topology data (left-right adjacency and front-rear drive relationships).

[0043] In a preferred embodiment, for multiple driving trajectory data, different lane groups are split and divided into regions by road boundary elements to achieve deduplication and fusion of the topology.

[0044] Specifically, road boundary feature data is introduced. This data can be vector polygon data from a high-precision map, representing the physical boundaries of roads, such as median strips, curbs, green belts, building boundaries, etc. These boundaries naturally divide the road network into multiple relatively independent driving areas (e.g., a main road and auxiliary roads in one direction, a guide lane area within an intersection, etc.). Using the road boundary polygons, the entire working area is divided into multiple non-overlapping lane areas. Each lane area is a logical unit, representing a continuous and unified lane layout space. All lane groups or lane segments are traversed. The spatial relationship between each lane group and the road boundary polygons is examined. If a lane group is completely within a boundary polygon, it is completely preserved within that area. If a lane group spans multiple boundary polygons, it is split into multiple independent sub-lane segments at its intersections with the boundaries, each sub-segment belonging to its respective boundary polygon area.

[0045] Based on the regional division of the front-end and rear-end topology construction, it is divided into intra-regional connections and cross-regional connections.

[0046] Topology construction within the same lane area: Perform existing connection relationship checks. Connect the current lane segment with its preceding and succeeding segments within the same area to establish a topological chain within the area.

[0047] Cross-regional topology construction: Identify lane segments that act as exit points of a region (whose endpoints are close to the region boundary) and lane segments that act as entrance points of adjacent regions (whose starting points are close to the same region boundary). Based on spatial proximity and directional consistency, establish cross-regional forward and backward topological connections from "exit segments" to "entrance segments". Since multiple trajectories may generate geometrically very similar lane segments within the same lane region, these segments need to be fused in this step. Lane segments with similar geometry and attributes are grouped using a clustering algorithm (such as DBSCAN). Each group retains only one geometrically optimal lane segment (e.g., the one with the highest fit to multiple trajectories) as a representative, and all relevant topological relationships are integrated onto this representative lane. Based on the global topology network after deduplication, fusion, and cross-regional connections, the final redundancy-free, high-precision vector lane line map is output.

[0048] In a preferred embodiment, in intersection areas lacking lane lines, connecting lanes are automatically generated to complete the road network based on the lanes associated with the preceding and following trajectory segments, and the vector lane line map is updated.

[0049] Specifically, within the intersection area defined by the road boundary, all lane segments serving as both entrances and exits are examined. If an entrance lane segment is not topologically connected to any exit lane segment, or an exit lane segment has no corresponding entrance connection, it is identified as a potential connection breakpoint. All vehicle trajectory data traversing the intersection area is analyzed to find trajectories that can connect these entrance and exit breakpoints. For example, a trajectory starting from "entrance breakpoint A," crossing the intersection, and ending at "exit breakpoint B" strongly suggests a passable connection from lane A to lane B. When multiple trajectories (exceeding a set threshold) exhibit the same connection pattern (A→B), this connection is confirmed as valid. For each confirmed valid connection (e.g., A→B), a connecting lane is automatically generated. Using a curve fitting algorithm (e.g., B-spline curve or circular arc fitting), a vector geometric line smoothly connecting the entrance breakpoint to the exit breakpoint is generated based on the associated vehicle trajectory points; this is the connecting lane. Assign attributes to the connecting lane, such as lane type = "virtual connecting lane" and confidence level = [calculated value based on the number of supporting trajectories]. Assign this connecting lane to the current intersection area, add all newly generated connecting lanes to the existing vector lane line dataset, and process these connecting lanes as valid nodes together with the entrance and exit lanes to form a complete topology chain.

[0050] In a preferred embodiment, for each trajectory segment, extracting vector lane markings spatially associated with that trajectory segment from the vector lane marking data according to preset spatial association conditions to form a marking dataset within that trajectory segment includes the following steps: defining the spatial scope of each trajectory segment and establishing a spatial index for the vector lane marking data; based on the spatial index, acquiring all vector lane markings that intersect with the spatial scope of the current trajectory segment or are located within a preset buffer distance to form a candidate marking set; traversing each vector lane marking in the candidate marking set and filtering it based on the preset spatial association conditions; and aggregating all vector lane markings filtered through the spatial association conditions to generate a marking dataset corresponding to the current trajectory segment.

[0051] Specifically, a spatial scope is defined for the currently processed trajectory segment. This scope is typically a geometric region that encloses the trajectory segment, such as the minimum bounding rectangle of the trajectory segment or a buffer polygon generated along the trajectory segment. Simultaneously, a spatial index is built for the entire input vector lane marking dataset. Efficient spatial index structures such as R-trees or Quadtrees are preferred to accelerate subsequent spatial queries. Based on the established spatial index, a preliminary spatial query is performed to retrieve all vector lane markings that intersect with the spatial scope of the current trajectory segment or are within its preset buffer distance. The purpose of this step is to quickly eliminate a large number of obviously irrelevant lane markings, narrowing the range of markings requiring fine-grained judgment to a candidate set. This significantly improves overall processing efficiency. Each vector lane marking in the candidate set obtained in the previous step is traversed, and for each candidate marking, a series of more stringent preset spatial association conditions are used for fine-grained filtering. All vector lane markings filtered through all the above fine-grained spatial association conditions are aggregated to generate a marking dataset uniquely corresponding to the current trajectory segment. This dataset will serve as the direct input for the next step, "lane group generation." This preferred embodiment employs a two-stage strategy of "preliminary index query -> fine-grained condition filtering," which balances processing efficiency and correlation accuracy, avoiding brute-force matching in massive amounts of data.

[0052] In a preferred embodiment, traversing each vector lane marking in the candidate marking set and screening based on the preset spatial association conditions includes the following steps: calculating the shortest spatial distance between the vector lane marking and all trajectory points in the current trajectory segment, and determining whether the shortest spatial distance is less than a first preset threshold; calculating the elevation values of the vector lane marking and the nearest point pair in the current trajectory segment, and determining whether the elevation difference between the two is less than a second preset threshold; calculating the angle between the average direction of the vector lane marking and the overall traveling direction of the current trajectory segment, and determining whether the angle is less than a third preset threshold; through spatial relationship analysis, determining whether there is a road boundary element barrier between the vector lane marking and the current trajectory segment; and determining that the vector lane marking passes the spatial association condition screening if and only if all of the shortest distance is less than the first preset threshold, the elevation difference is less than the second preset threshold, the angle is less than the third preset threshold, and there is no road boundary element barrier.

[0053] Specifically, traverse each vector lane marking in the candidate marking set obtained in the previous step. For each candidate marking, use a series of more stringent, multi-dimensional fusion preset spatial association conditions for fine screening. As a preferred embodiment, this screening process includes the following parallel judgment steps, and all conditions must be met simultaneously:

[0054] Judgment of planar spatial proximity: Calculate the shortest spatial distance (e.g., using Euclidean distance) between the vector lane marking and all trajectory points in the current trajectory segment, and determine whether the shortest spatial distance is less than a first preset threshold D (e.g., 5 meters). Let the vector lane marking L be defined by the point set Q = {q1, q2,..., q m}, and the current trajectory segment P = {p1, p2,..., p n} be defined by the point set, then the shortest spatial distance d is calculated as: where, (x pi , x qj ) and (y pi , y qj ) are the planar coordinates of the points p i and q j respectively. If d < D, the distance condition is satisfied.

[0055] Judgment of vertical spatial consistency: Calculate the elevation values of the vector lane marking and the nearest point pair in the current trajectory segment, and determine whether the elevation difference between the two is less than a second preset threshold H (e.g., 2 meters). Let the point pair that makes the distance the smallest be (p min , q min ), and their elevation values are respectively and then the elevation difference Δh is: If Δh < H, the elevation condition is satisfied. This condition is crucial for excluding lane markings that, although close in planar distance, are not on the same driving plane under or above elevated bridges or tunnels.

[0056] Judgment of the consistency of the traveling direction: Calculate the angle between the average direction vector of the vector lane marking and the overall traveling direction vector of the current trajectory segment, and determine whether this angle is less than the third preset threshold (for example, 30 degrees). Let the average direction vector of the current trajectory segment By calculating the vector from the first point p1 to the last point p n That is The average direction vector of the vector lane marking Similar calculation, that is Then the angle θ is: If θ < the third preset threshold (such as 30 degrees), the angle condition is satisfied. This condition ensures that the direction of the associated lane marking is basically consistent with the vehicle traveling direction, effectively excluding the interference of intersecting roads.

[0057] Judgment of physical connectivity: Through spatial relationship analysis (such as ray method or overlay analysis), judge whether there are road boundary elements (such as median strips, curbs, etc.) blocking between the vector lane marking and the current trajectory segment. This condition is a high-level logical judgment to ensure that even if the distance and direction conditions are met, physically isolated markings (such as oncoming lanes, auxiliary roads separated by green belts) will not be wrongly associated.

[0058] Only when a candidate vector lane marking simultaneously satisfies all the above conditions (that is, the shortest distance is less than the first preset threshold, the elevation difference is less than the second preset threshold, the angle is less than the third preset threshold, and there is no road boundary element blocking), it is determined that the vector lane marking passes the spatial association condition screening.

[0059] In a preferred embodiment, the combining the associated vector lane markings into a lane group includes the following steps: performing a parallelism judgment on the vector lane markings in the marking dataset, classifying the vector lane markings that are parallel to each other and have a spacing within the preset lane width range into the same candidate group; calculating the direction vectors of each vector lane marking in the candidate group, comparing them with the average direction of the current trajectory segment, and retaining the同向标线 as valid lane markings; sorting the valid lane markings according to the lateral position relative to the driving trajectory by the distance perpendicular to the trajectory driving direction, generating an ordered lane marking sequence; combining the ordered lane marking sequence into a lane group.

[0060] Specifically, the parallelism of all vector lane markings in the marking dataset corresponding to the current trajectory segment is determined pairwise. The angle between the direction vectors of two lane markings is calculated; if the angle is less than a preset parallelism tolerance (e.g., 5 degrees), they are considered parallel. For two vector lane markings L... a and L b Its direction vector and As calculated above, the included angle φ is: If φ < parallel tolerance, then it is determined to be parallel. For pairs of parallel markings, the average spacing between them is further calculated. Along a marking L a Sampling point set S a ={s1,s2,...,s k}, for each sampling point s i Calculate to another marking line L a shortest distance d i Then the average spacing for: if Within a preset lane width range (e.g., 2.5 meters to 4.5 meters), lane markings are grouped into the same candidate group. The direction vector of each lane marking within the candidate group is calculated, along with the average travel direction vector of the current trajectory segment. The direction vector of each marking is compared to the trajectory direction vector, and the included angle α is calculated. If α is less than the direction tolerance (e.g., 15 degrees), it is retained as a valid lane marking. Based on the lateral position of the valid lane markings relative to the driving trajectory, the valid lane markings are sorted according to their distance perpendicular to the trajectory's travel direction. Let the average direction vector of the trajectory... The unit vector is T^=(T x ,T y If ), then the unit vector perpendicular to T^ is N^=(-T y ,T x For each valid lane marking L, calculate the projected distance d of the vector from its center point c to the trajectory center point o in the N^ direction. lateral :d lateral =(co)·N^, according to d lateral The lane markings are sorted by size to generate an ordered sequence. This ordered sequence is then combined into a lane group. This sequence clearly reflects the actual distribution of lane lines from left to right or right to left.

[0061] In a preferred embodiment, the step of generating lane segments based on lane groups and establishing left-right adjacency topology relationships between lanes within the trajectory segment includes the following steps: In the lane group, for every two sequentially adjacent lane markings, a lane center line is generated between them, each lane center line representing a drivable lane; based on the ordered lane marking sequence, the left and right boundary markings of each lane center line are determined; for any two adjacent lane center lines, they share the same lane marking as a common boundary, and a left-right adjacency relationship is established based on the common boundary; all lane center lines generated within the current trajectory segment and their left-right adjacency topology relationships are combined to form lane segment data.

[0062] Specifically, within the generated lane groups, traverse their ordered sequence of lane markings. For each pair of sequentially adjacent lane markings L... i and L i+1 This generates a lane centerline between them. Assume L i and L i+1 Having already resampled to the same number of points, let L i The point sequence is A1, A2, ..., A n L i+1 If the point sequence is B1, B2, ..., B; then the centerline point C of the lane... j for: For j = 1, 2, ..., n, each lane centerline represents a drivable lane. Based on an ordered sequence of lane markings, the left and right boundary markings of each lane centerline are determined. For lanes L... i and L i+1 The generated lane centerline has a left boundary of L. i The right boundary is L. i+1 For any two adjacent lane center lines, they share the same lane marking as a common boundary. For example, L... i and L i+1 The generated lanes and those generated by L i+1 and L i+2 The generated lanes are adjacent and share the vector lane markings L. i+1 As a common boundary, left and right adjacency relationships are established based on this. All lane centerlines generated within the current trajectory segment and their left and right adjacency topologies are used to form lane segment data.

[0063] In a preferred embodiment, the step of determining the connection relationship between the currently generated lane segment and its preceding lane segment to construct a front-to-back lane topology network includes the following steps: spatially superimposing the starting end of the current lane segment and the ending end of the preceding lane segment, and calculating the spatial overlap of the corresponding lane markings at the two endpoints; comparing the number of lanes, lane type, and lane width attributes of the current lane segment and the preceding lane segment at the connection endpoints; if the number of lanes and the type of lanes of the current lane segment and the preceding lane segment are the same, and the spatial overlap of the corresponding lane markings at the endpoints is higher than a preset splicing threshold, it is determined to be a smooth splicing scenario; if the number of lanes at the connection endpoints changes, it is identified as a merging / diverging scenario; if the current lane segment and the preceding lane segment neither meet the smooth splicing conditions nor constitute a merging / diverging relationship, then topology breaking processing is performed at the connection point, and a preceding-to-follower topology link relationship is established between the spatially nearest lanes.

[0064] Specifically, the starting point of the current lane segment and the ending point of the front-wheel drive lane segment are spatially superimposed for analysis, and the spatial overlap of the corresponding lane markings at the two endpoints is calculated. Let E be the set of lane centerline endpoints at the starting point of the current lane segment. current ={e1,e2,...,e m The set of lane centerline endpoints at the termination point of the front-wheel drive lane segment is E. predecessor ={f1,f2,...,f n For each endpoint e i Find the nearest endpoint f j Calculate the distance d i =||e i -f j || and the angle θ between the directions i (Based on the tangent direction of the lane centerline at the endpoint). The spatial overlap score S can then be calculated as: Where σ is the scale parameter (e.g., 1 meter). If S is greater than the splicing threshold (e.g., 0.85), it is determined to be a smooth splicing scenario. The number of lanes, lane type, and lane width attributes at the connection endpoint are compared between the current lane segment and the preceding lane segment. If the current lane segment and the preceding lane segment have the same number and type of lanes, and the spatial overlap is higher than the preset splicing threshold, a one-to-one direct preceding-front-wheel drive connection is established. If the number of lanes at the connection endpoint changes, it is identified as a merging or splitting scenario. Specifically, when a lane marking of a preceding lane segment is detected to be connected to two lane markings of a subsequent lane segment, it is identified as a splitting scenario, and a virtual right lane marking is generated for the splitting lane; when two lane markings of a preceding lane segment are detected to be connected to a lane marking of a subsequent lane segment, it is identified as a merging scenario, and a virtual left lane marking is generated for the lane before merging. The generation of virtual lane markings can use curve fitting methods, such as Bézier curves, based on the driving trajectory points. If the current lane segment and the preceding lane segment neither meet the conditions for smooth splicing nor constitute a merging or splitting relationship, then topology breaking is performed at the connection point, and a preceding-follower topology link is established between the lanes that are spatially closest.

[0065] In a preferred embodiment, the step of identifying a merging or splitting scenario when the number of lanes at the connection endpoint changes specifically includes: when it is detected that one lane marking of a front-drive lane segment is connected to two lane markings of a subsequent lane segment, it is identified as a splitting scenario, and a virtual right lane marking is generated for the splitting lane; when it is detected that two lane markings of a front-drive lane segment are connected to one lane marking of a subsequent lane segment, it is identified as a merging scenario, and a virtual left lane marking is generated for the lane before merging.

[0066] Specifically, when a lane in a forward lane segment is detected to connect with two (or more) lanes in the current lane segment, it is identified as a traffic splitting scenario (e.g., a straight lane splits into a straight lane and a right-turn lane), and a one-to-many topological relationship is established between the forward lane and the two subsequent lanes. Virtual lane markings are created for the newly generated lanes to complete the lane geometry. The specific method is as follows: based on the driving trajectory density and road design specifications, a virtual, gradually changing lane boundary line is generated between the bifurcation point of the forward lane and the two subsequent lanes, making the geometric representation of the lane segmentation region natural and continuous. The virtual lane markings can be generated using Bézier curves, with control points determined based on trajectory points. Let the endpoint of the forward lane be P0, and the starting points of the two subsequent lanes be P1 and P2, respectively. Then, the virtual line from P0 to P1 or P2 uses a quadratic Bézier curve: B(t) = (1-t). 2 P0+2t(1-t)P c +t 2 P1, t∈[0,1]; where control point Pc The merging is achieved based on trajectory point fitting. When two (or more) lanes in a front-drive lane segment are detected to connect with one lane in the current lane segment, it is identified as a merging scenario (e.g., a secondary lane merging with a primary lane), and a many-to-one topological relationship is established between the two front-drive lanes and one subsequent lane. Virtual lane markings are created for the lane segment before merging to maintain geometric continuity. Specifically, in the area where two front-drive lanes gradually converge until they merge into one lane, virtual lane lines are generated to clearly mark the boundary of the merging area, ensuring that the map geometry is not broken.

[0067] Example 2

[0068] In a preferred embodiment, the step of outputting a vector lane line map based on the front and rear drive lane topology network includes the following steps: integrating lane geometry data and their internal topological relationships of all lane segments; integrating the front and rear drive topological connection relationships between all lane segments; associating the integrated lane geometry data with the topological relationships to generate lane line data containing complete topological information; and converting the lane line data into a standard vector map format and outputting it.

[0069] Specifically, all lane segments generated during the processing are aggregated, including their lane centerline geometry data (a polyline composed of discrete point sequences), lane boundary marking data (geometric definitions of left and right boundary lines), and the left and right adjacency topology within the segments; the left and right adjacency topology within the segments is established by constructing an adjacency matrix A. ij ,in: Integrate all front and rear drive lane topology connections established through connection relationship determination, and establish a front and rear drive topology connection matrix P. ij ,in: The integrated geometric data and topological relationships are validated for consistency to ensure that each lane segment is correctly linked to the topological network. Based on the lane centerline geometry and the corresponding lane width attribute, lane surface geometry or lane boundaries represented by double lines with actual width are generated. The lane surface geometry is generated using the following algorithm:

[0070]

[0071] Where w is the lane width, and the normal vector is obtained by rotating it 90 degrees through the tangent direction of the center line.

[0072] The process involves converting virtual lane markings into dashed lines or special symbols on a map and marking their virtual origin in the attributes; assigning a unique identifier to each lane element and recording its attribute information, including lane type, direction of travel, and speed limit; embedding topological relationship attributes for each lane element in the output vector lane line data, including: a list of left adjacent lane IDs, a list of right adjacent lane IDs, a list of front-drive lane IDs, and a list of successor lane IDs; establishing a lane-level topology relationship table to record the connection relationships between lanes throughout the network; organizing lane line data, topology relationship data, and metadata according to the standard format of a geographic information system; outputting the processing results as a vector map file in a specified format, along with a corresponding topology relationship description file; and creating a spatial index for the output vector map to support efficient spatial query and route planning applications.

[0073] Example 3

[0074] In a preferred embodiment, the step of splitting and dividing different lane groups into regions based on road boundary elements to achieve deduplication and fusion of the topology for multiple driving trajectory data includes the following steps: acquiring lane groups generated from multiple driving trajectory data and their corresponding trajectory segments; dividing continuous lane groups into different road regions based on road boundary elements; performing spatial overlay analysis on lane groups from different trajectories within the same road region to identify and merge similar lanes that overlap or are adjacent in geometric position; establishing lane topology connection relationships within and between regions to complete the fusion of the topology network.

[0075] Specifically, lane groups and their corresponding trajectory segments are generated from multiple driving trajectory data. Lane groups contain geometric, attribute, and topological information; trajectory segments contain time series and spatial location information. A correspondence table between lane groups and trajectory segments is established. Continuous lane groups are divided into different road regions based on road boundary elements, including but not limited to: curbs, median strips, and guardrails. A spatial overlay analysis method is used to determine the road region to which each lane group belongs, establishing a region index to improve subsequent processing efficiency. Within the same road region, spatial overlay analysis is performed on lane groups from different trajectories: calculating the spatial similarity between lane groups, including: geometric shape similarity, directional consistency, and positional overlap; setting a similarity threshold to identify and merge similar lanes with overlapping or adjacent geometric positions; and using a weighted average algorithm to generate the optimal lane geometry. Lane topological connections within and between regions are established; based on the merged lane geometry, left-right adjacency relationships are re-established, and a multi-trajectory voting mechanism is used to determine lane types and attributes. Lane correspondences at adjacent region boundaries are identified, and cross-regional front-to-back topological connections are established to ensure global consistency of the topological network. The quality of the merged topology network is evaluated: topological connectivity is checked, geometric continuity is verified, and the consistency of evaluation attributes is assessed. Based on the evaluation results, necessary optimization adjustments are made.

[0076] In a preferred embodiment, the deduplication and fusion of the topology structure further includes: performing a topology consistency check on the merged lanes to eliminate duplicate lane connection relationships; at the boundary of the lane group, resolving topology connection conflicts between lane segments from different trajectory sources based on the trajectory direction and lane matching degree; and for topology breaks caused by missing trajectories, generating virtual connecting lanes to complete the topology network based on the principles of spatial proximity and road continuity.

[0077] Example 4

[0078] In a preferred embodiment, the step of automatically generating connecting lanes to complete the road network and updating the vector lane line map in intersection areas lacking lane lines includes the following steps: identifying connection break areas in the topological network and defining them as intersection areas to be completed; determining the termination end of the preceding drive lane and the starting end of the following drive lane in the intersection area; generating a smoothly transitioning centerline of the connecting lane using a curve interpolation algorithm based on the termination direction of the preceding drive lane and the starting direction of the following drive lane; calculating the gradient width of the connecting lane using linear interpolation based on the lane width attributes of the preceding and following lanes; generating left and right boundary markings of the connecting lane based on the centerline and gradient width of the connecting lane; marking the generated connecting lane boundary markings as virtual lane markings; establishing the preceding and following topological relationships between the connecting lane and its preceding and following drive lanes; and updating the global topological network to include the newly generated connecting lane.

[0079] In a preferred embodiment, the step of generating a smoothly transitioning centerline of the connecting lane using a curve interpolation algorithm includes: fitting the centerline using a Bézier curve or spline curve algorithm based on the geometric orientation and direction angle of the front and rear lanes; ensuring that the generated centerline of the connecting lane satisfies the conditions of curvature continuity or tangent continuity with the front and rear lanes at the connection point; and when there are multiple driving trajectories, prioritizing the path with the highest trajectory density as the orientation reference for the centerline of the connecting lane.

[0080] Specifically, based on the geometry and direction angle of the front and rear lanes, a cubic Bézier curve is used for fitting: B(t) = (1-t) 3 P0+3t(1-t) 2 P1+3t 2 (1-t)P2+t 3 P3, t∈[0,1]; where the control points P0, P1 and P3 are determined as follows: P0 is the end point of the front drive lane, P3 is the start point of the subsequent lane, (Extending along the front-wheel drive lane) (Extending in the opposite direction along the subsequent lane); among them... Let be the unit tangent vector at the endpoint, and α and β be adjustment parameters. Ensure that the centerline of the generated connecting lanes satisfies the curvature continuity condition with the preceding and following lanes at the connection point: and Where κ0 and κ1 represent the curvatures of the preceding and following lanes at the connection point, respectively. When multiple driving trajectories exist, the path with the highest trajectory density is preferentially selected as the reference for the orientation of the centerline connecting lanes. Here, Paths is the set of all possible connection paths.

[0081] Example 5

[0082] An automated vector lane topology mapping system based on trajectory-driven architecture includes: a data receiving module for receiving driving trajectory data and vector lane marking data; a trajectory segmentation module for generating dividing lines at preset intervals along the driving trajectory direction to segment the driving trajectory data into multiple trajectory segments; an association processing module for extracting vector lane markings spatially associated with each trajectory segment from the vector lane marking data according to preset spatial association conditions, forming a marking dataset within the trajectory segment; a topology construction module for analyzing the logical relationships between the extracted vector lane markings within each trajectory segment, combining associated vector lane markings into lane groups, generating lane segments based on the lane groups, and establishing left-right adjacency topology relationships between lanes within the trajectory segment; a network generation module for determining the connection relationship between the currently generated lane segment and its preceding lane segment to construct a front-to-rear-drive lane topology network; and an output module for outputting a vector lane map based on the front-to-rear-drive lane topology network.

[0083] In a preferred embodiment, the trajectory fusion module is used to segment and divide different lane groups into regions using road boundary elements for multiple driving trajectory data, so as to achieve deduplication and fusion of the topology structure; the road network completion module is used to automatically generate connecting lanes to complete the road network in intersection areas lacking lane lines, based on the lanes associated with the trajectory segments, and update the vector lane line map.

[0084] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A trajectory-driven automated mapping method for vector lane topology, characterized in that, Includes the following steps: It receives driving trajectory data and vector lane marking data as input; Dividing lines are generated at preset intervals along the driving trajectory direction to divide the driving trajectory data into multiple trajectory segments; For each trajectory segment, based on preset spatial association conditions, vector lane markings that are spatially associated with that trajectory segment are extracted from the vector lane marking data to form a marking dataset within that trajectory segment; Within each trajectory segment, the logical relationships between the captured vector lane markings are analyzed, the associated vector lane markings are combined into lane groups, and lane segments are generated based on the lane groups. At the same time, the left and right adjacency topology relationships between lanes within the trajectory segment are established. The connection relationship between the currently generated lane segment and its front-wheel drive lane segment is determined in order to construct a front-wheel drive lane topology network; Based on the front and rear drive lane topology network, a vector lane line map is output.

2. The automatic mapping method for vector lane line topology based on trajectory driving according to claim 1, characterized in that, Also includes: For multiple driving trajectory data, different lane groups are split and regions are divided by road boundary elements to achieve deduplication and fusion of topology; In intersection areas lacking lane lines, connecting lanes are automatically generated to complete the road network based on the lanes associated with the preceding and following trajectory segments, and the vector lane line map is updated.

3. The method for automated topology mapping of vector lane lines based on trajectory driving according to claim 1, characterized in that, For each trajectory segment, based on preset spatial association conditions, vector lane markings spatially associated with that trajectory segment are extracted from the vector lane marking data to form a marking dataset within that trajectory segment, including the following steps: Define the spatial scope of each trajectory segment and create a spatial index for the vector lane marking data; Based on the spatial index, all vector lane markings that intersect with the spatial domain of the current trajectory segment or are located within a preset buffer distance are obtained, forming a candidate marking set; Traverse each vector lane marking in the candidate marking set and filter them based on the preset spatial association conditions; All vector lane markings filtered by the spatial association conditions are aggregated to generate a marking dataset corresponding to the current trajectory segment.

4. The method for automated topology mapping of vector lane lines based on trajectory driving according to claim 3, characterized in that, The process of traversing each vector lane marking in the candidate lane marking set and filtering based on the preset spatial association conditions includes the following steps: Calculate the shortest spatial distance between the vector lane marking and all trajectory points in the current trajectory segment, and determine whether the shortest spatial distance is less than a first preset threshold. Calculate the elevation values ​​of the vector lane markings and the nearest point pair in the current trajectory segment, and determine whether the elevation difference between the two is less than a second preset threshold. Calculate the angle between the average direction of the vector lane markings and the overall travel direction of the current trajectory segment, and determine whether the angle is less than a third preset threshold. By analyzing spatial relationships, it is determined whether there are road boundary elements obstructing the vector lane markings and the current trajectory segment; The vector lane marking is determined to pass the spatial association condition screening if and only if the following conditions are met simultaneously: the shortest distance is less than a first preset threshold, the elevation difference is less than a second preset threshold, the included angle is less than a third preset threshold, and there are no road boundary elements blocking the view.

5. The method for automated topology mapping of vector lane lines based on trajectory driving according to claim 1, characterized in that, The process of combining associated vector lane markings into lane groups includes the following steps: Parallelism is determined for the vector lane markings in the marking dataset, and vector lane markings that are parallel to each other and whose spacing is within the preset lane width range are divided into the same candidate group; Calculate the direction vector of each lane marking vector in the candidate group, compare it with the average direction of the current trajectory segment, and retain the lane markings with the same direction as the valid lane markings; Based on the lateral position of the effective lane markings relative to the driving trajectory, the effective lane markings are sorted according to the distance perpendicular to the driving direction of the trajectory to generate an ordered lane marking sequence. The ordered sequence of lane markings is combined into a lane group.

6. The method for automated topology mapping of vector lane lines based on trajectory driving according to claim 1, characterized in that, The process of generating lane segments based on lane groups and establishing the left-right adjacency topology between lanes within the trajectory segment includes the following steps: In the lane group, for every two sequentially adjacent lane markings, a lane center line is generated between them, and each lane center line represents a drivable lane. Based on the ordered lane marking sequence, the left and right boundary markings of each lane centerline are determined; For any two adjacent lane center lines, they share the same lane marking as a common boundary, and a left-right adjacency relationship is established based on the common boundary. The center lines of all lanes generated within the current trajectory segment and their adjacent topological relationships are used to form lane segment data.

7. The method for automated topology mapping of vector lane lines based on trajectory driving according to claim 1, characterized in that, The step of determining the connection relationship between the currently generated lane segment and its preceding lane segment to construct a front-to-rear lane topology network includes the following steps: The starting point of the current lane segment and the ending point of the front-wheel drive lane segment are spatially superimposed and analyzed to calculate the spatial overlap of the lane markings at the two endpoints. Compare the number of lanes, lane type, and lane width attributes at the connection endpoints of the current lane segment and the front-wheel drive lane segment; If the current lane segment and the front-wheel drive lane segment have the same number and type of lanes, and the spatial overlap of the corresponding lane markings at the endpoints is higher than the preset splicing threshold, then it is determined to be a smooth splicing scenario. If the number of lanes changes at the connection endpoint, it is identified as a merging or diverging scenario; If the current lane segment and the preceding lane segment neither meet the conditions for smooth splicing nor constitute a merging or splitting relationship, then topology breaking is performed at the connection point, and a preceding-follower topology link is established between the lanes that are spatially closest.

8. The method for automated topology mapping of vector lane lines based on trajectory driving according to claim 7, characterized in that, If the number of lanes at the connection endpoint changes, it is identified as a merging or diverging scenario, specifically including: When a lane marking in a front-drive lane segment is detected to be connected to two lane markings in a subsequent lane segment, it is identified as a diversion scenario, and a virtual right lane marking is generated for the diverted lane. When it is detected that two lane markings of the front drive lane segment are connected to one lane marking of the subsequent lane segment, it is identified as a merging scenario, and a virtual left lane marking is generated for the lane before the merging.

9. A trajectory-driven automated vector lane topology mapping system, characterized in that, include: The data receiving module is used to receive driving trajectory data and vector lane marking data; The trajectory segmentation module is used to generate dividing lines at preset intervals along the driving trajectory direction, dividing the driving trajectory data into multiple trajectory segments. The association processing module is used to extract the vector lane markings that are spatially associated with the trajectory segment from the vector lane marking data for each trajectory segment according to the preset spatial association conditions, and form the marking dataset within the trajectory segment. The topology building module is used to analyze the logical relationships between the captured vector lane markings within each trajectory segment, combine the associated vector lane markings into lane groups, generate lane segments based on the lane groups, and establish the left and right adjacency topology relationships between lanes within the trajectory segment. The network generation module is used to determine the connection relationship between the currently generated lane segment and its front-wheel drive lane segment, and to construct the front-wheel drive lane topology network. The output module is used to output a vector lane line map based on the front and rear drive lane topology network.

10. The trajectory-driven vector lane topology automated mapping system according to claim 9, characterized in that, Also includes: The trajectory fusion module is used to segment and divide different lane groups into regions using road boundary features for multiple driving trajectory data, so as to achieve deduplication and fusion of topology structure; The road network completion module is used to automatically generate connecting lanes to complete the road network in intersection areas lacking lane lines, based on the lanes associated with the preceding and following trajectory segments, and to update the vector lane line map.

Images