A light map construction method, system and electronic device

Abstract

The present application provides a kind of light map construction method, system and electronic equipment, obtain crowdsourcing data, and the crowdsourcing data is preprocessed, obtain target crowdsourcing data;Wherein, target crowdsourcing data at least includes the information of multiple target roads, each target road includes the multiple target lane lines of direction consistent and lane line complete;According to each target ground marking in each target road, generate multiple intersections;Obtain road semantic information, and according to each intersection, road semantic information and each target lane line in each target road, generate corresponding road level topology;According to each target lane line in each target road, generate the multiple lanes corresponding to each target road, and according to each lane and road level topology, generate corresponding lane level topology;For each lane, bind corresponding traffic sign, and according to lane level topology and each lane that is bound with traffic sign, construct corresponding light map.

Description

Technical Field

[0001] This invention relates to the field of intelligent driving technology, and more specifically, to a lightweight map construction method, system, and electronic device. Background Technology

[0002] With the increasing demand for intelligent driving technology, map data plays a crucial role in vehicle navigation, route planning, and decision support. Currently, high-precision maps can provide intelligent assisted driving with high-precision map positioning, rich traffic information, and environmental perception assistance. However, due to issues such as high acquisition costs, difficulties in data updates, and regulatory restrictions, the full-scale use of high-precision maps in intelligent assisted driving is not feasible. Therefore, to address these issues, the concept of "lightweight maps" can be proposed, leveraging high-precision maps and crowdsourced data provided by intelligent driving vehicles.

[0003] In existing technologies, lightweight maps are usually constructed by "simplifying" high-precision map data to reduce the precision or enrich the corresponding features, or by updating and processing high-precision maps with the help of crowdsourced data. It is evident that lightweight maps cannot be constructed without high-precision maps, and lightweight map construction methods that rely on high-precision maps are still constrained by cost and time. Summary of the Invention

[0004] In view of this, the present invention provides a lightweight map construction method, system and electronic device to solve the problems of high cost and long time when constructing lightweight maps due to reliance on high-precision maps in the prior art.

[0005] The first aspect of this application provides a lightweight map construction method, the method comprising:

[0006] Crowdsourced data is acquired and preprocessed to obtain target crowdsourced data; wherein, the target crowdsourced data includes information on at least multiple target roads, and each target road includes multiple target lane lines with consistent direction and complete lane lines;

[0007] Extract the target ground markings for each target road from the crowdsourced data, and generate intersections based on the target ground markings within each target road;

[0008] The semantic attributes of the road centerline of each target road are obtained according to the nearest neighbor method, and a road-level topology is generated based on the intersection, the target lane lines and the road centerline and their semantic attributes in each target road; wherein, the road centerline of the target road is determined based on the target lane lines in the target road.

[0009] Based on each target lane line within each target road, generate multiple lanes corresponding to each target road, and generate a corresponding lane-level topology based on each lane and the road-level topology;

[0010] Each lane is bound to a corresponding traffic sign, and a corresponding lightweight map is constructed based on the lane-level topology and each lane bound to a traffic sign.

[0011] Optionally, in the above-described lightweight map construction method, the step of acquiring crowdsourced data and preprocessing the crowdsourced data to obtain target crowdsourced data includes:

[0012] Acquire crowdsourced data, wherein the crowdsourced data includes at least multiple lane lines;

[0013] Traverse each lane line to determine at least one driving trajectory for each lane line;

[0014] Determine whether the direction of the lane line is correct based on at least one driving trajectory of the lane line;

[0015] If the direction of the lane line is incorrect, the direction of the lane line is corrected to obtain a lane line with the correct direction;

[0016] The target lane lines are grouped to obtain multiple roads; wherein, the target lane line refers to the lane line with the correct direction; the direction of each target lane line in each road is consistent;

[0017] For each of the roads, determine whether there are any target lane lines in the road that require lane line repair;

[0018] If present, the target lane lines within the road that require lane line repair are repaired to obtain the target road; wherein, the direction of each target lane line within the target road is consistent, correct, and the lane lines are complete.

[0019] If it does not exist, the road is identified as the target road.

[0020] Optionally, in the above-described lightweight map construction method, the semantic attributes of the road centerline of each target road are obtained according to the nearest neighbor method, and a road-level topology is generated based on the intersection, the target lane lines and road centerlines within each target road, and their semantic attributes, including:

[0021] For each target road, based on the road centerline and its semantic attributes, determine the adjacent lane lines that have a potential connection relationship with the road centerline of the target road in a single driving trajectory;

[0022] Based on the road centerline and its adjacent lane lines, a first topology connection line is generated, and the intersection associated with the target road is bound to the first topology connection line to obtain a first topology relationship;

[0023] Identify other adjacent lane lines that have potential connection relationships with the road centerline under other driving trajectories; wherein, the other driving trajectories are driving trajectories on the target lane line other than the single driving trajectory;

[0024] Based on the road centerline and its other adjacent lane lines, a second topology connection line is generated, and the intersection associated with the target road is bound to the second topology connection line to obtain a second topology relationship;

[0025] Based on the first and second topological relationships of each target road, a corresponding road-level topology is generated.

[0026] Optionally, in the above-described lightweight map construction method, for each target road, based on the road centerline and its semantic attributes, adjacent lane lines that have a potential connection relationship with the road centerline of the target road in a single driving trajectory are determined, including:

[0027] For each target road, the semantic attribute of the road centerline is assigned to the road centerline;

[0028] Determine the single driving trajectory of the road centerline with semantic attributes, and query the adjacent lane lines that have potential connection relationships with the road centerline under the single driving trajectory based on the single driving trajectory.

[0029] Optionally, in the above-described lightweight map construction method, a first topology connection line is generated based on the road centerline and its adjacent lane lines, and intersections associated with the target road are bound to the first topology connection line to obtain a first topology relationship, including:

[0030] From the adjacent lane lines that have a potential connection with the road centerline, select the first lane line that has a connection with the road centerline in a single driving trajectory;

[0031] A first topology connection line is generated based on the road centerline and the first lane lines that have a connection relationship with it.

[0032] By binding the intersection associated with the target road on the first topology connection line, a first topological relationship is established between the road centerline and the first lane line that is connected to it in a single driving trajectory.

[0033] Optionally, in the above-described lightweight map construction method, a second topology connection line is generated based on the road centerline and its other adjacent lane lines, and intersections associated with the target road are bound to the second topology connection line to obtain a second topology relationship, including:

[0034] Select a second lane line from other adjacent lane lines that have a potential connection with the road centerline under other driving trajectories;

[0035] A second topology connection line is generated based on the road centerline and the second lane lines that have a connection relationship with it.

[0036] By binding the intersection associated with the target road to the second topological connection line of the road centerline and the second lane line that has a connection relationship, a second topological relationship is established between the road centerline and the second lane line that has a connection relationship under other driving trajectories.

[0037] Optionally, in the above-described lightweight map construction method, determining the road centerline of the target road based on the target lane lines within the target road includes:

[0038] Identify the lane line inflection points on each target lane line within the target road;

[0039] Select any one of the lane line inflection points as the center of a circle, and gather all other lane line inflection points around the center of the circle to obtain the gathering point corresponding to the center of the circle;

[0040] The center line of the target road is identified from each of the target lane lines based on the aggregation point.

[0041] Optionally, in the above-described lightweight map construction method, the step of generating multiple lanes corresponding to each target road based on each target lane line within each target road, and generating a corresponding lane-level topology based on each lane and the road-level topology, includes:

[0042] Determine the lane lines that match the target lane lines within each of the target roads;

[0043] Based on the target lanes and their matching lane lines within each target road, generate multiple lanes corresponding to the target road;

[0044] For each lane, the lane centerline corresponding to the lane is determined based on the target lane line on the lane and each inflection point on the matching lane line.

[0045] The traffic markings of the lane are determined according to the direction of the lane centerline, and the lane and its traffic markings are bound together.

[0046] Generate a lane lateral sequence for the target road based on each lane of the target road;

[0047] Based on the road-level topology, identify other target roads that are associated with the target road from among the other target roads;

[0048] A lane-level topology is generated based on each lane bound to traffic markings, the lane lateral sequence corresponding to each target road, and each target road and its associated other target roads.

[0049] A second aspect of this application provides a lightweight map building system, the system comprising:

[0050] A preprocessing unit is used to acquire crowdsourced data and preprocess the crowdsourced data to obtain target crowdsourced data; wherein, the target crowdsourced data includes information on at least multiple target roads, and each target road includes multiple target lane lines with consistent direction and complete lane lines;

[0051] The intersection generation unit is used to extract the target ground markings of each target road from the crowdsourced data, and generate intersections based on the target ground markings in each target road.

[0052] A road-level topology generation unit is used to obtain the semantic attributes of the road centerline of each target road according to the nearest neighbor method, and generate a road-level topology based on the intersection, the target lane lines and the road centerline and their semantic attributes in each target road; wherein, the road centerline of the target road is determined based on the target lane lines in the target road.

[0053] The lane-level topology generation unit is used to generate multiple lanes corresponding to each target road based on each target lane line within each target road, and to generate a corresponding lane-level topology based on each lane and the road-level topology;

[0054] The lightweight map building unit is used to bind corresponding traffic signs to each lane and build a corresponding lightweight map based on the lane-level topology and each lane bound with traffic signs.

[0055] A third aspect of this application provides an electronic device, comprising:

[0056] Memory and processor;

[0057] The memory is used to store programs;

[0058] The processor is used to execute the program, which, when executed, is specifically used to implement the lightweight map construction method as described in any of the above.

[0059] This application provides a lightweight map construction method, system, and electronic device. The method involves acquiring crowdsourced data and preprocessing it to obtain target crowdsourced data. The target crowdsourced data includes information on at least multiple target roads, where the direction of each target lane line within each target road is consistent, and each target lane line is complete. Multiple intersections are generated based on the target lane lines within each target road. Road semantic information is acquired, and corresponding road-level topologies are generated based on each intersection, the road semantic information, and the target lane lines within each target road. Multiple lanes corresponding to each target road are generated based on the target lane lines within each target road, and corresponding lane-level topologies are generated based on each lane and the road-level topology. Corresponding traffic signs are bound to each lane, and a lightweight map is constructed based on the lane-level topology and the lanes bound with traffic signs. Therefore, the technical solution provided in this application only utilizes crowdsourced data in the process of building a lightweight map, without the need for high-precision maps. This not only solves the problems of high collection costs, difficult updates and maintenance, and poor freshness of high-precision maps, but also provides rich map elements, lane-level map information, and map freshness when the lightweight map is built using crowdsourced data. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0061] Figure 1 A flowchart illustrating a lightweight map construction method provided in an embodiment of this application;

[0062] Figure 2 An example diagram illustrating lane line repair provided in this application embodiment;

[0063] Figure 3 Example diagram of a method for generating road-level topology provided in an embodiment of this application;

[0064] Figure 4 An example diagram of a topology connection cable provided in an embodiment of this application;

[0065] Figure 5 An example diagram for generating a lane centerline provided in this application embodiment;

[0066] Figure 6 Example diagram of a lane-level topology method provided in this application embodiment;

[0067] Figure 7 This is a schematic diagram of a lightweight map building system provided in an embodiment of the present invention;

[0068] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0069] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0070] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0071] To better understand the technical terms used in this application, the following explanations are provided:

[0072] Lightweight maps: A map mode with accuracy between high-precision and standard-precision maps. Compared to high-precision maps, lightweight maps achieve lightweight map data by reducing data precision and non-core elements. While maintaining detailed lane-level information and necessary information such as traffic signs, they achieve efficient data transmission and reduce map production and maintenance costs by optimizing data processing algorithms and transmission protocols.

[0073] Crowdsourced data: mainly refers to road data scanned by LiDAR, panoramic cameras, etc. of intelligent driving vehicles, as well as real-time positioning data from inertial navigation systems (INS).

[0074] See Figure 1 The diagram illustrates a flowchart of a lightweight map construction method provided in an embodiment of this application. The lightweight map construction method specifically includes the following steps:

[0075] S101: Obtain crowdsourced data and preprocess the crowdsourced data to obtain target crowdsourced data; wherein, the target crowdsourced data includes information on at least multiple target roads, and each target road includes multiple target lane lines with consistent direction and complete lane lines.

[0076] In this embodiment of the application, the applicant has discovered that the lane line data provided by crowdsourced data is the foundation for building lightweight maps. However, due to the influence of sensors and the surrounding environment, the lane lines in the crowdsourced data may have problems such as incorrect direction or missing lane lines. Furthermore, the crowdsourced data cannot provide the road segment (road) to which each lane belongs. Therefore, after obtaining the crowdsourced data, the data can be preprocessed according to the direction and driving trajectory of each lane line in the crowdsourced data to correct and repair lane lines with incorrect direction or missing lane lines, thereby obtaining the corresponding target crowdsourced data. The target crowdsourced data includes at least multiple roads, and each road includes multiple target lane lines that are consistent in direction, correct, and have complete lane lines.

[0077] It should be noted that by preprocessing the crowdsourced data, the accuracy and completeness of the lightweight map built using the crowdsourced data can be guaranteed, while also improving the efficiency of lightweight map construction.

[0078] Optionally, the process of acquiring and preprocessing crowdsourced data to obtain target crowdsourced data can be as follows: Acquire crowdsourced data, which includes at least multiple lane lines; traverse each lane line to determine at least one driving trajectory for each lane line; determine whether the direction of the lane line is correct based on the at least one driving trajectory; if the direction of the lane line is incorrect, correct the direction of the lane line to obtain a lane line with the correct direction; group the target lane lines to obtain multiple roads; wherein the target lane line refers to a lane line with the correct direction, and the direction of each target lane line in each road is consistent; for each road, determine whether there are target lane lines that need lane line repair; if so, repair the target lane lines that need lane line repair in the road to obtain the target road; wherein the direction of each target lane line in the target road is consistent, correct, and the lane lines are complete; if not, determine the road as the target road.

[0079] As one implementation of this application, for each lane line, when there is only one corresponding driving trajectory for the lane line, the process of determining whether the direction of the lane line is correct based on the driving trajectory of the lane line can be as follows: determine whether the direction of the lane line and the direction of its driving trajectory are consistent; if they are consistent, the direction of the lane line can be considered correct; if they are inconsistent, the direction of the lane line can be considered incorrect.

[0080] As another implementation of this application, for each lane line, when there are multiple corresponding driving trajectories for the lane line, the process of determining whether the direction of the lane line is correct based on the driving trajectory of the lane line can be as follows: perform single-track separation on each driving trajectory of the lane line to obtain a single driving trajectory of the lane line; determine whether the direction of the lane line is consistent with the direction of its single driving trajectory; if the direction of the lane line is consistent with the direction of its single driving trajectory, determine that the direction of the lane line is correct; if the direction of the lane line is inconsistent with the direction of its single driving trajectory, determine that the direction of the lane line is incorrect.

[0081] In practical applications, since lane line direction errors are relatively random, for all lane lines in the crowdsourced data, each lane line can be traversed, and at least one driving trajectory for each lane line can be queried through the spatial buffer lookup method. For each lane line, if there are multiple corresponding driving trajectories for the lane line, the single driving trajectory can be separated according to the timestamp of each driving trajectory to separate the single driving trajectory corresponding to the lane line from each driving trajectory.

[0082] In this embodiment, after obtaining all lane lines with correct directions, each lane line with correct direction can be identified as a target lane line. This allows for the traversal of all target lane lines, filtering out multiple target lane lines that meet preset grouping conditions, and then combining these target lane lines into a road. The multiple target lane lines constituting the road include one current lane line and multiple other target lane lines. The preset grouping conditions are: the current lane line constituting the road has the same direction as the other target lane lines, and the distance between the current lane line and the other target lane lines is within a first distance range.

[0083] It should be noted that, for each lane line, if it is determined that the direction of the lane line is incorrect, the direction of the lane line can be corrected by reversing the coordinates of the midpoint of the lane line.

[0084] Optionally, the process of grouping the target lane lines to obtain multiple roads can be as follows: traverse each target lane line to determine one current lane line and multiple other lane lines; wherein, the current lane line and other lane lines are ungrouped target lane lines; from each other lane line, select at least one target other lane line that has the same direction as the current lane line and is within a first distance range, and combine the current lane line and at least one target other lane line into a road; mark the current lane line and at least one target other lane line as grouped target lane lines, and return to traverse each target lane line to determine one current lane line and multiple other lane lines, until there are no ungrouped target lane lines among the target lane lines.

[0085] It should be noted that a corresponding first distance range can be preset. For example, the first distance range can be 3.5m-3.7m. The first distance range can be set according to the actual application. This embodiment of the application does not limit it.

[0086] In practical applications, after obtaining multiple target lane lines, these spatially disordered target lane lines can be grouped according to spatial clustering and directional consistency to obtain multiple roads, each of which can be considered a road segment. Specifically, firstly, the target lane lines can be traversed to filter out ungrouped target lane lines, and any one of these ungrouped target lane lines is taken as the current lane line, with the rest as other lane lines. Secondly, spatial proximity search and directional consistency judgment are performed on the current lane line and each of the other lane lines to find at least one other target lane line that has the same direction as the current lane line and is within a first distance range. Finally, the current lane line and at least one other target lane line are combined into a road, and the current lane line and at least one other target lane line are marked as grouped target lane lines. At the same time, the traversal of each target lane line (at this time, the target lane lines are unmarked) is resumed to determine one current lane line and multiple other lane lines, until there are no ungrouped target lane lines among the target lane lines.

[0087] In this embodiment, after obtaining multiple roads, for any target lane line of each road, the target lane line can be used as a seed lane line, and it is determined whether there is a difference in the direction of the start and end positions of the seed lane line. If there is, the average direction of the start and end positions of the seed lane line can be calculated, and an extension line can be drawn at both ends (start and end positions) with the average direction. At the same time, a buffer of a preset distance is set to filter out other target lane lines that intersect with the extension line from other target lane lines of the road as candidate lane lines. Each candidate lane line is traversed, and the distance between the two nearest endpoints of the seed lane line and the candidate lane line is calculated. If there is a seed lane line and a candidate lane line with the shortest distance and the closest direction, it is considered that the two lane lines need to be repaired, and then these two lane lines can be repaired.

[0088] It should be noted that there are multiple lane shape points (nodes) on a lane line. That is to say, the starting position of the target lane line is the first lane shape point on the target lane line, and the ending position is the last lane shape point on the target lane line. The average direction can be calculated by taking the direction angle between two adjacent shape points. For example, if there are 10 lane shape points on the target lane line, the direction angle can be calculated 9 times, and then the average of the 9 direction angles is the average direction.

[0089] For example, see Figure 2 We can assume that L2 is the seed lane line, and that the directions of the start and end positions of L2 are different. In this case, we can calculate the average direction of the start and end positions of the seed lane line, and extend the line at both ends (start and end positions) with this average direction. At the same time, we set a buffer with a preset distance to select other target lane lines that intersect with the extended line as candidate lane lines from other target lane lines on the road. We traverse each candidate lane line and find that L4 is the shortest distance from L2 and the direction is the closest, that is, L4 and L2 need to be repaired. In order to reassemble the inflection points of the two lane lines (L2 and L4), the missing parts are filled in by interpolation and a new lane line is generated. L5 in the figure is the repaired lane line. After the inflection point reassembly, L2, L4 and L5 will be merged into one lane line.

[0090] S102: Generate multiple intersections based on the ground markings of each target road.

[0091] The target road markings for each target road are extracted from crowdsourced data.

[0092] In this embodiment, intersections are a crucial element of the map. Traffic conditions at intersections are complex and varied; therefore, to ensure the safe operation of autonomous vehicles, the locations of intersections need to be marked in the lightweight map. Although high-precision data contains the topological relationships between intersections and lane lines, acquiring such data is not only costly but also difficult to update. Therefore, this application utilizes crowdsourced data to construct a lightweight map, reducing costs and data update difficulty. However, crowdsourced data does not contain precise intersection information and cannot directly mark intersection locations. Therefore, after obtaining the target crowdsourced data, clustering can be used to identify multiple target ground markings such as zebra crossings and stop lines from the various target roads in the target crowdsourced data. Since zebra crossings and stop lines exist within intersections or at their boundaries, the initial intersection corresponding to each zebra crossing and stop line can be determined. Since zebra crossings and stop lines are spatially similar, the nearest neighbor matching algorithm can be used to determine zebra crossings and stop lines that are close in distance (the distance between zebra crossings and stop lines is within a preset distance range) from each zebra crossing and stop line. Two initial intersections with close zebra crossings and stop lines are identified as the same intersection, thereby enabling the selection of at least one intersection from each initial intersection.

[0093] S103: Obtain road semantic information and generate corresponding road-level topology based on intersections, road semantic information, and target lane lines within each target road.

[0094] In the specific execution step S103, after identifying at least one intersection, road semantic information can be further acquired, and the road centerline corresponding to each target road can be identified from each target lane line within each target road using the local master curve algorithm. Based on the road semantic information and the road centerline corresponding to each target road, the lane lines connected to each road centerline are determined. Finally, the corresponding road-level topology is constructed based on each intersection, the road centerline corresponding to each target road, and the lane lines with which they are connected.

[0095] It should be noted that the Local Principal Curve algorithm can extract the intrinsic structure and features of the data more effectively. The Local Principal Curve algorithm emphasizes finding a smooth one-dimensional curve that passes through the "middle" of the data distribution and satisfies "self-consistency". Therefore, the Local Principal Curve algorithm can identify the corresponding road centerline from each target lane line.

[0096] Optionally, the process of generating a road-level topology can be as follows: Obtain road semantic information; for each target road, identify the road centerline from the target lane lines based on the lane line inflection points; read the semantic attributes of the road centerline from the road semantic information and assign these attributes to the road centerline to obtain the target road centerline; determine the single-trip trajectory of the target road centerline and query the adjacent lane lines that have a potential connection with the target road centerline based on the single-trip trajectory; select the first lane line that has a connection with the target road centerline under the single-trip trajectory from the adjacent lane lines of the target road centerline; generate a corresponding topology connection line based on the target road centerline and the first lane line with a connection, and bind the corresponding intersection to the topology connection line to establish the connection between the target road centerline and its adjacent lane lines under the single-trip trajectory. The topological relationships between first lane lines with existing connections are established. For other driving trajectories, other adjacent lane lines with potential connections to the target road centerline are queried based on other driving trajectories of the target road centerline. These other driving trajectories are driving trajectories other than a single driving trajectory from the target lane line. Second lane lines with connections to the target road centerline under other driving trajectories are selected from each of the other adjacent lane lines of the target road centerline. Corresponding topology connection lines are generated based on the target road centerline and its connected second lane lines, and corresponding intersections are bound to these connection lines to establish the topological relationships between the target road centerline and its connected second lane lines under other driving trajectories. Based on the topological relationships between the target road centerline and its connected lane lines under all driving trajectories, a corresponding road-level topology is generated. Each lane line with a connection to the target road centerline includes first lane lines and second lane lines.

[0097] It should be noted that semantic attributes can include road name, road type, speed limit information, etc., but are not limited to these in this embodiment.

[0098] As a preferred embodiment of this application, the process of identifying the road centerline from each target lane line in the target road can be as follows: for each target road, identify the lane line inflection points on each target lane line in the target road; select any lane line inflection point as the center of a circle, and aggregate all other lane line inflection points with the center of the circle to obtain the aggregation point corresponding to the center of the circle; identify the road centerline from each target lane line based on the aggregation point.

[0099] In practical applications, the applicant's research revealed that the spatial distribution of lane line inflection points is disordered, and the road centerline is the median line among all lane line inflection points. Therefore, for each target road, the lane line inflection points on each target lane line within the target road can be identified first. Then, any lane line inflection point can be selected as the center of a circle, and a circle can be drawn with the center and a certain radius. Since the points within the circle can be considered to be clustered, the average value of each lane line inflection point within the circle can be calculated to obtain the cluster point corresponding to the center. Starting from the obtained cluster point, the target lane line inflection point with a preset step distance from the cluster point can be found among all lane line inflection points. This target lane line inflection point is then used as the center of the circle, and the process of drawing a circle with the center and a certain radius is repeated until the target lane line inflection point corresponding to the last cluster point is found. Finally, the median line among all target lane line inflection points is determined as the road centerline of the target road; where the road centerline is the target lane line within the target road.

[0100] In this embodiment, since the identified road centerline lacks semantic information, such as road name and road class, after identifying the road centerline, the nearest neighbor method is used to find roads in the nearby Open Street Map (OSM). Specifically, the OSM road corresponding to the road centerline is determined through constraints such as shape similarity and direction consistency. The semantic attributes corresponding to the OSM road are then read from the road semantic information and assigned as the semantic attributes of the road centerline, thus obtaining the target road centerline. The target road centerline is the road centerline assigned the corresponding semantic attributes.

[0101] In practical applications, after obtaining the centerline of the target road, the single-trip trajectory of the target road centerline can be further identified. This allows for the determination of lane lines connected to the target road centerline based on these single-trip trajectories. Furthermore, a corresponding road-level topology is generated based on each intersection, each target road centerline, and their connected lane lines. Specifically, as follows... Figure 3 As shown;

[0102] A1: Find all driving trajectories on the centerline of the target road through the spatial buffer, and separate the single driving trajectory of the centerline of the target road from all driving trajectories on the centerline of the target road.

[0103] It should be noted that each driving trajectory may be the trajectory of different vehicles at different times, and the direction of each driving trajectory may be slightly different. The different directions of driving trajectories indicate different connections between the center line of the target road and other lane lines. Each driving trajectory is separated into individual driving trajectories according to the timestamp corresponding to each driving trajectory, so as to separate the individual driving trajectory corresponding to the lane line from each driving trajectory. Since each driving trajectory has a unique ID, the driving trajectories of different vehicles at different times can be distinguished according to the ID of each driving trajectory. Therefore, if the separated individual driving trajectory has been driven multiple times on the center line of the target road, the individual driving trajectory of the center line of the target road can be further determined according to the time of each drive.

[0104] A2: Expand the single driving trajectory of the target road centerline to obtain the target single driving trajectory, and select at least one adjacent lane line that has a potential connection with the target road centerline from the set of road centerlines based on the target single driving trajectory.

[0105] It should be noted that the set of road centerlines includes the target road centerlines of all target roads except the target road centerline of the current target road.

[0106] In some embodiments, the single driving trajectory can be extended forward and backward by a preset distance (e.g., 50 meters) to obtain the target single driving trajectory. Then, the target single driving trajectory is used to perform a spatial proximity search on the set of road centerlines. Other target road centerlines whose distance from the target road centerline is within a preset distance range are identified as adjacent roads that have a potential connection relationship with the target lane centerline.

[0107] A3: Determine whether the trajectory between the centerline of the adjacent road and the centerline of the target road passes through other lane lines. If the trajectory between the adjacent road and the centerline of the target road passes through other lane lines, it is determined that the adjacent road is not connected to the centerline of the target road; if the trajectory between the adjacent road and the centerline of the target road does not pass through other lane lines, proceed to A4.

[0108] A4: Determine that the adjacent lane is the first lane line that is connected to the center line of the target road in a single driving trajectory, and determine the driving trajectory state between the center line of the target road and the first lane line that is connected to it.

[0109] It should be noted that since there must be a trajectory between two lane lines that are definitely connected, when the adjacent road is determined to be the first lane line that is connected to the center line of the target road, the order of passing through these two lane lines can be determined by the time it takes to pass through the center line of the target road and the first lane line that is connected. For example, the order of passing through these two lane lines can be determined based on the time of entering or exiting these two lane lines. Based on the determined order, the trajectory between the center line of the target road and the first lane line that is connected can be extracted, and the state of the trajectory can be calculated by calculating the cumulative heading angle, thus obtaining the corresponding driving trajectory state. The driving trajectory state can be straight, turning, or U-turn, etc., which is not limited in this embodiment of the application.

[0110] It should also be noted that using trajectories to determine the order in which two lane lines pass is not only simple and efficient, but also reduces the computational resources required for spatial relationships between road segments.

[0111] A5: Generate the topology connection line between the target road centerline and the first lane line that is connected to it, based on the driving trajectory status between the target road centerline and the first lane line that is connected to it.

[0112] In some embodiments, the target road centerline and the first lane line that are connected can be connected according to the driving trajectory state between the target road centerline and the first lane line that are connected to it to obtain the corresponding topology hook line.

[0113] For details, please refer to [link / reference]. Figure 4 Assuming the target road centerline is *perlink*, and the first lane connected to it is *nextlink*, if the travel trajectory of *perlink* and the connected *nextlink* is straight, the end point of *perlink* and the start point of *nextlink* can be directly connected to obtain the corresponding topology connection line. If the travel trajectory is turning, the end point of *perlink*, the start point of *nextlink*, and the intersection of the extensions of *perlink* and *nextlink* will be used as control points to connect the end point of *perlink* and the start point of *nextlink*, thus obtaining the corresponding topology connection line. From... Figure 4 As can be seen, the first lane line is the center line of the target road of another target road.

[0114] A6: Identify the intersections from each intersection that pass through the target road centerline and the first lane line that is connected to it, and bind the intersections that pass through the topological connection line to the topological connection line to establish the topological relationship between the target road centerline and the first lane line that is connected to it in a single driving trajectory. At the same time, record the ID of the target road centerline and its first lane line corresponding to the topological connection line.

[0115] A7: For other driving trajectories, query other adjacent lane lines that have potential connections with the center line of the target road based on other driving trajectories of the target road center line; among them, other driving trajectories are driving trajectories other than a single driving trajectory on the target lane line.

[0116] A8: Select the second lane line from the other adjacent lane lines of the target road centerline that has a connection relationship with the target road centerline under other driving trajectories.

[0117] A9: Obtain the ID of the second lane line, and based on the ID of the second lane line and the ID of the target road center line, determine whether a topology connection already exists between the target road center line and the second lane line; if it already exists, the second lane line can be ignored; if it does not exist, execute A10.

[0118] A10: Generate a corresponding topology connection line based on the target road centerline and its connected second lane lines, and bind the corresponding intersection on the topology connection line to establish the topology relationship between the target road centerline and its connected second lane lines under other driving trajectories.

[0119] It should be noted that the method for establishing the topological relationship between the target road centerline and the second lane line that is connected to it under other driving trajectories is the same as the method for establishing the topological relationship between the target road centerline and the first lane line that is connected to it under a single driving trajectory. Please refer to the contents shown in A5-A6. This embodiment of the application will not be described again here.

[0120] A11: Generate the corresponding road-level topology based on the topological relationships between the target road centerline and the lane lines that are connected to it under all driving trajectories.

[0121] Among them, the topological relationship between the centerline of the target road and the lane lines that are connected is the topological relationship between the centerline of the target road and the first or second lane line that is connected.

[0122] S104: Generate multiple lanes corresponding to each target road based on each target lane line within each target road, and generate corresponding lane-level topology based on each lane and road-level topology.

[0123] In the specific execution step S104, for each target road, multiple lanes corresponding to the target road can be generated based on each target lane line within the target road, wherein each lane consists of two matching target lane lines; the lane center line within each lane is identified, and corresponding traffic markings are bound to each lane based on the lane center line within each lane; finally, a lane-level topology is generated based on the road-level topology, each target road, and the lanes bound with traffic markings.

[0124] Optionally, the lane-level topology generation process can be as follows: For each target lane line within each target road, the target lane line is paired with other lanes to identify lane lines matching the target lane line from the target road, and multiple lanes are generated based on the target lane line and its matching lane lines; for each lane, the lane centerline corresponding to the lane is determined based on the target lane line and each inflection point on its matching lane line; the traffic markings of the lane are determined based on the direction of the lane centerline, and the lane and its traffic markings are bound together; the lanes corresponding to the target road are sorted by their lateral positions to obtain the lateral sequence of lanes corresponding to the target road; other target roads associated with the target road are identified from other target roads based on the road-level topology; and the corresponding lane-level topology is generated based on the lanes with bound traffic markings, the lateral sequence of lanes corresponding to each target road, and each target road and its associated other target roads.

[0125] In some embodiments, for each target lane line within a target road, the target lane line can be paired with other target lane lines within the target road using nearest neighbor search and spatial sorting to identify lane lines matching the target lane line from each of the target roads. The lane lines matching the target lane line are other target lane lines within the target road besides the target lane line itself.

[0126] In practical application, the applicant's research revealed that in a lightweight map, a lane is generated by two lane lines and a lane centerline formed by those lane lines. The lane centerline represents the geometric center line of the lane and does not exist in the actual road; its primary purpose is to provide a reference driving path for vehicles traveling within that lane. For details, please refer to [link to relevant documentation]. Figure 5For each target lane line ({L1, L2, L3, L4}) within each target road, the target lane lines are paired with other lane lines using nearest neighbor search and spatial sorting. That is, one target lane line is paired with another target lane line of suitable distance and similar length to form the two edge lines of a lane, resulting in {{L1, L2}, {L2, L3}, and {L3, L4}}. Taking {L1, L2} as an example, the inflection points of the two edge lines are {p1, p2, p3, p4, p5, p6, p... 7}, {q1,q2,p3,p4,q5,q6}. For the two edge lines {q1,q2,p3,p4,q5,q6}, perform nearest neighbor search to pair the inflection points, resulting in multiple inflection point pairs {{p1,q1}, {p2,q2}, {p3,q3}, {p4,q4}, {p5,q5}, {p6,q6}, {p7,q6}}. Calculate the average of the two inflection points in each pair to obtain the center inflection point of each pair. The first and last points of the two edge lines are fixed and no nearest neighbor search is performed; the center inflection point is calculated directly. Finally, connect all the center inflection points according to the driving directions of the two edge lines to obtain the lane centerline, R1, formed by the two edge lines L1 and L2.

[0127] It should be noted that the distance between the target lane line constituting a lane and its matching lane line is within a preset distance range, and the length difference is also within a preset length range. The preset distance range and preset length range can be set according to actual applications, and are not limited in this embodiment.

[0128] In some embodiments, since the road-level topology previously constructed establishes the final connectivity relationships between road networks, it can be used as prior information for constructing the lane-level topology. This means that the road-level topology can be used to identify other target roads associated with the current target road from among the various other target roads. Here, "other target roads" refers to all target roads other than the current target road.

[0129] Specifically, to facilitate understanding, the following explanation will be provided using examples, such as... Figure 6 As shown.

[0130] B1: For each target road, determine other target roads that have a topological relationship with the target road based on the road-level topology, and identify other target roads that have a topological relationship with the target road as other target roads that are associated with the target road.

[0131] It should be noted that the topological relationship between the target road and other target lanes refers to the topological relationship between the target road centerline and the first lane line that is connected to it in a single driving trajectory, or the topological relationship between the target road centerline and the second lane line that is connected to it in other driving trajectories; wherein, the first lane line or the second lane line is the target road centerline within other target roads.

[0132] Specifically, in road-level topology, the topological relationship between two roads can be expressed through topological joints and road nodes (intersections). Therefore, the way to find other target roads (forward or backward roads) that are related to the target road is as follows: find the topological connection line that connects to the start or end point of the target road's centerline, find the first lane line / second lane line that has the same end or start point as the topological connection line, and determine the target road to which the found first lane line / second lane line belongs as another target road that is related to the target road.

[0133] Suppose that during the forward search, a road is found and recorded as nextlink (indicating the road entered), then the target road is recorded as prelink (indicating the road exited). At this point, the topology connection line between prelink and nextlink can be extracted, and the driving state trajectory corresponding to this topology connection line can also be identified. The driving trajectory status can be straight, left turn, right turn, U-turn, etc.

[0134] B2: For each lane in the target road, obtain the direction of the lane centerline and determine the traffic markings of the lane based on the direction of the lane centerline, so as to bind the lane and its traffic markings.

[0135] In practical applications, for each lane within the target road, after identifying the lane centerline, a proximity search can be performed using that centerline to extract nearby traffic markings. Typically, multiple traffic markings may exist under a single lane, therefore, priority determination is necessary for the extracted markings. Specifically, if the extracted traffic markings are of the same type, no priority determination is needed. If the types are different, and the lane is an exit lane, the lane turning type can be determined based on the type of traffic marking near the exit lane, i.e., the traffic marking near the exit lane is identified as the final traffic marking for that lane. If it is an entry lane, the lane turning type can be determined based on the type of traffic marking near the entry lane, i.e., the traffic marking near the exit lane is identified as the final traffic marking for that lane. If no traffic markings are found for a lane, the lane type needs to be determined by searching for the driving trajectory passing through that lane, so that the traffic markings for that lane can be determined based on the lane type. If no corresponding driving trajectory exists for a lane, the lane's turning attribute is empty, and the lane can be ignored; that is, the lane will not be included in the topology analysis.

[0136] For example, suppose there are three types of traffic markings for the center line of a lane, ordered from largest to smallest distance from the exit lane: straight, left turn, and left turn plus U-turn; then the priority of these three markings is left turn plus U-turn, left turn, and straight, and finally the traffic markings of the left turn plus U-turn type can be used as the final traffic markings for that lane.

[0137] B3: The relative positional relationship between lanes within the road is determined by the vector product method, so as to sort the lanes laterally and obtain the lane lateral sequence.

[0138] For example, if the prelink and nextlink are connected in a straight line, then the straight lane under the prelink must be connected to the lane in the nextlink that is in the same lateral position. Otherwise, a lane connection misalignment problem will occur. Therefore, by sorting the lanes laterally, the connection misalignment problem between lanes can be avoided.

[0139] B4: Generate the corresponding lane-level topology based on each lane bound to traffic markings, the lateral sequence of lanes corresponding to each target road, and each target road and its associated other target roads.

[0140] In practical applications, see Figure 4When the topology connection between the prelink and the nextlink is a straight-through connection, the straight-through lane under the prelink connects to the corresponding lane under the nextlink; when the topology connection between the prelink and the nextlink is a right-turn connection, the lane with right-turn attribute or the rightmost straight-through lane under the prelink connects to the lane in the nextlink; when the topology connection between the prelink and the nextlink is a left-turn connection, the lane with left-turn attribute under the prelink connects to the lane in the nextlink.

[0141] S105: Bind the corresponding traffic sign to each lane, and build the corresponding lightweight map based on the lane-level topology and each lane with the bound traffic sign.

[0142] In this embodiment of the application, traffic signs corresponding to each lane can be obtained from crowdsourced data through spatial lookup, and traffic signs for each lane can be bound to the corresponding lane, so as to construct a corresponding lightweight map based on the lane-level topology and each lane bound with traffic signs.

[0143] It should be noted that traffic signs refer to information from crowdsourced data, such as speed limits, weight limits, and no-entry signs; traffic markings refer to ground markings from crowdsourced data, such as traffic arrows (straight ahead, right turn, etc.), stop lines, and zebra crossings. Traffic signs and markings provide important semantic information for the lightweight map. This semantic information needs to be bound to the corresponding lanes and transmitted to the autonomous vehicle through lane information to ensure the safe and stable driving of the autonomous vehicle.

[0144] It should also be noted that dot-shaped traffic signs are usually associated with one road and one lane, but line-shaped traffic markings may be associated with multiple roads and lanes.

[0145] This application provides a lightweight map construction method. It obtains target crowdsourced data by acquiring and preprocessing the crowdsourced data. The target crowdsourced data includes information on at least multiple target roads, where the direction of each target lane line is consistent and the lane lines are complete. Multiple intersections are generated based on the target lane lines of each target road. Road semantic information is acquired, and corresponding road-level topologies are generated based on each intersection, road semantic information, and target lane lines of each target road. Multiple lanes are generated for each target road based on the target lane lines, and corresponding lane-level topologies are generated based on each lane and the road-level topology. Corresponding traffic signs are bound to each lane, and a lightweight map is constructed based on the lane-level topology and the lanes bound to traffic signs. Therefore, the technical solution provided in this application utilizes only crowdsourced data in the lightweight map construction process, without the need for high-precision maps. This not only solves the problems of high acquisition costs, difficult updates and maintenance, and poor timeliness of high-precision maps, but also provides rich map elements, lane-level map information, and map timeliness for lightweight maps constructed using crowdsourced data.

[0146] Based on the lightweight map construction method provided in the above embodiments of this application, correspondingly, the present application also provides a lightweight map construction system, such as... Figure 7 As shown, this lightweight map building system includes:

[0147] The preprocessing unit 71 is used to acquire crowdsourced data and preprocess the crowdsourced data to obtain target crowdsourced data; wherein, the target crowdsourced data includes information on at least multiple target roads, and each target road includes multiple target lane lines with the same direction and complete lane lines;

[0148] The intersection generation unit 72 is used to generate multiple intersections based on the target lane lines within each target road.

[0149] The road-level topology generation unit 73 is used to acquire road semantic information and generate corresponding road-level topology based on each intersection, road semantic information and each target lane line in each target road;

[0150] The lane-level topology generation unit 74 is used to generate multiple lanes corresponding to each target road based on each target lane line in each target road, and to generate the corresponding lane-level topology based on each lane and the road-level topology.

[0151] The lightweight map building unit 75 is used to bind corresponding traffic signs to each lane and build corresponding lightweight maps based on the lane-level topology and each lane bound with traffic signs.

[0152] The specific principles and execution processes of each unit in the lightweight map construction system disclosed in the above embodiments of this application are the same as those in the lightweight map construction method disclosed in the above embodiments of this application. Please refer to the corresponding parts in the lightweight map construction method disclosed in the above embodiments of this application, and they will not be repeated here.

[0153] This application provides a lightweight map construction method. It obtains target crowdsourced data by acquiring and preprocessing the crowdsourced data. The target crowdsourced data includes information on at least multiple target roads, where the direction of each target lane line is consistent and the lane lines are complete. Multiple intersections are generated based on the target lane lines of each target road. Road semantic information is acquired, and corresponding road-level topologies are generated based on each intersection, road semantic information, and target lane lines of each target road. Multiple lanes are generated for each target road based on the target lane lines, and corresponding lane-level topologies are generated based on each lane and the road-level topology. Corresponding traffic signs are bound to each lane, and a lightweight map is constructed based on the lane-level topology and the lanes bound to traffic signs. Therefore, the technical solution provided in this application utilizes only crowdsourced data in the lightweight map construction process, without the need for high-precision maps. This not only solves the problems of high acquisition costs, difficult updates and maintenance, and poor timeliness of high-precision maps, but also provides rich map elements, lane-level map information, and map timeliness for lightweight maps constructed using crowdsourced data.

[0154] Optional, the preprocessing unit includes:

[0155] The first acquisition unit is used to acquire crowdsourced data, wherein the crowdsourced data includes at least multiple lane lines;

[0156] The first determining unit is used to traverse each lane line to determine at least one driving trajectory for each lane line;

[0157] The first judgment unit is used to determine whether the direction of the lane line is correct based on at least one driving trajectory of the lane line;

[0158] The direction correction unit is used to correct the direction of the lane lines if the direction of the lane lines is incorrect, so as to obtain lane lines with the correct direction.

[0159] A grouping unit is used to group the target lane lines into multiple roads; wherein, the target lane line refers to the lane line with the correct direction; the direction of each target lane line in each road is consistent;

[0160] The second judgment unit is used to determine, for each road, whether there are target lane lines that need to be repaired.

[0161] The lane line repair unit is used to repair the target lane lines that need to be repaired within the road, if they exist, to obtain the target road; wherein, the direction of each target lane line within the target road is consistent, correct, and the lane lines are complete.

[0162] The second determining unit is used to determine the road as the target road if it does not exist.

[0163] Optionally, the first judgment unit includes:

[0164] The third judgment unit is used to determine whether the direction of the lane line is consistent with the direction of its driving trajectory if there is a driving trajectory for the lane line.

[0165] The third determining unit is used to determine that the direction of the lane line is correct if the direction of the lane line is consistent with the direction of its driving trajectory.

[0166] The single-trip trajectory separation unit is used to separate each travel trajectory into a single-trip trajectory if there are multiple travel trajectories for the lane line, so as to obtain the single-trip trajectory corresponding to the lane line.

[0167] The fourth judgment unit is used to determine whether the direction of the lane line is consistent with the direction of its single driving trajectory;

[0168] The fourth determining unit is used to determine that the direction of the lane line is correct if the direction of the lane line is consistent with the direction of its single driving trajectory.

[0169] Optional grouping units include:

[0170] The fifth determination unit is used to traverse each target lane line to determine one current lane line and multiple other lane lines; wherein, the current lane line and other lane lines are ungrouped target lane lines;

[0171] A road composition unit is used to select at least one target other lane line from each other lane line that is in the same direction as the current lane line and is within a first distance range, and to combine the current lane line and at least one target other lane line into a road.

[0172] The marking unit is used to mark the current lane line and at least one other target lane line as grouped target lane lines, and then return to execute the fifth determination unit until there are no ungrouped target lane lines among the target lane lines.

[0173] Optional, road-level topology generation units include:

[0174] The second acquisition unit is used to acquire road semantic information;

[0175] The road centerline recognition unit is used to identify the road centerline from each target lane line in each target road based on the lane line inflection points in each target lane line within the target road.

[0176] The assignment unit is used to read the semantic attributes of the road centerline from the road semantic information and assign the semantic attributes of the road centerline to the road centerline to obtain the target road centerline;

[0177] The first query unit is used to determine the single driving trajectory of the target road centerline and query the adjacent lane lines that have a potential connection with the target road centerline based on the single driving trajectory of the target road centerline.

[0178] The first filtering unit is used to filter out the first lane line that has a connection relationship with the center line of the target road in a single driving trajectory from each adjacent lane line of the target road center line.

[0179] The first establishment unit is used to generate a corresponding topology connection line based on the target road centerline and the first lane line that has a connection relationship with it, and to bind the corresponding intersection on the topology connection line to establish the topology relationship between the target road centerline and the first lane line that has a connection relationship with it in a single driving trajectory.

[0180] The second query unit is used to query other adjacent lane lines that have a potential connection with the center line of the target road based on other driving trajectories of the target road center line; wherein, other driving trajectories are driving trajectories other than the single driving trajectory of the target lane line.

[0181] The second filtering unit is used to filter out the second lane lines that have a connection relationship with the center line of the target road under other driving trajectories from each of the other adjacent lane lines of the target road center line.

[0182] The second establishment unit is used to generate corresponding topology connection lines based on the target road centerline and its connected second lane lines, and to bind corresponding intersections on the topology connection lines of the target road centerline and its connected second lane lines to establish the topology relationship between the target road centerline and its connected second lane lines under other driving trajectories.

[0183] The road-level topology generation subunit is used to generate the corresponding road-level topology based on the topological relationship between the target road centerline and the lane lines that are connected to it under all driving trajectories; wherein, the topological relationship between the target road centerline and the lane lines that are connected is the topological relationship between the target road centerline and the first lane line or the second lane line that are connected.

[0184] Optionally, the road centerline recognition unit includes:

[0185] The lane line inflection point recognition unit is used to identify lane line inflection points on each target lane line within the target road for each target road.

[0186] The aggregation unit is used to select any lane line inflection point from all lane line inflection points as the center, and aggregate all other lane line inflection points with the center to obtain the aggregation point corresponding to the center.

[0187] The road centerline identification subunit is used to identify the road centerline within the target road from each target lane line based on the aggregation point.

[0188] Optionally, the lane-level topology generation unit includes:

[0189] The lane generation unit is used to perform lane matching for each target lane line in each target road, so as to identify the lane line that matches the target lane line from the target road, and generate the corresponding lane based on the target lane line and its matching lane line.

[0190] The lane centerline determination unit is used to determine the lane centerline corresponding to each lane based on the target lane line on the lane and the inflection points on the matching lane line.

[0191] The binding unit is used to determine the traffic markings of the lane according to the direction of the lane center line and bind the lane and its traffic markings.

[0192] The lane transverse position sorting unit is used to sort the transverse positions of each lane corresponding to the target road to obtain the transverse sequence of lanes corresponding to the target road.

[0193] The associated road identification unit is used to identify other target roads that are associated with the target road from among other target roads based on the road-level topology.

[0194] The lane-level topology generation subunit is used to generate the corresponding lane-level topology based on each lane bound to traffic markings, the lane lateral sequence corresponding to each target road, and each target road and its associated other target roads.

[0195] This application provides an electronic device, such as... Figure 8 As shown, the electronic device includes a processor 801 and a memory 802. The memory 802 is used to store the program code and related data of the lightweight map construction method provided in any of the above embodiments. The processor 801 is used to call the program instructions in the memory 802 to execute the steps shown in the lightweight map construction method in the above embodiments.

[0196] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0197] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0198] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0199] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A lightweight map construction method, characterized in that, The method includes: Crowdsourced data is acquired and preprocessed to obtain target crowdsourced data; wherein, the target crowdsourced data includes information on at least multiple target roads, and each target road includes multiple target lane lines with consistent direction and complete lane lines; Extract the target ground markings for each target road from the crowdsourced data, and generate intersections based on the target ground markings within each target road; The semantic attributes of the road centerline of each target road are obtained according to the nearest neighbor method, and a road-level topology is generated based on the intersection, the target lane lines and the road centerline and their semantic attributes in each target road; wherein, the road centerline of the target road is determined based on the target lane lines in the target road. Based on each target lane line within each target road, generate multiple lanes corresponding to each target road, and generate a corresponding lane-level topology based on each lane and the road-level topology; Each lane is bound to a corresponding traffic sign, and a corresponding lightweight map is constructed based on the lane-level topology and each lane bound to a traffic sign.

2. The method according to claim 1, characterized in that, The process of acquiring crowdsourced data and preprocessing the crowdsourced data to obtain target crowdsourced data includes: Acquire crowdsourced data, wherein the crowdsourced data includes at least multiple lane lines; Traverse each lane line to determine at least one driving trajectory for each lane line; Determine whether the direction of the lane line is correct based on at least one driving trajectory of the lane line; If the direction of the lane line is incorrect, the direction of the lane line is corrected to obtain a lane line with the correct direction; The target lane lines are grouped to obtain multiple roads; wherein, the target lane line refers to the lane line with the correct direction; the direction of each target lane line in each road is consistent; For each of the roads, determine whether there are any target lane lines in the road that require lane line repair; If present, the target lane lines within the road that require lane line repair are repaired to obtain the target road; wherein, the direction of each target lane line within the target road is consistent, correct, and the lane lines are complete. If it does not exist, the road is identified as the target road.

3. The method according to claim 1, characterized in that, The semantic attributes of the road centerline of each target road are obtained using the nearest neighbor method, and a road-level topology is generated based on the intersections, target lane lines within each target road, and road centerlines and their semantic attributes, including: For each target road, based on the road centerline and its semantic attributes, determine the adjacent lane lines that have a potential connection relationship with the road centerline of the target road in a single driving trajectory; Based on the road centerline and its adjacent lane lines, a first topology connection line is generated, and the intersection associated with the target road is bound to the first topology connection line to obtain a first topology relationship; Identify other adjacent lane lines that have potential connection relationships with the road centerline under other driving trajectories; wherein, the other driving trajectories are driving trajectories on the target lane line other than the single driving trajectory; Based on the road centerline and its other adjacent lane lines, a second topology connection line is generated, and the intersection associated with the target road is bound to the second topology connection line to obtain a second topology relationship; Based on the first and second topological relationships of each target road, a corresponding road-level topology is generated.

4. The method according to claim 3, characterized in that, For each target road, based on the road centerline and its semantic attributes, adjacent lane lines that have a potential connection with the road centerline of the target road in a single driving trajectory are determined, including: For each target road, the semantic attribute of the road centerline is assigned to the road centerline; Determine the single driving trajectory of the road centerline with semantic attributes, and query the adjacent lane lines that have potential connection relationships with the road centerline under the single driving trajectory based on the single driving trajectory.

5. The method according to claim 3, characterized in that, Based on the road centerline and its adjacent lane lines, a first topology connection line is generated, and intersections associated with the target road are bound to the first topology connection line to obtain a first topology relationship, including: From the adjacent lane lines that have a potential connection with the road centerline, select the first lane line that has a connection with the road centerline in a single driving trajectory; A first topology connection line is generated based on the road centerline and the first lane lines that have a connection relationship with it. By binding the intersection associated with the target road on the first topology connection line, a first topological relationship is established between the road centerline and the first lane line that is connected to it in a single driving trajectory.

6. The method according to claim 3, characterized in that, Based on the road centerline and its other adjacent lane lines, a second topology connection line is generated, and intersections associated with the target road are bound to the second topology connection line to obtain a second topology relationship, including: Select a second lane line from other adjacent lane lines that have a potential connection with the road centerline under other driving trajectories; A second topology connection line is generated based on the road centerline and the second lane lines that have a connection relationship with it. By binding the intersection associated with the target road to the second topological connection line of the road centerline and the second lane line that has a connection relationship, a second topological relationship is established between the road centerline and the second lane line that has a connection relationship under other driving trajectories.

7. The method according to claim 1, characterized in that, Determining the road centerline of the target road based on the target lane lines within the target road includes: Identify the lane line inflection points on each target lane line within the target road; Select any one of the lane line inflection points as the center of a circle, and gather all other lane line inflection points around the center of the circle to obtain the gathering point corresponding to the center of the circle; The center line of the target road is identified from each of the target lane lines based on the aggregation point.

8. The method according to claim 1, characterized in that, The step of generating multiple lanes corresponding to each target road based on each target lane line within each target road, and generating a corresponding lane-level topology based on each lane and the road-level topology, includes: Determine the lane lines that match the target lane lines within each of the target roads; Based on the target lanes and their matching lane lines within each target road, generate multiple lanes corresponding to the target road; For each lane, the lane centerline corresponding to the lane is determined based on the target lane line on the lane and each inflection point on the matching lane line. The traffic markings of the lane are determined according to the direction of the lane centerline, and the lane and its traffic markings are bound together. Generate a lane lateral sequence for the target road based on each lane of the target road; Based on the road-level topology, identify other target roads that are associated with the target road from among the other target roads; A lane-level topology is generated based on each lane bound to traffic markings, the lane lateral sequence corresponding to each target road, and each target road and its associated other target roads.

9. A lightweight map building system, characterized in that, The system includes: A preprocessing unit is used to acquire crowdsourced data and preprocess the crowdsourced data to obtain target crowdsourced data; wherein, the target crowdsourced data includes information on at least multiple target roads, and each target road includes multiple target lane lines with consistent direction and complete lane lines; The intersection generation unit is used to extract the target ground markings of each target road from the crowdsourced data, and generate intersections based on the target ground markings in each target road. A road-level topology generation unit is used to obtain the semantic attributes of the road centerline of each target road according to the nearest neighbor method, and generate a road-level topology based on the intersection, the target lane lines and the road centerline and their semantic attributes in each target road; wherein, the road centerline of the target road is determined based on the target lane lines in the target road. The lane-level topology generation unit is used to generate multiple lanes corresponding to each target road based on each target lane line within each target road, and to generate a corresponding lane-level topology based on each lane and the road-level topology; The lightweight map building unit is used to bind corresponding traffic signs to each lane and build a corresponding lightweight map based on the lane-level topology and each lane bound with traffic signs.

10. An electronic device, characterized in that, include: Memory and processor; The memory is used to store programs; The processor is used to execute the program, which, when executed, is specifically used to implement the lightweight map construction method as described in any one of claims 1 to 8.

Images