Map building methods, devices, equipment, storage media and program products

Abstract

This application provides a map construction method, apparatus, device, storage medium, and program product. The method includes: determining initial trajectory data and performing road matching processing on the initial trajectory data to obtain target trajectory data; determining lane information of the drivable road where the target trajectory data is located based on the trajectory data distribution in a first direction; wherein the lane information is used to indicate at least one lane; determining the lane centerline of each lane in a second direction based on the target trajectory data, and determining the lane width of each lane based on the lane centerlines of adjacent lanes; acquiring trajectory image data and performing lane line recognition processing on the trajectory image data to obtain lane line information; determining road attribute information based on the lane information, lane centerline, lane width, and lane line information; wherein the road attribute information is used to construct a high-precision map. This method can achieve low-cost, low-consumption, and high-efficiency map construction.

Description

Technical Field

[0001] This application relates to the fields of high-precision mapping technology and intelligent driving technology, and in particular to a map construction method, device, equipment, storage medium and program product. Background Technology

[0002] In autonomous driving systems, high-precision maps are a key support for environmental perception, path planning, and decision control. Therefore, high requirements are placed on the real-time performance and accuracy of high-precision map construction.

[0003] Traditional high-precision map building methods mainly rely on specialized data collection vehicles, such as those equipped with specialized LiDAR (e.g., high-beam LiDAR) and high-precision positioning equipment (e.g., GNSS / IMU).

[0004] This implementation method is costly, slow to update, and involves complex processing of the data collected by the data collection vehicle, making it difficult to meet the real-time and accuracy requirements for high-precision map construction. Summary of the Invention

[0005] This application provides a map building method, apparatus, device, storage medium, and program product, which can achieve low-cost, low-consumption, and high-efficiency map building.

[0006] In a first aspect, embodiments of this application provide a map construction method, including:

[0007] Determine the initial trajectory data and perform road matching processing on the initial trajectory data to obtain the target trajectory data;

[0008] Based on the trajectory data distribution of the target trajectory data in the first direction, the lane information of the drivable road where the target trajectory data is located is determined; wherein, the lane information is used to indicate at least one lane;

[0009] Based on the target trajectory data, the lane centerline of each lane in the second direction is determined, and the lane width of each lane is determined based on the lane centerlines of the adjacent lanes.

[0010] Acquire trajectory image data and perform lane line recognition processing on the trajectory image data to obtain lane line information;

[0011] Based on the lane information, the lane centerline, the lane width, and the lane line information, road attribute information is determined; wherein, the road attribute information is used to construct a high-precision map.

[0012] In one possible implementation, the first direction indicates the cross-sectional direction of the drivable road; determining the lane information of the drivable road where the target trajectory data is located, based on the trajectory data distribution of the target trajectory data in the first direction, includes:

[0013] Based on the probability density curve of the target trajectory data in the cross-sectional direction of the drivable road indicated in the first direction, the peak region of the probability density curve is determined.

[0014] Based on the peak region of the probability density curve, the lane information of the drivable road where the target trajectory data is located is determined.

[0015] In one possible implementation, determining the peak region of the probability density curve based on the probability density curve of the target trajectory data in the cross-sectional direction of the drivable road indicated by the first direction includes:

[0016] Determine the road classification of the drivable road where the target trajectory data is located;

[0017] Based on the road grade, determine the probability density threshold for matching;

[0018] Based on the probability density threshold and the probability density curve, the peak region of the probability density curve is determined.

[0019] In one possible implementation, the second direction indicates the tangential direction of the drivable road; determining the lane centerline of each lane in the second direction based on the target trajectory data includes:

[0020] Based on a preset sliding window, determine the center point data of each lane in the target trajectory data in the tangential direction of the drivable road indicated by the second direction;

[0021] The center point data is subjected to curve fitting to obtain the lane centerline.

[0022] In one possible implementation, lane line recognition processing is performed on the trajectory image data to obtain lane line information, including:

[0023] The trajectory image data is processed for lane line recognition to determine the initial position and type of the lane lines.

[0024] If the lane line type is a preset type, the initial position of the lane line is corrected according to the monocular vision geometric model to obtain the final position of the lane line;

[0025] The lane line information is determined based on the final position of the lane line and the lane line type.

[0026] In one possible implementation, determining the initial trajectory data includes:

[0027] Acquire the collected trajectory data and perform low-pass filtering on the collected trajectory data to obtain filtered trajectory data;

[0028] The filtered trajectory data is processed to remove outliers, resulting in the initial trajectory data.

[0029] In one possible implementation, the target trajectory data includes trajectory data in at least one trajectory travel direction; before determining the lane information of the drivable road where the target trajectory data is located based on the trajectory data distribution in the first direction, the method further includes:

[0030] Determine the travel direction of the trajectory included in the target trajectory data;

[0031] Based on the direction of travel of the trajectory, the target trajectory data is grouped to obtain grouped target trajectory data.

[0032] Secondly, embodiments of this application provide a map building apparatus, comprising:

[0033] A matching unit is used to determine initial trajectory data and perform road matching processing on the initial trajectory data to obtain target trajectory data;

[0034] The first determining unit is configured to determine the lane information of the drivable road where the target trajectory data is located based on the trajectory data distribution of the target trajectory data in a first direction; wherein the lane information is used to indicate at least one lane.

[0035] The second determining unit is used to determine the lane centerline of each lane in the second direction based on the target trajectory data, and to determine the lane width of each lane based on the lane centerlines of adjacent lanes.

[0036] The recognition unit is used to acquire trajectory image data and perform lane line recognition processing on the trajectory image data to obtain lane line information;

[0037] The third determining unit is used to determine road attribute information based on the lane information, the lane centerline, the lane width, and the lane line information; wherein the road attribute information is used to construct a high-precision map.

[0038] In one possible implementation, the first direction indicates the cross-sectional direction of the drivable road; in this case, the first determining unit is configured to:

[0039] Based on the probability density curve of the target trajectory data in the cross-sectional direction of the drivable road indicated in the first direction, the peak region of the probability density curve is determined.

[0040] Based on the peak region of the probability density curve, the lane information of the drivable road where the target trajectory data is located is determined.

[0041] In one possible implementation, the first determining unit is configured to:

[0042] Determine the road classification of the drivable road where the target trajectory data is located;

[0043] Based on the road grade, determine the probability density threshold for matching;

[0044] Based on the probability density threshold and the probability density curve, the peak region of the probability density curve is determined.

[0045] In one possible implementation, the second direction indicates the tangential direction of the drivable road; in this case, the second determining unit is configured to:

[0046] Based on a preset sliding window, determine the center point data of each lane in the target trajectory data in the tangential direction of the drivable road indicated by the second direction;

[0047] The center point data is subjected to curve fitting to obtain the lane centerline.

[0048] In one possible implementation, the identification unit is used for:

[0049] The trajectory image data is processed for lane line recognition to determine the initial position and type of the lane lines.

[0050] If the lane line type is a preset type, the initial position of the lane line is corrected according to the monocular vision geometric model to obtain the final position of the lane line;

[0051] The lane line information is determined based on the final position of the lane line and the lane line type.

[0052] In one possible implementation, the matching unit is used for:

[0053] Acquire the collected trajectory data and perform low-pass filtering on the collected trajectory data to obtain filtered trajectory data;

[0054] The filtered trajectory data is processed to remove outliers, resulting in the initial trajectory data.

[0055] In one possible implementation, the target trajectory data includes trajectory data in at least one trajectory travel direction; before determining the lane information of the drivable road where the target trajectory data is located based on the trajectory data distribution in the first direction, the device is further configured to:

[0056] Determine the travel direction of the trajectory included in the target trajectory data;

[0057] Based on the direction of travel of the trajectory, the target trajectory data is grouped to obtain grouped target trajectory data.

[0058] Thirdly, embodiments of this application provide a computer device, including: a memory and a processor;

[0059] The memory stores computer-executed instructions;

[0060] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0061] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0062] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0063] The map construction method, apparatus, device, storage medium, and program product provided in this application can determine initial trajectory data and perform road matching processing on the initial trajectory data to obtain target trajectory data. Then, based on the trajectory data distribution of the target trajectory data in a first direction, the lane information of the drivable road where the target trajectory data is located is determined. At this time, the lane information can be determined simply and efficiently through the trajectory data distribution in the first direction. Afterwards, based on the target trajectory data, the lane centerline of each lane in a second direction can be determined. At this time, the accurate lane centerline can be extracted through lane constraints. In the above embodiments, the efficient extraction of lane centerlines can be achieved based on trajectory fitting algorithms without relying on prior maps, without relying on point cloud data, reducing the equipment requirements for collecting trajectory data, and thus reducing hardware costs. Next, the lane width of each lane can be determined based on the lane centerlines of adjacent lanes to obtain richer road information. Then, trajectory image data can be acquired, and lane line recognition processing can be performed on the trajectory image data to obtain lane line information. At this time, lane line recognition can be achieved through low-precision trajectory image data, which not only reduces the dependence on high-precision equipment and further reduces hardware costs, but also reduces the consumption of computing resources. Finally, road attribute information can be determined based on lane information, lane centerline, lane width, and lane line information. At this point, high-precision lane geometry information can be extracted from low-precision data, enabling low-cost, low-consumption, and high-efficiency construction when building high-precision maps based on road attribute information. Attached Figure Description

[0064] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0065] Figure 1 A schematic flowchart illustrating a map construction method provided in an embodiment of this application;

[0066] Figure 2 A flowchart illustrating another map construction method provided in this application embodiment;

[0067] Figure 3 This application provides a schematic diagram of the trajectory distribution of target trajectory data in an embodiment.

[0068] Figure 4 A schematic diagram of a high-precision map constructed based on road attribute information, provided as an embodiment of this application;

[0069] Figure 5 This is a schematic diagram of the structure of a map building device provided in an embodiment of this application;

[0070] Figure 6This is a schematic diagram of the structure of a computer device provided in an embodiment of this application.

[0071] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0072] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0073] The term "and / or" in this article merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A exists alone, A and B exist simultaneously, and B exists alone.

[0074] In addition, the term "at least one" in this document means any combination of at least two of any one or more of a plurality of elements, such as including at least one of A, B, and C, and may mean including any one or more elements selected from the set consisting of A, B, and C.

[0075] First, let me explain the terms used in this application:

[0076] GNSS: Global Navigation Satellite System.

[0077] IMU: Inertial Measurement Unit.

[0078] SLAM: Simultaneous Localization and Mapping.

[0079] BEV: Bird's Eye View.

[0080] RTK: Real-Time Kinematic, real-time dynamic positioning.

[0081] GPS: Global Positioning System.

[0082] In autonomous driving systems, high-precision maps are a key support for environmental perception, path planning, and decision control. Therefore, high requirements are placed on the real-time performance and accuracy of high-precision map construction.

[0083] Traditional high-precision map construction methods mainly rely on specialized data collection vehicles, such as those equipped with specialized LiDAR (e.g., high-beam LiDAR) and high-precision positioning equipment (e.g., GNSS / IMU). The procurement costs of specialized LiDAR and high-precision positioning equipment are high (e.g., millions of dollars). In addition, the data collection vehicles require special modifications, and a professional data collection and processing team is needed, making high-precision map collection a capital-intensive, asset-heavy business.

[0084] Furthermore, the data collection process using specialized vehicles is quite time-consuming. To cover a large area (e.g., nationwide) of road network, a large number of vehicles are needed for long-term, traversal data collection, which creates efficiency bottlenecks and further slows down the updating of high-precision maps. Moreover, real-world roads are constantly changing, with road construction, lane adjustments, and traffic rule changes. Therefore, relying on a limited fleet of specialized vehicles for periodic re-collection cannot keep up with the dynamic changes in roads, causing map data to become outdated very easily, thus posing safety risks to autonomous driving systems that rely on it for decision-making.

[0085] Therefore, the limitations of traditional map-building methods in terms of cost, efficiency, and timeliness have become the core bottleneck restricting the large-scale application and development of high-precision maps, thus necessitating the exploration of new cartographic technologies with lower costs and faster updates.

[0086] Research has revealed that BEV-based perception models are a key technology for achieving low-cost, high-precision map construction. BEV perception models can fuse and infer images from multiple cameras around the vehicle within a unified bird's-eye view coordinate system, providing a powerful tool for low-cost mapping. This reduces the reliance on sensors for mapping from "expensive LiDAR + high-precision positioning equipment" to "ordinary cameras + low-cost positioning equipment," thereby lowering hardware costs.

[0087] Building a globally consistent high-precision map based on the BEV perception model presents two major challenges: absolute accuracy and scalability / consistency.

[0088] The map output by the BEV perception model is relative to the vehicle's own coordinate system. Relying on a single perception by a single vehicle cannot solve the problem of the absolute geographical coordinate accuracy of the map. The global position of a local map generated by a single vehicle may drift by several meters.

[0089] The conventional solution involves crowdsourcing data and backend optimization. This means collecting local maps generated by a large number of vehicles (vehicles equipped with BEV perception capabilities) traveling on the same road segment, finding loop closures and matching relationships between them, and using SLAM technology for global optimization. This eliminates drift in individual trajectories, resulting in a globally consistent map with sufficiently high absolute accuracy. BEVs provide high-quality local map "materials," while crowdsourcing fusion and backend optimization act as the "glue" that stitches these "materials" together into a precise global map.

[0090] Furthermore, to ensure that the local maps output by the BEV perception model maintain consistent quality and scale across different vehicles, times, lighting conditions, and weather conditions, the backend optimization algorithm / model needs strong generalization capabilities. Simultaneously, processing massive amounts of crowdsourced data and performing global optimization in the cloud also requires substantial computing resources and efficient processing algorithms.

[0091] Based on this, this application provides a map construction method, which mainly outputs the number of lanes and road surface based on the fitting results of collected trajectory data (e.g., trajectory data collected by RTK, i.e., RTK trajectory data). It also combines low-precision trajectory images and uses image recognition technology to identify the changes in the virtual and real attributes of lane lines, thereby determining the road attribute data. This reduces the collection cost and enables low-cost, automated updates of lane-level map data.

[0092] At this point, on the one hand, the cost of RTK trajectory data acquisition is relatively low, mainly concentrated in hardware. A professional RTK setup (base station + rover) typically costs around 100,000 yuan, while an entry-level network RTK solution only costs tens of thousands of yuan. Furthermore, its software is easy to operate, and labor and operational costs are low, making costs clear and controllable, thus reducing overall expenses. On the other hand, trajectory data fitting and image recognition technologies can reduce computational resource consumption and improve processing efficiency.

[0093] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0094] Figure 1 This is a flowchart illustrating a map construction method provided in an embodiment of this application, such as... Figure 1 As shown, the method includes:

[0095] S101. Determine the initial trajectory data and perform road matching processing on the initial trajectory data to obtain the target trajectory data.

[0096] In one example, initial trajectory data can refer to crowdsourced trajectory data collected by RTK-based devices. In this case, the initial trajectory data can be understood as single-trip trajectory data, and complete, continuous, and consistent initial trajectory data can be determined from single-trip trajectory data.

[0097] In one example, the initial trajectory data may include, but is not limited to, information such as timestamps, coordinates, velocity, and heading.

[0098] In one example, road matching can be performed based on the spatial information of the initial trajectory data (e.g., spatial information may include, but is not limited to, the aforementioned timestamps, coordinates, speed, and heading information) to associate the initial trajectory data with the corresponding roads in the road network model, thereby obtaining the target trajectory data.

[0099] Optionally, the initial trajectory data can be processed by road matching based on a Hidden Markov Model (HMM) to obtain the target trajectory data.

[0100] S102. Based on the distribution of the target trajectory data in the first direction, determine the lane information of the drivable road where the target trajectory data is located; wherein, the lane information is used to indicate at least one lane.

[0101] In one example, the first direction is used to indicate the direction that can determine the number of lanes on the drivable road. In this case, the first direction can indicate the direction that forms a first angle with the drivable road, for example, the first angle can be 90 degrees, or it can be 80 degrees, etc.

[0102] In one example, lane information may include, but is not limited to, the center position of the lane and the number of lanes.

[0103] S103. Based on the target trajectory data, determine the lane centerline of each lane in the second direction, and determine the lane width of each lane based on the lane centerlines of the adjacent lanes.

[0104] In one example, the second direction is used to indicate the direction that can determine the center line of the lane of the drivable road. In this case, the second direction can indicate the direction that forms a second angle with the drivable road. For example, the second angle can be 180 degrees or 0 degrees.

[0105] S104. Acquire trajectory image data and perform lane line recognition processing on the trajectory image data to obtain lane line information.

[0106] In one example, trajectory image data can be understood as image data associated with the initial trajectory data. In this case, the trajectory image data can be understood as image data collected by a regular camera installed in the vehicle.

[0107] In one example, lane line recognition can be performed on trajectory image data using an image recognition algorithm.

[0108] Optionally, the image recognition algorithm can be any one of edge detection algorithms, morphological algorithms, and spectral analysis algorithms. There is no limitation on the type of image recognition algorithm, as long as it can be implemented.

[0109] In one example, lane line information can be used to indicate whether a lane line is a solid line or a dashed line.

[0110] S105. Determine road attribute information based on lane information, lane centerline, lane width, and lane line information; among which, road attribute information is used to construct a high-precision map.

[0111] As described above, this embodiment of the application can determine initial trajectory data and perform road matching processing on the initial trajectory data to obtain target trajectory data. Then, based on the trajectory data distribution of the target trajectory data in the first direction, the lane information of the drivable road where the target trajectory data is located is determined. At this time, the lane information can be determined simply and efficiently through the trajectory data distribution in the first direction. Afterwards, the lane centerline of each lane in the second direction can be determined based on the target trajectory data. At this time, the accurate lane centerline can be extracted through lane constraints. In the above embodiment, the efficient extraction of lane centerlines can be achieved based on trajectory fitting algorithms without relying on prior maps, without relying on point cloud data, reducing the equipment requirements for collecting trajectory data, and thus reducing hardware costs. Next, the lane width of each lane can be determined based on the lane centerlines of adjacent lanes to obtain richer road information. Then, trajectory image data can be acquired and lane line recognition processing can be performed on the trajectory image data to obtain lane line information. At this time, lane line recognition can be achieved through low-precision trajectory image data, which not only reduces the dependence on high-precision equipment and further reduces hardware costs, but also reduces the consumption of computing resources. Finally, road attribute information can be determined based on lane information, lane centerline, lane width, and lane line information. At this point, high-precision lane geometry information can be extracted from low-precision data, enabling low-cost, low-consumption, and high-efficiency construction when building high-precision maps based on road attribute information.

[0112] Figure 2 A flowchart illustrating another map construction method provided in this application embodiment is shown below. Figure 2 As shown, in this embodiment... Figure 1 Based on the examples, the map construction method is described in detail, which includes:

[0113] S201. Acquire the collected trajectory data and perform low-pass filtering on the collected trajectory data to obtain filtered trajectory data.

[0114] In one example, the collected trajectory data can be understood as trajectory data collected based on RTK.

[0115] In one example, the acquired trajectory data can be low-pass filtered.

[0116] Optionally, the low-pass filter can be, but is not limited to, a Kalman filter. In this case, the acquired trajectory data can be smoothed by the Kalman filter, which can remove high-frequency jitter while preserving the true macroscopic shape such as turning.

[0117] S202. Perform outlier removal on the filtered trajectory data to obtain the initial trajectory data.

[0118] In one example, outliers can be understood as trajectory data that deviates significantly from the main path; for example, outliers could be trajectory data that deviates into buildings.

[0119] Optionally, in this embodiment of the application, outliers in the filtered trajectory data can be identified through statistical analysis. For example, the degree of deviation of the trajectory data can be determined by calculating the Z-score of the trajectory data points, thereby identifying outliers.

[0120] In the above implementation, low-pass filtering can preserve low-frequency signals, suppress high-frequency noise, and effectively eliminate instantaneous drift, thus ensuring that the filtered trajectory data reflects the true road conditions. Subsequently, outlier removal further eliminates interference factors from the filtered trajectory data. The synergistic effect of these two methods ensures that the initial trajectory data retains true road characteristics while possessing higher data quality, providing stable and reliable data support for subsequent trajectory matching and fitting.

[0121] S203. Perform road matching processing on the initial trajectory data to obtain the target trajectory data.

[0122] In one example, this step can be referred to the content described in S101 above, and will not be repeated in detail here.

[0123] In one possible implementation, the target trajectory data may include trajectory data in at least one trajectory travel direction.

[0124] For example, the target trajectory data may include one, two, or three trajectory directions, etc. There is no limitation on the trajectory directions here, as long as it is feasible. For example, the trajectory directions may include, but are not limited to: from A to B, from B to A, and from A to C, etc.

[0125] Based on this, before determining the lane information of the drivable road where the target trajectory data is located according to the trajectory data distribution in the first direction, the embodiments of this application may first perform the steps described in S204 to S205 below.

[0126] S204. Determine the travel direction of the trajectory included in the target trajectory data.

[0127] S205. Based on the trajectory travel direction, group the target trajectory data to obtain the grouped target trajectory data.

[0128] For example, if the target trajectory data includes trajectory travel directions of: from A to B and from B to A, then the obtained grouped target trajectory data may include target trajectory data from A to B and target trajectory data from B to A.

[0129] In the above embodiments, the target trajectory data can be grouped according to the trajectory travel direction, which can avoid interference between trajectory data in different directions, thereby improving the accuracy of subsequent trajectory fitting based on the target trajectory data.

[0130] In this embodiment of the application, after determining the grouped target trajectory data, the target trajectory data under each trajectory travel direction can be processed in the following two ways: firstly, lanes can be divided laterally based on density estimation; secondly, a centerline can be generated longitudinally based on a sliding window. See the process described below for details.

[0131] Optionally, if the first direction indicates the cross-sectional direction of the drivable road, the lane information of the drivable road where the target trajectory data is located can be determined according to the following steps S206 to S207.

[0132] S206. Based on the probability density curve of the target trajectory data in the cross-sectional direction of the drivable road indicated in the first direction, determine the peak region of the probability density curve.

[0133] In one example, a probability density curve can be obtained by projecting the target trajectory data onto a first direction and then performing density estimation on the projected target trajectory data.

[0134] Optionally, in this embodiment of the application, the density of the projected target trajectory data can be estimated according to the kernel density estimation algorithm to obtain the probability density curve.

[0135] S207. Based on the peak region of the probability density curve, determine the lane information of the drivable road where the target trajectory data is located.

[0136] Lane information is used to indicate at least one lane.

[0137] In one example, the probability density curve may include one or more peak regions. In this case, the maximum value of each peak region corresponds to the center position of a lane, and the number of peak regions is the number of lanes.

[0138] In the above embodiments, the probability density curve of the target trajectory data in the cross-sectional direction of the drivable road can be determined by cross-sectional projection and density estimation. The lane information can be determined based on the peak area of ​​the probability density curve. This can effectively resist the interference of trajectory jitter and anomalies, and improve the scene adaptability of lane division, thereby determining stable and reliable lane information.

[0139] In one possible implementation, when determining the peak region of the probability density curve based on the probability density curve of the drivable road cross-sectional direction indicated by the target trajectory data in the first direction, the embodiments of this application may first determine the road level of the drivable road where the target trajectory data is located, then determine the matching probability density threshold based on the road level, and finally determine the peak region of the probability density curve based on the probability density threshold and the probability density curve.

[0140] In one example, the road level of the drivable road where the target trajectory data is located is associated with the road type. For example, the road level of a highway can be level 1, the road level of a national highway can be level 2, and the road level of a village road can be level 7. Here, there are no restrictions on the road type and the corresponding road level, as long as it can be achieved.

[0141] In one example, different probability density thresholds can be set for different road grades, which can not only reduce the interference of non-lane peak areas, but also avoid missing the peak areas of lane indicators, thus enabling accurate peak area delineation.

[0142] For example, for a highway of road class 1, a smaller probability density threshold can be set (e.g., 0.6 times the average of the maximum values ​​in all peak regions) to identify the entire peak region; for a village road of road class 7, a larger probability density threshold can be set (e.g., 0.8 times the average of the maximum values ​​in all peak regions) to avoid interference from small peak regions formed by target trajectory data under non-lane conditions.

[0143] Optionally, the second direction can indicate the tangential direction of the drivable road; in this case, when determining the lane centerline of each lane in the second direction based on the target trajectory data, refer to the steps described in S208 to S209 below.

[0144] S208. Based on the preset sliding window, determine the center point data of each lane in the target trajectory data under the tangent direction of the drivable road indicated in the second direction.

[0145] In one example, the preset sliding window can be a rectangular window. In this case, the target trajectory data can be divided into multiple segments along the tangent direction of the drivable road where the target trajectory data is located, based on the rectangular window. At the same time, the target trajectory data under each lane can be determined based on the lane to which the target trajectory data belongs in the first direction. Then, the lane center point data can be determined based on the target trajectory data belonging to the same lane within each segment (i.e., each preset sliding window).

[0146] In one example, the center point data can be determined based on the median of the target trajectory data belonging to the same lane within each preset sliding window.

[0147] Optionally, in this embodiment, a preset sliding window can be set for each lane to determine the center point data of each lane in the second direction. The number of preset sliding windows is not limited, and can be implemented as needed.

[0148] S209. Perform curve fitting on the center point data to obtain the lane centerline.

[0149] Optionally, the centerline of the lane can be obtained by curve fitting the center point data using a cubic spline curve.

[0150] In the above embodiments, a preset sliding window can be used to determine the center point data of the lane based on statistical analysis data (e.g., median) of the local target trajectory data. This can suppress noise interference in the target trajectory data, resist target trajectory data jitter, and thus improve the stability and accuracy of the lane center line determined based on the center point data.

[0151] S210. Determine the lane width of each lane based on the lane center lines of adjacent lanes.

[0152] In one example, the lane width can be determined based on the average width of the center lines of adjacent lanes.

[0153] In one possible implementation, embodiments of this application can obtain a first principal component direction by performing principal component analysis on the target trajectory data. At this point, the first principal component direction can be determined as the tangent direction of the drivable road, i.e., the second direction. Then, the perpendicular direction of the second direction is determined as the cross-sectional direction of the drivable road, i.e., the first direction.

[0154] For example, see Figure 3 , Figure 3 This application provides a schematic diagram of the trajectory distribution of target trajectory data, as shown in the embodiments of this application. Figure 3 As shown, on a drivable road, multiple target trajectory data can be included. At this time, lane information can be determined based on the target trajectory data in the first direction (i.e., the cross-sectional direction of the drivable road), and the lane centerline can be determined based on the target trajectory data in the second direction of the drivable road (i.e., the tangential direction of the drivable road). See the process described above for details, which will not be elaborated here.

[0155] S211. Acquire trajectory image data, perform lane line recognition processing on the trajectory image data, and determine the initial position and type of lane line.

[0156] In one example, the initial position of the lane line can be understood as the position corresponding to the target trajectory data associated with the trajectory image data, that is, the GPS position.

[0157] In one example, the lane line type can be solid line and / or dashed line. In this case, if the lane line type is both solid and dashed, the locations in the trajectory image data that include the changes between solid and dashed lines are determined.

[0158] S212. If the lane line type is a preset type, the initial position of the lane line is corrected according to the monocular vision geometric model to obtain the final position of the lane line.

[0159] In one example, the preset type can indicate solid line type and dashed line type.

[0160] In one example, the monocular vision geometric model can be seen in the following formula (1).

[0161] (1)

[0162] Wherein, D represents the position correction distance determined by the monocular vision geometric model; f represents the camera focal length, which is the calibrated camera intrinsic parameter; H represents the height of the camera lens above the ground (an external parameter that needs to be measured manually); y represents the pixel ordinate of the object's ground point in the trajectory image data (where the top of the trajectory image data is 0, increasing downwards); y0 represents the pixel position of the horizon (or the ordinate of the principal point, where the principal point is the intersection of the optical axis and the imaging plane) in the image.

[0163] In this embodiment of the application, the position correction distance D can be determined according to the monocular vision geometric model. Then, the final position of the lane line corresponding to the preset type of lane line can be determined according to the sum of the position correction distance D and the initial position of the lane line.

[0164] For example, if the camera focal length f is 1200 pixels, the camera height H is 1.2 meters, the horizon position y0 in the image is 300 pixels, and the pixel ordinate of the position where the dashed and solid lines change is 600 pixels, then we can determine that D = 4.8 meters. At this time, we can determine that the position where the dashed and solid lines of the lane change is the position determined by adding the initial position of the lane line to D.

[0165] S213. Determine lane line information based on the final position and type of lane line.

[0166] In the above implementation, the absolute position error of low-precision trajectories can be corrected by a monocular visual geometric model, thereby converting image-level observations into accurate estimates of real-world positions through a physical model. This enables high-precision positioning of lane line change points under different camera configurations and image quality conditions, thus compensating for the insufficient accuracy of crowdsourced GPS.

[0167] S214. Determine road attribute information based on lane information, lane centerline, lane width, and lane line information; among which, road attribute information is used to construct a high-precision map.

[0168] See Figure 4 , Figure 4 This is a schematic diagram of a high-precision map constructed based on road attribute information, provided in an embodiment of this application. Figure 4 As shown, it can be based on single-trip trajectory data (such as...) Figure 4 The dots represent trajectory data within the starting / ending range. The labels on the dots indicate the type of single-trip trajectory data (e.g., "2" indicates trajectory data under a variable lane, "9" and "10" indicate trajectory data at an intersection), as well as pre-set road network models (such as...). Figure 4As shown by the blue lines (where the arrows within the blue lines indicate the direction of travel), based on the process described above, the number of lanes and lane widths are determined, and combined with the actual road boundaries (such as...). Figure 4 (As shown by the red lines in the image), the lanes included in the high-precision map are identified, such as... Figure 4 As shown by the green lines in the image.

[0169] As described above, the embodiments of this application provide a multi-layered, self-correcting intelligent processing pipeline for low-precision crowdsourced data. Specifically, the quality of the initial trajectory data is ensured through a preprocessing process (including the aforementioned low-pass filtering and outlier removal); road matching processing of the initial trajectory data using an Hidden Markov Model (HMM) provides accurate road-level spatial context; lane information and lane centerline inference methods based on kernel density estimation and median statistics exhibit strong robustness to data noise, thus stably reconstructing lane-level geometry; and by fusing image recognition and visual ranging models, the relative accuracy of image pixels is combined with the ambiguity of GPS absolute positioning, enabling the location correction of key road elements (e.g., changes in the position of dashed and solid lines). These embodiments can efficiently and accurately extract high-value, high-precision map information from noisy, low-cost data, thereby achieving a balance between cost and accuracy in map construction.

[0170] Figure 5 This is a schematic diagram of the structure of a map building device provided in an embodiment of this application, as shown below. Figure 5 As shown, the map building apparatus 50 provided in this embodiment includes:

[0171] The matching unit 501 is used to determine the initial trajectory data and perform road matching processing on the initial trajectory data to obtain the target trajectory data.

[0172] The first determining unit 502 is used to determine the lane information of the drivable road where the target trajectory data is located based on the trajectory data distribution of the target trajectory data in the first direction; wherein the lane information is used to indicate at least one lane.

[0173] The second determining unit 503 is used to determine the lane centerline of each lane in the second direction based on the target trajectory data, and to determine the lane width of each lane based on the lane centerlines of the adjacent lanes.

[0174] The recognition unit 504 is used to acquire trajectory image data and perform lane line recognition processing on the trajectory image data to obtain lane line information.

[0175] The third determining unit 505 is used to determine road attribute information based on lane information, lane center line, lane width and lane line information; wherein, the road attribute information is used to construct a high-precision map.

[0176] In one possible implementation, the first direction indicates the cross-sectional direction of the drivable road; in this case, the first determining unit 502 is used to:

[0177] Based on the probability density curve of the drivable road cross-section indicated by the target trajectory data in the first direction, determine the peak region of the probability density curve.

[0178] Based on the peak region of the probability density curve, determine the lane information of the drivable road where the target trajectory data is located.

[0179] In one possible implementation, the first determining unit 502 is configured to:

[0180] Determine the road classification of the drivable road where the target trajectory data is located;

[0181] Determine the probability density threshold for matching based on the road grade;

[0182] Based on the probability density threshold and the probability density curve, determine the peak region of the probability density curve.

[0183] In one possible implementation, the second direction indicates the tangential direction of the drivable road; in this case, the second determining unit 503 is used to:

[0184] Based on the preset sliding window, determine the center point data of each lane in the target trajectory data under the tangent direction of the drivable road indicated in the second direction;

[0185] Curve fitting is performed on the center point data to obtain the lane centerline.

[0186] In one possible implementation, the identification unit 504 is used for:

[0187] Lane line recognition processing is performed on the trajectory image data to determine the initial position and type of the lane lines;

[0188] If the lane line type is a preset type, the initial position of the lane line is corrected according to the monocular vision geometric model to obtain the final position of the lane line;

[0189] Determine lane line information based on the final position and type of lane lines.

[0190] In one possible implementation, the matching unit 501 is used for:

[0191] The collected trajectory data is acquired and low-pass filtered to obtain filtered trajectory data.

[0192] The filtered trajectory data is processed to remove outliers, resulting in the initial trajectory data.

[0193] In one possible implementation, the target trajectory data includes trajectory data in at least one trajectory travel direction; before determining the lane information of the drivable road where the target trajectory data is located based on the trajectory data distribution in the first direction, the device is further configured to:

[0194] Determine the travel direction of the trajectory included in the target trajectory data;

[0195] Based on the direction of travel, the target trajectory data is grouped to obtain the grouped target trajectory data.

[0196] The map building device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0197] Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Figure 6 As shown, the computer device 60 provided in this embodiment includes at least one processor 601 and a memory 602. Optionally, the computer device 60 further includes a communication component 603. The processor 601, memory 602, and communication component 603 are connected via a bus 604.

[0198] In a specific implementation, at least one processor 601 executes computer execution instructions stored in memory 602, causing at least one processor 601 to perform the above-described method.

[0199] The specific implementation process of processor 601 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0200] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0201] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0202] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0203] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0204] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0205] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0206] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0207] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0208] 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0209] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0210] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0211] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0212] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A map construction method, characterized in that, include: Determine the initial trajectory data and perform road matching processing on the initial trajectory data to obtain the target trajectory data; Based on the trajectory data distribution of the target trajectory data in the first direction, the lane information of the drivable road where the target trajectory data is located is determined; wherein, the lane information is used to indicate at least one lane; Based on the target trajectory data, the lane centerline of each lane in the second direction is determined, and the lane width of each lane is determined based on the lane centerlines of the adjacent lanes. Acquire trajectory image data and perform lane line recognition processing on the trajectory image data to obtain lane line information; Based on the lane information, the lane centerline, the lane width, and the lane line information, road attribute information is determined; wherein, the road attribute information is used to construct a high-precision map.

2. The method according to claim 1, characterized in that, The first direction indicates the cross-sectional direction of the drivable road; based on the trajectory data distribution of the target trajectory data in the first direction, the lane information of the drivable road where the target trajectory data is located is determined, including: Based on the probability density curve of the target trajectory data in the cross-sectional direction of the drivable road indicated in the first direction, the peak region of the probability density curve is determined. Based on the peak region of the probability density curve, the lane information of the drivable road where the target trajectory data is located is determined.

3. The method according to claim 2, characterized in that, Based on the probability density curve of the target trajectory data in the cross-sectional direction of the drivable road indicated by the first direction, determining the peak region of the probability density curve includes: Determine the road classification of the drivable road where the target trajectory data is located; Based on the road grade, determine the probability density threshold for matching; Based on the probability density threshold and the probability density curve, the peak region of the probability density curve is determined.

4. The method according to claim 1, characterized in that, The second direction indicates the tangential direction of the drivable road; based on the target trajectory data, determining the lane centerline of each lane in the second direction includes: Based on a preset sliding window, determine the center point data of each lane in the target trajectory data in the tangential direction of the drivable road indicated by the second direction; The center point data is subjected to curve fitting to obtain the lane centerline.

5. The method according to claim 1, characterized in that, The trajectory image data is processed for lane line recognition to obtain lane line information, including: The trajectory image data is processed for lane line recognition to determine the initial position and type of the lane lines. If the lane line type is a preset type, the initial position of the lane line is corrected according to the monocular vision geometric model to obtain the final position of the lane line; The lane line information is determined based on the final position of the lane line and the lane line type.

6. The method according to claim 1, characterized in that, Determine the initial trajectory data, including: Acquire the collected trajectory data and perform low-pass filtering on the collected trajectory data to obtain filtered trajectory data; The filtered trajectory data is processed to remove outliers, resulting in the initial trajectory data.

7. The method according to any one of claims 1-6, characterized in that, The target trajectory data includes trajectory data in at least one trajectory travel direction; before determining the lane information of the drivable road where the target trajectory data is located based on the trajectory data distribution in the first direction, the method further includes: Determine the travel direction of the trajectory included in the target trajectory data; Based on the direction of travel of the trajectory, the target trajectory data is grouped to obtain grouped target trajectory data.

8. A map building device, characterized in that, include: A matching unit is used to determine initial trajectory data and perform road matching processing on the initial trajectory data to obtain target trajectory data; The first determining unit is configured to determine the lane information of the drivable road where the target trajectory data is located based on the trajectory data distribution of the target trajectory data in a first direction; wherein the lane information is used to indicate at least one lane. The second determining unit is used to determine the lane centerline of each lane in the second direction based on the target trajectory data, and to determine the lane width of each lane based on the lane centerlines of adjacent lanes. The recognition unit is used to acquire trajectory image data and perform lane line recognition processing on the trajectory image data to obtain lane line information. The third determining unit is used to determine road attribute information based on the lane information, the lane centerline, the lane width, and the lane line information; wherein the road attribute information is used to construct a high-precision map.

9. A computer device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium / computer program product, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7; The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1-7.

Images