Lane boundary determination method and apparatus, vehicle, medium, and program product

Abstract

The embodiment of the application provides a kind of lane boundary determination method, device, vehicle, medium and program product, which obtains the multi-source lane boundary data collected by vehicle, and the multi-source lane boundary data is the data obtained by identifying lane boundary based on multiple sensors;According to the multi-source lane boundary data, at least one lane boundary data set of the same position is determined, each lane boundary data set includes at least one candidate lane boundary data;For each lane boundary data set, according to the preset lane boundary rule library, the lane boundary data set is corrected to obtain the target lane boundary data applied to control vehicle, and the lane boundary rule library is constructed based on the marking feature of standard lane boundary, the spatial topology feature of standard lane boundary, the lane boundary combination feature of standard lane boundary and priority feature.The technical scheme can more accurately realize the identification of lane boundary information.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and more particularly to a method, apparatus, vehicle, medium, and program product for determining lane boundaries. Background Technology

[0002] With the rapid development of autonomous driving technology, the vehicle's real-time perception of the road environment has become a core element in ensuring driving safety and efficiency. In complex and ever-changing road environments, lane boundary recognition is the foundation for autonomous driving systems to achieve key functions such as lane keeping, path planning, and lane change decisions.

[0003] In existing autonomous driving systems, lane boundary recognition mainly relies on raw data from sensors such as cameras, LiDAR, and millimeter-wave radar, and directly outputs the position and type of lane lines through image recognition algorithms or point cloud segmentation technology.

[0004] However, the above solutions often suffer from insufficient recognition accuracy or lack of logical verification, leading to incorrect lane boundary information and potentially causing misjudgments by the autonomous driving system. Summary of the Invention

[0005] This application provides a method, apparatus, vehicle, medium, and program product for determining lane boundaries, in order to achieve a more accurate technical effect of lane boundary information recognition.

[0006] In a first aspect, embodiments of this application provide a method for determining lane boundaries, including:

[0007] Acquire multi-source lane boundary data collected by the vehicle, wherein the multi-source lane boundary data is obtained based on the identification of lane boundaries by multiple sensors;

[0008] Based on the multi-source lane boundary data, at least one set of lane boundary data at the same location is determined, and each set of lane boundary data includes at least one candidate lane boundary data.

[0009] For each lane boundary data set, the lane boundary data set is corrected according to a preset lane boundary rule base to obtain target lane boundary data for controlling the vehicle. The lane boundary rule base is constructed based on the marking features of standard lane boundaries, the spatial topological features of standard lane boundaries, the lane boundary combination features of standard lane boundaries, and priority features. The priority features include: the reliability priority of sensors under different environmental conditions, and the recognition priority of different types of markings at the same location.

[0010] In one or more embodiments, the lane marking features include: visual features of the lane markings and geometric parameters of the lane markings;

[0011] The lane boundary combination features include: lane boundary combination methods preset for different road types and different traffic scenarios;

[0012] The spatial topological features include: the relative positional relationship between lane boundaries, the parallelism of lane lines in the same direction, and the curvature of lane lines.

[0013] In one or more embodiments, the step of correcting the lane boundary data set according to a preset lane boundary rule base to obtain target lane boundary data for controlling the vehicle includes:

[0014] For at least one candidate lane boundary data in the lane boundary data set, determine whether it is consistent with the lane marking features, the spatial topology features, and the lane boundary combination features, respectively.

[0015] If inconsistent candidate lane boundary data exist, the lane boundary data set is corrected according to the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data.

[0016] In one or more embodiments, the step of correcting the lane boundary data set based on the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data includes:

[0017] If the confidence value corresponding to the inconsistent candidate lane boundary data is greater than the first preset threshold, the lane boundary data set is corrected according to the marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data.

[0018] If the confidence value corresponding to the inconsistent candidate lane boundary data is less than the first preset threshold, the lane boundary data set is corrected according to the priority feature to obtain the target lane boundary data.

[0019] In one or more embodiments, correcting the lane boundary data set according to the priority feature to obtain the target lane boundary data includes:

[0020] In the set of lane boundary data, the first candidate lane boundary data corresponding to the sensor with the highest reliability priority, excluding the inconsistent candidate lane boundary data, is determined.

[0021] If the first candidate lane boundary data is consistent with the marking features, the spatial topology features, and the lane boundary combination features, the first candidate lane boundary data is determined as the target lane boundary data;

[0022] In one or more embodiments, correcting the lane boundary data set according to the priority feature to obtain the target lane boundary data includes:

[0023] Based on the preset weights of the corresponding sensors in the lane boundary data set, at least one candidate lane boundary data in the lane boundary data set is fused to obtain the target lane boundary data. The preset weights are determined based on the reliability priority and / or the identification priority.

[0024] In one or more embodiments, the step of correcting the lane boundary data set based on the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data includes:

[0025] The lane boundary data set is input into the probability model to obtain the probability values ​​corresponding to at least one candidate lane boundary data. The probability model is obtained by training a Bayesian network based on the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features.

[0026] The candidate lane boundary data with the highest probability value is used as the target lane boundary data.

[0027] In one or more embodiments, the multi-source lane boundary data includes: multiple lane boundary data and a confidence value corresponding to each lane boundary data;

[0028] Accordingly, before determining at least one set of lane boundary data at the same location based on the multi-source lane boundary data, the method further includes:

[0029] The multiple lane boundary data are preprocessed to obtain preprocessed multiple lane boundary data. The preprocessing includes: transformation to the same coordinate system, time synchronization processing, and spatial registration processing.

[0030] Based on the confidence level value corresponding to each lane boundary data, the preprocessed lane boundary data is filtered to obtain the filtered lane boundary data.

[0031] Secondly, embodiments of this application provide a lane boundary determination device, comprising:

[0032] The acquisition module is used to acquire multi-source lane boundary data collected by the vehicle. The multi-source lane boundary data is obtained based on the identification of lane boundaries by multiple sensors.

[0033] The determination module is used to determine at least one set of lane boundary data at the same location based on the multi-source lane boundary data, and each set of lane boundary data includes at least one candidate lane boundary data.

[0034] The processing module is used to perform correction processing on each lane boundary data set according to a preset lane boundary rule base to obtain target lane boundary data for controlling the vehicle. The lane boundary rule base is constructed based on the marking features of standard lane boundaries, the spatial topological features of standard lane boundaries, the lane boundary combination features of standard lane boundaries, and priority features. The priority features include: the reliability priority of sensors under different environmental conditions, and the recognition priority of different types of markings at the same location.

[0035] In one or more embodiments, the lane marking features include: visual features of the lane markings and geometric parameters of the lane markings;

[0036] The lane boundary combination features include: lane boundary combination methods preset for different road types and different traffic scenarios;

[0037] The spatial topological features include: the relative positional relationship between lane boundaries, the parallelism of lane lines in the same direction, and the curvature of lane lines.

[0038] In one or more embodiments, the processing module performs correction processing on the lane boundary data set according to a preset lane boundary rule base to obtain target lane boundary data for controlling the vehicle, specifically for:

[0039] For at least one candidate lane boundary data in the lane boundary data set, determine whether it is consistent with the lane marking features, the spatial topology features, and the lane boundary combination features, respectively.

[0040] If inconsistent candidate lane boundary data exist, the lane boundary data set is corrected according to the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data.

[0041] In one or more embodiments, the processing module corrects the lane boundary data set based on the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data, specifically for:

[0042] If the confidence value corresponding to the inconsistent candidate lane boundary data is greater than the first preset threshold, the lane boundary data set is corrected according to the marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data.

[0043] If the confidence value corresponding to the inconsistent candidate lane boundary data is less than the first preset threshold, the lane boundary data set is corrected according to the priority feature to obtain the target lane boundary data.

[0044] In one or more embodiments, the processing module corrects the lane boundary data set according to the priority feature to obtain the target lane boundary data, specifically for:

[0045] In the set of lane boundary data, the first candidate lane boundary data corresponding to the sensor with the highest reliability priority, excluding the inconsistent candidate lane boundary data, is determined.

[0046] If the first candidate lane boundary data is consistent with the marking features, the spatial topology features, and the lane boundary combination features, the first candidate lane boundary data is determined as the target lane boundary data;

[0047] In one or more embodiments, the processing module corrects the lane boundary data set according to the priority feature to obtain the target lane boundary data, specifically for:

[0048] Based on the preset weights of the corresponding sensors in the lane boundary data set, at least one candidate lane boundary data in the lane boundary data set is fused to obtain the target lane boundary data. The preset weights are determined based on the reliability priority and / or the identification priority.

[0049] In one or more embodiments, the processing module corrects the lane boundary data set based on the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data, specifically for:

[0050] The lane boundary data set is input into the probability model to obtain the probability values ​​corresponding to at least one candidate lane boundary data. The probability model is obtained by training a Bayesian network based on the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features.

[0051] The candidate lane boundary data with the highest probability value is used as the target lane boundary data.

[0052] In one or more embodiments, the multi-source lane boundary data includes: multiple lane boundary data and a confidence value corresponding to each lane boundary data;

[0053] Accordingly, before determining at least one set of lane boundary data at the same location based on the multi-source lane boundary data, the processing module is further configured to:

[0054] The multiple lane boundary data are preprocessed to obtain preprocessed multiple lane boundary data. The preprocessing includes: transformation to the same coordinate system, time synchronization processing, and spatial registration processing.

[0055] Based on the confidence level value corresponding to each lane boundary data, the preprocessed lane boundary data is filtered to obtain the filtered lane boundary data.

[0056] Thirdly, embodiments of this application provide a vehicle, including: a memory and a processor;

[0057] The memory stores computer-executed instructions;

[0058] 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.

[0059] 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.

[0060] 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.

[0061] The lane boundary determination method, device, vehicle, medium, and program product provided in this application embodiment acquire multi-source lane boundary data collected by the vehicle. The multi-source lane boundary data is data obtained by identifying lane boundaries based on multiple sensors. Based on the multi-source lane boundary data, at least one set of lane boundary data at the same location is determined. Each set of lane boundary data includes at least one candidate lane boundary data. For each set of lane boundary data, a correction process is performed on the set of lane boundary data according to a preset lane boundary rule base to obtain target lane boundary data applied to control the vehicle. The lane boundary rule base is constructed based on the marking features of standard lane boundaries, the spatial topological features of standard lane boundaries, the lane boundary combination features of standard lane boundaries, and priority features. The priority features include the reliability priority of sensors under different environmental conditions and the recognition priority of different types of markings at the same location. This technical solution acquires multi-source lane boundary data collected by vehicles, comprehensively utilizing information from different sensors to improve the redundancy and coverage of lane boundary perception. Furthermore, it determines a set of lane boundary data for the same location based on multi-source data, facilitating the alignment and fusion of recognition results from different sources for the same spatial location, providing a reliable data foundation for subsequent correction. For each lane boundary data set, correction processing is performed according to a pre-defined lane boundary rule base. This rule base integrates the marking features, spatial topological features, combined features, and priority features of standard lane boundaries, thus identifying and correcting abnormal boundaries caused by sensor noise, occlusion, or false detections. In particular, the priority features include the reliability priority of sensors under different environmental conditions, enabling the correction process to dynamically select the most reliable data source based on current weather, lighting, and other environmental factors, thereby improving the algorithm's environmental adaptability. Simultaneously, the recognition priority of different types of markings at the same location resolves decision-making problems when there are conflicts between solid lines, dashed lines, and double solid lines, ensuring that the output target lane boundary conforms to actual traffic rules. The final target lane boundary data used to control the vehicle has higher accuracy, effectively supporting stable lateral control and safe driving in complex road environments. Attached Figure Description

[0062] 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.

[0063] Figure 1 A schematic diagram of the architecture and signal flow of the lane boundary determination system provided in this application;

[0064] Figure 2 Flowchart of the method for determining lane boundaries provided in this application Figure 1 ;

[0065] Figure 3 Flowchart of the method for determining lane boundaries provided in this application Figure 2 ;

[0066] Figure 4 Flowchart of the method for determining lane boundaries provided in this application Figure 3 ;

[0067] Figure 5 Flowchart of the method for determining lane boundaries provided in this application Figure 4 ;

[0068] Figure 6 A schematic diagram of the structure of the lane boundary determination device provided in this application;

[0069] Figure 7 This is a schematic diagram of the vehicle structure provided in an embodiment of this application.

[0070] The accompanying drawings have illustrated 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 specific embodiments. Detailed Implementation

[0071] 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.

[0072] Lane boundary recognition is a fundamental component of the field of autonomous driving environmental perception and vehicle control technology, and is widely used in the road traffic processes of passenger vehicles' autonomous driving, assisted driving, and unmanned delivery vehicles.

[0073] In practical applications, vehicles are typically equipped with multiple sensors such as cameras, LiDAR, and millimeter-wave radar to continuously collect information on road markings, road edges, and the surrounding traffic environment. The collected results are then input into the perception and control link to support functions such as lane keeping, lane change decision-making, trajectory planning, and lateral control. Especially in complex scenarios such as urban roads, highways, ramps, tunnel entrances and exits, and in rain, snow, or at night, vehicles need to stably identify lane boundaries at the same location to avoid establishing control strategies based on erroneous boundary information.

[0074] In existing solutions, lane boundary determination typically relies on the recognition results from multiple sensors, which are then fused to obtain the final boundary information. For example, cameras focus on extracting the color, texture, and morphological features of lane markings from images; LiDAR focuses on identifying road edges and lane marking contours based on point cloud reflections and geometric structures; and millimeter-wave radar can provide some spatial perception information. A simple fusion process is then performed to obtain the final lane boundary. However, this approach has the following significant limitations in complex environments:

[0075] On the one hand, different sensors are affected by the environment to varying degrees. Factors such as nighttime glare, rainwater adhesion, snow cover, road reflection, and worn road markings can all cause deviations in the recognition results of a certain type of sensor, leading to discrepancies in the boundary type, boundary location, or even the existence of a boundary at the same location.

[0076] On the other hand, most existing fusion methods focus on the confidence level of numerical data synthesis, lacking constraints on the regular attributes of the lane boundaries themselves, making it difficult to determine whether the candidate results meet the marking characteristics, spatial topological characteristics and combination relationships that standard markings should have.

[0077] In other words, when conflicting candidate lane boundary data appear simultaneously at the same location, using fixed weights or simple voting strategies may result in outputs inconsistent with actual road rules, thus affecting the vehicle's path planning and control stability. Therefore, how to obtain accurate, stable, and rule-consistent target lane boundary data even when multi-source recognition results differ or even conflict has become a problem that current technologies urgently need to overcome.

[0078] To address the technical problems existing in the prior art, the inventors of this application propose the following concept: During actual vehicle operation, the recognition results of different sensors (such as cameras, millimeter-wave radar, and lidar) for the same lane boundary often differ or even conflict. Furthermore, the reliability of a single sensor drops sharply in specific environments such as rain, snow, and backlighting, leading to inconsistent, abrupt, or lost lane boundary data. Directly using this data for control could pose safety risks. Therefore, we can draw upon common sense driving principles—lane lines have fixed color widths, solid and dashed lines have specific arrangement rules, and double yellow lines do not suddenly become single white lines—to abstract these physical and traffic semantics into standard lane edge marking features, spatial topological features, and combined features. We also introduce priority features (such as camera priority in sunny weather, radar priority in rainy weather, and solid lines priority over dashed lines) to construct a rule base. In this way, for the data set at each location, the vehicle no longer simply averages or votes, but uses the rule base for consistency correction: outliers that violate the basic topology are removed, reliable information sources are selected according to environmental priority, and type conflicts are resolved according to marking priority. Finally, the target lane boundary data that can be directly used for vehicle lateral control is output.

[0079] and then, Figure 1 The schematic diagram of the architecture and signal flow of the lane boundary determination system provided in this application is as follows: Figure 1 As shown, the lane boundary determination system includes an input layer, a processing layer, and an output layer.

[0080] One possible signal flow direction could be:

[0081] In the input layer, multiple source sensors, such as cameras and LiDAR, collect data that is the raw lane boundary recognition data.

[0082] In the processing layer, data preprocessing includes coordinate transformation and spatiotemporal registration, and low-confidence filtering; candidate lane boundary generation includes preliminary fusion; logical consistency verification includes type, combination, and topology (lane boundary rule base); conflict detection and elimination includes priority filtering (lane boundary rule base); and boundary correction and supplementation.

[0083] In the output layer: corrected lane boundary information; autonomous driving applications: path planning & control.

[0084] According to this method embodiment for determining system construction, the following technical effects can be achieved:

[0085] (1) Significantly improves lane boundary recognition accuracy: Compared with simple image recognition (70% accuracy), the accuracy is improved by more than 10% (reaching 78% and above);

[0086] Improvements: Introduce a standards-based rule base for logical correction and priority filtering;

[0087] Cause: The rule base provides objective and authoritative judgment criteria, which can effectively filter out erroneous identifications caused by sensor noise, obstruction, bad weather, etc., correct unreasonable identification results, and make rule-based intelligent selections for multi-source conflicting data, thereby making up for the shortcomings of relying solely on sensor identification capabilities.

[0088] (2) Enhance the robustness of the system to complex environments: It can still maintain high lane boundary recognition stability under conditions such as poor lighting, bad weather or worn lane markings;

[0089] Improvements: Instead of relying solely on raw sensor data, reasoning is performed by combining prior knowledge (standard rules);

[0090] Cause: The knowledge in the rule base (such as the standard combination of markings and geometric characteristics) is not affected by the environment. Even if the quality of sensor data deteriorates, the rules can effectively constrain and guide the recognition results, reducing the negative impact of environmental factors.

[0091] (3) Improve the safety of autonomous driving decision-making: provide more reliable basic data for path planning and vehicle control, and reduce the risk of autonomous driving decision-making errors caused by lane recognition errors;

[0092] Improvements: Output lane boundary information that conforms to traffic rules;

[0093] Cause: The corrected lane boundaries strictly adhere to the standards, ensuring that the autonomous driving system correctly understands the road traffic rules. For example, it will not misinterpret a solid line prohibiting lane changes as a dashed line, thereby avoiding inappropriate lane changing behavior.

[0094] (4) Reduce reliance on a single sensor or a specific recognition algorithm: improve the system’s generalization ability and adaptability, and reduce the need for high-end sensors or large-scale labeled data;

[0095] Improvements: Correction is based on general rules, rather than algorithms that rely heavily on the recognition capabilities of specific sensors or specific training data;

[0096] Reason for its existence: The rule base is built based on common standards and has universality. Even if the sensor or recognition algorithm is changed, as long as basic lane boundary candidate information can be output, the rule base can play a corrective role, thereby improving the system's flexibility and economy.

[0097] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with specific embodiments. The subject of this technical solution is a vehicle, such as a control unit in a vehicle.

[0098] The following 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 be described below with reference to the accompanying drawings.

[0099] Figure 2 Flowchart of the method for determining lane boundaries provided in this application Figure 1 ,like Figure 2 As shown, the method includes:

[0100] Step 21: Acquire multi-source lane boundary data collected by the vehicle;

[0101] Among them, multi-source lane boundary data is data obtained by identifying lane boundaries based on multiple sensors;

[0102] In this step, lane boundaries on the road are identified using sensors such as cameras, lidar, and millimeter-wave radar installed on the vehicle.

[0103] Optionally, the multi-source lane boundary data includes: multiple lane boundary data and the confidence score corresponding to each lane boundary data.

[0104] For example, for a camera, image recognition algorithms (such as Convolutional Neural Network (CNN) or Transformer model) can be used to output lane boundary data such as pixel coordinates of lane boundaries, lane line type (such as solid line, dashed line, double yellow line, etc.), and corresponding confidence scores (such as ratings); the data format output by the camera can be a pixel-level mask or a vectorized line segment (including parameters such as start point, end point, curvature).

[0105] For LiDAR, point cloud segmentation and clustering algorithms are used to output lane boundary data such as 3D point cloud coordinates and reflection intensity, as well as confidence scores. The LiDAR data format is a set of point clouds, with each point cloud point accompanied by reflection intensity and confidence scores.

[0106] For millimeter-wave radar, information on road edges or large obstacles is provided, indirectly assisting in lane boundary determination, and its data can be used as a supplement to fusion.

[0107] In the process of acquiring multi-source lane boundary data, each lane boundary data is accompanied by a confidence score value to indicate the reliability of the recognition result; the confidence score value can range from 0 to 1, and the higher the value, the more reliable the recognition result.

[0108] For example, the confidence level of lane lines output by a camera can reach 0.95 under good lighting conditions, but may drop to 0.60 under backlight or nighttime conditions.

[0109] Step 22: Based on the multi-source lane boundary data, determine at least one set of lane boundary data at the same location;

[0110] Each lane boundary data set includes: at least one candidate lane boundary data set;

[0111] In this step, all lane boundary data from different sensors (such as cameras, lidar, and millimeter-wave radar) are uniformly converted to the same vehicle coordinate system or world coordinate system.

[0112] Coordinate transformation can be achieved using pre-calibrated sensor extrinsic matrices (e.g., rotation matrix and translation vector), ensuring that data from different sensors are spatially aligned.

[0113] Furthermore, spatial clustering is performed on the converted lane boundary data. A clustering algorithm based on Euclidean distance can be used to group lane boundary identification results that are spatially adjacent (for example, distance thresholds can be set according to the calibration, such as 0.3 meters laterally and 1.0 meters longitudinally) into the same lane boundary data set. Each lane boundary data set contains a candidate lane boundary data corresponding to a lane boundary in the actual road, such as a lane line.

[0114] During the clustering process, each candidate lane boundary data can retain the following attributes: geometric parameters (e.g., line segment start point, end point, curvature), lane boundary type (e.g., solid line, dashed line, double yellow line, etc.), source sensor identifier, and the confidence value output by that sensor.

[0115] For example, in a certain road segment, the camera identifies the left lane line as a white dashed line with a confidence level of 0.85; the LiDAR identifies the lane line at the same location as a white dashed line with a confidence level of 0.78. Because these two identification results are spatially close, they are grouped into the same lane boundary data set, forming a set containing two candidate lane boundary data (camera and LiDAR).

[0116] Step 23: For each lane boundary data set, perform correction processing on the lane boundary data set according to the preset lane boundary rule base to obtain the target lane boundary data applied to control the vehicle.

[0117] The lane boundary rule base is constructed based on the marking features of standard lane boundaries, the spatial topological features of standard lane boundaries, the lane boundary combination features of standard lane boundaries, and priority features.

[0118] In this step, based on the preset lane boundary rule base, logical consistency checks and conflict resolution are performed on at least one candidate lane boundary data in each lane boundary data set, and finally, target lane boundary data that meets the preset standards and can be used for vehicle control is output.

[0119] Optionally, lane marking features include: visual features of lane markings and geometric parameters of lane markings;

[0120] In this implementation, 1) the lane marking features can be defined by visual features (e.g., color, width, solid / dark style) and geometric parameters (e.g., dashed line segment length, interval length).

[0121] For example, the length of a white dashed line segment is 600cm and the interval is 900cm; the width of each solid line in a double yellow line is 15cm and the distance between the two solid lines is 15cm.

[0122] 2) Lane boundary combination features include: preset lane boundary combination methods for different road types and different traffic scenarios;

[0123] In this implementation, the lane boundary combination feature can specify the allowed lane boundary combination methods under different road types (e.g., highways, urban arterial roads, secondary arterial roads, and branch roads) and different traffic scenarios (e.g., intersections, entrances and exits, curves, and ramps).

[0124] For example, the central median strip on highways should have double solid yellow lines or single solid yellow lines on both sides; solid white lines, dashed white lines, or dashed and solid lines may exist between lanes traveling in the same direction (lane changes are allowed on the side with the dashed line).

[0125] 3) Spatial topological features include: the relative positional relationship between lane boundaries, the parallelism of lane lines in the same direction, and the curvature of lane lines.

[0126] In this implementation, spatial topological features can describe the relative positional relationships, parallelism requirements, and continuity requirements between lane boundaries.

[0127] For example, lane lines in the same direction should be roughly parallel (e.g., with an angle of less than 5 degrees); the width of adjacent lanes should be between 2.5 meters and 3.75 meters; the curvature of lane lines should change smoothly, with a curvature change rate not exceeding 0.01 / m².

[0128] 4) Priority features include: the reliability priority of sensors under different environmental conditions, and the recognition priority of different types of markings at the same location.

[0129] In this implementation, reliability priority defines the reliability priority of different sensors under different environmental conditions.

[0130] For example, in good lighting and clear weather, cameras have a higher priority in recognizing colors and solid / dark lines than lidar; when heavy rain or fog causes cameras to malfunction or degrade in performance, lidar's reliability becomes the primary data source; millimeter-wave radar has the best stability in adverse weather conditions and can be used as a backup reference in extreme situations.

[0131] In this implementation, the recognition priority is defined to prioritize the recognition of different types of markings at the same location.

[0132] For example, when recognizing double yellow lines, if one sensor recognizes it as a double solid yellow line and the other as a single dashed yellow line, the double solid yellow line has a higher rule compliance (more in line with national standards) and its priority is set higher; solid lines usually have a higher priority than dashed lines because solid lines have a stronger constraint on driving behavior.

[0133] Furthermore, after obtaining the target lane boundary data, the target lane boundary data can be output to other modules of the autonomous driving system, such as path planning modules, decision control modules, and high-precision map construction and updating modules.

[0134] These modules perform lateral vehicle control (e.g., keeping the lane centered), lane change decisions, and navigation path generation based on more accurate lane boundary information.

[0135] In addition, it can record the original multi-source lane boundary data, the correction process, and the comparison of results before and after correction, which can be used for performance evaluation and optimization of subsequent models and related algorithms.

[0136] When the correction effect is poor in certain scenarios or when specific rule conflicts are frequently triggered, feedback can be sent to the cloud, indicating that the lane boundary rule base may need to be updated or the sensor recognition algorithm optimized, so as to update the vehicle and make the vehicle's subsequent application of lane boundary determination methods more efficient.

[0137] Optionally, the lane boundary rule base contains lane marking features, spatial topology features, lane boundary combination features, and priority features, which can be stored in the form of structured data (e.g., XML, JSON) or knowledge graphs and can be upgraded according to a unified standard.

[0138] In addition, the lane boundary rule base supports modular updates. When a high recognition error rate is detected in a specific scenario, the lane marking combination rules or priority strategies in the rule base can be dynamically adjusted.

[0139] For example, in urban road intersection scenarios, if the misjudgment rate of "combination of single yellow dashed line and single yellow solid line" is found to be high, the combination rule in the lane boundary rule library can be updated to "only combination of white solid line and white dashed line is allowed".

[0140] The lane boundary determination method provided in this application involves acquiring multi-source lane boundary data collected by a vehicle. This multi-source lane boundary data is obtained by identifying lane boundaries using multiple sensors. Based on the multi-source lane boundary data, at least one set of lane boundary data at the same location is determined. Each set of lane boundary data includes at least one candidate lane boundary data. For each set of lane boundary data, a correction process is performed according to a preset lane boundary rule base to obtain target lane boundary data for controlling the vehicle. The lane boundary rule base is constructed based on the marking features of standard lane boundaries, the spatial topological features of standard lane boundaries, the lane boundary combination features of standard lane boundaries, and priority features. The priority features include the reliability priority of sensors under different environmental conditions and the recognition priority of different types of markings at the same location. This technical solution acquires multi-source lane boundary data collected by vehicles, comprehensively utilizing information from different sensors to improve the redundancy and coverage of lane boundary perception. Furthermore, it determines a set of lane boundary data for the same location based on multi-source data, facilitating the alignment and fusion of recognition results from different sources for the same spatial location, providing a reliable data foundation for subsequent correction. For each lane boundary data set, correction processing is performed according to a pre-defined lane boundary rule base. This rule base integrates the marking features, spatial topological features, combined features, and priority features of standard lane boundaries, thus identifying and correcting abnormal boundaries caused by sensor noise, occlusion, or false detections. In particular, the priority features include the reliability priority of sensors under different environmental conditions, enabling the correction process to dynamically select the most reliable data source based on current weather, lighting, and other environmental factors, thereby improving the algorithm's environmental adaptability. Simultaneously, the recognition priority of different types of markings at the same location resolves decision-making problems when there are conflicts between solid lines, dashed lines, and double solid lines, ensuring that the output target lane boundary conforms to actual traffic rules. The final target lane boundary data used to control the vehicle has higher accuracy, effectively supporting stable lateral control and safe driving in complex road environments.

[0141] Based on the above embodiments, Figure 3 Flowchart of the method for determining lane boundaries provided in this application Figure 2 ,like Figure 3 As shown, in step 23, for each lane boundary data set, the following can be performed:

[0142] Step 31: For at least one candidate lane boundary data in the lane boundary data set, determine whether it is consistent with the marking features, spatial topology features, and lane boundary combination features respectively;

[0143] In this step, each candidate lane boundary data in the candidate lane boundary set is traversed, and each candidate lane boundary and its combination with other candidate lane boundaries is checked one by one to see if they conform to the rules in the rule base.

[0144] Example 1 (Checking lane boundary combination features): If a lane boundary combination in the same direction appears in the lane boundary data set, which is "a single yellow dashed line on the left and a single yellow solid line immediately adjacent to the right", it is determined that the combination violates the combination rule in the rule base that "lanes in the same direction usually use white markings". Therefore, the candidate is marked as logically inconsistent.

[0145] Example 2 (Checking Spatial Topology Features): If the geometric parameters of a candidate lane boundary show that its curvature changes drastically within a short distance (e.g., from curvature 0 to curvature 0.5 within 1 meter), exceeding the "smooth transition" threshold (e.g., the rate of change of curvature threshold) defined in the spatial topology features, then the candidate lane boundary is judged to violate the spatial topology features and is marked as inconsistent.

[0146] Optionally, a consistent judgment could be:

[0147] 1) Consistency of lane marking features: Compare the color, dashed / solid pattern, and segment length of the candidate lane boundary data with the lane marking parameters (e.g., the segment length of a white dashed line is 600cm, and the interval is 900cm). If the deviation exceeds the threshold (e.g., the segment length deviation exceeds 10%), it is marked as inconsistent.

[0148] 2) Spatial topology consistency: Verify the relative positional relationship between lane lines (e.g., parallelism, curvature). For example, if the angle between two lane lines is greater than 5 degrees, it is marked as inconsistent.

[0149] 3) Combination Consistency: Verify whether the combination of lane boundaries conforms to the combination rules in the rule base (e.g., the combination of double solid yellow lines and single dashed yellow lines is prohibited on urban roads).

[0150] If the combination is invalid, it is marked as inconsistent.

[0151] Step 32: If inconsistent candidate lane boundary data exist, the lane boundary data set is corrected based on the marking features, spatial topology features, lane boundary combination features, and priority features to obtain the target lane boundary data.

[0152] In this step, if inconsistent candidate lane boundary data exists, conflict detection, correction, and completion are performed based on lane marking features, spatial topology features, and lane boundary combination features; and filtering, correction, and completion are performed based on priority features.

[0153] Optionally, one possible implementation of step 32 could be:

[0154] Step 1: Input the lane boundary data set into the probability model to obtain the probability values ​​corresponding to at least one candidate lane boundary data. The probability model is obtained by training a Bayesian network based on the lane marking features, spatial topology features, lane boundary combination features, and priority features.

[0155] In this implementation, the knowledge (marking features, spatial topology features, combination features, priority features) in the lane boundary rule base is pre-converted into nodes and conditional probability tables of a Bayesian network. Based on historical labeled data, the conditional probability tables of the Bayesian network are trained using maximum likelihood estimation or Bayesian estimation methods to ensure that the network can reflect the logical relationships in the lane boundary rule base.

[0156] The nodes of this Bayesian network can represent variables such as "sensor type", "weather conditions", "identification type", and "spatial location relationship", while the conditional probability table of the node encodes the logical relationships and priorities in the rule base.

[0157] Then, the lane boundary data sets are used as evidence and input into the trained Bayesian network. The network calculates the posterior probability, i.e. the probability value, of each candidate lane boundary data set as the true boundary through Bayesian inference (e.g., joint tree algorithm, variational inference).

[0158] Step 2: Select the candidate lane boundary data with the highest probability value as the target lane boundary data.

[0159] In this implementation, the candidate lane boundary data with the highest posterior probability value is selected as the target lane boundary data.

[0160] The lane boundary determination method provided in this application determines whether at least one candidate lane boundary data in the lane boundary data set is consistent with the marking features, spatial topology features, and lane boundary combination features. If inconsistent candidate lane boundary data exists, the lane boundary data set is corrected according to the marking features, spatial topology features, lane boundary combination features, and priority features to obtain the target lane boundary data. This technical solution systematically identifies anomalous data from at least one candidate lane boundary data in the lane boundary dataset, determining whether it matches lane marking features, spatial topology features, and lane boundary combination features. This improves the comprehensiveness and accuracy of detection by systematically identifying anomalous data from three dimensions: visual attributes, location correlation, and overall structure. If inconsistent candidate lane boundary data exists, further correction is performed based on lane marking features, spatial topology features, lane boundary combination features, and priority features. This ensures that the correction process not only relies on multi-dimensional feature constraints but also makes reasonable decisions based on priority when features conflict. The final target lane boundary data, while retaining the original valid information, corrects erroneous boundaries that do not match multiple feature types, thereby improving the completeness, continuity, and logical consistency of the lane boundary extraction results. This provides a more reliable data foundation for subsequent high-precision map construction or autonomous driving path planning.

[0161] Based on the above embodiments, Figure 4 Flowchart of the method for determining lane boundaries provided in this application Figure 3 ,like Figure 4 As shown, step 32 may include:

[0162] Step 41: If the confidence value corresponding to inconsistent candidate lane boundary data is greater than the first preset threshold, the lane boundary data set is corrected according to the marking features, spatial topology features, lane boundary combination features, and priority features to obtain the target lane boundary data.

[0163] In this step, when the confidence value corresponding to inconsistent candidate lane boundary data is greater than a first preset threshold (e.g., 0.8), it indicates that the candidate lane boundary data itself has high reliability, and its inconsistency may be due to local noise or slight deviation. At this time, the data is corrected according to the standard parameters in the lane boundary rule base.

[0164] For example, if the identified dashed line segment is 1.8 meters long, while the standard length defined in the lane boundary rule library for the line marking features is 2.0 meters, and the confidence level of the identification result is 0.9, then its length parameter should be adjusted to 2.0 meters within a reasonable range (such as ±0.3 meters).

[0165] Step 42: If the confidence value corresponding to the inconsistent candidate lane boundary data is less than the first preset threshold, the lane boundary data set is corrected according to the priority feature to obtain the target lane boundary data.

[0166] In this step, when the confidence value corresponding to inconsistent candidate lane boundary data is less than the first preset threshold, it indicates that the reliability of the candidate lane boundary data is low. At this time, conflict resolution can be performed according to the priority features of the lane boundary rule base to obtain the target lane boundary data.

[0167] Optionally, one possible implementation of step 42 could be:

[0168] Step 1: In the lane boundary data set, identify the first candidate lane boundary data corresponding to the sensor with the highest reliability priority, excluding inconsistent candidate lane boundary data.

[0169] In this implementation, candidate lane boundary data that are inconsistent and have low confidence values ​​are excluded from the lane boundary dataset.

[0170] From the remaining candidate lane boundary data, the candidate lane boundary data provided by the sensor with the highest reliability priority is selected as the first candidate lane boundary data.

[0171] For example, in rainy or foggy weather, if the priority of LiDAR is defined as higher than that of camera in the lane boundary rule base, then the recognition result of LiDAR will be adopted first.

[0172] Step 2: If the first candidate lane boundary data is consistent with the marking features, spatial topology features, and lane boundary combination features, the first candidate lane boundary data is determined as the target lane boundary data.

[0173] In this implementation, it is checked whether the first candidate lane boundary data is consistent with the marking features, spatial topology features, and lane boundary combination features; if they are consistent, the first candidate lane boundary data is directly determined as the target lane boundary data.

[0174] Optionally, another possible implementation of step 42 is: according to the preset weights of the corresponding sensors in the lane boundary data set, at least one candidate lane boundary data in the lane boundary data set is fused to obtain the target lane boundary data, wherein the preset weights are determined based on reliability priority and / or identification priority.

[0175] In this implementation, weights are dynamically assigned to different sensors based on priority features in the lane boundary rule base.

[0176] For example, when judging the color of a line, the preset weight of the camera is set to 0.9, and the preset weight of the LiDAR is 0.1; when judging the geometric position, the preset weight of the LiDAR is set higher than that of the camera; at the same time, different preset weights are assigned to different types of recognition results according to the priority of the line type (e.g., double solid yellow lines have higher priority than single dashed yellow lines).

[0177] Next, all candidate lane boundary data in the lane boundary dataset are weighted and fused, for example, by weighted averaging or weighted voting, to calculate the parameters (e.g., location, type, etc.) of the final target lane boundary data.

[0178] Furthermore, the correction process also includes: handling discontinuous or missing boundaries.

[0179] 1) Discontinuity handling: For lane boundaries in candidate lane boundary data that are broken or discontinuous due to occlusion, sensor noise or other reasons, reasonable completion is performed based on the continuity rules in the lane boundary rule base (e.g., lane lines should extend with smooth curves) and the surrounding confirmed boundary information.

[0180] For example, if dashed line segments A and C are detected, but the middle segment B is missing, then segment B is generated by interpolation between segments A and C according to the dashed line interval length specified in the lane boundary rule base.

[0181] 2) Missing lane completion: When a lane boundary is not detected by any sensor (e.g., it is obscured by the vehicle in front), the possible location of the missing lane line is inferred and completed based on the parallelism rules in the lane boundary rule library and the position of the adjacent lane lines.

[0182] The lane boundary determination method provided in this application corrects the lane boundary data set based on marking features, spatial topology features, lane boundary combination features, and priority features when the confidence value of inconsistent candidate lane boundary data is greater than a first preset threshold, thereby obtaining target lane boundary data. If the confidence value of inconsistent candidate lane boundary data is less than the first preset threshold, the lane boundary data set is corrected based on priority features to obtain target lane boundary data. This technical solution, by integrating marking features, spatial topology features, lane boundary combination features, and priority features for correction when the confidence value of inconsistent candidate lane boundary data is greater than the first preset threshold, can fully utilize the multi-dimensional effective information contained in high-confidence data, thereby maximizing the structural consistency and detail accuracy of lane boundaries during the correction process. If the confidence value of inconsistent candidate lane boundary data is less than the first preset threshold, correction is performed only based on priority features, which can avoid interference from low-quality features on the correction results when data reliability is low, thereby reducing the risk of miscorrection and improving processing efficiency, thus improving the overall credibility and generation efficiency of target lane boundary data.

[0183] Based on the above embodiments, Figure 5 Flowchart of the method for determining lane boundaries provided in this application Figure 4 ,like Figure 5 As shown, before step 22, the following may also be included:

[0184] Step 51: Preprocess the multiple lane boundary data to obtain preprocessed multiple lane boundary data;

[0185] The preprocessing includes: transformation to the same coordinate system, time synchronization processing, and spatial registration processing;

[0186] In this step, the following preprocessing operations are performed on multiple lane boundary data:

[0187] 1) Transform to the same coordinate system: Transform all lane boundary data from different sensors (cameras, lidar, millimeter-wave radar) to a unified vehicle coordinate system (with the center of the vehicle's rear axle or the center of the Inertial Measurement Unit (IMU) as the origin) or a world coordinate system (such as the Universal Transverse Mercator (UTM) coordinate system) through the sensor extrinsic parameter matrix (rotation matrix and translation vector);

[0188] 2) Time synchronization: Time alignment of sensor data from different sampling frequencies, for example, by using interpolation or timestamp nearest neighbor matching, to ensure that the data used for fusion corresponds to the same time.

[0189] 3) Spatial registration: By calibrating the sensor parameters, ensure that the points or pixels projected by different sensors onto the same space can correspond accurately, thus eliminating spatial deviation.

[0190] Step 52: Based on the confidence value corresponding to each lane boundary data, filter the preprocessed lane boundary data to obtain filtered lane boundary data.

[0191] In this step, the preprocessed lane boundary data is filtered based on the confidence level value corresponding to each lane boundary data.

[0192] Specifically, a confidence threshold T1 (e.g., 0.5) can be set; all preprocessed recognition results are iterated through, and recognition elements with confidence values ​​lower than T1 are judged as "obviously abnormal recognition results" and filtered out (discarded). Only data with confidence values ​​greater than or equal to T1 are retained for subsequent candidate lane boundary generation and correction processing.

[0193] The lane boundary determination method provided in this application preprocesses multiple lane boundary data to obtain preprocessed multiple lane boundary data. The preprocessing includes: transformation to the same coordinate system, time synchronization processing, and spatial registration processing. Based on the confidence value corresponding to each lane boundary data, the preprocessed multiple lane boundary data is filtered to obtain filtered multiple lane boundary data. This technical solution preprocesses multiple lane boundary data, including transforming them to the same coordinate system, time synchronization, and spatial registration. This eliminates inconsistencies in coordinate references, time deviations, and spatial misalignments caused by different sensors or different acquisition times, thus providing consistent and comparable input data for subsequent processing. Based on the confidence level of each lane boundary data point, the preprocessed lane boundary data is filtered to remove low-confidence, abnormal, or unreliable data while retaining high-confidence, valid boundary information, thereby reducing noise interference and improving overall data quality. The resulting filtered lane boundary data not only achieves temporal and spatial alignment and fusion of multi-source data but also optimizes the reliability of the dataset through confidence level filtering, providing a more accurate and robust data foundation for subsequent lane boundary fusion, correction, or modeling.

[0194] Based on the above embodiments, Figure 6 A schematic diagram of the structure of the lane boundary determination device provided in this application is shown below. Figure 6As shown, the device includes:

[0195] The acquisition module 61 is used to acquire multi-source lane boundary data collected by the vehicle. The multi-source lane boundary data is data obtained by identifying lane boundaries based on multiple sensors.

[0196] The determination module 62 is used to determine at least one set of lane boundary data at the same location based on multi-source lane boundary data, wherein each set of lane boundary data includes at least one candidate lane boundary data.

[0197] The processing module 63 is used to perform correction processing on each lane boundary data set according to a preset lane boundary rule base to obtain target lane boundary data for controlling the vehicle. The lane boundary rule base is constructed based on the marking features of standard lane boundaries, the spatial topological features of standard lane boundaries, the lane boundary combination features of standard lane boundaries, and priority features. The priority features include: the reliability priority of sensors under different environmental conditions, and the recognition priority of different types of markings at the same location.

[0198] In one or more embodiments, lane marking features include: visual features of lane markings and geometric parameters of lane markings;

[0199] Lane boundary combination features include: preset lane boundary combination methods for different road types and different traffic scenarios;

[0200] Spatial topological features include: the relative positional relationship between lane boundaries, the parallelism of lane lines in the same direction, and the curvature of lane lines.

[0201] In one or more embodiments, the processing module 63 performs correction processing on the lane boundary data set according to a preset lane boundary rule base to obtain target lane boundary data for controlling the vehicle, specifically for:

[0202] For at least one candidate lane boundary data in the lane boundary dataset, determine whether it is consistent with the marking features, spatial topology features, and lane boundary combination features, respectively.

[0203] If inconsistent candidate lane boundary data exist, the lane boundary data set is corrected based on lane marking features, spatial topology features, lane boundary combination features, and priority features to obtain the target lane boundary data.

[0204] In one or more embodiments, the processing module 63 corrects the lane boundary data set based on lane marking features, spatial topology features, lane boundary combination features, and priority features to obtain target lane boundary data, specifically for:

[0205] If the confidence value corresponding to inconsistent candidate lane boundary data is greater than the first preset threshold, the lane boundary data set is corrected according to the marking features, spatial topology features, lane boundary combination features, and priority features to obtain the target lane boundary data.

[0206] If the confidence value of inconsistent candidate lane boundary data is less than the first preset threshold, the lane boundary data set is corrected according to the priority feature to obtain the target lane boundary data.

[0207] In one or more embodiments, the processing module 63 corrects the lane boundary data set according to priority features to obtain target lane boundary data, specifically for:

[0208] In the lane boundary data set, the first candidate lane boundary data corresponding to the sensor with the highest reliability priority is determined, excluding inconsistent candidate lane boundary data.

[0209] If the first candidate lane boundary data is consistent with the marking features, spatial topology features, and lane boundary combination features, the first candidate lane boundary data is determined as the target lane boundary data;

[0210] In one or more embodiments, the processing module 63 corrects the lane boundary data set according to priority features to obtain target lane boundary data, specifically for:

[0211] Based on the preset weights of the corresponding sensors in the lane boundary data set, at least one candidate lane boundary data in the lane boundary data set is fused to obtain the target lane boundary data. The preset weights are determined based on reliability priority and / or identification priority.

[0212] In one or more embodiments, the processing module 63 corrects the lane boundary data set based on lane marking features, spatial topology features, lane boundary combination features, and priority features to obtain target lane boundary data, specifically for:

[0213] The lane boundary data set is input into the probability model to obtain the probability values ​​corresponding to at least one candidate lane boundary data. The probability model is obtained by training a Bayesian network based on lane marking features, spatial topology features, lane boundary combination features, and priority features.

[0214] The candidate lane boundary data with the highest probability value is used as the target lane boundary data.

[0215] In one or more embodiments, the multi-source lane boundary data includes: multiple lane boundary data and a confidence value corresponding to each lane boundary data;

[0216] Accordingly, before determining at least one set of lane boundary data at the same location based on multi-source lane boundary data, processing module 63 is also used to:

[0217] Multiple lane boundary data are preprocessed to obtain preprocessed multiple lane boundary data. The preprocessing includes: transformation to the same coordinate system, time synchronization processing, and spatial registration processing.

[0218] Based on the confidence level value corresponding to each lane boundary data, the preprocessed lane boundary data is filtered to obtain the filtered lane boundary data.

[0219] It should be noted that the division of the various modules in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical element, or they can be physically separated. Furthermore, these modules can be implemented entirely in software through processing element calls, or entirely in hardware. Alternatively, some modules can be implemented through processing element calls in software, while others can be implemented in hardware. Moreover, these modules can be integrated together or implemented independently. The processing element here can be an integrated circuit with signal processing capabilities. During implementation, each step of the above method or each of the above modules can be completed through the integrated logic circuits in the hardware of the processor element or through software instructions.

[0220] As can be seen from the above, the lane boundary determination device provided in this embodiment can execute the method provided in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be described in detail here.

[0221] Figure 7 This is a structural schematic diagram of a vehicle provided in an embodiment of this application. Figure 7 As shown, the vehicle provided in this embodiment includes at least one processor 71 and a memory 72.

[0222] Optionally, the vehicle also includes a communication component 73.

[0223] The processor 71, memory 72, and communication component 73 are connected via bus 74.

[0224] In a specific implementation, at least one processor 71 executes computer execution instructions stored in memory 72, causing at least one processor 71 to perform the above-described method.

[0225] The specific implementation process of processor 71 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0226] 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.

[0227] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0228] 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.

[0229] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0230] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0231] 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.

[0232] 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.

[0233] 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.

[0234] 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.

[0235] 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.

[0236] 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.

[0237] 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.

[0238] 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 method for determining lane boundaries, characterized in that, include: Acquire multi-source lane boundary data collected by the vehicle, wherein the multi-source lane boundary data is obtained based on the identification of lane boundaries by multiple sensors; Based on the multi-source lane boundary data, at least one set of lane boundary data at the same location is determined, and each set of lane boundary data includes at least one candidate lane boundary data. For each lane boundary data set, the lane boundary data set is corrected according to a preset lane boundary rule base to obtain target lane boundary data for controlling the vehicle. The lane boundary rule base is constructed based on the marking features of standard lane boundaries, the spatial topological features of standard lane boundaries, the lane boundary combination features of standard lane boundaries, and priority features. The priority features include: the reliability priority of sensors under different environmental conditions, and the recognition priority of different types of markings at the same location.

2. The method according to claim 1, characterized in that, The lane marking features include: the visual features of the lane markings and the geometric parameters of the lane markings; The lane boundary combination features include: lane boundary combination methods preset for different road types and different traffic scenarios; The spatial topological features include: the relative positional relationship between lane boundaries, the parallelism of lane lines in the same direction, and the curvature of lane lines.

3. The method according to claim 1 or 2, characterized in that, The step of correcting the lane boundary data set according to a preset lane boundary rule base to obtain target lane boundary data for controlling the vehicle includes: For at least one candidate lane boundary data in the lane boundary data set, determine whether it is consistent with the lane marking features, the spatial topology features, and the lane boundary combination features, respectively. If inconsistent candidate lane boundary data exist, the lane boundary data set is corrected according to the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data.

4. The method according to claim 3, characterized in that, The step of correcting the lane boundary data set based on the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data includes: If the confidence value corresponding to the inconsistent candidate lane boundary data is greater than the first preset threshold, the lane boundary data set is corrected according to the marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data. If the confidence value corresponding to the inconsistent candidate lane boundary data is less than the first preset threshold, the lane boundary data set is corrected according to the priority feature to obtain the target lane boundary data.

5. The method according to claim 4, characterized in that, The step of correcting the lane boundary data set according to the priority feature to obtain the target lane boundary data includes: In the set of lane boundary data, the first candidate lane boundary data corresponding to the sensor with the highest reliability priority, excluding the inconsistent candidate lane boundary data, is determined. If the first candidate lane boundary data is consistent with the lane marking features, the spatial topology features, and the lane boundary combination features, the first candidate lane boundary data is determined as the target lane boundary data.

6. The method according to claim 4, characterized in that, The step of correcting the lane boundary data set according to the priority feature to obtain the target lane boundary data includes: Based on the preset weights of the corresponding sensors in the lane boundary data set, at least one candidate lane boundary data in the lane boundary data set is fused to obtain the target lane boundary data. The preset weights are determined based on the reliability priority and / or the identification priority.

7. The method according to claim 3, characterized in that, The step of correcting the lane boundary data set based on the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features to obtain the target lane boundary data includes: The lane boundary data set is input into the probability model to obtain the probability values ​​corresponding to at least one candidate lane boundary data. The probability model is obtained by training a Bayesian network based on the lane marking features, the spatial topology features, the lane boundary combination features, and the priority features. The candidate lane boundary data with the highest probability value is used as the target lane boundary data.

8. The method according to claim 1 or 2, characterized in that, The multi-source lane boundary data includes: multiple lane boundary data and a confidence value corresponding to each lane boundary data; Accordingly, before determining at least one set of lane boundary data at the same location based on the multi-source lane boundary data, the method further includes: The multiple lane boundary data are preprocessed to obtain preprocessed multiple lane boundary data. The preprocessing includes: transformation to the same coordinate system, time synchronization processing, and spatial registration processing. Based on the confidence level value corresponding to each lane boundary data, the preprocessed lane boundary data is filtered to obtain the filtered lane boundary data.

9. A device for determining lane boundaries, characterized in that, include: The acquisition module is used to acquire multi-source lane boundary data collected by the vehicle. The multi-source lane boundary data is obtained based on the identification of lane boundaries by multiple sensors. The determination module is used to determine at least one set of lane boundary data at the same location based on the multi-source lane boundary data, and each set of lane boundary data includes at least one candidate lane boundary data. The processing module is used to perform correction processing on each lane boundary data set according to a preset lane boundary rule base to obtain target lane boundary data for controlling the vehicle. The lane boundary rule base is constructed based on the marking features of standard lane boundaries, the spatial topological features of standard lane boundaries, the lane boundary combination features of standard lane boundaries, and priority features. The priority features include: the reliability priority of sensors under different environmental conditions, and the recognition priority of different types of markings at the same location.

10. A vehicle / computer-readable storage medium / computer program product, characterized in that, The vehicle includes: a memory and a 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 1-8; 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 1-8; The computer program product includes a computer program that, when executed by a processor, is used to implement the method as described in any one of claims 1-8.

Images