A driving memory method, device and terminal equipment

Abstract

The application discloses a kind of memory driving method, device and terminal equipment, belong to memory driving field, method is: obtaining lane center line set and lane center line head-tail point set;Obtain the tangential direction set of lane center line head-tail point set, based on the tangential direction set of lane center line head-tail point set and lane center line head-tail point set constructs projection broken line set;Based on any two adjacent projection broken lines in projection broken line set, the lane center line set is segmented to obtain several center line broken line segment subsets, to obtain center line broken line segment set based on all center line broken line segment subsets;Select target center line broken line segment subset from center line broken line segment set based on current vehicle trajectory;The distance between the distance between current vehicle trajectory and each center line broken line segment in target center line broken line segment subset determines main lane to carry out memory driving based on main lane, to improve the reliability of memory driving.

Description

Technical Field

[0001] This invention relates to the field of memory driving, and more particularly to a memory driving method, apparatus and terminal equipment. Background Technology

[0002] In the field of intelligent driving, memory-based driving technology is gaining increasing attention. The core of memory formation is to collect environmental data for a specific route through vehicle sensors, such as cameras, lidar, and millimeter-wave radar, to build and store a high-precision environmental model. When driving the same route again, the model is quickly matched with real-time perception results to achieve more stable lane keeping, following, and steering operations.

[0003] Currently, existing technical solutions for achieving memory-based driving mainly include: First, generating topological maps end-to-end using deep learning methods, but this method suffers from high annotation costs, weak model generalization ability, poor interpretability, and high computational pressure when deployed on the vehicle side; Second, relying on high-precision maps provided by traditional map providers, however, such maps suffer from untimely updates, difficulty in ensuring freshness, and high licensing costs. Summary of the Invention

[0004] This invention provides a method, apparatus, and terminal device for remembering driving records, which can solve the problems in the prior art and improve the reliability of remembering driving records.

[0005] This invention provides a method for remembering driving records, comprising:

[0006] Obtain the set of lane centerlines, and determine the set of start and end points of the lane centerlines based on the set of lane centerlines;

[0007] Obtain the set of tangential directions of the start and end points of the lane centerline corresponding to the set of start and end points of the lane centerline, and construct a set of projection break lines based on the set of tangential directions of the start and end points of the lane centerline and the set of start and end points of the lane centerline.

[0008] The lane centerline set is divided based on any two adjacent projected break lines in the projected break line set to obtain several centerline break line segment subsets, and a centerline break line segment set is obtained based on all the centerline break line segment subsets.

[0009] Obtain the current vehicle trajectory;

[0010] Based on the current vehicle trajectory, a subset of target centerline broken segments is selected from the set of centerline broken segments;

[0011] The main lane is determined based on the distance between the current vehicle trajectory and each centerline break segment in the target centerline break segment subset, so as to perform memory driving based on the main lane.

[0012] In the above scheme, by acquiring the set of lane centerlines and determining the set of tangential directions of the start and end points of the centerlines, as well as the set of the start and end points of the lane centerlines, a precise set of projected break lines is constructed that can effectively focus on the special and complex lane positions represented by the set of the start and end points of the lane centerlines. Based on the set of projected break lines, the set of lane centerlines is divided into multiple continuous centerline break line segment subsets, effectively segmenting the lane. This decomposes the problem of dealing with the entire lane in traditional technology into a series of centerline break line segment subsets. Then, combined with the current vehicle trajectory, the main lane most relevant to the current driving state is selected from the set of centerline break line segments for memory driving. This effectively improves the accuracy of lane recognition and matching, more accurately describes the local direction, curvature, and width changes of the lane, and effectively overcomes the matching deviation caused by the accumulation of overall fitting errors in long lanes. At the same time, this segmented structure provides flexible units for the dynamic construction of the entire lane, making it easier to handle complex scenarios such as bifurcations and merging, enhancing the ability to handle changes in road morphology, and ultimately effectively improving the reliability of memory driving in complex road environments.

[0013] Further, obtaining the set of lane centerlines and determining the set of start and end points of the lane centerlines based on the set of lane centerlines includes:

[0014] Get the set of lane lines;

[0015] For any lane line in the set of lane lines, obtain the set of adjacent lane lines of the lane line, and perform a candidate left and right boundary adjacent lane line filtering action on the set of adjacent lane lines to obtain the candidate left and right boundary adjacent lane line set corresponding to the lane line.

[0016] Perform a target left and right boundary adjacent lane line filtering action on the candidate left and right boundary adjacent lane line set to obtain the target left and right boundary adjacent lane line set corresponding to the lane line;

[0017] The lane centerline corresponding to the lane line is determined based on the lane line and the set of adjacent lane lines on the left and right boundaries of the target.

[0018] Based on all the lane centerlines corresponding to the lane line set, construct a lane centerline set corresponding to the lane line set, and determine the set of the beginning and end points of the lane centerlines based on the lane centerline set.

[0019] In the above scheme, by starting from the original set of lane lines, the target left and right boundary adjacent lane line sets are selected for each lane line, and the lane center line is generated accordingly. This ensures that the constructed set of lane center lines has clear physical boundary basis, provides an accurate geometric basis for the subsequent segmentation and application of lane center lines, and improves the credibility and reliability of the entire lane model construction.

[0020] Further, for any lane line in the lane line set, the set of adjacent lane lines of that lane line is obtained, and a candidate left and right boundary adjacent lane line filtering action is performed on the adjacent lane line set to obtain the candidate left and right boundary adjacent lane line set corresponding to that lane line. The candidate left and right boundary adjacent lane line filtering action includes:

[0021] For any lane line in the set of lane lines, obtain the current lane line start point, current lane line end point and current sampling point tangential direction sequence of the lane line;

[0022] The tangential direction of the current lane line starting point is obtained based on the tangential direction sequence of the current lane line starting point and the current sampling point, and the tangential direction of the current lane line ending point is obtained based on the tangential direction sequence of the current lane line ending point.

[0023] Obtain the set of adjacent lane lines corresponding to the lane line. For any adjacent lane line in the set of adjacent lane lines, obtain the tangential direction sequence of the adjacent sampling points of the adjacent lane line. Based on all the tangential direction sequences of the adjacent sampling points in the set of adjacent lane lines, obtain the set of tangential direction sequences of the adjacent sampling points of the lane line.

[0024] Project the starting point of the current lane line corresponding to the lane line onto the set of adjacent lane lines to obtain the set of projected lane line starting points, and project the ending point of the current lane line corresponding to the lane line onto the set of adjacent lane lines to obtain the set of projected lane line ending points;

[0025] Based on the set of starting points and ending points of the projected lane lines corresponding to the adjacent lane lines, the set of tangential directions of the starting points and the set of tangential directions of the projected lane lines are obtained by filtering from the set of tangential direction sequences of adjacent sampling points.

[0026] Based on a preset included angle threshold, the tangential direction of the current lane line starting point, the tangential direction of the current lane line ending point, the set of tangential directions of the projected lane line starting point, and the set of tangential directions of the projected lane line ending point, the included angle comparison result set corresponding to the lane line is determined;

[0027] Based on a preset distance threshold, the current lane line start point, the current lane line end point, the set of projected lane line start points, and the set of projected lane line end points, determine the set of distance comparison results corresponding to the lane line;

[0028] Based on the set of angle comparison results and the set of distance comparison results, a set of candidate left and right boundary adjacent lane lines corresponding to the lane line is selected from the set of adjacent lane lines.

[0029] In the above scheme, the starting point and ending point of the current lane line are projected onto the set of adjacent lane lines, and the directional angle and actual spatial distance at the projection point are calculated comprehensively. A dual threshold of angle and distance is set for filtering, thereby obtaining the set of adjacent lane lines on the left and right boundaries of the candidate lanes. This multi-dimensional constraint mechanism effectively filters out invalid adjacent relationships due to spatial proximity but inconsistent directions or abnormal widths, ensuring that the initially selected set of adjacent lane lines on the left and right boundaries has the physical rationality to constitute the same lane, providing accurate input for subsequent precise matching.

[0030] Further, the step of determining the set of angle comparison results corresponding to the lane line based on a preset included angle threshold, the tangential direction of the current lane line starting point, the tangential direction of the current lane line ending point, the set of tangential directions of the projected lane line starting point, and the set of tangential directions of the projected lane line ending point includes:

[0031] The set of angles between adjacent starting points is obtained based on the set of tangential directions of the current lane line starting point and the set of tangential directions of the projected lane line starting point; and the set of angles between adjacent ending points is obtained based on the set of tangential directions of the current lane line ending point and the set of tangential directions of the projected lane line ending point.

[0032] Based on the preset included angle threshold, the included angle set of adjacent starting points and the included angle set of adjacent ending points, the included angle comparison result set of the lane line is determined.

[0033] In the above scheme, by specifically calculating the set of included angles between adjacent starting points and the set of included angles between adjacent ending points, and comparing them with a preset included angle threshold, a quantitative assessment of the local parallelism of lane lines is achieved.

[0034] Further, the candidate set of adjacent lane lines at the left and right boundaries includes a set of candidate adjacent lane lines at the left boundary and a set of candidate adjacent lane lines at the right boundary. The step of performing a target adjacent lane line filtering operation on the candidate set of adjacent lane lines at the left and right boundaries to obtain the target adjacent lane line set corresponding to that lane line includes:

[0035] Perform a target left boundary adjacent lane line filtering action on the candidate left boundary adjacent lane line set corresponding to the lane line to obtain the target left boundary adjacent lane line set corresponding to the lane line.

[0036] Perform a target right boundary adjacent lane line filtering action on the candidate right boundary adjacent lane line set corresponding to the lane line to obtain the target right boundary adjacent lane line set corresponding to the lane line.

[0037] Based on the set of adjacent lane lines on the left boundary of the target and the set of adjacent lane lines on the right boundary of the target corresponding to the lane line, construct the set of adjacent lane lines on the left and right boundaries of the target corresponding to the lane line.

[0038] In the above scheme, by dividing the candidate left and right boundary adjacent lane line sets into the candidate left boundary adjacent lane line set and the candidate right boundary adjacent lane line set and performing corresponding filtering actions respectively, the asymmetry and independence of the left and right adjacent lane line relationship are handled, ensuring that the left and right boundaries determined for each lane line are the optimal solutions for their respective sides, thereby constructing a balanced and accurate lane line pairing relationship.

[0039] Further, the step of performing a target left boundary adjacent lane line filtering action on the candidate left boundary adjacent lane line set corresponding to the lane line to obtain the target left boundary adjacent lane line set corresponding to the lane line, the target left boundary adjacent lane line filtering action includes:

[0040] When there is only one candidate left boundary adjacent lane line in the set of candidate left boundary adjacent lane lines corresponding to the lane line, the set of candidate left boundary adjacent lane lines is taken as the target left boundary adjacent lane line set.

[0041] When there are multiple candidate left-boundary adjacent lane lines in the set of candidate left-boundary adjacent lane lines corresponding to the lane line, obtain the projection segment interval set of the candidate left-boundary adjacent lane line set in the lane line, and divide the candidate left-boundary adjacent lane line set into an overlapping left-boundary adjacent lane line set and a non-overlapping left-boundary adjacent lane line set based on the projection segment interval set.

[0042] Based on the lane line and the set of adjacent lane lines of the overlapping left boundary, the set of left boundary lane line widths corresponding to the lane line is obtained, and the set of adjacent lane lines of the overlapping left boundary is filtered based on the preset width and the set of left boundary lane line widths to obtain a subset of adjacent lane lines of the overlapping left boundary.

[0043] Select the subset of projection line segment intervals corresponding to the subset of adjacent lane lines of the overlapping left boundary from the set of projection line segment intervals;

[0044] Based on the projection segment interval subset, the overlapping left boundary adjacent lane line subset corresponding to the lane line is updated until the projection segment interval set corresponding to the overlapping left boundary adjacent lane line subset has no interval overlap. Then, the target left boundary adjacent lane line set corresponding to the lane line is obtained based on the overlapping left boundary adjacent lane line subset and the non-overlapping left boundary adjacent lane line set.

[0045] In the above scheme, for complex scenarios where a lane line corresponds to multiple candidate lane lines with overlapping projected segment intervals, the overlapping area is identified by introducing the concept of projected segment intervals. Preliminary filtering based on a preset width is then performed to arbitrate lane line occupancy conflicts. This automatically resolves multiple boundary issues caused by perceptual ambiguity or complex topology, such as brief parallelism, ensuring that within any tracking length interval, a lane line belongs to at most the left boundary of one lane, thus generating clear and conflict-free lane line occupancy relationships.

[0046] Further, the step of updating the overlapping left boundary adjacent lane line subset based on the projection line segment interval subset for the lane line corresponding to the lane line continues until there is no interval overlap in the projection line segment interval set corresponding to the overlapping left boundary adjacent lane line subset. Then, based on the overlapping left boundary adjacent lane line subset and the non-overlapping left boundary adjacent lane line set, the target left boundary adjacent lane line set corresponding to the lane line is obtained. The overlapping left boundary adjacent lane line subset update action includes:

[0047] From the subset of adjacent lane lines with overlapping left boundaries corresponding to the lane line, select an adjacent lane line with overlapping left boundaries according to a preset selection rule. Based on the projection segment interval of the adjacent lane line with overlapping left boundaries and the projection segment interval of other adjacent lane lines with overlapping left boundaries in the subset of adjacent lane lines with overlapping left boundaries, obtain several adjacent lane lines with overlapping left boundaries to be processed corresponding to the adjacent lane line with overlapping left boundaries from the subset of adjacent lane lines with overlapping left boundaries.

[0048] Find the starting point of the adjacent lane line of the overlapping left boundary;

[0049] For any one of the several overlapping left boundary adjacent lane lines to be processed, obtain the starting point of the left boundary adjacent lane line of the overlapping left boundary adjacent lane line to be processed, so as to obtain the set of starting points of the left boundary adjacent lane lines of the lane line based on all the starting points of the left boundary adjacent lane lines of the several overlapping left boundary adjacent lane lines to be processed.

[0050] Based on the current lane line start point, the start points of adjacent lane lines at the overlapping left boundary, and the set of start points of adjacent lane lines at the left boundary to be processed, the start point comparison result corresponding to the lane line is obtained, and the subset of adjacent lane lines at the overlapping left boundary is updated based on the start point comparison result.

[0051] In the above scheme, during the iterative process of resolving overlapping conflicts in projection intervals, a comparison rule based on the starting position of lane lines is introduced to provide a clear basis for arbitration. This rule tends to retain lane lines adjacent to the left boundary of the overlapping lanes with an earlier starting point, which usually corresponds to the actual continuity of the lanes. By dynamically updating the subset of lane lines adjacent to the left boundary of the overlapping lanes, all interval overlaps are eventually eliminated, so that each element of the set of lane lines adjacent to the target left boundary has an exclusive interval on that lane line without interference, thus ensuring the reliability of the lane topology.

[0052] Further, the step of determining the main lane based on the distance between the current vehicle trajectory and each centerline break segment in the target centerline break segment subset, so as to perform memory driving based on the main lane, includes:

[0053] Obtain the distance between the current vehicle trajectory and each target centerline break segment in the target centerline break segment subset to obtain the trajectory centerline distance set;

[0054] Obtain the target centerline break segment corresponding to the trajectory centerline distance with the smallest value in the trajectory centerline distance set, and use it as the main lane for memory driving based on the main lane.

[0055] In the above scheme, by calculating the distance from the target vehicle trajectory point to each target centerline break segment in the target centerline break segment subset and selecting the target centerline break segment corresponding to the minimum value as the main lane, an intuitive and robust main lane determination criterion is provided. This method directly links the actual driving trajectory of the vehicle with a high-precision lane geometry model, so that the main lane selection is always closely bound to the real-time pose of the vehicle, providing a stable and reliable reference benchmark for subsequent control.

[0056] Another embodiment of the present invention also provides a memory driving device, comprising:

[0057] The basic data acquisition module is used to acquire a set of lane centerlines and determine the set of start and end points of the lane centerlines based on the set of lane centerlines.

[0058] The projection break line set construction module is used to obtain the set of tangential directions of the start and end points of the lane centerline corresponding to the set of start and end points of the lane centerline, and construct the projection break line set based on the set of tangential directions of the start and end points of the lane centerline and the set of start and end points of the lane centerline.

[0059] The centerline break segment set acquisition module is used to divide the lane centerline set based on any two adjacent projected break lines in the projected break line set to obtain several centerline break segment subsets, so as to obtain the centerline break segment set based on all the centerline break segment subsets.

[0060] The current vehicle trajectory acquisition module is used to acquire the current vehicle trajectory;

[0061] The target centerline broken segment subset acquisition module is used to select a target centerline broken segment subset from the centerline broken segment set based on the current vehicle trajectory;

[0062] The main lane determination module is used to determine the main lane based on the distance between the current vehicle trajectory and each centerline break segment in the target centerline break segment subset, so as to perform memory driving based on the main lane.

[0063] Another embodiment of the present invention provides a terminal device, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the steps of a memory driving method as described in the present invention. Attached Figure Description

[0064] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0065] Figure 1 This is a flowchart illustrating a memory driving method provided in an embodiment of the present invention;

[0066] Figure 2 This is a schematic diagram of the structure of a memory driving device provided in an embodiment of the present invention;

[0067] Figure 3 This is a schematic diagram of the structure of the centerline broken line segment set provided in an embodiment of the present invention;

[0068] Figure 4 This is a schematic diagram of the structure of the candidate left boundary adjacent lane line set provided in an embodiment of the present invention;

[0069] Example of attached diagrams: 1: Lane centerline; 2: Projected break line; 3: Start point of lane centerline among the start and end points of lane centerline; 4: End point of lane centerline among the start and end points of lane centerline; 5: Lane line; 6: Example lane line; 7: Lane line adjacent to right boundary of candidate A; 8: Lane line adjacent to right boundary of candidate B; 9: Lane line adjacent to right boundary of candidate C. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0071] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0072] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0073] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0074] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0075] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0076] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0077] See Figure 1 To address the aforementioned problems in the prior art, an embodiment of the present invention provides a memory driving method, comprising:

[0078] Step S1: Obtain the set of lane centerlines, and determine the set of start and end points of the lane centerlines based on the set of lane centerlines;

[0079] Step S2: Obtain the set of tangential directions of the start and end points of the lane centerline corresponding to the set of tangential directions of the start and end points of the lane centerline, and construct the set of projection break lines based on the set of tangential directions of the start and end points of the lane centerline and the set of start and end points of the lane centerline.

[0080] Step S3: Divide the lane centerline set based on any two adjacent projected break lines in the projected break line set to obtain several centerline break line segment subsets, and obtain the centerline break line segment set based on all centerline break line segment subsets.

[0081] Step S4: Obtain the current vehicle trajectory;

[0082] Step S5: Select a subset of target centerline broken segments from the set of centerline broken segments based on the current vehicle trajectory;

[0083] Step S6: Determine the main lane based on the distance between the current vehicle trajectory and each centerline break segment in the target centerline break segment subset, so as to perform memory driving based on the main lane.

[0084] In the above scheme, by acquiring the set of lane centerlines and determining the set of tangential directions of the start and end points of the centerlines, as well as the set of the start and end points of the lane centerlines, a precise set of projected break lines is constructed that can effectively focus on the special and complex lane positions represented by the set of the start and end points of the lane centerlines. Based on the set of projected break lines, the set of lane centerlines is divided into multiple continuous centerline break line segment subsets, effectively segmenting the lane. This decomposes the problem of dealing with the entire lane in traditional technology into a series of centerline break line segment subsets. Then, combined with the current vehicle trajectory, the main lane most relevant to the current driving state is selected from the set of centerline break line segments for memory driving. This effectively improves the accuracy of lane recognition and matching, more accurately describes the local direction, curvature, and width changes of the lane, and effectively overcomes the matching deviation caused by the accumulation of overall fitting errors in long lanes. At the same time, this segmented structure provides flexible units for the dynamic construction of the entire lane, making it easier to handle complex scenarios such as bifurcations and merging, enhancing the ability to handle changes in road morphology, and ultimately effectively improving the reliability of memory driving in complex road environments.

[0085] It should be noted that the specific generation process of obtaining the lane centerline set can be based on the preceding perception and mapping process. The set of lane centerline start and end points includes the set of lane centerline start points and the set of lane centerline end points. The construction of the projection break line set based on the set of tangential directions of the centerline start and end points and the set of lane centerline start and end points refers to generating virtual line segments to divide the continuous lane centerlines at specific locations according to the spatial positions and tangential directions of the set of lane centerline start points and the set of lane centerline end points. These virtual line segments are used as the projection break line set to achieve lane alignment, ensuring that the effective driving segments of different lanes are aligned end-to-end in the direction of vehicle movement. The subset of centerline break line segments constitutes the basic lane unit for path planning. The method of determining the main lane based on the distance between the current vehicle trajectory and each centerline break segment in the target centerline break segment subset specifically refers to calculating the vertical projection distance from the trajectory point to each centerline break segment, and determining the lane corresponding to the nearest centerline break segment as the main lane. This serves as a reference benchmark for vehicle lateral control and path tracking in the lane unit corresponding to the target centerline break segment subset, thereby controlling the vehicle to perform memory driving.

[0086] In another embodiment, obtaining the set of lane centerlines and determining the set of start and end points of the lane centerlines based on the set of lane centerlines includes:

[0087] Get the set of lane lines;

[0088] For any lane line in the set of lane lines, obtain the set of adjacent lane lines of the lane line, and perform a candidate left and right boundary adjacent lane line filtering action on the set of adjacent lane lines to obtain the candidate left and right boundary adjacent lane line set corresponding to the lane line.

[0089] Perform a target left and right boundary adjacent lane line filtering action on the candidate left and right boundary adjacent lane line set to obtain the target left and right boundary adjacent lane line set corresponding to the lane line;

[0090] The lane centerline corresponding to the lane line is determined based on the lane line and the set of adjacent lane lines on the left and right boundaries of the target.

[0091] Based on all the lane centerlines corresponding to the lane line set, construct a lane centerline set corresponding to the lane line set, and determine the set of the beginning and end points of the lane centerlines based on the lane centerline set.

[0092] It should be noted that the process involves acquiring multi-sensor fusion data at the front end, generating ordered lane line points for each frame based on learning-based perception technology, and then performing corresponding preprocessing. Specifically, using existing SLAM technology, the lane lines of each frame are fused into a global coordinate system. SLAM technology primarily provides a world coordinate system and the camera pose for each frame. The lane lines extracted from each frame are in the camera coordinate system and can be obtained through... Transform the lane line points into the world coordinate system. Among them, It refers to the coordinates of a point in the world coordinate system. The vehicle's position is provided by SLAM. These are the point coordinates in the camera coordinate system. The above formula is a commonly used point coordinate transformation formula in SLAM. Using this formula, the coordinates of the ordered points of the lane line captured by the camera at different times can be uniformly transformed into a world coordinate system. Then, by filtering and fitting, the overlapping parts of consecutive frames are removed, and a complete lane line is output. At the same time, the relationship between the left and right lane lines is output. It can be understood that the lane lines are generally parallel to the direction of the car's movement. In the image, with the direction of the car's movement as the reference direction, each lane line is sorted from left to right, so that the left and right relationship of each line can be known. Furthermore, for the ordered points of each lane line, a discrete line structure is established. That is, within each lane line, the tracking length from 0 to L can be obtained, where L is the total length of the lane line, and the difference in direction between every two adjacent ordered points on the lane line is obtained. In the offline sense, this represents the tangent direction of the line at that point. With these discrete tangent directions, and by using the tracking length position, the tangent direction at any position on the line can be interpolated, thus transforming discrete points into continuous lines. After the above processing, all lane lines are combined into the lane line set.

[0093] Next, all lane lines in the lane line set are traversed. For any lane line in the lane line set, its adjacent lines with left-right relationships are found and included as the adjacent lane line set for that lane line. It is understandable that due to shooting errors, recognition errors, unclear lane lines leading to lane line breaks, complex road sections, etc., a lane line may have several adjacent lane lines on its left or right boundaries. Therefore, there may be cases where a lane line has multiple adjacent lane lines, thus obtaining its corresponding adjacent lane line set. For any lane line in the lane line set, the adjacent lane line set for that lane line is obtained, and a candidate left-right boundary adjacent lane line filtering action is performed on the adjacent lane line set. The candidate left-right boundary adjacent lane line filtering action aims to initially filter out several adjacent lane lines that may form the same lane with the current lane line from geometrically adjacent lane lines, thereby obtaining the candidate left-right boundary adjacent lane line set for that lane line. The target left-right boundary adjacent lane line filtering action, based on the candidate left-right boundary adjacent lane line set, further determines the final target left-right boundary adjacent lane line set that is paired with the current lane line through further geometric constraints. Then, a set of lane center lines corresponding to the set of lane lines is constructed based on all the lane center lines corresponding to the set of lane lines. Specifically, the determination of the lane center line is generated by calculating the midpoint of the corresponding position between the set of lane lines and the set of lane lines adjacent to their respective target left and right boundaries.

[0094] In another embodiment, for any lane line in the lane line set, the process of obtaining the set of adjacent lane lines for that lane line and performing a candidate left and right boundary adjacent lane line filtering action on the set of adjacent lane lines to obtain the candidate left and right boundary adjacent lane line set corresponding to that lane line includes:

[0095] For any lane line in the set of lane lines, obtain the current lane line start point, current lane line end point and current sampling point tangential direction sequence of the lane line;

[0096] The tangential direction of the current lane line starting point is obtained based on the tangential direction sequence of the current lane line starting point and the current sampling point, and the tangential direction of the current lane line ending point is obtained based on the tangential direction sequence of the current lane line ending point.

[0097] Obtain the set of adjacent lane lines corresponding to the lane line. For any adjacent lane line in the set of adjacent lane lines, obtain the tangential direction sequence of the adjacent sampling points of the adjacent lane line. Based on all the tangential direction sequences of the adjacent sampling points in the set of adjacent lane lines, obtain the set of tangential direction sequences of the adjacent sampling points of the lane line.

[0098] Project the starting point of the current lane line corresponding to the lane line onto the set of adjacent lane lines to obtain the set of projected lane line starting points, and project the ending point of the current lane line corresponding to the lane line onto the set of adjacent lane lines to obtain the set of projected lane line ending points;

[0099] Based on the set of starting points and ending points of the projected lane lines corresponding to the adjacent lane lines, the set of tangential directions of the starting points and the set of tangential directions of the projected lane lines are obtained by filtering from the set of tangential direction sequences of adjacent sampling points.

[0100] Based on a preset included angle threshold, the tangential direction of the current lane line starting point, the tangential direction of the current lane line ending point, the set of tangential directions of the projected lane line starting point, and the set of tangential directions of the projected lane line ending point, the included angle comparison result set corresponding to the lane line is determined;

[0101] Based on a preset distance threshold, the current lane line start point, the current lane line end point, the set of projected lane line start points, and the set of projected lane line end points, determine the set of distance comparison results corresponding to the lane line;

[0102] Based on the set of angle comparison results and the set of distance comparison results, a set of candidate left and right boundary adjacent lane lines corresponding to the lane line is selected from the set of adjacent lane lines.

[0103] It should be noted that the attribute data corresponding to the lane line set includes the set of lane line sampling point tangential direction sequences. This means that each lane line corresponds to a lane line sampling point tangential direction sequence. It's important to note that because the discrete points have been treated as continuous lines in the above processing, the sampling points mentioned in the text specifically refer to the position points corresponding to any tracking length on any lane line. The set of lane line sampling point tangential direction sequences refers to the set of lane line sampling point tangential direction sequences formed by the tangential direction sequences of all lane lines. It's important to note that the lane line sampling point tangential direction sequence corresponding to any lane line is used to represent the value calculated based on the vector difference between any two adjacent sampling points on that lane line. For any lane line in the lane line set, the lane line sampling point tangential direction sequence corresponding to that lane line is used as the current sampling point tangential direction sequence. Based on the current lane line start point, the current lane line start point tangential direction is selected from the current sampling point tangential direction sequence; based on the current lane line end point, the current lane line end point tangential direction is obtained from the current sampling point tangential direction sequence. Simultaneously, the set of adjacent lane lines corresponding to the current lane line is obtained. For any adjacent lane line in the set, the tangential direction sequence of the lane line sampling points of that adjacent lane line is obtained as the tangential direction sequence of the adjacent sampling points. Based on all the tangential direction sequences of the adjacent sampling points in the set of adjacent lane lines, the set of tangential direction sequences of the adjacent sampling points of the current lane line is obtained. Furthermore, the attribute data corresponding to the lane line set also includes a set of lane line start points and a set of lane line end points. For any lane line in the set, the lane line start point corresponding to that lane line is taken as the current lane line start point, and the lane line end point corresponding to that lane line is taken as the current lane line end point.

[0104] The process of projecting the current lane start and end points of the current lane line onto the set of adjacent lane lines refers to performing a vertical projection calculation from point to line segment. Specifically, first, find the two sampling points closest to the current lane start point on each adjacent lane line in the set of adjacent lane lines, connect these two sampling points to form a line segment, obtain the vertical projection from the current lane start point to this line segment, and use this as the starting point of the projected lane line. Obtain the projected lane start points corresponding to all adjacent lane lines to obtain the set of projected lane start points. Similarly, find the two sampling points closest to the current lane end point on each adjacent lane line in the set of adjacent lane lines, connect these two sampling points to form a line segment, obtain the vertical projection from the current lane end point to this line segment, and use this as the ending point of the projected lane line. Obtain the projected lane end points corresponding to all adjacent lane lines to obtain the set of projected lane end points.

[0105] Then, based on the set of projected lane line start points and the set of projected lane line end points, the set of tangential direction sequences of adjacent sampling points is queried to obtain the set of tangential direction sequences of projected lane line start points and the set of tangential direction sequences of projected lane lines. Next, using a preset angle threshold, the tangential direction sequences of the current lane line start point and the current lane line end point, the set of tangential direction sequences of the projected lane line start point and the set of tangential direction sequences of the projected lane line end point, it is determined whether the direction of the current lane line is approximately parallel to that of any adjacent lane line in the adjacent lane line set at its start or end point. For example, the preset angle threshold can be set to 90 degrees to obtain the set of angle comparison results between the current lane line and any adjacent lane line in the adjacent lane line set. Finally, using a preset distance threshold, the current lane line start point, the current lane line end point, the set of projected lane line start points, and the set of projected lane line end points, it is determined whether the lane width calculated based on projection of the current lane line and any adjacent lane line in the adjacent lane line set is within a reasonable range, such as 2.5 meters to 4 meters, to obtain the set of distance comparison results between the current lane line and any adjacent lane line in the adjacent lane line set. Only adjacent lane lines that simultaneously satisfy both the direction angle constraint and the width distance constraint in the set of angle comparison results and the set of distance comparison results will be included in the candidate set of adjacent lane lines on the left and right boundaries.

[0106] In another embodiment, determining the set of angle comparison results corresponding to the lane line based on a preset included angle threshold, the tangential direction of the current lane line start point, the tangential direction of the current lane line end point, the set of tangential directions of the projected lane line start point, and the set of tangential directions of the projected lane line end point includes:

[0107] The set of angles between adjacent starting points is obtained based on the set of tangential directions of the current lane line starting point and the set of tangential directions of the projected lane line starting point; and the set of angles between adjacent ending points is obtained based on the set of tangential directions of the current lane line ending point and the set of tangential directions of the projected lane line ending point.

[0108] Based on the preset included angle threshold, the included angle set of adjacent starting points and the included angle set of adjacent ending points, the included angle comparison result set of the lane line is determined.

[0109] It should be noted that the specific process of determining the angle comparison result set corresponding to the lane line based on the preset angle threshold, the tangential direction of the current lane line starting point, the tangential direction of the current lane line ending point, the set of tangential directions of the projected lane line starting point, and the set of tangential directions of the projected lane line ending point is as follows: Obtain the set of angle differences between the tangential direction of the current lane line starting point and the set of tangential directions of the projected lane line starting point, as the set of adjacent starting point angles; obtain the set of angle differences between the tangential direction of the current lane line ending point and the set of tangential directions of the projected lane line ending point, as the set of adjacent ending point angles; the preset angle threshold is preferably 90°; compare the set of adjacent starting point angles with the preset angle threshold to obtain the starting point angle comparison result set; compare the set of adjacent ending point angles with the preset angle threshold to obtain the ending point angle comparison result set; and based on the set of starting point angle comparison result set and the set of ending point angle comparison result set, obtain the angle comparison result set.

[0110] Furthermore, the specific process of determining the distance comparison result set corresponding to the lane line based on the preset distance threshold, the current lane line start point, the current lane line end point, the set of projected lane line start points, and the set of projected lane line end points is as follows: Obtain the distance difference between the current lane line start point and the set of projected lane line start points to obtain the start point distance difference set; obtain the distance difference between the current lane line end point and the set of projected lane line end points to obtain the end point distance difference set; and obtain the distance comparison result set based on the start point distance difference set, the end point distance difference set, and the preset distance threshold. Then, based on the angle comparison result set and the distance comparison result set, select the candidate left and right boundary adjacent lane line sets corresponding to the lane line from the adjacent lane line set. Specifically, only adjacent lane lines that meet the following conditions in the angle comparison result set (where the angle between adjacent start points is less than the preset angle threshold and the angle between adjacent end points is less than the preset angle threshold), and also meet the condition that the start point distance difference is less than the preset distance threshold and the end point distance difference is less than the preset distance threshold in the distance comparison result set, can be used to construct the candidate left and right boundary adjacent lane line sets. In most real-world situations, the left and right lane lines are parallel, including their beginning and end positions. This lane line should also be nearly parallel to its normal adjacent lane lines in most cases; even at the endpoints of lane branching or merging, the angle between adjacent endpoints should not exceed 90 degrees. Therefore, distance constraints and angle constraints can eliminate abnormal situations.

[0111] In another embodiment, the candidate set of adjacent lane lines at the left and right boundaries includes a set of candidate adjacent lane lines at the left boundary and a set of candidate adjacent lane lines at the right boundary. The step of performing a target adjacent lane line filtering operation on the candidate set of adjacent lane lines at the left and right boundaries to obtain the target adjacent lane line set corresponding to that lane line includes:

[0112] Perform a target left boundary adjacent lane line filtering action on the candidate left boundary adjacent lane line set corresponding to the lane line to obtain the target left boundary adjacent lane line set corresponding to the lane line.

[0113] Perform a target right boundary adjacent lane line filtering action on the candidate right boundary adjacent lane line set corresponding to the lane line to obtain the target right boundary adjacent lane line set corresponding to the lane line.

[0114] Based on the set of adjacent lane lines on the left boundary of the target and the set of adjacent lane lines on the right boundary of the target corresponding to the lane line, construct the set of adjacent lane lines on the left and right boundaries of the target corresponding to the lane line.

[0115] It should be noted that for any lane line in the lane line set, after removing abnormal adjacent lane lines through the candidate left and right boundary adjacent lane line filtering action to obtain the corresponding candidate left and right boundary adjacent lane line set, it is necessary to further perform the target left and right boundary adjacent lane line filtering action on the corresponding candidate left and right boundary adjacent lane line set to remove overlapping adjacent lane lines, thus obtaining the target left and right boundary adjacent lane line set corresponding to that lane line. It is understood that for any lane line in the lane line set, its candidate left and right boundary adjacent lane line set includes a candidate left boundary adjacent lane line set and a candidate right boundary adjacent lane line set. The target left boundary adjacent lane line filtering action is performed on the candidate left boundary adjacent lane line set to obtain the target left boundary adjacent lane line set corresponding to that lane line, and the target right boundary adjacent lane line filtering action is performed on the candidate right boundary adjacent lane line set corresponding to that lane line to obtain the target right boundary adjacent lane line set corresponding to that lane line. The target left boundary adjacent lane line filtering action and the target right boundary adjacent lane line filtering action have the same principle but opposite direction. The purpose is to resolve the situation where a lane line may correspond to multiple overlapping candidate left or right boundary adjacent lane lines. The target left boundary adjacent lane line filtering action and the target right boundary adjacent lane line filtering action analyze the overlap of projection intervals on the candidate left and right boundary adjacent lane line sets, and perform deduplication and uniqueness determination according to preset rules. Finally, the most suitable target left boundary adjacent lane line set and target right boundary adjacent lane line set without interval overlap are selected for the lane line. Then, based on the target left boundary adjacent lane line set and target right boundary adjacent lane line set corresponding to the lane line, the target left and right boundary adjacent lane line set corresponding to the lane line is constructed.

[0116] In another embodiment, the step of performing a target left-boundary adjacent lane line filtering action on the candidate left-boundary adjacent lane line set corresponding to the lane line to obtain the target left-boundary adjacent lane line set corresponding to the lane line includes:

[0117] When there is only one candidate left boundary adjacent lane line in the set of candidate left boundary adjacent lane lines corresponding to the lane line, the set of candidate left boundary adjacent lane lines is taken as the target left boundary adjacent lane line set.

[0118] When there are multiple candidate left-boundary adjacent lane lines in the set of candidate left-boundary adjacent lane lines corresponding to the lane line, obtain the projection segment interval set of the candidate left-boundary adjacent lane line set in the lane line, and divide the candidate left-boundary adjacent lane line set into an overlapping left-boundary adjacent lane line set and a non-overlapping left-boundary adjacent lane line set based on the projection segment interval set.

[0119] Based on the lane line and the set of adjacent lane lines of the overlapping left boundary, the set of left boundary lane line widths corresponding to the lane line is obtained, and the set of adjacent lane lines of the overlapping left boundary is filtered based on the preset width and the set of left boundary lane line widths to obtain a subset of adjacent lane lines of the overlapping left boundary.

[0120] Select the subset of projection line segment intervals corresponding to the subset of adjacent lane lines of the overlapping left boundary from the set of projection line segment intervals;

[0121] Based on the projection segment interval subset, the overlapping left boundary adjacent lane line subset corresponding to the lane line is updated until the projection segment interval set corresponding to the overlapping left boundary adjacent lane line subset has no interval overlap. Then, the target left boundary adjacent lane line set corresponding to the lane line is obtained based on the overlapping left boundary adjacent lane line subset and the non-overlapping left boundary adjacent lane line set.

[0122] It should be noted that, in most cases, for any lane line in the set of lane lines, there will only be one candidate left-boundary adjacent lane line in the set of candidate left-boundary adjacent lane lines corresponding to that lane line. In this case, the set of candidate left-boundary adjacent lane lines is taken as the target left-boundary adjacent lane line set. However, sometimes lane line breaks or repetitions may occur due to problems such as lane line image acquisition errors, unclear lane line traces, or complex roads. Therefore, there may be multiple candidate left-boundary adjacent lane lines in the set of candidate left-boundary adjacent lane lines corresponding to that lane line.

[0123] When there are multiple candidate left-boundary adjacent lane lines in the set of candidate left-boundary adjacent lane lines corresponding to the lane line, the projection segment interval set of the candidate left-boundary adjacent lane line set on the lane line is obtained. The projection segment interval refers to the tracking length range occupied by the candidate left-boundary adjacent lane line on the lane line. Based on the value range in the projection segment interval set, the candidate left boundary adjacent lane line set is divided into an overlapping left boundary adjacent lane line set and a non-overlapping left boundary adjacent lane line set. For example, lane lines a, b, and c are all in the candidate left boundary adjacent lane line set of this lane line. Assuming that lane line a occupies the tracking length of this lane line [1,20] in the projection segment interval of this lane line, and lane line b occupies the tracking length of this lane line [5,15] in the projection segment interval of this lane line, then it can be seen that the overlapping area of ​​the two is in [5,15], which belongs to the overlapping left boundary adjacent lane line set. However, assuming that lane line a occupies the tracking length of this lane line [1,20] in the projection segment interval of this lane line, and lane line b occupies the tracking length of this lane line [5,30] in the projection segment interval of this lane line, then it can be seen that the overlapping area of ​​the two is in [5,20], and lane lines a and b still belong to the overlapping left boundary adjacent lane line set.

[0124] The set of starting points of lane lines corresponding to the set of adjacent lane lines at the overlapping left boundary is vertically projected onto the lane line, and the corresponding width is obtained. This width is used as the set of left boundary lane line widths corresponding to the set of adjacent lane lines at the overlapping left boundary. Then, a filter is performed based on a preset width. It is understood that the preset width can be manually set based on the standard lane width and a certain tolerance value, or it can be obtained from the set of left boundary lane line widths, with the minimum width as a preset multiple, preferably 2 times, as the preset width. The set of adjacent lane lines at the overlapping left boundary that are less than or equal to the preset width is preferentially retained, resulting in a subset of adjacent lane lines at the overlapping left boundary, thereby removing those that do not conform to common sense. Correspondingly, a subset of projection segment intervals corresponding to the subset of adjacent lane lines at the overlapping left boundary is selected from the set of projection segment intervals. Based on this subset of projection segment intervals, an update action is performed on the subset of adjacent lane lines at the overlapping left boundary corresponding to the lane line. This update action is an iterative process designed to resolve conflicts arising from overlapping projection intervals. When multiple overlapping left-boundary adjacent lane lines in the overlapping left-boundary adjacent lane line subset have overlapping projection segment intervals, it means that they are competing to become the left boundary of the same lane line. The rules must be arbitrated to retain one and remove the projection segment intervals of the other competitors from the overlapping left-boundary adjacent lane line subset until there is no longer any interval overlap in the projection segment interval set of all retained overlapping left-boundary adjacent lane line subsets on the lane line. At this time, the updated overlapping left-boundary adjacent lane line subset and the non-overlapping left-boundary adjacent lane line set are summarized to obtain the target left-boundary adjacent lane line set corresponding to the lane line.

[0125] In another embodiment, the step of updating the overlapping left boundary adjacent lane line subset based on the projection line segment interval subset for the lane line corresponding to the lane line continues until there is no interval overlap in the projection line segment interval set corresponding to the overlapping left boundary adjacent lane line subset. Then, based on the overlapping left boundary adjacent lane line subset and the non-overlapping left boundary adjacent lane line set, the target left boundary adjacent lane line set corresponding to the lane line is obtained. The overlapping left boundary adjacent lane line subset update action includes:

[0126] From the subset of adjacent lane lines with overlapping left boundaries corresponding to the lane line, select an adjacent lane line with overlapping left boundaries according to a preset selection rule. Based on the projection segment interval of the adjacent lane line with overlapping left boundaries and the projection segment interval of other adjacent lane lines with overlapping left boundaries in the subset of adjacent lane lines with overlapping left boundaries, obtain several adjacent lane lines with overlapping left boundaries to be processed corresponding to the adjacent lane line with overlapping left boundaries from the subset of adjacent lane lines with overlapping left boundaries.

[0127] Find the starting point of the adjacent lane line of the overlapping left boundary;

[0128] For any one of the several overlapping left boundary adjacent lane lines to be processed, obtain the starting point of the left boundary adjacent lane line of the overlapping left boundary adjacent lane line to be processed, so as to obtain the set of starting points of the left boundary adjacent lane lines of the lane line based on all the starting points of the left boundary adjacent lane lines of the several overlapping left boundary adjacent lane lines to be processed.

[0129] Based on the current lane line start point, the start points of adjacent lane lines at the overlapping left boundary, and the set of start points of adjacent lane lines at the left boundary to be processed, the start point comparison result corresponding to the lane line is obtained, and the subset of adjacent lane lines at the overlapping left boundary is updated based on the start point comparison result.

[0130] It should be noted that the preset selection rule is to select sequentially according to the vehicle's forward direction. When the selected overlapping left boundary adjacent lane line projection segment interval does not overlap with the projection segment intervals of other overlapping left boundary adjacent lane lines in the overlapping left boundary adjacent lane line subset, the next overlapping left boundary adjacent lane line is selected from the overlapping left boundary adjacent lane line subset along the vehicle's forward direction until the selected overlapping left boundary adjacent lane line projection segment interval overlaps with the projection segment intervals of other overlapping left boundary adjacent lane lines in the overlapping left boundary adjacent lane line subset. Then, based on the overlapping left boundary adjacent lane line projection segment interval and the projection segment intervals of other overlapping left boundary adjacent lane lines in the overlapping left boundary adjacent lane line subset, several overlapping left boundary adjacent lane lines to be processed are obtained from the overlapping left boundary adjacent lane line subset.

[0131] The term "overlapping left boundary adjacent lane line to be processed" refers to other overlapping left boundary adjacent lane lines in the subset that have interval overlap with the currently selected overlapping left boundary adjacent lane line. For any one of the several overlapping left boundary adjacent lane lines to be processed, the starting point of the lane line corresponding to the overlapping left boundary adjacent lane line to be processed is selected from the set of lane line starting points as the starting point of the overlapping left boundary adjacent lane line to be processed. Based on all the starting points of the overlapping left boundary adjacent lane lines to be processed, the set of starting points of the overlapping left boundary adjacent lane lines to be processed for that lane line is obtained. At the same time, the starting point of the lane line corresponding to the overlapping left boundary adjacent lane line is selected from the set of lane line starting points as the starting point of the overlapping left boundary adjacent lane line.

[0132] Based on the current lane line start point, the start points of adjacent lane lines at the overlapping left boundary, and the set of start points of adjacent lane lines at the left boundary to be processed, the start point comparison result corresponding to the lane line is obtained. Specifically, with the current lane line start point as the reference, the start points of adjacent lane lines at the overlapping left boundary and the set of start points of adjacent lane lines at the left boundary to be processed are vertically projected onto the lane line, and then the distance values ​​between their projection points and the current lane line start point are obtained as the start point comparison result.

[0133] Then, based on the starting point comparison results, the subset of adjacent lane lines at the overlapping left boundary is updated: that is, the adjacent lane line at the smallest distance among several adjacent lane lines at the overlapping left boundary to be processed is retained, and the other adjacent lane lines at the overlapping left boundary to be processed are removed from the subset of adjacent lane lines at the overlapping left boundary. This iterative process continues until all interval overlaps are eliminated. It is understood that the selection action for adjacent lane lines at the target right boundary is similar to the selection action for adjacent lane lines at the target left boundary, except that the object it processes is the set of candidate adjacent lane lines at the right boundary corresponding to the lane line, and the result is the set of adjacent lane lines at the target right boundary corresponding to the lane line.

[0134] The following example illustrates how to obtain a subset of overlapping right boundary adjacent lane lines by performing a filtering action on the right boundary of the target lane. (Refer to...) Figure 4 The leftmost lane line is taken as any one of the lane lines in the set of lane lines and named as example lane line 6. Its candidate right boundary adjacent lane line set includes candidate right boundary adjacent lane line 7, candidate right boundary adjacent lane line 8, and candidate right boundary adjacent lane line 9. It can be seen that the three corresponding projected line segment intervals overlap, and therefore are all divided into overlapping right boundary adjacent lane line sets. The corresponding non-overlapping right boundary adjacent lane line sets contain 0 elements. Assuming that the right boundary lane line width of candidate right boundary adjacent lane line 9 is greater than the preset width, candidate right boundary adjacent lane line 9 is filtered out. At this time, only candidate right boundary adjacent lane line 7 and candidate right boundary adjacent lane line 8 remain in the overlapping right boundary adjacent lane line subset. Then, according to the preset selection rules, an overlapping right boundary adjacent lane line is selected, and candidate right boundary adjacent lane line 7 is selected. Its corresponding unprocessed overlapping right boundary adjacent lane lines include candidate right boundary adjacent lane line 8. Based on the current lane line starting point of example lane line 6, the starting point of overlapping left boundary adjacent lane line (i.e., the lane line starting point of candidate right boundary adjacent lane line 7), and the set of unprocessed right boundary adjacent lane line starting points (i.e., the unprocessed right boundary adjacent lane line starting point of candidate right boundary adjacent lane line 8), the starting point comparison result of the lane line is obtained, that is, the lane line starting point of candidate right boundary adjacent lane line 7 is projected onto example lane line 6, as shown. Figure 4The dashed line extending from lane line 7 adjacent to the right boundary of candidate A to example lane line 6 is the same, and the starting point of lane line 8 adjacent to the right boundary of candidate B is projected onto example lane line 6, as shown. Figure 4 The dashed line extending from candidate B's right-boundary adjacent lane line 8 to example lane line 6 shows that, based on the starting point comparison, candidate A's right-boundary adjacent lane line 7 is closer to the current starting point of example lane line 6. Therefore, the subset of overlapping right-boundary adjacent lane lines is updated, retaining candidate A's right-boundary adjacent lane line 7 and removing candidate B's right-boundary adjacent lane line 8. Then, the subset of overlapping right-boundary adjacent lane lines and the set of non-overlapping right-boundary adjacent lane lines are combined to obtain the target right-boundary adjacent lane line set corresponding to this lane line.

[0135] Furthermore, after obtaining the lane line and the corresponding set of lane lines adjacent to the left and right boundaries of the target, the center line on the left side of the lane line is determined based on the lane line and the set of lane lines adjacent to the left boundary of the target, and the center line on the right side of the lane line is determined based on the lane line and the set of lane lines adjacent to the right boundary of the target. Combining the center lines on the left and right sides, all lane center lines corresponding to the lane line are determined. Based on all the lane center lines corresponding to the set of lane lines, a set of lane center lines corresponding to the set of lane lines is constructed, and the set of start and end points of the lane center lines is determined based on the set of lane center lines.

[0136] Furthermore, regarding step S2: obtaining the set of tangential directions of the start and end points of the lane centerline corresponding to the set of start and end points of the lane centerline, and constructing a set of projection break lines based on the set of tangential directions of the start and end points of the lane centerline and the set of start and end points of the lane centerline, specifically:

[0137] Because it's necessary to ensure that adjacent lane lines are broken at the start and end points of each lane line to prepare for subsequent road construction, a set of lane centerlines is obtained. Based on this set, a set of lane centerline start and end points is determined, including the set of lane centerline start points and the set of lane centerline end points. A set of tangential directions corresponding to the start and end points of each lane centerline is also obtained. For each start and end point, the tangential direction does not refer to the direction connecting the start and end points, but rather to the tangential direction between the start point and its adjacent point, and the tangential direction between the end point and its adjacent point. Line segments are then extended perpendicularly to their respective tangential directions, using the set of lane centerline start points and the set of lane centerline end points as references, to form a set of projected break lines.

[0138] Furthermore, for step S3: the lane centerline set is divided based on any two adjacent projected break lines in the projected break line set to obtain several centerline break line segment subsets, and a centerline break line segment set is obtained based on all the centerline break line segment subsets, specifically:

[0139] Then, based on the projected break line set, the lane centerline set, or the lane line sets on both sides of the lane centerline set, are subjected to uniform break processing. During uniform break, for the end of each lane line, the in-line tracking length is found by projecting onto the adjacent lane line, and a break is performed at that point. This process continues recursively until all adjacent lane lines are broken. The uniform break is not for removal, but rather to break the lane lines into different smaller segments, such as... Figure 3 As shown, the red line indicates the break point. Make a clean cut, making the adjacent lane line the same length as the current lane line. Cut the original long lane line into smaller segments. The subsequent steps will be... Figure 3 A lane group between two red lines is defined as a road. Figure 3 The black lines correspond to lane lines, and the orange lines correspond to lane center lines. This results in several subsets of center line break segments, which are then used to create a set of center line break segments. Each center line break segment is bound to left-right adjacent relationships and front-back relationships; the front-back relationship indicates belonging to the same lane center line. A subset of center line break segments can be used to construct a road segment: based on segments perpendicular to the direction of travel, a road segment is formed. These segments perpendicular to the direction of travel are the segments perpendicular to all the lane lines on either side of the lane center line, and thus also the subset of center line break segments. See... Figure 3 The section of road is between the two red lines. (Reference) Figure 3 The orange line is lane centerline 1, and the adjacent black line is lane line 5. The red line is projection break line 2, which is formed by extending line segments perpendicular to the tangential direction of the lane centerline starting point 3 and the lane centerline ending point 4, respectively, as references. It can be understood that... Figure 3 The orange line is divided by the red line, forming a center line break segment. All center line break segments between any two adjacent projected break lines constitute a center line break segment subset. All such center line break segment subsets are summarized into a center line break segment set.

[0140] Furthermore, for steps S4-6: obtaining the current vehicle trajectory; selecting a target centerline break segment subset from the set of centerline break segments based on the current vehicle trajectory; determining the main lane based on the distance between the current vehicle trajectory and each centerline break segment in the target centerline break segment subset, so as to perform memory driving based on the main lane. Specifically:

[0141] Based on the current vehicle trajectory, a target centerline break segment subset is selected from the set of centerline break segments. That is, the current vehicle trajectory is selected to be located in which centerline break segment subset, and this subset is used as the target centerline break segment subset. In the road corresponding to the target centerline break segment subset, the centerline break segment closest to the current vehicle trajectory is found and used as the main lane.

[0142] In another embodiment, determining the main lane based on the distance between the current vehicle trajectory and each centerline break segment in the target centerline break segment subset, and then performing memory driving based on the main lane, includes:

[0143] Obtain the distance between the current vehicle trajectory and each target centerline break segment in the target centerline break segment subset to obtain the trajectory centerline distance set;

[0144] Obtain the target centerline break segment corresponding to the trajectory centerline distance with the smallest value in the trajectory centerline distance set, and use it as the main lane for memory driving based on the main lane.

[0145] It should be noted that the specific calculation method for obtaining the distance between the current vehicle trajectory and each target centerline break segment can be: calculating the vertical projection distance from the point corresponding to the current vehicle trajectory or the representative point of the trajectory to each target centerline break segment. The subset of target centerline break segments typically corresponds to a road where the vehicle is currently located or about to enter, which consists of a set of horizontally parallel lanes aligned end-to-end through a break operation. The target centerline break segment corresponding to the trajectory centerline distance with the smallest value in the set of trajectory centerline distances is obtained and used as the main lane. After determining the main lane, the vehicle's lateral control, lane keeping, and other decision-making and planning modules can use the centerline of the main lane as a reference path for tracking.

[0146] Furthermore, in constructing a complete topology map, a virtual lane generation step can be included. This involves generating a virtual lane centerline based on the endpoints of the preceding and following lanes at locations where physical lane connections are interrupted but driving semantics remain continuous, ensuring the connectivity of the topology network. It also includes steps for constructing lane nodes (LNodes) and road nodes (RNodes) to explicitly record the preceding and following connections between lanes and roads, forming a topology map usable for global path search. Specifically, after determining the main lanes and constructing the basic lanes and roads, virtual lane generation and topology node creation steps are required to form a memory-based driving topology map usable for global path planning. Virtual lane generation refers to the process where, when two preceding and following lanes sharing the same lane line are detected, but this shared lane line is not fully occupied by any actual lane as a boundary in the connecting section, geometric interpolation is performed between the endpoints of the two lanes to generate a virtual lane centerline, and its width is calculated. This achieves seamless lane connection at the topology level, ensuring the connectivity of the drivable network. Furthermore, to describe the traffic logic between roads and lanes, topology nodes need to be created: First, based on the occupancy of the target vehicle's trajectory within each road, the sequential connections between roads are determined, and road nodes (RNodes) are created accordingly. Each RNode records the set of information about roads that can enter and exit from that node. Second, based on the sequential connections between roads, the sequential connections between adjacent lane lines belonging to different roads are further determined, and lane nodes (LNodes) are created accordingly. Each LNode records the set of information about lanes that can enter and exit from that lane. Finally, the virtual lanes, RNodes, and LNodes together constitute a hierarchical topology map with clearly defined connections. The memory-based driving method uses this topology map for global path planning and guidance.

[0147] Based on the above method embodiments, corresponding apparatus embodiments are provided;

[0148] like Figure 2 As shown, one embodiment of the present invention provides a memory driving device, comprising:

[0149] The basic data acquisition module is used to acquire a set of lane centerlines and determine the set of start and end points of the lane centerlines based on the set of lane centerlines.

[0150] The projection break line set construction module is used to obtain the set of tangential directions of the start and end points of the lane centerline corresponding to the set of start and end points of the lane centerline, and construct the projection break line set based on the set of tangential directions of the start and end points of the lane centerline and the set of start and end points of the lane centerline.

[0151] The centerline break segment set acquisition module is used to divide the lane centerline set based on any two adjacent projected break lines in the projected break line set to obtain several centerline break segment subsets, so as to obtain the centerline break segment set based on all the centerline break segment subsets.

[0152] The current vehicle trajectory acquisition module is used to acquire the current vehicle trajectory;

[0153] The target centerline broken segment subset acquisition module is used to select a target centerline broken segment subset from the centerline broken segment set based on the current vehicle trajectory;

[0154] The main lane determination module is used to determine the main lane based on the distance between the current vehicle trajectory and each centerline break segment in the target centerline break segment subset, so as to perform memory driving based on the main lane.

[0155] It is understood that the above-described device embodiments correspond to the method embodiments of the present invention, and can implement the memory driving method provided by any of the above-described method embodiments of the present invention.

[0156] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0157] Based on the above-described embodiment of the memory driving method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a memory driving method according to any embodiment of the present invention.

[0158] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.

[0159] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0160] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0161] The modules / units integrated in the device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0162] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for remembering driving directions, characterized in that, include: Obtain the set of lane centerlines, and determine the set of start and end points of the lane centerlines based on the set of lane centerlines; Obtain the set of tangential directions of the start and end points of the lane centerline corresponding to the set of start and end points of the lane centerline, and construct a set of projection break lines based on the set of tangential directions of the start and end points of the lane centerline and the set of start and end points of the lane centerline. The lane centerline set is divided based on any two adjacent projected break lines in the projected break line set to obtain several centerline break line segment subsets, and a centerline break line segment set is obtained based on all the centerline break line segment subsets. Obtain the current vehicle trajectory; Based on the current vehicle trajectory, a subset of target centerline broken segments is selected from the set of centerline broken segments; The main lane is determined based on the distance between the current vehicle trajectory and each centerline break segment in the target centerline break segment subset, so as to perform memory driving based on the main lane.

2. The memory driving method according to claim 1, characterized in that, The step of obtaining the set of lane centerlines and determining the set of start and end points of the lane centerlines based on the set of lane centerlines includes: Get the set of lane lines; For any lane line in the set of lane lines, obtain the set of adjacent lane lines of the lane line, and perform a candidate left and right boundary adjacent lane line filtering action on the set of adjacent lane lines to obtain the candidate left and right boundary adjacent lane line set corresponding to the lane line. Perform a target left and right boundary adjacent lane line filtering action on the candidate left and right boundary adjacent lane line set to obtain the target left and right boundary adjacent lane line set corresponding to the lane line; The lane centerline corresponding to the lane line is determined based on the lane line and the set of adjacent lane lines on the left and right boundaries of the target. Based on all the lane centerlines corresponding to the lane line set, construct a lane centerline set corresponding to the lane line set, and determine the set of the beginning and end points of the lane centerlines based on the lane centerline set.

3. The memory driving method according to claim 2, characterized in that, For any lane line in the lane line set, the adjacent lane line set of that lane line is obtained, and a candidate left and right boundary adjacent lane line filtering action is performed on the adjacent lane line set to obtain the candidate left and right boundary adjacent lane line set corresponding to that lane line. The candidate left and right boundary adjacent lane line filtering action includes: For any lane line in the set of lane lines, obtain the current lane line start point, current lane line end point and current sampling point tangential direction sequence of the lane line; The tangential direction of the current lane line starting point is obtained based on the tangential direction sequence of the current lane line starting point and the current sampling point, and the tangential direction of the current lane line ending point is obtained based on the tangential direction sequence of the current lane line ending point. Obtain the set of adjacent lane lines corresponding to the lane line. For any adjacent lane line in the set of adjacent lane lines, obtain the tangential direction sequence of the adjacent sampling points of the adjacent lane line. Based on all the tangential direction sequences of the adjacent sampling points in the set of adjacent lane lines, obtain the set of tangential direction sequences of the adjacent sampling points of the lane line. Project the starting point of the current lane line corresponding to the lane line onto the set of adjacent lane lines to obtain the set of projected lane line starting points, and project the ending point of the current lane line corresponding to the lane line onto the set of adjacent lane lines to obtain the set of projected lane line ending points; Based on the set of starting points and ending points of the projected lane lines corresponding to the adjacent lane lines, the set of tangential directions of the starting points and the set of tangential directions of the projected lane lines are obtained by filtering from the set of tangential direction sequences of adjacent sampling points. Based on a preset included angle threshold, the tangential direction of the current lane line starting point, the tangential direction of the current lane line ending point, the set of tangential directions of the projected lane line starting point, and the set of tangential directions of the projected lane line ending point, the included angle comparison result set corresponding to the lane line is determined; Based on a preset distance threshold, the current lane line start point, the current lane line end point, the set of projected lane line start points, and the set of projected lane line end points, determine the set of distance comparison results corresponding to the lane line; Based on the set of angle comparison results and the set of distance comparison results, a set of candidate left and right boundary adjacent lane lines corresponding to the lane line is selected from the set of adjacent lane lines.

4. The memory driving method according to claim 3, characterized in that, The method for determining the set of angle comparison results corresponding to the lane line based on a preset included angle threshold, the tangential direction of the current lane line start point, the tangential direction of the current lane line end point, the set of tangential directions of the projected lane line start point, and the set of tangential directions of the projected lane line end point includes: The set of angles between adjacent starting points is obtained based on the set of tangential directions of the current lane line starting point and the set of tangential directions of the projected lane line starting point; and the set of angles between adjacent ending points is obtained based on the set of tangential directions of the current lane line ending point and the set of tangential directions of the projected lane line ending point. Based on the preset included angle threshold, the set of included angles of adjacent starting points and the set of included angles of adjacent ending points, the set of included angle comparison results for the lane line is determined.

5. The memory driving method according to claim 3, characterized in that, The candidate left and right boundary adjacent lane line set includes a candidate left boundary adjacent lane line set and a candidate right boundary adjacent lane line set. The step of performing a target left and right boundary adjacent lane line filtering operation on the candidate left and right boundary adjacent lane line set to obtain the target left and right boundary adjacent lane line set corresponding to that lane line includes: Perform a target left boundary adjacent lane line filtering action on the candidate left boundary adjacent lane line set corresponding to the lane line to obtain the target left boundary adjacent lane line set corresponding to the lane line. Perform a target right boundary adjacent lane line filtering action on the candidate right boundary adjacent lane line set corresponding to the lane line to obtain the target right boundary adjacent lane line set corresponding to the lane line. Based on the set of adjacent lane lines on the left boundary of the target and the set of adjacent lane lines on the right boundary of the target corresponding to the lane line, construct the set of adjacent lane lines on the left and right boundaries of the target corresponding to the lane line.

6. The memory driving method according to claim 5, characterized in that, The step of performing a target left boundary adjacent lane line filtering action on the candidate left boundary adjacent lane line set corresponding to the lane line to obtain the target left boundary adjacent lane line set corresponding to the lane line includes: When there is only one candidate left boundary adjacent lane line in the set of candidate left boundary adjacent lane lines corresponding to the lane line, the set of candidate left boundary adjacent lane lines is taken as the target left boundary adjacent lane line set. When there are multiple candidate left-boundary adjacent lane lines in the set of candidate left-boundary adjacent lane lines corresponding to the lane line, obtain the projection segment interval set of the candidate left-boundary adjacent lane line set in the lane line, and divide the candidate left-boundary adjacent lane line set into an overlapping left-boundary adjacent lane line set and a non-overlapping left-boundary adjacent lane line set based on the projection segment interval set. Based on the lane line and the set of adjacent lane lines of the overlapping left boundary, the set of left boundary lane line widths corresponding to the lane line is obtained, and the set of adjacent lane lines of the overlapping left boundary is filtered based on the preset width and the set of left boundary lane line widths to obtain a subset of adjacent lane lines of the overlapping left boundary. Select the subset of projection line segment intervals corresponding to the subset of adjacent lane lines of the overlapping left boundary from the set of projection line segment intervals; Based on the projection segment interval subset, the overlapping left boundary adjacent lane line subset corresponding to the lane line is updated until the projection segment interval set corresponding to the overlapping left boundary adjacent lane line subset has no interval overlap. Then, the target left boundary adjacent lane line set corresponding to the lane line is obtained based on the overlapping left boundary adjacent lane line subset and the non-overlapping left boundary adjacent lane line set.

7. The memory driving method according to claim 6, characterized in that, The process of updating the overlapping left boundary adjacent lane line subset based on the projection line segment interval subset for the lane line corresponding to the lane line continues until there is no interval overlap in the projection line segment interval set corresponding to the overlapping left boundary adjacent lane line subset. Then, based on the overlapping left boundary adjacent lane line subset and the non-overlapping left boundary adjacent lane line set, the target left boundary adjacent lane line set corresponding to the lane line is obtained. The overlapping left boundary adjacent lane line subset update action includes: From the subset of adjacent lane lines with overlapping left boundaries corresponding to the lane line, select an adjacent lane line with overlapping left boundaries according to a preset selection rule. Based on the projection segment interval of the adjacent lane line with overlapping left boundaries and the projection segment interval of other adjacent lane lines with overlapping left boundaries in the subset of adjacent lane lines with overlapping left boundaries, obtain several adjacent lane lines with overlapping left boundaries to be processed corresponding to the adjacent lane line with overlapping left boundaries from the subset of adjacent lane lines with overlapping left boundaries. Find the starting point of the adjacent lane line of the overlapping left boundary; For any one of the several overlapping left boundary adjacent lane lines to be processed, obtain the starting point of the left boundary adjacent lane line of the overlapping left boundary adjacent lane line to be processed, so as to obtain the set of starting points of the left boundary adjacent lane lines of the lane line based on all the starting points of the left boundary adjacent lane lines of the several overlapping left boundary adjacent lane lines to be processed. Based on the current lane line start point, the start points of adjacent lane lines at the overlapping left boundary, and the set of start points of adjacent lane lines at the left boundary to be processed, the start point comparison result corresponding to the lane line is obtained, and the subset of adjacent lane lines at the overlapping left boundary is updated based on the start point comparison result.

8. The memory driving method according to claim 2, characterized in that, The step of determining the main lane based on the distance between the current vehicle trajectory and each centerline break segment in the target centerline break segment subset, and then performing memory driving based on the main lane, includes: Obtain the distance between the current vehicle trajectory and each target centerline break segment in the target centerline break segment subset to obtain the trajectory centerline distance set; Obtain the target centerline break segment corresponding to the trajectory centerline distance with the smallest value in the trajectory centerline distance set, and use it as the main lane for memory driving based on the main lane.

9. A memory driving device, characterized in that, include: The basic data acquisition module is used to acquire a set of lane centerlines and determine the set of start and end points of the lane centerlines based on the set of lane centerlines. The projection break line set construction module is used to obtain the set of tangential directions of the start and end points of the lane centerline corresponding to the set of start and end points of the lane centerline, and construct the projection break line set based on the set of tangential directions of the start and end points of the lane centerline and the set of start and end points of the lane centerline. The centerline break segment set acquisition module is used to divide the lane centerline set based on any two adjacent projected break lines in the projected break line set to obtain several centerline break segment subsets, so as to obtain the centerline break segment set based on all the centerline break segment subsets. The current vehicle trajectory acquisition module is used to acquire the current vehicle trajectory; The target centerline broken segment subset acquisition module is used to select a target centerline broken segment subset from the centerline broken segment set based on the current vehicle trajectory; The main lane determination module is used to determine the main lane based on the distance between the current vehicle trajectory and each centerline break segment in the target centerline break segment subset, so as to perform memory driving based on the main lane.

10. A terminal device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements a memory-based driving method as described in any one of claims 1-8.

Images