Automatic mapping method and system for autonomous vehicle

Abstract

The invention relates to the technical field of information processing, and particularly discloses an automatic mapping method and system for an autonomous vehicle. An automatic mapping method for an autonomous vehicle comprises the following steps: constructing a generation and preprocessing model of a point cloud map, and fusing image, point cloud and position information to obtain a complete lane line map; single-frame lane information extraction is carried out on the lane line map, lane information around the vehicle at each moment in the vehicle driving process is calculated, and the lane information comprises lane line information and lane center line information; and carrying out multi-frame lane association and post-processing on the lane line information and the lane center line information, and carrying out inter-frame matching association and alignment on the lane information generated by a single frame to obtain a vector map for an automatic driving vehicle. The existing lane line information in the display environment is given, and the concept of lanes and the topological connection relation between the lanes can be represented.

Description

Technical Field

[0001] The present invention relates to the field of information processing technology, and in particular to an automatic mapping method and system for an autonomous driving vehicle. Background Art

[0002] At present, with the development of unmanned driving technology, the application of unmanned driving technology is becoming more and more extensive. High-precision vector maps are defined relative to ordinary maps. They provide map information with higher precision and more dimensions, and have become an indispensable part of unmanned driving technology. CN116452852A proposes a similar high-precision vector map construction method. The grayscale enhancement-based method used converts the intensity information of the point cloud into grayscale information, and better realizes the conversion of three-dimensional point clouds into two-dimensional projection images. Compared with direct processing of three-dimensional point clouds, dimension reduction is achieved, which is conducive to improving processing speed, and the grayscale image lane markings are obvious, which is conducive to the subsequent image processing effect; then the canny edge detection operator and dynamic adaptive threshold segmentation used can identify possible markings as much as possible, and at the same time are robust to noise points, with good processing effects, and low omission and false detection rates. The lane marking classification based on the template knowledge base has good classification effect, and the extracted lane markings can basically be correctly classified. Based on the extracted lane markings calibrated with high-precision GPS data, the vectorization map for the autonomous driving navigation map is drawn, and the effect is good.

[0003] However, this method can only be used to construct a vector map of lane lines, and does not have the ability to describe the connection between lane lines, such as what type the left and right lines of a lane are, or whether a certain dashed line is connected to a non-changeable white line area. In the final analysis, it is only an objective description and fitting of the position of lane lines in the real environment, and it does not convert it into the "lane" information used by autonomous driving. Summary of the invention

[0004] In view of the above problems, the first purpose of the present invention is to provide an automated mapping method for autonomous driving vehicles. In the process of constructing vectorized maps for various types of autonomous driving vehicles, automated map generation greatly improves the speed and quality of mapping, which is conducive to the large-scale promotion of autonomous driving vehicles. This method still has excellent results in complex scenes such as ramp merging and overpasses. It not only provides the existing lane line information in the display environment, but also can well represent the concept of lanes and the topological connection relationship between lanes, which is conducive to the deployment of autonomous driving systems in real environments at a lower cost and on a larger scale.

[0005] A second object of the present invention is to provide an automated mapping system for autonomous vehicles.

[0006] To achieve the first objective, the first technical solution of the present invention is: an automatic mapping method for an autonomous driving vehicle, comprising:

[0007] Step S01: construct a point cloud map generation and preprocessing model, fuse the image, point cloud and location information, and obtain a complete lane line map;

[0008] Step S02: extracting single-frame lane information from the lane line map, and calculating lane information around the vehicle at each moment during the vehicle's driving process, wherein the lane information includes lane line information and lane centerline information;

[0009] Step S03: Perform multi-frame lane association and post-processing on the lane line information and the lane centerline information, perform inter-frame matching association and alignment on the lane information generated by a single frame, and obtain a vector map for the autonomous driving vehicle.

[0010] Preferably, the step S01 includes: obtaining position information and point cloud, obtaining the position of the vehicle in each frame of the image according to the position information, removing dynamic objects from the point cloud corresponding to each frame of the image, and performing point cloud stitching to obtain a global point cloud.

[0011] Preferably, the global point cloud is obtained by stitching point clouds based on position information and point clouds acquired multiple times.

[0012] Preferably, the step S01 also includes extracting lane lines based on the global point cloud, projecting points in the global point cloud onto the image using visual perception information, and if the projection falls within a lane line perception area in the image, the point belongs to the lane line.

[0013] Preferably, step S02 includes obtaining a local map from the lane line map, selecting any frame, clustering all point clouds whose distance from the frame position is less than a predetermined threshold, obtaining candidate point cloud clusters and filtering them, and obtaining candidate lane line information for lane construction.

[0014] Preferably, the step S02 further comprises: calculating the eigenvector and eigenvalue of each point cloud cluster, sorting the eigenvalues ​​and verifying the filtering conditions, and taking the eigenvalues ​​satisfying the filtering conditions as candidate lane line information;

[0015] The filtering condition includes a first filtering condition, wherein the first filtering condition is that the eigenvalues ​​λ0>>λ1 and λ0>>λ2. If the eigenvalues ​​satisfy the first filtering condition, the eigenvector corresponding to the maximum eigenvalue is selected. If the angle between the eigenvector and the driving direction of the vehicle is less than a threshold, it is determined that the filtering condition is satisfied and is used as candidate lane line information.

[0016] Preferably, the filtering condition also includes a second filtering condition, and the second filtering condition is the eigenvalue λ0>>λ2 and λ1>>λ2. If the eigenvalue satisfies the second filtering condition, the point cloud cluster is divided into two parts, and the eigenvector and eigenvalue of the point cloud cluster after the division are calculated. If the eigenvalue satisfies the first filtering condition, the eigenvalue is used as the candidate lane line information.

[0017] Preferably, the step S03 further includes:

[0018] Minimize the longitudinal position difference of all lanes between two frames. If the longitudinal position difference is less than a set threshold, connect the lanes between the two frames.

[0019] Perform cross-frame detection on the unconnected lanes. If the longitudinal position difference between the lane without successor in the Nth frame and the lane without predecessor in the N+ith frame is less than the set threshold, the lanes between the two frames are connected.

[0020] Wherein, N is the ordinal number of the frame and i is a positive integer.

[0021] Preferably, the step S03 further includes:

[0022] The lanes are checked. If the number of lanes or the lane attributes change between the i-th frame and the i+1-th frame, it is determined that the lane information changes in the current frame, and the connected lanes are interrupted at the position of the current frame to form two lanes, and the lane attributes are marked.

[0023] To achieve the second objective, the second technical solution of the present invention is: an automatic mapping system for an autonomous driving vehicle, comprising:

[0024] Point cloud map generation module: used to fuse images, point clouds and location information to obtain a complete lane map;

[0025] Single-frame lane information extraction module: used to extract single-frame lane information from the lane line map, and calculate lane information around the vehicle at each moment during the vehicle's driving process, wherein the lane information includes lane line information and lane centerline information;

[0026] Vector map generation module: used to perform multi-frame lane association and post-processing on the lane line information and lane centerline information, perform inter-frame matching association and alignment on the lane information generated by a single frame, and obtain a vector map for an autonomous driving vehicle.

[0027] Beneficial effects of the above technical solution:

[0028] The automated mapping method and system for autonomous driving vehicles provided by the present invention use laser point cloud, camera image and vehicle GPS as input, and output a high-precision map for autonomous driving with lane centerline and lane line connection relationship. In the process of constructing vectorized maps for various types of autonomous driving vehicles, automated map generation greatly improves the speed and quality of map construction, which is conducive to the large-scale promotion of autonomous driving vehicles. This method still has excellent results in complex scenes such as ramp merging and overpasses. Compared with the existing automatic vector map generation algorithm, this patent can not only give the existing lane line information in the display environment, but also can well represent the concept of lane and the topological connection relationship between lanes, thereby improving the degree of automation in the mapping algorithm process. It greatly reduces the manpower and time consumption required in the process of constructing autonomous driving maps, which is conducive to the deployment of autonomous driving systems in real environments at a lower cost and on a larger scale. It solves the problems of time-consuming and labor-intensive construction of high-precision lane maps in the prior art, low accuracy, and the inability of traditional automated mapping systems to effectively describe the relationship between lanes. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0030] Figure 1 A flow chart of an automated mapping method for an autonomous driving vehicle provided by one embodiment of the present invention;

[0031] Figure 2 A flow chart of point cloud map generation and preprocessing provided by one embodiment of the present invention;

[0032] Figure 3 A schematic diagram of projection of a point cloud to an image provided by an embodiment of the present invention;

[0033] Figure 4 A lane map with attribute information provided by an embodiment of the present invention;

[0034] Figure 5 A schematic diagram of stitching multiple point clouds provided by an embodiment of the present invention;

[0035] Figure 6 A single-frame lane information extraction flow chart provided for one embodiment of the present invention;

[0036] Figure 7 A flowchart of multi-frame lane association and post-processing provided for one embodiment of the present invention.

[0037] Figure 8 A schematic diagram of a smoothing filtering effect provided by an embodiment of the present invention;

[0038] Fig. 9 A schematic diagram of waypoint downsampling provided by an embodiment of the present invention;

[0039] Fig.10 This is a schematic diagram of the application effect of the processing method provided by the present invention. DETAILED DESCRIPTION

[0040] The following is a further detailed description of the implementation methods of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than an exhaustive list of all the embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0041] The terms "first", "second", etc. (if any) in the specification and claims are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way are interchangeable where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0042] It should be understood that the term "and / or" used in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0043] Embodiment 1

[0044] An embodiment of the present invention provides an automated mapping method for an autonomous driving vehicle. The specific process is as follows: Figure 1As shown. The automatic mapping method for autonomous driving vehicles provided by the present invention includes the generation and preprocessing of point cloud maps, single-frame lane information extraction, and multi-frame lane association and post-processing. Among them, the generation and preprocessing of point cloud maps are used to fuse the images, point clouds and position information collected on the vehicle to generate a lane line map with attributes for each point. Single-frame lane information extraction is used to calculate the lane information around the vehicle at each moment during the vehicle's driving process based on the lane line map that has been generated. Multi-frame lane association and post-processing are mainly responsible for associating and aligning the lane information generated by a single frame for post-processing such as smoothing or optimization.

[0045] 1. Generation and preprocessing of point cloud maps

[0046] Figure 2 The generation and preprocessing flow chart of a point cloud map provided by an embodiment of the present invention specifically includes:

[0047] Point cloud stitching module: It is used to record the position of the vehicle in real time during driving and collect the point cloud returned by the laser radar, and then optimize the position of the vehicle to obtain the position P of the vehicle in each frame. n At the same time, the part of the vehicle with high reflectivity after removing dynamic objects in each frame of point cloud is recorded as O n , then the point cloud generated after splicing can be expressed as P = ΣP n ×O n , a global point cloud is generated according to this method.

[0048] Lane line radar point cloud map extraction module: After obtaining the global point cloud, it is necessary to extract the part belonging to the lane line from the point cloud. This method uses visual perception information for screening, obtains visual perception information on the image in each frame, and projects the point cloud on the image. If the point falls in the corresponding lane line perception area on the image, it is considered that the point belongs to the lane line. Figure 3 The figure is a schematic diagram of the projection of point cloud to image, showing the projection of point cloud onto the image. The blue points are radar points with high reflectivity, and the green points are the perception results of lane lines. Figure 4 The lane line map with attribute information shows the lane line map with attributes, where green represents dotted lines, red represents solid lines, and blue represents non-lane lines.

[0049] Multi-map fusion module: used to generate a complete description point cloud after multiple acquisitions and then stitching, to obtain a complete lane line map. Due to the limited perception range of the radar or the occlusion of the lane, it is often impossible to ensure that every lane line falls on it in one driving. Therefore, it may be necessary to collect multiple times and then stitch them together to generate a complete description point cloud. i represents the point cloud collected in the i-th time and aligned to the same coordinate system, then the point cloud synthesized by multiple passes can be expressed as O = ΣOi , Figure 5 This is a schematic diagram of the stitching of multiple point clouds, showing the effect of multi-pass point cloud synthesis. All lane lines are well reflected.

[0050] 2. Single-frame lane information extraction

[0051] After obtaining the complete lane map, this method will extract lane information based on the vehicle's driving trajectory, such as Figure 6 The single-frame lane information extraction flow chart provided by an embodiment of the present invention specifically includes:

[0052] Get the local map from the complete lane map. For each frame, get all the point clouds with a distance less than r from the current position in the complete lane point cloud and transform them to the vehicle coordinate system of the current frame. After obtaining the point cloud near the current frame, cluster the point cloud to obtain N candidate point cloud clusters, and filter them through morphology to obtain lane candidates.

[0053] The specific filtering process includes: Step 1, calculate the eigenvector and eigenvalue of each point cloud cluster, sort the eigenvalues, and then λ0, λ1, and λ2 are adjacent eigenvalues ​​arranged in descending order. If λ0>>λ1 and λ0>>λ2, it proves that the current point cloud cluster is a linear feature and proceeds to the next step. Step 2, considering that the direction vector of the lane line is often parallel to the driving direction of the vehicle, based on the eigenvector calculated in step 1, the eigenvector corresponding to the maximum eigenvalue is taken out. If the angle between the vector and the driving direction of the vehicle is less than the threshold θ, it is considered to have passed the morphological filtering. Step 3, considering some special lane line structures, such as bifurcation and merging, some special cases will be processed additionally. For example, if a point cloud cluster satisfies λ0>>λ2 and λ1>>λ2 when calculating the eigenvalue, its longitudinal centerline is found for binary division. If the point cloud after binary division meets conditions 1) and 2), it is considered to be the entrance or exit of the ramp on the highway, and it is also added to the lane line candidate.

[0054] After obtaining lane line candidates, lane construction is performed based on these candidate lane line information. Lane line information is selected based on a scoring function, and the scoring function is:

[0055] loss=α*(lane width-standard width)+β*lane length+γ*(width-road_width)

[0056] The scoring function consists of three parts: the width difference between the lane and the standard lane, the length of the lane line, and the difference between the sum of the widths of all lanes and the total width. All possible combinations of candidate lane lines are arranged and their losses are calculated respectively, and finally the combination with the minimum loss is selected as the lane line information of the current frame.

[0057] After obtaining the lane lines of each frame, we start to calculate the lane center lines of each frame, sort them according to the coordinates of each lane line in the lateral position of the vehicle, and use the adjacent lane lines to calculate the lane center lines. Then, we remove abnormal lanes based on the angles between the lanes, and finally generate the lane lines and lane information of a single frame.

[0058] 3. Multi-frame lane association and post-processing

[0059] After obtaining the lanes and lane lines of each frame, it is necessary to associate the lanes and lane lines of multiple frames to generate a lane and lane line map with topological relationships. The specific flow chart is as follows: Figure 7 The multi-frame lane association and post-processing flow chart is shown in the figure. It includes:

[0060] First, the lane lines between frames are connected based on the Hungarian algorithm, and the lane matching problem between two frames is converted into minimizing the connection cost of all lanes between two frames. Its physical meaning is that the center point of a lane line in the n-1th frame is projected to the longitudinal position difference between a lane line in the nth frame and the nth frame. It can be expressed as: in,

[0061] Among them, cost is the longitudinal position difference, is the i-th line of the n-th frame, y is, P n-1 is the pose of the n-1th frame, P n is the n-th frame pose, is the j-th line of the n-1-th frame, where n is a positive integer.

[0062] Preferably, for special cases, such as a double lane on the side of a lane, when the cost is less than a certain threshold, the two lanes between the two frames will be directly connected to form a one-to-two structure. At the same time, considering that some frames may lose lanes, cross-frame detection will also be performed on lanes that are not connected in each frame. If the cost of the non-successor lane in the Nth frame and the non-successor lane in the N+ith frame is less than a certain threshold, the two lanes will be directly connected, and reverse interpolation will be performed to obtain the lane center position in each frame from N to N+i. The problem of lane number change is solved, where N is the ordinal number of the frame and i is a positive integer.

[0063] Secondly, after obtaining the relationship between lanes, it is necessary to extract the lane change information and store it in the graph. The lanes are checked starting from the 0th frame. If the number of lanes or the lane attributes change between the i-th frame and the i+1-th frame, it is considered that the lane information changes in the current frame. The connected lanes are interrupted at the position of the current frame to form two lanes, and their subsequent lane attributes are marked.

[0064] Then, each lane line is smoothed and filtered, and the hampel filter is used for the corresponding smoothing filter operation. For a certain lane, a sliding window (iN, i+N) is constructed in the i-th frame, and the distance from the point of each frame in this window to the line connecting the adjacent frames is calculated. If the distance from the i-th frame point to the line constructed by the i-1, i+1 frame point is greater than the average distance in the window plus a certain threshold ε, then this point is removed. The schematic diagram of the smoothing filter effect is shown below: Figure 8 shown.

[0065] Finally, the Douglas thinning method is used to downsample the line, reducing the number of points by deleting unimportant points while trying to maintain the shape characteristics of the curve. Starting from the two endpoints of the curve, calculate the perpendicular distance from each point to the straight line connecting the two endpoints, and find the point farthest from the straight line (i.e., the maximum deviation point). If the deviation of the maximum deviation point is less than the preset tolerance value, it is considered that the curve segment can be represented by a straight line segment and the other points can be deleted; if the deviation is greater than the tolerance, the maximum deviation point is retained and the remaining curve segments are recursively simplified. Effectively simplify the curve, reduce the amount of calculation, and maintain the main features of the curve. The schematic diagram of waypoint downsampling is shown in the figure. Fig. 9 In places with large curvature such as ramps, the road points on the lane lines are denser and the road points on the straight roads are relatively sparse.

[0066] Fig.10 This is a schematic diagram of the application effect of the processing method provided by the present invention. It shows the actual effect of the present method. It can be seen from the figure that the present algorithm still has excellent effect in complex scenes such as ramp merging and overpasses. Compared with the existing automatic vector map generation algorithm, the present invention not only displays the existing lane line information in the environment, but also can well represent the concept of lanes and the topological connection relationship between lanes, thereby improving the degree of automation in the mapping algorithm process. It reduces the manpower and time consumption required in the process of building an autonomous driving map, which is conducive to the deployment of the autonomous driving system in a real environment at a lower cost and on a larger scale.

[0067] The present invention also provides an automated mapping system for autonomous driving vehicles, including a point cloud map generation module, a single-frame lane information extraction module and a vector map generation module. Among them, the point cloud map generation module is used to fuse the image, point cloud and location information to obtain a complete lane line map. The single-frame lane information extraction module is used to extract single-frame lane information from the lane line map, calculate the lane information around the vehicle at each moment during the vehicle's driving process, and the lane information includes lane line information and lane centerline information. The vector map generation module is used to perform multi-frame lane association and post-processing on the lane line information and lane centerline information, and perform inter-frame matching association and alignment on the lane information generated by a single frame to obtain a vector map for autonomous driving vehicles.

[0068] The automated mapping method and system for autonomous driving vehicles provided by the present invention use laser point cloud, camera image and vehicle GPS as input, and output a high-precision map for autonomous driving with lane centerline and lane line connection relationship. In the process of constructing vectorized maps for various types of autonomous driving vehicles, automated map generation greatly improves the speed and quality of map construction, which is conducive to the large-scale promotion of autonomous driving vehicles. This method still has excellent results in complex scenes such as ramp merging and overpasses. Compared with the existing automatic vector map generation algorithm, this patent can not only give the existing lane line information in the display environment, but also can well represent the concept of lane and the topological connection relationship between lanes, thereby improving the degree of automation in the mapping algorithm process. It greatly reduces the manpower and time consumption required in the process of constructing autonomous driving maps, which is conducive to the deployment of autonomous driving systems in real environments at a lower cost and on a larger scale. It solves the problems of time-consuming and labor-intensive construction of high-precision lane maps in the prior art, low accuracy, and the inability of traditional automated mapping systems to effectively describe the relationship between lanes.

[0069] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not limitations on the implementation methods of the present invention. For ordinary technicians in the relevant field, other different forms of changes or modifications can be made on the basis of the above description. It is impossible to list all the implementation methods here. All obvious changes or modifications derived from the technical solution of the present invention are still within the protection scope of the present invention.

Claims

1. An automated mapping method for an autonomous driving vehicle, characterized in that: include: Step S01: construct a point cloud map generation and preprocessing model, fuse the image, point cloud and location information, and obtain a complete lane map; Step S02: extracting single-frame lane information from the lane line map, and calculating lane information around the vehicle at each moment during the vehicle's driving process, wherein the lane information includes lane line information and lane centerline information; Step S03: Perform multi-frame lane association and post-processing on the lane line information and the lane centerline information, perform inter-frame matching association and alignment on the lane information generated by a single frame, and obtain a vector map for the autonomous driving vehicle.

2. The automatic mapping method for an autonomous driving vehicle according to claim 1, characterized in that: The step S01 includes: obtaining position information and point cloud, obtaining the position of the vehicle in each frame of the image according to the position information, removing dynamic objects from the point cloud corresponding to each frame of the image, and stitching the point cloud to obtain a global point cloud.

3. The automatic mapping method for an autonomous driving vehicle according to claim 2, characterized in that: The global point cloud is obtained by stitching point clouds based on position information and point clouds acquired multiple times.

4. The automatic mapping method for an autonomous driving vehicle according to claim 2, characterized in that: The step S01 also includes extracting lane lines based on the global point cloud, projecting points in the global point cloud onto the image using visual perception information, and if the projection falls within a lane line perception area in the image, the point belongs to the lane line.

5. The automatic mapping method for an autonomous driving vehicle according to claim 1, characterized in that: The step S02 includes obtaining a local map from the lane line map, selecting any frame, clustering all point clouds whose distance from the frame position is less than a predetermined threshold, obtaining candidate point cloud clusters and filtering them, and obtaining candidate lane line information for lane construction.

6. The automatic mapping method for an autonomous driving vehicle according to claim 5, characterized in that: The step S02 further includes: calculating the eigenvector and eigenvalue of each point cloud cluster, sorting the eigenvalues ​​and verifying the filtering conditions, and using the eigenvalues ​​that meet the filtering conditions as candidate lane line information; The filtering condition includes a first filtering condition, the first filtering condition is that the eigenvalues ​​λ0>>λ1 and λ0>>λ2, if the eigenvalues ​​satisfy the first filtering condition, then a eigenvector corresponding to the maximum eigenvalue is selected, and if the angle between the eigenvector and the driving direction of the vehicle is less than a threshold, then it is determined that the filtering condition is satisfied and the lane is used as candidate lane line information; Among them, λ0, λ1, and λ2 are adjacent eigenvalues ​​arranged in descending order.

7. The automatic mapping method for an autonomous driving vehicle according to claim 6, characterized in that: The filtering condition also includes a second filtering condition, and the second filtering condition is that the eigenvalues ​​λ0>>λ2 and λ1>>λ2. If the eigenvalues ​​satisfy the second filtering condition, the point cloud cluster is divided into two parts, and the eigenvectors and eigenvalues ​​of the point cloud cluster after the division are calculated. If the eigenvalues ​​satisfy the first filtering condition, the eigenvalues ​​are used as candidate lane line information.

8. The automatic mapping method for an autonomous driving vehicle according to claim 1, characterized in that: The step S03 further includes: Minimize the longitudinal position difference of all lanes between two frames. If the longitudinal position difference is less than a set threshold, connect the lanes between the two frames. Perform cross-frame detection on the unconnected lanes. If the longitudinal position difference between the lane without successor in the Nth frame and the lane without predecessor in the N+ith frame is less than the set threshold, the lanes between the two frames are connected. Wherein, N is the ordinal number of the frame and i is a positive integer.

9. The automatic mapping method for an autonomous driving vehicle according to claim 8, characterized in that: The step S03 further includes: The lanes are checked. If the number of lanes or the lane attributes change between the i-th frame and the i+1-th frame, it is determined that the lane information changes in the current frame, and the connected lanes are interrupted at the position of the current frame to form two lanes, and the lane attributes are marked.

10. An automated mapping system for an autonomous driving vehicle, characterized in that: include: Point cloud map generation module: used to fuse images, point clouds and location information to obtain a complete lane map; Single-frame lane information extraction module: used to extract single-frame lane information from the lane line map, and calculate lane information around the vehicle at each moment during the vehicle's driving process, wherein the lane information includes lane line information and lane centerline information; Vector map generation module: used to perform multi-frame lane association and post-processing on the lane line information and lane centerline information, perform inter-frame matching association and alignment on the lane information generated by a single frame, and obtain a vector map for an autonomous driving vehicle.

Images