Method for crowd-sourced lane line fusion map updating, vehicle-mounted device and storage medium

Abstract

The application discloses a crowd-sourced lane line fusion map updating method, a vehicle-mounted device and a storage medium. The crowd-sourced lane line fusion map updating method comprises the following steps: acquiring any original lane line polyline point cloud in an original lane line map and at least one candidate lane line polyline point cloud matched with the any original lane line polyline point cloud in a crowd-sourced lane line map; fusing the any original lane line polyline point cloud and the at least one candidate lane line polyline point cloud to obtain a fused lane line polyline point cloud; performing a nonlinear optimization processing operation on the fused lane line polyline point cloud to obtain a target optimized lane line polyline point cloud, and updating the corresponding original lane line polyline point cloud by using the target optimized lane line polyline point cloud, so as to realize updating of the original lane line map. The scheme can successfully update the original lane line map and improve the updating efficiency and the map precision.

Description

Technical Field

[0001] This application relates to the field of high-precision map technology, and in particular to a method, vehicle-mounted device and storage medium for map updating by crowdsourced lane line fusion. Background Technology

[0002] In fields such as autonomous driving and high-precision mapping, crowdsourced data collection and mapping, which features high update frequency, low computational load, and fast transmission, is becoming increasingly popular.

[0003] For the same area, different crowdsourced data collection vehicles passing through the area can generate multiple local maps of that area, or the same crowdsourced data collection vehicle passing through the area multiple times can also generate multiple local maps of that area. However, the start and end points of the lane lines in the local map may differ from the corresponding start and end points of the lane lines in the original map, making it impossible to directly update the lane lines in the original map. Summary of the Invention

[0004] This application provides at least one method, in-vehicle device, and storage medium for crowdsourced lane line fusion map updates to address the aforementioned problems.

[0005] The first aspect of this application provides a map update method for crowdsourced lane line fusion, the method comprising: obtaining a polyline point cloud of any original lane line in the original lane line map and at least one candidate lane line polyline point cloud in the crowdsourced lane line map that matches the polyline point cloud of the original lane line.

[0006] The point cloud of any one original lane line and the point cloud of at least one candidate lane line are fused to obtain a fused lane line point cloud.

[0007] The fused lane line polyline point cloud is subjected to nonlinear optimization processing to obtain a target optimized lane line polyline point cloud. The target optimized lane line polyline point cloud is then used to update the corresponding original lane line polyline point cloud, thereby updating the original lane line map.

[0008] The step of obtaining at least one candidate lane line polyline point cloud that matches any one of the original lane line polyline point clouds includes: obtaining the point-to-line distance value between any one of the original lane line polyline point clouds and the candidate lane line polyline point clouds in the crowdsourced lane line map, so as to obtain at least one point-to-line distance value corresponding to each original lane line polyline point cloud.

[0009] Obtain the directional distance value between any original lane line polyline point cloud and the candidate lane line polyline point cloud in the crowdsourced lane line map, so as to obtain at least one directional distance value corresponding to each original lane line polyline point cloud.

[0010] In response to the point-to-line distance value being less than a preset point-to-line distance threshold and the direction distance value being less than a preset direction distance threshold, the corresponding candidate lane line polyline point cloud is obtained, so that each original lane line polyline point cloud is matched to obtain at least one candidate lane line polyline point cloud.

[0011] The step of obtaining the point-to-line distance value between any original lane line polyline point cloud and the candidate lane line polyline point cloud in the crowdsourced lane line map includes: obtaining any original point in the original lane line polyline point cloud and two candidate points in any candidate lane line polyline point cloud that are close to the original point.

[0012] The point-to-line distance value of any original point is characterized by the vertical distance between the original point and the line connecting the two corresponding candidate points, thereby obtaining the point-to-line distance value corresponding to each original point in the original lane line polyline point cloud.

[0013] The point-to-line distance between the original lane line polyline point cloud and the candidate lane line polyline point cloud is the average of the point-to-line distances corresponding to all original points in the original lane line polyline point cloud, thereby obtaining the point-to-line distance between the original lane line polyline point cloud and at least one candidate lane line polyline point cloud.

[0014] The step of obtaining the directional distance value between any original lane line polyline point cloud and the candidate lane line polyline point cloud in the crowdsourced lane line map includes: obtaining a first vector formed by the original points at both ends of the original lane line polyline point cloud, and a second vector formed by the candidate points at both ends of any candidate lane line polyline point cloud.

[0015] The directional distance between the original lane line polyline point cloud and the candidate lane line polyline point cloud is characterized by the angle between the first vector and the second vector, thereby obtaining the directional distance between the original lane line polyline point cloud and at least one of the candidate lane line polyline point clouds.

[0016] The fusion of any original lane line polyline point cloud and at least one candidate lane line polyline point cloud includes: obtaining candidate points in at least one candidate lane line polyline point cloud that overlap with the original lane line polyline point cloud, and all original points in the original lane line polyline point cloud, to form a point set;

[0017] Perform semantic averaging and curve fitting operations on the point set to obtain a fitted curve;

[0018] The fitted curve is resampled to obtain the fused lane line polyline point cloud.

[0019] The fused lane line polyline point cloud includes a front lane line polyline point cloud and a rear lane line polyline point cloud; the nonlinear optimization processing operation on the fused lane line polyline point cloud includes: obtaining three consecutive fusion points of the front lane line polyline point cloud that are close to the rear lane line polyline point cloud, and three consecutive fusion points of the rear lane line polyline point cloud that are close to the front lane line polyline point cloud.

[0020] The first nonlinear optimization is performed on the six fusion points to obtain the optimized polyline point cloud of the front lane line and the optimized polyline point cloud of the rear lane line;

[0021] A second nonlinear optimization is performed on the optimized polygonal point cloud of the front lane line and the optimized polygonal point cloud of the rear lane line respectively to obtain the target optimized lane line polygonal point cloud.

[0022] Specifically, in response to the fact that the endpoint of the optimized front lane line point cloud near the optimized rear lane line point cloud does not coincide with the endpoint of the optimized rear lane line point cloud near the optimized front lane line point cloud, an overlap operation is performed on the two non-overlapping endpoints to connect the optimized front lane line point cloud and the optimized rear lane line point cloud, thereby obtaining the target optimized lane line point cloud.

[0023] The first nonlinear optimization of the six fusion points includes: obtaining the first error loss of the six fusion points based on the distance condition, smoothing condition, consistency condition and same point condition between the six fusion points and the first reference sampling point;

[0024] In response to the first error loss satisfying a preset error condition, a front lane line optimized polyline point cloud and a rear lane line optimized polyline point cloud are obtained; and / or, a second nonlinear optimization is performed on the front lane line optimized polyline point cloud and the rear lane line optimized polyline point cloud, respectively, including: obtaining a second error loss of the optimized points in the front lane line optimized polyline point cloud based on the distance condition, smoothing condition and consistency condition between the optimized points in the front lane line optimized polyline point cloud and the second reference sampling point;

[0025] Based on the distance conditions, smoothing conditions, and consistency conditions between the optimized points and the third reference sampling points in the optimized polyline point cloud of the rear lane line, the third error loss of the optimized points in the optimized polyline point cloud of the rear lane line is obtained.

[0026] In response to the second error loss and the third error loss satisfying the preset error condition, the target optimized lane line polyline point cloud is obtained.

[0027] A second aspect of this application provides an autonomous driving method, comprising:

[0028] Obtain a lane line map; wherein the lane line map is obtained using the crowdsourced lane line fusion map update method described in the first aspect above;

[0029] Based on the lane map, driving path planning is performed to achieve autonomous driving.

[0030] A third aspect of this application provides an in-vehicle device including a memory and a processor coupled to each other, the processor being configured to execute program instructions stored in the memory to implement the crowdsourced lane line fusion map update method of the first aspect and the autonomous driving method of the second aspect.

[0031] The fourth aspect of this application provides a non-volatile computer-readable storage medium for storing program instructions, which, when executed by a processor, are used to implement the crowdsourced lane line fusion map update method in the first aspect and the autonomous driving method in the second aspect.

[0032] The above-mentioned scheme obtains the point cloud of any original lane line and at least one candidate lane line point cloud that matches the original lane line point cloud. It then fuses the original lane line point cloud and the candidate lane line point cloud to obtain a fused lane line point cloud. Furthermore, it performs nonlinear optimization on the fused lane line point cloud to obtain a target optimized lane line point cloud. This target optimized lane line point cloud is then used to update the corresponding original lane line point cloud, thereby updating the original lane line map. The scheme of this application can successfully update the original lane line map and improve update efficiency and map accuracy.

[0033] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this application. Attached Figure Description

[0034] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the technical solutions of this application.

[0035] Figure 1 This is a flowchart illustrating the crowdsourced lane line fusion map update method in this application embodiment;

[0036] Figure 2 This is a schematic diagram of the original lane line polyline point cloud matching in the embodiments of this application;

[0037] Figure 3 This is a schematic diagram of the distance values ​​between points and lines in an embodiment of this application;

[0038] Figure 4(a) is a schematic diagram of the first distribution of fusion points in an embodiment of this application;

[0039] Figure 4(b) is a schematic diagram of the second distribution of fusion points in an embodiment of this application;

[0040] Figure 5 This is a diagram showing the effect of the first nonlinear optimization result in the embodiments of this application;

[0041] Figure 6(a) is a first effect diagram of the map update of crowdsourced lane line fusion in an embodiment of this application;

[0042] Figure 6(b) is a second effect diagram of the map update of crowdsourced lane line fusion in an embodiment of this application;

[0043] Figure 7 This is a flowchart illustrating the autonomous driving method in an embodiment of this application;

[0044] Figure 8 This is a schematic diagram of the structure of the vehicle-mounted device in the embodiments of this application;

[0045] Figure 9 This is a schematic diagram of the structure of a non-volatile computer-readable storage medium in an embodiment of this application. Detailed Implementation

[0046] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be particularly noted that the following embodiments are for illustrative purposes only and do not limit the scope of the application. Similarly, the following embodiments are only some, not all, embodiments of the present application, and all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present application.

[0047] In this application, the reference to "embodiment" means that a specific 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.

[0048] In this document, 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, or B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects are in an "or" relationship. Furthermore, "many" in this document means two or more. Additionally, the term "at least one" in this document means any combination of at least two of any one or more of a plurality of objects. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C. Furthermore, the terms "first," "second," and "third" in this application are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features.

[0049] As mentioned above, the start and end points of lane lines in the local map may differ from the corresponding start and end points of lane lines in the original map, making it impossible to directly update the lane lines in the original map.

[0050] Therefore, this application provides a method, vehicle-mounted device, and storage medium for crowdsourced lane line fusion map updates.

[0051] Please see Figure 1 , Figure 1 This is a flowchart illustrating the crowdsourced lane line fusion map update method in this application's embodiments. It should be noted that if substantially the same result is achieved, the method of this application does not necessarily require further elaboration. Figure 1 The illustrated process sequence is limited. This method can be applied to in-vehicle devices with computing and other functions. These devices can execute this method by receiving information collected by sensor devices, such as millimeter-wave radar, lidar, or cameras equipped in autonomous vehicles. During the autonomous vehicle's operation, the sensor devices perceive the dynamic real-world scene surrounding the vehicle, including lane markings on the road, and then annotate these lane markings. Figure 1 As shown, the map update method for crowdsourced lane line fusion includes the following steps:

[0052] S11. Obtain the point cloud of any original lane line in the original lane line map and at least one candidate lane line point cloud that matches any original lane line point cloud in the crowdsourced lane line map.

[0053] Point cloud data can be collected using radar sensors, for example, by mounting the radar sensor on a mobile device. This mobile device can be an automated mobile device, such as a robot or an autonomous vehicle.

[0054] In some embodiments, the radar sensor may be a lidar sensor, such as a mechanical lidar, semi-solid-state lidar, or solid-state lidar. In one embodiment, the radar sensor may be any radar device that provides point cloud data and is used in autonomous driving to meet perception accuracy requirements.

[0055] Understandably, the original lane line map is a lane line map representing a certain area that needs to be updated, and it can be the lane line map obtained from the last update; crowdsourced lane line maps can be generated by collecting data from autonomous vehicles driving on the road.

[0056] In one application scenario, an autonomous vehicle travels on a road. When it reaches the same area as the original lane line map, it acquires point cloud data of the road using radar sensors mounted on the vehicle. This generates a crowdsourced lane line map for the same area as the original lane line map. The crowdsourced lane line map includes multiple candidate lane line polyline point clouds, which are used to update the original lane line polyline point clouds in the original lane line map. In some embodiments, the autonomous vehicles passing through the same area may be the same vehicle or different vehicles. Each time the vehicle passes through the area, regardless of whether it is the same vehicle or not, a crowdsourced lane line map for that area can be generated based on the perceived point cloud data to update the original lane line map.

[0057] During the operation of autonomous vehicles, the point cloud data of the road sensed by radar sensors typically includes three-dimensional lane lines (polylines), which can be used to obtain candidate lane line polyline point clouds for updating the original lane line polyline point cloud. The polyline point cloud is the same as the polyline point cloud.

[0058] In the crowdsourced lane map, the candidate lane line polyline point clouds representing the same lane line have at least one identical ID. Understandably, during autonomous driving, radar sensors perceive lane lines and obtain corresponding point cloud data. Since a lane line is relatively long, the radar sensor cannot represent a complete lane line in a single frame of point cloud data; it needs to perceive multiple frames of point cloud data to fully represent a lane line. Therefore, lane lines need to be ID-labeled, and the point cloud data representing the same lane line have the same ID; that is, at least one candidate lane line polyline point cloud representing the same lane line has the same ID.

[0059] The candidate lane line polyline point clouds in the crowdsourced lane line map are matched with the original lane line polyline map in the original lane line map. This ensures that any original lane line polyline point cloud in the original lane line map can be matched with at least one candidate lane line polyline point cloud, which is then used to update the corresponding original lane line polyline point cloud. The specific matching method can be selected according to actual usage requirements and is not specifically limited.

[0060] S12. Fuse any original lane line polyline point cloud with at least one candidate lane line polyline point cloud to obtain a fused lane line polyline point cloud.

[0061] The original lane line polyline point cloud and at least one candidate lane line polyline point cloud are fused to obtain a fused lane line polyline point cloud. In essence, each original lane line polyline point cloud corresponds to at least one candidate lane line polyline point cloud. The original lane line polyline point cloud is then fused with the corresponding candidate lane line polyline point clouds to obtain the fused lane line polyline point cloud. The specific fusion method can be chosen based on actual usage requirements and is not specifically limited.

[0062] S13. Perform nonlinear optimization processing on the fused lane line polyline point cloud to obtain the target optimized lane line polyline point cloud, and use the target optimized lane line polyline point cloud to update the corresponding original lane line polyline point cloud, thereby updating the original lane line map.

[0063] A nonlinear optimization operation is performed on the fused lane line polyline point cloud to obtain a target optimized lane line polyline point cloud. This target optimized lane line polyline point cloud is then used to update any corresponding original lane line polyline point cloud, thereby updating the original lane line map. In essence, performing nonlinear optimization on the obtained fused lane line polyline point cloud yields a smooth target optimized lane line polyline point cloud, which is then used to update the corresponding original lane line polyline point cloud, thus updating the original lane line map.

[0064] The above-mentioned scheme obtains the point cloud of any original lane line and at least one candidate lane line point cloud that matches the original lane line point cloud. It then fuses the original lane line point cloud and the candidate lane line point cloud to obtain a fused lane line point cloud. Furthermore, it performs nonlinear optimization on the fused lane line point cloud to obtain a target optimized lane line point cloud. This target optimized lane line point cloud is then used to update the corresponding original lane line point cloud, thereby updating the original lane line map. The scheme of this application can successfully update the original lane line map and improve update efficiency and map accuracy.

[0065] As described above, at least one candidate lane line polyline point cloud matching any original lane line polyline point cloud is obtained from the crowdsourced lane line map. In one embodiment of this application, obtaining at least one candidate lane line polyline point cloud matching any original lane line polyline point cloud includes: obtaining the point-to-line distance value between any original lane line polyline point cloud and the candidate lane line polyline point cloud in the crowdsourced lane line map, to obtain at least one point-to-line distance value corresponding to each original lane line polyline point cloud; obtaining the directional distance value between any original lane line polyline point cloud and the candidate lane line polyline point cloud in the crowdsourced lane line map, to obtain at least one directional distance value corresponding to each original lane line polyline point cloud; in response to a point-to-line distance value being less than a preset point-to-line distance threshold and a directional distance value being less than a preset directional distance threshold, obtaining the corresponding candidate lane line polyline point cloud, thereby ensuring that each original lane line polyline point cloud is matched with at least one candidate lane line polyline point cloud.

[0066] Obtain the point-to-point distance between any original lane line polyline point cloud and each candidate lane line polyline point cloud in the crowdsourced lane line map, thus obtaining at least one point-to-point distance value corresponding to each original lane line polyline point cloud. In essence, obtaining the point-to-point distance between any original lane line polyline point cloud and each candidate lane line polyline point cloud in the crowdsourced lane line map yields at least one point-to-point distance value corresponding to that original lane line polyline point cloud. This process is repeated to obtain the point-to-point distance between each original lane line polyline point cloud in the original lane line map and each candidate lane line polyline point cloud in the crowdsourced lane line map, thereby obtaining at least one point-to-point distance value corresponding to each original lane line polyline point cloud in the original lane line map.

[0067] Obtain the directional distance values ​​between any original lane line polyline point cloud and the candidate lane line polyline point clouds in the crowdsourced lane line map, thus obtaining at least one directional distance value corresponding to each original lane line polyline point cloud. In essence, obtaining the directional distance value between any original lane line polyline point cloud and each candidate lane line polyline point cloud in the crowdsourced lane line map yields at least one directional distance value corresponding to that original lane line polyline point cloud. This process is repeated to obtain the directional distance values ​​between each original lane line polyline point cloud in the original lane line map and each candidate lane line polyline point cloud in the crowdsourced lane line map, thereby obtaining at least one directional distance value corresponding to each original lane line polyline point cloud in the original lane line map.

[0068] In response to a point-to-line distance value being less than a preset point-to-line distance threshold and a directional distance value being less than a preset directional distance threshold, corresponding candidate lane line polyline point clouds are obtained, ensuring that each original lane line polyline point cloud is matched with at least one candidate lane line polyline point cloud. This can be understood as obtaining the point-to-line distance value between any original lane line polyline point cloud and a candidate lane line polyline point cloud, and also obtaining the directional distance value between them. That is, a point-to-line distance value and a directional distance value are calculated for each original lane line polyline point cloud and a candidate lane line polyline point cloud, respectively. If the point-to-line distance value between the candidate lane line polyline point cloud and the original lane line polyline point cloud is less than a preset point-to-line distance threshold and the directional distance value is less than a preset directional distance threshold, then the candidate lane line polyline point cloud and the original lane line polyline point cloud are successfully matched, ensuring that each original lane line polyline point cloud can be matched with at least one candidate lane line polyline point cloud. The preset point-to-line distance threshold can be set according to actual usage needs without specific limitations; the preset directional distance threshold can also be set according to actual usage needs without specific limitations.

[0069] For example, Figure 2 This is a schematic diagram of the scene matching of the original lane line polyline point cloud in the embodiments of this application, such as... Figure 2 As shown, the original lane line point cloud map includes one original lane line polyline point cloud (m1), and the crowdsourced lane line map includes four candidate lane line polyline point clouds (n1, n2, n3, and n4). Obtain the point-to-line distance value x11 between m1 and n1 in the original lane line polyline point cloud, obtain the point-to-line distance value x12 between m1 and n2, obtain the point-to-line distance value x13 between m1 and n3, and obtain the point-to-line distance value x14 between m1 and n4 in the original lane line polyline point cloud, thus obtaining the four point-to-line distance values ​​x11, x12, x13, and x14 corresponding to m1 in the original lane line polyline point cloud; obtain the directional distance value y11 between m1 and n1, obtain the directional distance value y12 between m1 and n2, obtain the directional distance value y13 between m1 and n3, and obtain the directional distance value y14 between m1 and n4 in the original lane line polyline point cloud, thus obtaining the four directional distance values ​​y11, y12, y13, and y14 corresponding to the original lane line polyline point cloud.

[0070] If the point-to-line distance value x11 is less than the preset point-to-line distance threshold and the direction distance value y11 is less than the preset direction distance threshold, then the corresponding candidate lane line polyline point cloud n1 is successfully matched with the original lane line polyline point cloud m1; if the point-to-line distance value x12 is less than the preset point-to-line distance threshold and the direction distance value y12 is less than the preset direction distance threshold, then the corresponding candidate lane line polyline point cloud n2 is successfully matched with the original lane line polyline point cloud m1; if the point-to-line distance value x13 is less than the preset point-to-line distance threshold and the direction distance value y13 is less than the preset direction distance threshold, then the corresponding candidate lane line polyline point cloud n3 is successfully matched with the original lane line polyline point cloud m1; if the point-to-line distance value x14 is greater than the preset point-to-line distance threshold but the direction distance value y14 is less than the preset direction distance threshold, then the corresponding candidate lane line polyline point cloud n4 is not matched with the original lane line polyline point cloud m1. Thus, the original lane line polyline point cloud m1 is matched to obtain three candidate lane line polyline point clouds, namely candidate lane line polyline point clouds n1, n2 and n3.

[0071] In one embodiment of this application, obtaining the point-to-line distance value between any original lane line polyline point cloud and the candidate lane line polyline point cloud in a crowdsourced lane line map includes: obtaining any original point in the original lane line polyline point cloud and two candidate points in any candidate lane line polyline point cloud that are close to the original point; the point-to-line distance value of any original point is characterized by the vertical distance between the original point and the corresponding two candidate points, thereby obtaining the point-to-line distance value corresponding to each original point in the original lane line polyline point cloud; the point-to-line distance value between the original lane line polyline point cloud and the candidate lane line polyline point cloud is the average of the point-to-line distance values ​​corresponding to all original points in the original lane line polyline point cloud, thereby obtaining the point-to-line distance value between the original lane line polyline point cloud and at least one candidate lane line polyline point cloud.

[0072] Obtain any original point in the original lane line polyline point cloud and two candidate points close to that original point in any candidate lane line polyline point cloud. It can be understood that each original lane line polyline point cloud includes multiple original points, and each candidate lane line polyline point cloud includes multiple candidate points. For each original point, two candidate points close to that original point can be found in the candidate lane line polyline point cloud.

[0073] Figure 3 This is a schematic diagram of the distance values ​​between points and lines in an embodiment of this application, such as... Figure 3 As shown, the point-to-line distance value of any original point is represented by the perpendicular distance between the original point and the line connecting its two corresponding candidate points, thus obtaining the point-to-line distance value corresponding to each original point in the original lane line polyline point cloud. It can be understood that, for any original point, the perpendicular distance d from the original point to the line connecting its two corresponding candidate points is... iThis is the point-to-line distance value of the original point, thus obtaining the point-to-line distance value corresponding to each original point in the original lane line polyline point cloud.

[0074] The point-to-line distance between the original lane line polyline point cloud and the candidate lane line polyline point cloud is the average of the point-to-line distances corresponding to all original points in the original lane line polyline point cloud. This average value is then used to obtain the point-to-line distance between the original lane line polyline point cloud and at least one candidate lane line polyline point cloud. In essence, after obtaining the point-to-line distance for each original point in the original lane line polyline point cloud, the average of the point-to-line distances for all original points in that original lane line polyline point cloud is calculated. The resulting average value is the point-to-line distance between the original lane line polyline point cloud and the corresponding candidate lane line polyline point cloud, thus obtaining the point-to-line distance between the original lane line polyline point cloud and at least one candidate lane line polyline point cloud.

[0075] For example, the original lane line polyline point cloud includes five original points: m11, m12, m13, m14, and m15. We obtain the original point m11 and two candidate points n11 and n12 that are close to it from the candidate lane line polyline point cloud. Connecting the two candidate points n11 and n12, the point-to-line distance value d1 of the original point m11 is the perpendicular distance between the original point m11 and the line connecting the candidate points n11 and n12. Similarly, we obtain the point-to-line distance values ​​d2, d3, d4, and d5 for the original point m12, m13, m14, and m15, respectively. Therefore, the point-to-line distance values ​​of the original lane line polyline point cloud are the average of d1, d2, d3, d4, and d5. The average point-to-line distance value can be calculated using the following formula:

[0076]

[0077] Where, d p2l This represents the point-to-line distance value of the original lane line polyline point cloud, where i represents the original point in the original lane line polyline point cloud, and d represents the distance between the points. i represents the point-to-line distance value of the original point i, and N represents the total number of original points in the original lane line polyline point cloud.

[0078] In one embodiment of this application, obtaining the directional distance value between any original lane line polyline point cloud and the candidate lane line polyline point cloud in the crowdsourced lane line map includes: obtaining a first vector formed by the original points at both ends of the original lane line polyline point cloud, and a second vector formed by the candidate points at both ends of any candidate lane line polyline point cloud; the directional distance value between the original lane line polyline point cloud and the candidate lane line polyline point cloud is characterized by the angle between the first vector and the second vector, thereby obtaining the directional distance value between the original lane line polyline point cloud and at least one candidate lane line polyline point cloud.

[0079] Obtain a first vector composed of the original points at both ends of the original lane line polyline point cloud, and a second vector composed of the candidate points at both ends of any candidate lane line polyline point cloud. It is understood that each original lane line polyline point cloud includes multiple original points, with one original point at each end; the two original points at each end constitute the first vector. Similarly, each candidate lane line polyline point cloud includes multiple candidate points, with one candidate point at each end; the two candidate points at each end constitute the second vector.

[0080] The directional distance between the original lane line polyline point cloud and the candidate lane line polyline point cloud is represented by the angle between the first vector and the second vector, thus obtaining the directional distance between the original lane line polyline point cloud and at least one candidate lane line polyline point cloud. Understandably, after obtaining the first and second vectors, the angle between them is calculated, and the result represents the directional distance between the original lane line polyline point cloud and the candidate lane line polyline point cloud, thus obtaining the directional distance between the original lane line polyline point cloud and at least one candidate lane line polyline point cloud. The angle between the first and second vectors can be calculated using the following formula:

[0081]

[0082] Where, d dir The vector represents the directional distance value of the original lane line polyline point cloud, (x1, y1) represents the first vector, and (x2, y2) represents the second vector.

[0083] In one embodiment of this application, fusing any original lane line polyline point cloud and at least one candidate lane line polyline point cloud includes: obtaining candidate points in at least one candidate lane line polyline point cloud that overlap with the original lane line polyline point cloud, and all original points in the original lane line polyline point cloud, to form a point set; performing semantic averaging and curve fitting operations on the point set to obtain a fitted curve; and performing a resampling operation on the fitted curve to obtain the fused lane line polyline point cloud.

[0084] Obtain candidate points from at least one candidate lane line polyline point cloud that overlap with the original lane line polyline point cloud, along with all original points in the original lane line polyline point cloud, to form a point set. It is understood that an original lane line polyline point cloud corresponds to at least one candidate lane line polyline point cloud, but each candidate lane line polyline point cloud does not necessarily overlap entirely with the original lane line polyline point cloud. That is, when projecting each candidate lane line polyline point cloud onto the original lane line polyline point cloud, only a portion of the candidate lane line polyline point cloud may overlap with the original lane line polyline point cloud. Obtain the candidate points in the overlapping portion and all original points in the original lane line polyline point cloud to form a point set.

[0085] The point set is subjected to semantic averaging and curve fitting operations to obtain a fitted curve. In essence, performing semantic averaging and curve fitting on the obtained point set involves first performing semantic averaging on all points within the set, and then performing curve fitting, which involves plotting a curve based on all points to obtain the fitted curve. The curves used in the curve fitting operation include, but are not limited to, exponential function fitting curves, power function fitting curves, and hyperbolic fitting curves; the choice is based on the actual application requirements and is not specifically limited.

[0086] The fitted curve is resampled to obtain a fused lane polyline point cloud. This resampling operation involves equidistant resampling of the fitted curve to obtain multiple resampled points. Extrapolation and interpolation operations are then performed on these resampled points. For example, a predetermined number of extrapolation points are inserted outside the resampled points, and a predetermined number of interpolation points are inserted between adjacent resampled points. This results in multiple resampled points, multiple extrapolation points, and multiple interpolation points, ultimately generating a new lane polyline, i.e., a fused lane polyline point cloud.

[0087] In one embodiment of this application, the fused lane line polyline point cloud includes a front lane line polyline point cloud and a rear lane line polyline point cloud; the nonlinear optimization processing operation on the fused lane line polyline point cloud includes: obtaining three consecutive fusion points of the front lane line polyline point cloud close to the rear lane line polyline point cloud, and three consecutive fusion points of the rear lane line polyline point cloud close to the front lane line polyline point cloud; performing a first nonlinear optimization on the six fusion points to obtain an optimized front lane line polyline point cloud and an optimized rear lane line polyline point cloud; and performing a second nonlinear optimization on the optimized front lane line polyline point cloud and the optimized rear lane line polyline point cloud respectively to obtain a target optimized lane line polyline point cloud.

[0088] The fused lane line polyline point cloud includes the front lane line polyline point cloud and the rear lane line polyline point cloud. It can be understood that the fused lane line polyline point cloud includes multiple fused lane line polyline point clouds arranged in sequence, where the front lane line polyline point cloud and the rear lane line polyline point cloud represent two adjacent lane line polyline point clouds.

[0089] Obtain three consecutive fusion points from the front lane line polyline point cloud that are close to the rear lane line polyline point cloud, and three consecutive fusion points from the rear lane line polyline point cloud that are close to the front lane line polyline point cloud. It can be understood that the front and rear lane line polyline point clouds are two adjacent lane line polyline point clouds, where the front lane line polyline point cloud contains multiple fusion points, and the rear lane line polyline point cloud contains multiple fusion points. Select three consecutive fusion points from the front lane line polyline point cloud that are close to the rear lane line polyline point cloud, and select three consecutive fusion points from the rear lane line polyline point cloud that are close to the front lane line polyline point cloud, to obtain six consecutive fusion points.

[0090] Figure 4(a) is a schematic diagram of the first distribution of fusion points in an embodiment of this application, and Figure 4(b) is a schematic diagram of the second distribution of fusion points in an embodiment of this application. As shown in Figures 4(a) and 4(b), the distribution of the three fusion points (r11, r12, r13) of the front lane line broken line point cloud r1 and the three fusion points (r21, r22, r23) of the rear lane line broken line point cloud r2 may have two situations.

[0091] Figure 5 This is a diagram showing the effect of the first nonlinear optimization result in the embodiments of this application, such as... Figure 5 As shown, the first nonlinear optimization is performed on the six fusion points to obtain the optimized polygonal point cloud r1' for the front lane line and r2' for the rear lane line. It can be understood that the six fusion points, such as r11, r12, r13, r21, r22, and r23, are treated as a whole, and the first nonlinear optimization operation is performed on these six fusion points to smooth them, making them closer together to obtain multiple optimized points. These optimized points are then used to update the six fusion points, resulting in the updated polygonal point cloud for the front lane line (r1') and the updated polygonal point cloud for the rear lane line (r2').

[0092] A second nonlinear optimization is performed on both the front and rear lane line optimized polyline point clouds to obtain the target optimized lane line polyline point cloud. Understandably, after updating the fusion points using the optimized points to obtain the front lane line optimized polyline point cloud, a second nonlinear optimization operation is performed on it to smooth all points in the front lane line optimized polyline point cloud. Similarly, after updating the fusion points using the optimized points to obtain the rear lane line optimized polyline point cloud, a second nonlinear optimization operation is performed on it to smooth all points in the rear lane line optimized polyline point cloud. The front and rear lane line optimized polyline point clouds, after completing the second nonlinear optimization, can be smoothly connected to each other, thus obtaining the target optimized lane line polyline point cloud.

[0093] Figure 6(a) is a first effect diagram of the map update of the crowdsourced lane line fusion in the embodiment of this application, and Figure 6(b) is a second effect diagram of the map update of the crowdsourced lane line fusion in the embodiment of this application. As shown in Figure 6(a) and Figure 6(b), Figure 6(a) represents the fused lane line polyline point cloud, and Figure 6(b) represents the target optimized lane line polyline point cloud after performing nonlinear optimization processing on the fused lane line polyline point cloud, which can completely represent each lane line.

[0094] As described above, a second nonlinear optimization is performed on the front lane line optimized polyline point cloud and the rear lane line optimized polyline point cloud respectively to obtain the target optimized lane line polyline point cloud. In one embodiment of this application, in response to the fact that the endpoints of the front lane line optimized polyline point cloud near the rear lane line optimized polyline point cloud and the endpoints of the rear lane line optimized polyline point cloud near the front lane line optimized point cloud do not coincide, an overlap operation is performed on the two non-coincident endpoints to connect the front lane line optimized polyline point cloud and the rear lane line optimized polyline point cloud, thereby obtaining the target optimized lane line polyline point cloud.

[0095] Since the endpoints of the optimized point cloud of the front lane line close to the optimized point cloud of the rear lane line do not coincide with the endpoints of the optimized point cloud of the rear lane line close to the optimized point cloud of the front lane line, it is understandable that after performing a second nonlinear optimization operation on the optimized point clouds of the front and rear lane lines respectively, the two optimized point clouds of the lane lines will still be disconnected from each other, that is, the endpoints of the two optimized point clouds of the lane lines do not coincide.

[0096] A coincidence operation is performed on two non-overlapping endpoints to connect the optimized lane line point clouds of the front and rear lane lines, thus obtaining the target optimized lane line point cloud. In essence, if the endpoints of two optimized lane line point clouds that are close to each other do not overlap, a coincidence operation is performed on these two non-overlapping endpoints, forcing them to overlap. This smoothly connects the optimized lane line point clouds of the front and rear lane lines, resulting in the target optimized lane line point cloud.

[0097] In one embodiment of this application, a first nonlinear optimization is performed on six fusion points, including: obtaining a first error loss for the six fusion points based on distance conditions, smoothing conditions, consistency conditions, and same-point conditions between the six fusion points and a first reference sampling point; in response to the first error loss satisfying a preset error condition, a front lane line optimized polyline point cloud and a rear lane line optimized polyline point cloud are obtained; and / or, a second nonlinear optimization is performed on the front lane line optimized polyline point cloud and the rear lane line optimized polyline point cloud, respectively, including: obtaining a second error loss for the optimized points in the front lane line optimized polyline point cloud based on distance conditions, smoothing conditions, and consistency conditions between the optimized points in the front lane line optimized polyline point cloud and a second reference sampling point; obtaining a third error loss for the optimized points in the rear lane line optimized polyline point cloud based on distance conditions, smoothing conditions, and consistency conditions between the optimized points in the rear lane line optimized polyline point cloud and a third reference sampling point; in response to the second error loss and the third error loss satisfying a preset error condition, a target optimized lane line polyline point cloud is obtained.

[0098] Based on the distance, smoothing, consistency, and same-point conditions between the six fusion points and the first reference sampling point, the first error loss of the six fusion points is obtained. It can be understood that the six fusion points can be considered as a sequence of fusion points. This sequence is used as the current sampling point sequence. Based on the current sampling point sequence and a preset step size, the first reference sampling point sequence is obtained. The preset step size refers to a pre-set transformation parameter. It can be understood as the magnitude of a change in a numerical value. For example, for a certain operation on parameter e, assuming the preset step size is f, an operation is performed on e, then e+f is assigned to e, and the operation is performed again with the new e value, and so on. The preset step size can be a specific number, vector, etc., and can be set according to the actual situation or the solution result of the relevant expression; there is no specific limitation.

[0099] The distance condition is represented by the distance constraint value between the fused points in the fused point sequence and the first reference sampling points in the first reference sampling point sequence. The smoothing condition is represented by the smoothing constraint value between the first reference sampling points in the first reference sampling point sequence. The consistency condition is represented by the consistency constraint value between the first reference sampling points in the first reference sampling point sequence. The same-point condition is represented by the same-point constraint value between the two endpoints of the two lane line polyline point clouds in the fused point sequence. Based on the corresponding distance constraint value, smoothing constraint value, consistency constraint value, and same-point constraint value, the first error loss between the fused point sequence and the first reference sampling point sequence is calculated. The first error loss refers to the constraint conditions between the fused point sequence and the first reference sampling point sequence formed by the distance constraint value, smoothing constraint value, consistency constraint value, and same-point constraint value.

[0100] Specifically, the first physical distance between the fusion point sequence and the corresponding first reference sampling point sequence is calculated to obtain the distance constraint value. The distance constraint value is a parameter used to constrain the distance between the fusion point sequence and the first reference sampling point sequence. It can be understood as a parameter used to constrain the deviation of the first reference sampling point sequence from the fusion point sequence. The distance constraint value is determined by the first physical distance and its corresponding weight. For example, the physical distance between the fusion point and the corresponding first reference sampling point can be calculated using the position coordinates of the fusion point and the first reference sampling point, and then using the distance calculation formula.

[0101] The smoothness between three adjacent first reference sampling points in the first reference sampling point sequence is calculated sequentially to obtain a smoothing constraint value. The smoothing constraint value is a parameter that constrains the degree of smoothness between the first reference sampling points in the first reference sampling point sequence. Specifically, the smoothness between any three adjacent first reference sampling points in the first reference sampling point sequence is calculated according to the smoothness calculation formula. Based on the multiple smoothness values ​​contained in the first reference sampling point sequence, the smoothing constraint value of the first reference sampling point sequence is calculated. In one embodiment, the smoothing constraint value is determined by the smoothness and its corresponding weight.

[0102] The consistency degree between four adjacent first reference sampling points in the first reference sampling point sequence is calculated sequentially to obtain the consistency constraint value. The consistency constraint value is a parameter that constrains the degree of consistency among the first reference sampling points in the first reference sampling point sequence. Specifically, the consistency degree between any four adjacent first reference sampling points in the first reference sampling point sequence is calculated according to the consistency degree calculation formula. Based on the multiple consistency degrees contained in the first reference sampling point sequence, the consistency constraint value of the first reference sampling point sequence is calculated. In one embodiment, the consistency constraint value is determined by the consistency degree and its corresponding weight.

[0103] The similarity degree between the two endpoints of the two lane line polyline point clouds in the fusion point sequence is calculated to obtain the similarity constraint value. It can be understood that among the six fusion points in the fusion point sequence, two are endpoints of the two lane line polyline point clouds. The similarity constraint value refers to a parameter that constrains the proximity between the two endpoints; that is, the two endpoints are closest when they are at the same point. Specifically, assuming the two endpoints are p1(t1, v1) and p2(t2, v2), the similarity degree between the two endpoints can be calculated using the formulas z1 = t2 - t1 and z2 = v2 - v1. Based on the similarity degree, the similarity constraint value between the two endpoints is calculated. In one embodiment, the similarity constraint value is determined by the similarity degree and its corresponding weight. In practice, the weight corresponding to the similarity degree can be increased to make the two endpoints relatively smoothly approach and overlap.

[0104] In response to the first error loss meeting the preset error condition, the optimized polyline point cloud of the front lane line and the optimized polyline point cloud of the rear lane line are obtained. Understandably, the first reference sampling point sequence is used as the current sampling point sequence. The step of calculating the first reference sampling point sequence based on the current sampling point sequence and a preset step size is executed until the first error loss reaches the preset error condition, obtaining optimized points. These optimized points are then used to update the six fused points, resulting in the optimized polyline point cloud of the front lane line and the optimized polyline point cloud of the rear lane line.

[0105] Based on the distance, smoothing, and consistency conditions between the optimized points in the optimized polyline point cloud of the front lane line optimization and the second reference sampling points, the second error loss of the optimized points in the optimized polyline point cloud of the front lane line optimization is obtained. It can be understood that the points in the optimized polyline point cloud of the front lane line optimization are used as the current sampling point sequence, and based on the current sampling point sequence and a preset step size, the second reference sampling point sequence is obtained.

[0106] The distance condition is represented by the distance constraint value between points in the current sampling point sequence and the second reference sampling point in the second reference sampling point sequence. The smoothing condition is represented by the smoothing constraint value between the second reference sampling points in the second reference sampling point sequence. The consistency condition is represented by the consistency constraint value between the second reference sampling points in the second reference sampling point sequence. Based on the corresponding distance constraint value, smoothing constraint value, and consistency constraint value, the second error loss between the current sampling point sequence and the second reference sampling point sequence is calculated. The second error loss refers to the constraint condition between the current sampling point sequence and the second reference sampling point sequence formed by the distance constraint value, smoothing constraint value, and consistency constraint value.

[0107] Based on the distance, smoothing, and consistency conditions between the optimized points and the third reference sampling points in the optimized polyline point cloud of the rear lane line optimization, the third error loss of the optimized points in the rear lane line optimization polyline point cloud is obtained. It can be understood that the points in the optimized polyline point cloud of the rear lane line optimization are used as the current sampling point sequence, and based on the current sampling point sequence and a preset step size, the third reference sampling point sequence is obtained.

[0108] The distance condition is represented by the distance constraint value between points in the current sampling point sequence and the third reference sampling point in the third reference sampling point sequence. The smoothing condition is represented by the smoothing constraint value between the third reference sampling points in the third reference sampling point sequence. The consistency condition is represented by the consistency constraint value between the third reference sampling points in the third reference sampling point sequence. Based on the corresponding distance constraint value, smoothing constraint value, and consistency constraint value, the third error loss between the current sampling point sequence and the third reference sampling point sequence is calculated. The third error loss refers to the constraint condition between the current sampling point sequence and the third reference sampling point sequence formed by the distance constraint value, smoothing constraint value, and consistency constraint value.

[0109] In response to the second and third error losses satisfying preset error conditions, the target optimized lane line polyline point cloud is obtained. Understandably, the process involves using the second reference sampling point sequence as the current sampling point sequence, returning the step of calculating the second reference sampling point sequence based on the current sampling point sequence and a preset step size, and executing this process until the second error loss reaches the preset error condition, thus obtaining the forward lane line optimization points. These forward lane line optimization points are then used to update the points in the forward lane line optimized polyline point cloud. Similarly, the process involves using the third reference sampling point sequence as the current sampling point sequence, returning the step of calculating the third reference sampling point sequence based on the current sampling point sequence and a preset step size, and executing this process until the third error loss reaches the preset error condition, thus obtaining the subsequent lane line optimization points. These subsequent lane line optimization points are then used to update the points in the subsequent lane line optimized polyline point cloud, thereby obtaining the target optimized lane line polyline point cloud.

[0110] Please see Figure 7 , Figure 7 This is a flowchart illustrating the autonomous driving method in an embodiment of this application.

[0111] The autonomous driving method includes the following steps:

[0112] S21. Obtain a lane line map; wherein the lane line map is obtained using the crowdsourced lane line fusion map update method in the above embodiments.

[0113] A lane line map is obtained using the crowdsourced lane line fusion map update method described in the above embodiments. For example, the method involves obtaining the point cloud of any original lane line in the original lane line map and at least one candidate lane line point cloud from the crowdsourced lane line map that matches the point cloud of the original lane line; fusing the point cloud of the original lane line and the at least one candidate lane line point cloud to obtain a fused lane line point cloud; and performing a nonlinear optimization operation on the fused lane line point cloud to obtain a target optimized lane line point cloud. This target optimized lane line point cloud is then used to update the corresponding original lane line point cloud, thereby updating the original lane line map.

[0114] S22. Based on the lane line map, perform driving route planning to achieve autonomous driving.

[0115] Based on the lane map, drive path planning is performed for autonomous vehicles to achieve autonomous driving.

[0116] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0117] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of the vehicle-mounted device in an embodiment of this application. The vehicle-mounted device 800 includes a memory 801 and a processor 802 coupled to each other. The processor 802 is used to execute program instructions stored in the memory 801 to implement the steps in the above-described crowdsourced lane line fusion map update method embodiment and the steps in the above-described autonomous driving method embodiment. In a specific implementation scenario, the vehicle-mounted device 800 may include, but is not limited to, a microcomputer or a server.

[0118] Specifically, processor 802 controls itself and memory 801 to implement the steps in the above-described crowdsourced lane line fusion map update method embodiment or the steps in the above-described autonomous driving method embodiment. Processor 802 can also be called a CPU (Central Processing Unit), and may be an integrated circuit chip with signal processing capabilities. Processor 802 can also be a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor can be a microprocessor or any conventional processor. Furthermore, processor 802 can be implemented using integrated circuit chips.

[0119] Please see Figure 9 , Figure 9 This is a schematic diagram of the structure of a non-volatile computer-readable storage medium in an embodiment of this application. The computer-readable storage medium 900 is used to store program instructions 901. When executed by the processor 802, the program instructions 901 are used to implement the steps in the above-described crowdsourced lane line fusion map update method embodiment or the steps in the above-described autonomous driving method embodiment.

[0120] The description of the various embodiments above tends to emphasize the differences between the various embodiments. The similarities or similarities between them can be referred to, and for the sake of brevity, they will not be repeated here.

[0121] In the several embodiments provided in this application, it should be understood that the disclosed methods and related devices can be implemented in other ways. For example, the related device implementations described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication disconnection shown or discussed may be indirect coupling or communication disconnection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0122] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0123] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A crowdsourced lane line fusion map update method, characterized in that, include: Obtain the polyline point cloud of any original lane line in the original lane line map and at least one candidate lane line polyline point cloud in the crowdsourced lane line map that matches the polyline point cloud of the original lane line. The point sets of any original lane line polyline point cloud and at least one candidate lane line polyline point cloud are fused to obtain a fused lane line polyline point cloud. The point set includes candidate points in the candidate lane line polyline point cloud that overlap with the original lane line polyline point cloud and all original points in the original lane line polyline point cloud. The fused lane line polyline point cloud includes a front lane line polyline point cloud and a rear lane line polyline point cloud. The fused lane line polyline point cloud is subjected to nonlinear optimization processing to obtain a target optimized lane line polyline point cloud, which is then used to update the corresponding original lane line polyline point cloud, thereby updating the original lane line map. The nonlinear optimization processing operation on the fused lane line polyline point cloud includes: The first nonlinear optimization is performed on the six fusion points in the front lane line polyline point cloud and the rear lane line polyline point cloud to obtain the front lane line optimized polyline point cloud and the rear lane line optimized polyline point cloud. A second nonlinear optimization is performed on the optimized polygonal point cloud of the front lane line and the optimized polygonal point cloud of the rear lane line respectively to obtain the target optimized lane line polygonal point cloud.

2. The method according to claim 1, characterized in that, Obtaining at least one candidate lane line polyline point cloud that matches any one of the original lane line polyline point clouds includes: Determine the point-to-line distance value between any original lane line polyline point cloud and the candidate lane line polyline point cloud in the crowdsourced lane line map, so as to obtain at least one point-to-line distance value corresponding to each original lane line polyline point cloud. Obtain the directional distance value between any original lane line polyline point cloud and the candidate lane line polyline point cloud in the crowdsourced lane line map, so as to obtain at least one directional distance value corresponding to each original lane line polyline point cloud. In response to the point-to-line distance value being less than a preset point-to-line distance threshold and the direction distance value being less than a preset direction distance threshold, the corresponding candidate lane line polyline point cloud is obtained, so that each original lane line polyline point cloud is matched to obtain at least one candidate lane line polyline point cloud.

3. The method according to claim 2, characterized in that, Obtaining the point-to-line distance value between any of the original lane line polyline point clouds and the candidate lane line polyline point clouds in the crowdsourced lane line map includes: Obtain any original point in the original lane line polyline point cloud and two candidate points in any candidate lane line polyline point cloud that are close to the original point. The point-to-line distance value of any original point is characterized by the vertical distance between the original point and the line connecting the two corresponding candidate points, thereby obtaining the point-to-line distance value corresponding to each original point in the original lane line polyline point cloud. The point-to-line distance between the original lane line polyline point cloud and the candidate lane line polyline point cloud is the average of the point-to-line distances corresponding to all original points in the original lane line polyline point cloud, thereby obtaining the point-to-line distance between the original lane line polyline point cloud and at least one candidate lane line polyline point cloud.

4. The method according to claim 2, characterized in that, Obtaining the directional distance value between any of the original lane line polyline point clouds and the candidate lane line polyline point clouds in the crowdsourced lane line map includes: Obtain a first vector formed by the original points at both ends of the original lane line polyline point cloud, and a second vector formed by the candidate points at both ends of any candidate lane line polyline point cloud. The directional distance between the original lane line polyline point cloud and the candidate lane line polyline point cloud is characterized by the angle between the first vector and the second vector, thereby obtaining the directional distance between the original lane line polyline point cloud and at least one of the candidate lane line polyline point clouds.

5. The method according to claim 1, characterized in that, The point sets of any original lane line polyline point cloud and at least one candidate lane line polyline point cloud are fused, including: Perform semantic averaging and curve fitting operations on the point set to obtain a fitted curve; The fitted curve is resampled to obtain the fused lane line polyline point cloud.

6. The method according to claim 1, characterized in that, The six fusion points include three consecutive fusion points of the front lane line polyline point cloud that are close to the rear lane line polyline point cloud, and three consecutive fusion points of the rear lane line polyline point cloud that are close to the front lane line polyline point cloud.

7. The method according to claim 6, characterized in that, In response to the fact that the endpoint of the optimized front lane line polyline point cloud near the optimized rear lane line polyline point cloud does not coincide with the endpoint of the optimized rear lane line polyline point cloud near the optimized front lane line point cloud, an overlap operation is performed on the two non-overlapping endpoints to connect the optimized front lane line polyline point cloud and the optimized rear lane line polyline point cloud, thereby obtaining the target optimized lane line polyline point cloud.

8. The method according to claim 6, characterized in that, The first nonlinear optimization is performed on the six fusion points, including: Based on the distance conditions, smoothing conditions, consistency conditions, and same-point conditions between the six fusion points and the first reference sampling point, the first error loss of the six fusion points is obtained. In response to the first error loss satisfying a preset error condition, optimized polyline point clouds of the front lane line and optimized polyline point clouds of the rear lane line are obtained; and / or A second nonlinear optimization is performed on the optimized polyline point cloud of the front lane line and the optimized polyline point cloud of the rear lane line, respectively, including: Based on the distance conditions, smoothing conditions, and consistency conditions between the optimized points in the optimized polyline point cloud of the front lane line and the second reference sampling point, the second error loss of the optimized points in the optimized polyline point cloud of the front lane line is obtained. Based on the distance conditions, smoothing conditions, and consistency conditions between the optimized points and the third reference sampling points in the optimized polyline point cloud of the rear lane line, the third error loss of the optimized points in the optimized polyline point cloud of the rear lane line is obtained. In response to the second error loss and the third error loss satisfying the preset error condition, the target optimized lane line polyline point cloud is obtained.

9. An autonomous driving method, characterized in that, include: Obtain a lane line map; wherein the lane line map is obtained using the crowdsourced lane line fusion map update method according to any one of claims 1-8; Based on the lane map, driving path planning is performed to achieve autonomous driving.

10. A vehicle-mounted device, characterized in that, It includes a memory and a processor coupled to each other, the processor being used to execute program instructions stored in the memory to implement the crowdsourced lane line fusion map update method as described in any one of claims 1-8 or the autonomous driving method as described in claim 9.

11. A non-volatile computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store program instructions, which, when executed by a processor, are used to implement the crowdsourced lane line fusion map update method as described in any one of claims 1-8 or the autonomous driving method as described in claim 9.

Images