Lane line point cloud map construction method and device, equipment, storage medium and program

Abstract

The application discloses a lane line point cloud map construction method and device, equipment, a storage medium and a program. Positioning information, original point cloud and image lane line detection results are acquired, the acquired information is time-synchronized, and in the case that lane line information of an actual lane line is contained in the image lane line detection result corresponding to the original point cloud, lane line prediction is performed according to the lane line information, a predicted lane line is obtained, the actual lane line and the predicted lane line are discretized, a two-dimensional preview frame corresponding to a discrete point is generated, coarse extraction is performed on the original point cloud based on the two-dimensional preview frame, a third point cloud is obtained, fine extraction is performed on the third point cloud based on a point cloud reflection intensity peak feature, a lane line point cloud is obtained, and a lane line point cloud map is generated according to the lane line point cloud of the original point cloud and the positioning information. According to the embodiment, the lane line detection range is expanded through lane line prediction, and the accuracy of the detected lane line point cloud is improved through the combination of coarse extraction and fine extraction.

Description

Technical Field

[0001] This application belongs to the field of perception technology for intelligent driving, and particularly relates to a method, device, equipment, storage medium and program for constructing lane line point cloud maps. Background Technology

[0002] Lane detection is a technology used to provide lane boundaries for autonomous driving, determining the constraints between the vehicle's current position and surrounding lanes. Existing lane detection technologies generally utilize images, employing feature extraction or deep learning to detect lane information around the vehicle, or extract lane point clouds using the reflection intensity of LiDAR 3D point clouds.

[0003] However, traditional lane detection methods based on images or LiDAR point clouds suffer from low detection accuracy. Furthermore, image-based lane detection often relies on the perception range of a single frame image, while LiDAR-based lane detection relies on the perception range of a single frame point cloud. Therefore, both methods have limited detection range. Summary of the Invention

[0004] This application provides a method, apparatus, device, storage medium, and program for constructing lane line point cloud maps, which can accurately detect lane lines and expand the detection range of lane lines.

[0005] In a first aspect, embodiments of this application provide a method for constructing a lane line point cloud map, including:

[0006] Obtain positioning information, at least one frame of raw point cloud data acquired by LiDAR, and lane line detection results from the image.

[0007] Synchronize the location information, at least one frame of original point cloud, and the lane line detection results in the image in time.

[0008] For each frame of the original point cloud, assuming time synchronization is complete and the corresponding image lane detection result contains lane line information of at least one actual lane line, lane line prediction is performed based on the lane line information of the actual lane line to obtain the predicted lane line.

[0009] For each lane line in both the actual and predicted lane lines, the lane line is discretized to obtain the discrete points corresponding to the lane line.

[0010] For each discrete point, assign a two-dimensional preview bounding box to the discrete point.

[0011] Extract the points within the 2D preview bounding box from the original point cloud to obtain the third point cloud.

[0012] Point cloud extraction is performed on the third point cloud based on the peak characteristics of the point cloud reflection intensity, resulting in the lane line point cloud in the original point cloud.

[0013] Generate a corresponding lane line point cloud map based on the location information and the lane line point cloud from at least one frame of the original point cloud.

[0014] As one possible implementation, before extracting the point cloud located within the two-dimensional preview box from the original point cloud to obtain the third point cloud, the method also includes:

[0015] The original point cloud is segmented, and points whose position coordinates do not meet the preset first coordinate condition are removed to obtain the first point cloud.

[0016] Obtain the 3D target detection results obtained from the original point cloud.

[0017] Based on the 3D target detection results, dynamic point clouds in the first point cloud are removed to obtain the second point cloud.

[0018] Extract the points within the two-dimensional preview boxes from the original point cloud to obtain the third point cloud, including:

[0019] Extract the points within the 2D preview bounding box from the second point cloud to obtain the third point cloud.

[0020] The first coordinate condition includes:

[0021] x min ≤x0≤x max

[0022] y min ≤y0≤y max

[0023] z min ≤z0≤z max

[0024] In the formula, x min This represents the minimum coordinate value of the X-axis of the set lidar center coordinate system. max This represents the maximum coordinate value of the X-axis of the set lidar center coordinate system, y min This represents the minimum Y-axis coordinate value of the set lidar center coordinate system. max This represents the maximum Y-axis coordinate value of the set lidar center coordinate system, z. min This represents the minimum Z-axis coordinate value of the set lidar center coordinate system. max This represents the maximum coordinate value of the Z-axis in the central coordinate system of the lidar, x0 represents the coordinate value of the point cloud on the X-axis in the central coordinate system of the lidar, y0 represents the coordinate value of the point cloud on the Y-axis in the central coordinate system of the lidar, and z0 represents the coordinate value of the point cloud on the Z-axis in the central coordinate system of the lidar.

[0025] As one possible implementation, based on the 3D target detection results, dynamic point clouds are removed from the first point cloud to obtain a second point cloud, including:

[0026] The 3D target detection result includes 3D bounding box information, which includes the 3D coordinates, pose information, and size information of the center point of the 3D bounding box. Based on the 3D coordinates and pose information of the center point, the position coordinates of the first point cloud are transformed to the coordinate system of the center point of the 3D bounding box.

[0027] Determine the second coordinate condition based on the dimensional information.

[0028] In the first point cloud, those points whose position coordinates in the coordinate system of the center point of the 3D detection box satisfy the second coordinate condition are removed to obtain the second point cloud.

[0029] The second coordinate condition is as follows:

[0030]

[0031]

[0032]

[0033] Where L3 represents the length of the 3D detection bounding box included in the size information, W3 represents the width of the 3D detection bounding box included in the size information, and H3 represents the height of the 3D detection bounding box included in the size information. This represents the X-axis coordinate value in the coordinate system of the center point of the 3D detection box. This represents the Y-axis coordinate value in the coordinate system of the center point of the 3D detection box. This represents the Z-axis coordinate value in the coordinate system of the center point of the 3D detection frame.

[0034] As one possible implementation, lane line prediction is performed based on the actual lane line information to obtain the predicted lane lines, including:

[0035] The lane width is predicted by analyzing the Y-axis coordinates of the actual lane lines in the vehicle's rear axle center coordinate system, as contained in the lane line information.

[0036] Based on the predicted lane width and the coordinates of the first closest lane line point and the first farthest lane line point (from the actual lane line to the rear axle center) contained in the lane line information, the coordinates of the second closest lane line point and the second farthest lane line point (from the predicted lane line to the rear axle center) are determined.

[0037] The inclination angle of the actual lane lines contained in the lane line information is determined as the inclination angle of the predicted lane lines.

[0038] The actual lane line width contained in the lane line information is used to determine the predicted lane line width.

[0039] The fitting equation for the predicted lane line is determined based on the predicted lane line inclination angle.

[0040] The predicted lane line is obtained by drawing a line based on the position coordinates of the second nearest lane line point, the position coordinates of the second farthest lane line point, the inclination angle of the predicted lane line, the line width of the predicted lane line, and the fitting equation of the predicted lane line.

[0041] As one possible implementation, based on the lane width prediction value and the position coordinates of the first closest lane line point closest to the vehicle's rear axle center and the first farthest lane line point farthest from the vehicle's rear axle center in the actual lane lines contained in the lane line information, the position coordinates of the second closest lane line point closest to the vehicle's rear axle center and the second farthest lane line point farthest from the vehicle's rear axle center in the predicted lane lines are determined, including:

[0042] The X-axis coordinate value of the first nearest lane line point is used as the X-axis coordinate value of the second nearest lane line point.

[0043] Based on the Y-axis coordinates of the first nearest lane line point and the predicted lane width, determine the Y-axis coordinates of the second nearest lane line point that is closest to the center of the vehicle's rear axle within the predicted lane line.

[0044] The X-axis coordinate value of the first farthest lane line point is used as the X-axis coordinate value of the second farthest lane line point.

[0045] Based on the Y-axis coordinates of the first farthest lane line point and the predicted lane width, determine the Y-axis coordinates of the second farthest lane line point that is closest to the center of the vehicle's rear axle in the predicted lane line.

[0046] As one possible implementation, the lane lines are discretized to obtain the discrete points corresponding to the lane lines, including:

[0047] Obtain the preset first ratio value.

[0048] Based on the first ratio value, the X-axis coordinates of the lane line point closest to the center of the vehicle's rear axle, and the X-axis coordinates of the lane line point farthest from the center of the vehicle's rear axle, the upper limit of the X-axis coordinates of the discrete points corresponding to the lane line is calculated according to a preset discrete range calculation formula.

[0049] The X-axis coordinate of the lane line point closest to the center of the vehicle's rear axle is used as the lower limit of the X-axis coordinate of the discrete point corresponding to the lane line.

[0050] The interval formed by the upper and lower limits of the X-axis coordinates is used as the range of X-axis coordinate values ​​for the discrete points corresponding to the lane lines.

[0051] Based on the preset fitting equations for the discretization step length and lane lines, the lines within the X-axis coordinate range of the lane lines are discretized to obtain the discrete points corresponding to the lane lines.

[0052] The formula for calculating the discrete range is as follows:

[0053]

[0054] In the formula, This represents the upper limit of the X-axis coordinate of the discrete point corresponding to lane line i. This represents the X-axis coordinate of the point on lane line i that is closest to the center of the vehicle's rear axle. The X-axis coordinate of lane line i represents the point on lane line i that is farthest from the center of the rear axle of the vehicle, and α represents the first proportional value.

[0055] One possible implementation is to assign a two-dimensional preview bounding box to discrete points, including:

[0056] The position coordinates of the discrete points are used as the center coordinates of their corresponding two-dimensional preview boxes.

[0057] Use the distance from the step size as the length of the two-dimensional preview box.

[0058] The width of the two-dimensional preview box is obtained by multiplying the preset lane width expansion coefficient by the width of the lane line to which the discrete point belongs.

[0059] Generate a rotation matrix for a two-dimensional preview frame based on the inclination angle of the lane line to which the discrete point belongs.

[0060] Assign a two-dimensional preview frame to a discrete point, along with its center coordinates, length, width, and rotation matrix.

[0061] As one possible implementation, point cloud extraction is performed on the second point cloud based on the two-dimensional preview bounding box to obtain the third point cloud, including:

[0062] For each point in the second point cloud, based on the center coordinates of the two-dimensional preview bounding box and the rotation matrix, the point's position coordinates are transformed to the center coordinate system of the two-dimensional preview bounding box.

[0063] The third coordinate condition is determined based on the length and width of the two-dimensional preview frame.

[0064] Extract the points from the second point cloud whose transformed coordinates meet the third coordinate condition.

[0065] The point cloud composed of the extracted points is used as the third point cloud.

[0066] The third coordinate conditions include:

[0067]

[0068]

[0069] In the formula, L2 represents the length of the two-dimensional preview box, and W2 represents the width of the two-dimensional preview box. This represents the X-axis coordinate of the second point cloud in the center coordinate system of the two-dimensional preview box. This represents the Y-axis coordinate value of the second point cloud in the center coordinate system of the two-dimensional preview frame.

[0070] As one possible implementation, point cloud extraction is performed on the third point cloud based on the peak features of the point cloud reflection intensity to obtain the lane line point cloud in the original point cloud, including:

[0071] For each point in the third point cloud, determine the point's pitch angle, horizontal azimuth angle, and reflection intensity characterization value.

[0072] Based on the pitch angle of the points, the third point cloud is divided into multiple point cloud clusters. Within the same point cloud cluster, the difference in pitch angle between any two points is less than a preset difference threshold.

[0073] For each point cloud cluster, arrange the points contained in the cluster in ascending order of horizontal azimuth angle.

[0074] Based on the reflection intensity characterization value of each point in the point cloud cluster, the matching rising and falling edges in the point cloud cluster are determined. The rising edge consists of multiple points whose reflection intensity characterization values ​​show a continuous upward trend, and the falling edge consists of multiple points whose reflection intensity characterization values ​​show a continuous downward trend.

[0075] If the matching rising and falling edges meet the preset lane line conditions, extract the points that meet the preset extraction conditions from the matching rising and falling edges as lane line points.

[0076] The point cloud composed of lane line points extracted from the point cloud cluster is used as the lane line point cloud in the original point cloud.

[0077] Lane line conditions include:

[0078] The difference between the Y-axis coordinate value at the end of the rising edge and the Y-axis coordinate value at the end of the falling edge is less than a preset lane width threshold, and the difference between the reflection intensity values ​​at the start and end of the falling edge is greater than a preset lane reflection intensity difference threshold.

[0079] Extraction criteria include:

[0080] The reflection intensity characterization value is the highest, and the height value is less than the preset height threshold.

[0081] As one possible implementation, the characterization value of the reflection intensity at a point includes:

[0082] After determining the points in the third point cloud and arranging them in ascending order of horizontal azimuth, the points located within the neighborhood of each point are...

[0083] Calculate the mean reflection intensity of the point and its neighborhood.

[0084] The mean value of the reflection intensity is determined as the characterization value of the reflection intensity at a point.

[0085] Secondly, embodiments of this application also provide a lane line point cloud map construction device, comprising:

[0086] The acquisition module is used to acquire positioning information, at least one frame of raw point cloud data collected by the LiDAR, and lane line detection results from the image.

[0087] The synchronization module is used to synchronize the positioning information, at least one frame of raw point cloud data, and the lane detection results in the image over time.

[0088] The lane prediction module is used to predict lane lines for each frame of the original point cloud, provided that time synchronization is completed and the corresponding image lane detection result contains lane line information of at least one actual lane line.

[0089] The discretization module is used to discretize each lane line in both the actual and predicted lane lines, obtaining the discrete points corresponding to the lane lines.

[0090] The preview bounding box determination module is used to assign a two-dimensional preview bounding box to each discrete point.

[0091] The coarse extraction module is used to extract the points within the two-dimensional preview boxes in the original point cloud, resulting in a third point cloud.

[0092] The fine extraction module is used to extract point clouds from the third point cloud based on the peak features of the point cloud reflection intensity, thereby obtaining the lane line point cloud from the original point cloud.

[0093] The map generation module is used to generate a corresponding lane line point cloud map based on the location information and the lane line point cloud in at least one frame of the original point cloud.

[0094] Thirdly, embodiments of this application also provide an electronic device, the device including: a processor and a memory storing computer program instructions.

[0095] When the processor executes computer program instructions, it implements a lane line point cloud map construction method as described in the first aspect.

[0096] Fourthly, embodiments of this application also provide a computer-readable storage medium, characterized in that the computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the lane line point cloud map construction method as described in the first aspect.

[0097] Fifthly, embodiments of this application also provide a computer program product, characterized in that, when the instructions in the computer program product are executed by the processor of an electronic device, the electronic device performs the lane line point cloud map construction method as described in the first aspect.

[0098] The lane line point cloud map construction method, apparatus, device, storage medium, and program of this application embodiment acquire positioning information, at least one frame of original point cloud, and image lane line detection results. After time synchronization of the acquired information, and given that the actual lane line information is contained in the image lane line detection results corresponding to the original point cloud, lane line prediction is performed based on the lane line information to obtain predicted lane lines. The actual lane lines and predicted lane lines are discretized to obtain discrete points, and two-dimensional preview boxes corresponding to the discrete points are generated. Based on the two-dimensional preview boxes, coarse extraction is performed on the original point cloud to obtain a third point cloud. Based on the peak features of the point cloud reflection intensity, fine extraction is performed on the third point cloud to obtain the lane line point cloud in the original point cloud. A lane line point cloud map is generated based on the lane line point cloud of at least one frame of original point cloud and positioning information. According to this embodiment, the lane line detection range is expanded by lane line prediction, and the accuracy of the detected lane line point cloud is improved by combining coarse and fine extraction. Attached Figure Description

[0099] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0100] Figure 1 This is a flowchart illustrating a lane line point cloud map construction method provided in an embodiment of this application.

[0101] Figure 2 This is a schematic diagram of lane lines provided in an embodiment of this application.

[0102] Figure 3 This is a schematic diagram of a lane line point cloud map construction method provided in another embodiment of this application.

[0103] Figure 4 This is a schematic diagram of the lane line point cloud map construction device provided in the embodiments of this application.

[0104] Figure 5 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0105] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0106] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0107] Traditional image-based lane detection methods can only detect two-dimensional lane lines. The three-dimensional information of lane lines is highly dependent on the accuracy of the camera's inverse perspective transformation calibration, and accurate elevation information is often unavailable, resulting in poor detection accuracy. Traditional LiDAR-based lane detection methods often contain a large number of non-lane noise points in the detected lane line point cloud, leading to low detection accuracy. Furthermore, as mentioned earlier, traditional image-based or LiDAR point cloud-based lane detection methods also suffer from limited detection range.

[0108] In view of this, in order to solve the problems existing in the prior art, this application provides a method, apparatus, device, storage medium and program for constructing lane line point cloud maps.

[0109] The method for constructing lane line point cloud maps provided in the embodiments of this application will be described below.

[0110] See Figure 1 This is a flowchart illustrating a lane line point cloud map construction method provided in an embodiment of this application. Figure 1 As shown, the method may include the following steps S11-S18.

[0111] S11. Obtain positioning information, at least one frame of original point cloud data acquired by LiDAR, and lane line detection results from the image.

[0112] The embodiments of this application can be used in vehicles to generate lane line point cloud maps around the vehicle.

[0113] The location information can be the vehicle's location information obtained by the onboard positioning system. Simultaneously, an onboard LiDAR can be used to collect surrounding point clouds, which are then used as the raw point cloud. An image acquisition device is then used to acquire images of the vehicle's surroundings. Lane line detection is performed on these images to obtain the image lane line detection result. If the image contains lane lines (hereinafter referred to as actual lane lines), the image lane line detection result will include the lane line information of the actual lane lines.

[0114] Among them, any mature image-based lane detection technology can be used for lane detection in the image, and there are no restrictions on it.

[0115] S12. Synchronize the positioning information, at least one frame of original point cloud, and the lane line detection results in the image in time.

[0116] During time synchronization, the timestamp of the original point cloud can be used as a reference, and the corresponding image lane detection results and positioning information can be matched according to the reference timestamp. The timestamps corresponding to the original point cloud, positioning information, and image lane detection results are uniformly assigned by the host computer.

[0117] S13. For each frame of the original point cloud, if time synchronization is completed and the corresponding image lane line detection result contains lane line information of at least one actual lane line, lane line prediction is performed based on the lane line information of the actual lane line to obtain the predicted lane line.

[0118] The lane line detection result of the image corresponding to the original point cloud is the lane line detection result of the image synchronized with the original point cloud in time.

[0119] As can be seen from the foregoing, existing lane line detection methods have the problem of limited detection range. In order to solve this problem, this application embodiment predicts lane lines based on actual lane lines, thereby expanding the lane line detection range.

[0120] Because the embodiments of this application are based on the lane line information of the actual lane lines to achieve lane line prediction, before performing lane line prediction, it is first determined whether the lane line detection result of the image contains the lane line information of the actual lane lines. If it is determined that the lane line information contains at least one actual lane line, then lane line prediction is performed to obtain the predicted lane line.

[0121] If the lane line detection results in the image do not contain lane line information of the actual lane lines, lane line prediction will not be performed.

[0122] S14. For each lane line in the actual lane lines and the predicted lane lines, the lane lines are discretized to obtain the discrete points corresponding to the lane lines.

[0123] S15. For each discrete point, assign a two-dimensional preview box to the discrete point.

[0124] S16. Extract the point cloud located within the two-dimensional preview box from the original point cloud to obtain the third point cloud.

[0125] In steps S14-S16, the discrete points are obtained by discretizing the lane lines, and the two-dimensional preview box is also obtained based on the discrete points. Therefore, it can be determined that the range enclosed by the two-dimensional preview box includes the local lane lines and / or a part of the area around the lane lines. Thus, it can be further determined that the point cloud located within the two-dimensional preview box is related to the lane lines. The point cloud within the two-dimensional preview box may be the point cloud on the lane lines or the point cloud around the lane lines. The point cloud located outside the two-dimensional preview box can be determined to have a large deviation from the lane lines. Therefore, by extracting the point cloud located within the two-dimensional preview box from the original point cloud, the point cloud with a large deviation from the lane lines can be removed, thereby obtaining a third point cloud that is related to the lane lines.

[0126] However, this method often only achieves coarse extraction of lane line point clouds, because the resulting third point cloud may contain some point clouds around the lane lines. Therefore, in order to further improve the accuracy of the final lane line point cloud, S17 is executed.

[0127] S17. Extract the third point cloud from the point cloud based on the peak characteristics of the point cloud reflection intensity to obtain the lane line point cloud in the original point cloud.

[0128] The color of lane lines on a road is usually different from the color of other objects on the road (such as the road surface and roadbed). The color of lane lines is usually lighter than that of other objects, so the reflection intensity of lane line point clouds is usually stronger than that of other object point clouds on the road. Based on this, embodiments of this application perform more refined extraction of lane line point clouds according to the emission intensity of the point clouds.

[0129] The lane lines are the lightest in color on the road, so the lane line point cloud has the highest reflection intensity. Therefore, the reflection intensity peak feature of the third point cloud can be extracted, and the lane line point cloud can be accurately identified based on the reflection intensity peak feature of the point cloud.

[0130] S18. Generate a corresponding lane line point cloud map based on the positioning information and the lane line point cloud in at least one frame of the original point cloud.

[0131] The lane line point cloud contained in each frame of the original point cloud can be accurately extracted using the above method. Based on the positioning information synchronized in step S12, all lane line point clouds contained in at least one frame of the original point cloud can be transformed to a unified coordinate system to generate a lane line point cloud map.

[0132] The lane line point cloud map construction method provided in this application involves acquiring location information, at least one frame of original point cloud, and image lane line detection results. After time synchronization of the acquired information, and given that the image lane line detection results corresponding to the original point cloud contain lane line information of the actual lane lines, lane line prediction is performed based on the lane line information to obtain predicted lane lines. The actual lane lines and predicted lane lines are discretized to obtain discrete points, and two-dimensional preview boxes corresponding to the discrete points are generated. Based on the two-dimensional preview boxes, coarse extraction is performed on the original point cloud to obtain a third point cloud. Based on the peak features of the point cloud reflection intensity, fine extraction is performed on the third point cloud to obtain the lane line point cloud in the original point cloud. A lane line point cloud map is generated based on the lane line point cloud of at least one frame of original point cloud and location information. According to this embodiment, lane line prediction expands the lane line detection range, and the combination of coarse and fine extraction improves the accuracy of the detected lane line point cloud.

[0133] In some embodiments, considering that the original point cloud is a 3D point cloud, it includes other point clouds besides the road surface point cloud, and the detection of lane line point clouds typically only requires the road surface point cloud, not the entire point cloud. Therefore, to reduce the amount of data processing, the following steps can be performed before executing S16:

[0134] The original point cloud is segmented, and points whose position coordinates do not meet the preset first coordinate condition are removed to obtain the first point cloud.

[0135] Obtain the 3D target detection results obtained from the original point cloud.

[0136] The second point cloud is obtained by removing the dynamic point cloud from the first point cloud based on the 3D target detection results.

[0137] The first coordinate condition includes:

[0138] x min ≤x0≤x max

[0139] y min ≤y0≤y max

[0140] z min ≤z0≤z max

[0141] In the formula, x min This represents the minimum coordinate value of the X-axis of the set lidar center coordinate system. max This represents the maximum coordinate value of the X-axis of the set lidar center coordinate system, y min This represents the minimum Y-axis coordinate value of the set lidar center coordinate system. max This represents the maximum Y-axis coordinate value of the set lidar center coordinate system, z. min This represents the minimum Z-axis coordinate value of the set lidar center coordinate system. max This represents the maximum coordinate value of the Z-axis in the central coordinate system of the lidar, x0 represents the coordinate value of the point cloud on the X-axis in the central coordinate system of the lidar, y0 represents the coordinate value of the point cloud on the Y-axis in the central coordinate system of the lidar, and z0 represents the coordinate value of the point cloud on the Z-axis in the central coordinate system of the lidar.

[0142] In the above formula, x min x max y min y max z min and z max The values ​​can be set according to the actual situation. Based on the above values, the point cloud region required for lane line point cloud detection can be extracted from the original point cloud, and some useless point clouds can be removed, thereby reducing the amount of data processing.

[0143] After obtaining the first point cloud, considering that it may contain some 3D targets such as vehicles and pedestrians, and that the point clouds corresponding to these 3D targets would interfere with lane line point cloud detection, the dynamic point clouds corresponding to the 3D targets in the first point cloud are further removed to obtain the second point cloud.

[0144] The point cloud corresponding to the 3D target can be determined based on the 3D target detection results obtained by performing 3D target detection on the original point cloud.

[0145] The 3D target detection result contains information corresponding to at least one 3D detection box, wherein the information corresponding to the 3D detection box includes the 3D coordinates t of the center point of the 3D detection box. bbox Attitude information R bbox And size information, which may include the length L, width W and height H of the frame.

[0146] t bbox and R bbox These all represent the position coordinates and orientation of the 3D detection box in the coordinate system of the rear axle center of the vehicle. Therefore, before removing the dynamic point cloud from the first point cloud based on the 3D target detection results, a point cloud transformation is performed on the first point cloud to transform its position coordinates from the LiDAR center coordinate system to the coordinate system of the rear axle center of the vehicle.

[0147] For point cloud transformation, the host computer's radar calibration algorithm can provide the pose transformation relationship between the vehicle's rear axle center coordinate system and the lidar center coordinate system. The transformation relationship is shown below:

[0148]

[0149] In the formula, R represents the relative attitude, t represents the relative position, and (x,y,z) represents the position coordinates of the point in the coordinate system of the lidar center. ′ ,y ′ ,z ′ () indicates the position coordinates of the point in the coordinate system of the rear axle center of the vehicle.

[0150] After transforming the coordinates of each point in the first point cloud from the coordinate system of the LiDAR center to the coordinate system of the rear axle center of the vehicle, the dynamic point cloud in the first point cloud can be removed based on the 3D target detection results to obtain the second point cloud.

[0151] When removing dynamic point clouds from the first point cloud based on the 3D target detection results to obtain the second point cloud, the following steps can be performed:

[0152] Based on the 3D coordinates and attitude information of the center point of the 3D detection frame, the position coordinates of the first point cloud are transformed from the vehicle rear axle center coordinate system to the center point coordinate system of the 3D detection frame.

[0153] The second coordinate condition is determined based on the size information of the 3D detection frame.

[0154] The point cloud that satisfies the second coordinate condition in the coordinate system of the center point of the 3D detection box in the first point cloud is removed to obtain the second point cloud.

[0155] When transforming the position coordinates of the first point cloud from the vehicle rear axle center coordinate system to the center point coordinate system of the 3D detection box based on the center point's 3D coordinates and attitude information, the transformation can be performed using the following method:

[0156]

[0157] In the formula, This indicates the position coordinates of the point in the coordinate system of the center point of the 3D detection box.

[0158] The second coordinate condition is as follows:

[0159]

[0160]

[0161]

[0162] Where L3 represents the length of the 3D detection bounding box included in the size information, W3 represents the width of the 3D detection bounding box included in the size information, and H3 represents the height of the 3D detection bounding box included in the size information. This represents the X-axis coordinate value in the coordinate system of the center point of the 3D detection box. This represents the Y-axis coordinate value in the coordinate system of the center point of the 3D detection box. This represents the Z-axis coordinate value in the coordinate system of the center point of the 3D detection frame.

[0163] If the transformed position coordinates of a point satisfy the second coordinate condition mentioned above, then the point is considered to be a dynamic point cloud, and thus the point is removed from the first point cloud; otherwise, the point is retained in the first point cloud.

[0164] If multiple 3D bounding boxes can be detected in a single frame of point cloud, dynamic point cloud culling is performed on each 3D bounding box to remove the dynamic point cloud within each box. This dynamic point cloud culling for multiple 3D bounding boxes can typically be implemented using GPU-based multithreading.

[0165] The requirement emphasizes that after dynamic point cloud removal, the three-dimensional coordinates of each point in the second point cloud will be expressed using the vehicle's rear axle center coordinate system.

[0166] Based on the above processing, in S16, the third point cloud can be obtained by directly extracting the point cloud located within the two-dimensional preview box from the second point cloud. Compared to directly extracting the third point cloud from the original point cloud, this reduces the amount of data processing.

[0167] In some embodiments, due to the limited field of view when determining lane lines based on image detection, and the existence of missed detections in image lane line detection algorithms, the number of lane lines detected in a single frame is relatively small. To enhance the coverage of lane lines in the image, lateral prediction can be made based on the two lane lines closest to the vehicle on the left and right, thereby obtaining the predicted lane lines.

[0168] The lane line information contained in the image lane line detection results may include the following:

[0169] The coordinates of the position of the lane line point closest to the center of the vehicle's rear axle in the actual lane line (hereinafter referred to as the first closest lane line point).

[0170] The coordinates of the position of the lane line point farthest from the center of the rear axle of the vehicle in the actual lane line (hereinafter referred to as the first farthest lane line point).

[0171] The actual lane markings fit equations with multiple terms, and the norm is: In the formula, i represents the ID of the lane line.

[0172] The actual lane line inclination angle is determined by the fact that the lane line equation must intersect the Y-axis in the vehicle's rear axle center coordinate system. Solving for the equation of the tangent line at that intersection point in the lane line equation gives the inclination angle, which is the angle between this tangent line and the X-axis in the vehicle's rear axle center coordinate system.

[0173] The actual lane line width, that is, the actual physical width value of the lane line.

[0174] Based on the lane line information of the actual lane lines mentioned above, the following steps can be performed when predicting lane lines in S13:

[0175] The lane width is predicted by analyzing the Y-axis coordinate of the actual lane line in the vehicle's rear axle center coordinate system, which is contained in the lane line information.

[0176] Based on the predicted lane width and the coordinates of the first closest lane line point and the first farthest lane line point (from the actual lane line to the rear axle center) contained in the lane line information, the coordinates of the second closest lane line point and the second farthest lane line point (from the predicted lane line to the rear axle center) are determined.

[0177] The inclination angle of the actual lane lines contained in the lane line information is determined as the inclination angle of the predicted lane lines.

[0178] The actual lane line width contained in the lane line information is used to determine the predicted lane line width.

[0179] The fitting equation for the predicted lane line is determined based on the predicted lane line inclination angle.

[0180] The predicted lane line is obtained by drawing a line based on the position coordinates of the second nearest lane line point, the position coordinates of the second farthest lane line point, the inclination angle of the predicted lane line, the line width of the predicted lane line, and the fitting equation of the predicted lane line.

[0181] Specifically, when determining the coordinates of the second closest lane line point and the second farthest lane line point in the predicted lane line that are closest to the rear axle center of the vehicle, based on the lane width prediction value and the coordinates of the first closest lane line point and the first farthest lane line point in the actual lane line that are farthest from the rear axle center of the vehicle, in the predicted lane line information, the following steps can be performed:

[0182] The X-axis coordinate value of the first nearest lane line point is used as the X-axis coordinate value of the second nearest lane line point.

[0183] Based on the Y-axis coordinates of the first nearest lane line point and the predicted lane width, determine the Y-axis coordinates of the second nearest lane line point that is closest to the center of the vehicle's rear axle within the predicted lane line.

[0184] The X-axis coordinate value of the first farthest lane line point is used as the X-axis coordinate value of the second farthest lane line point.

[0185] Based on the Y-axis coordinates of the first farthest lane line point and the predicted lane width, determine the Y-axis coordinates of the second farthest lane line point that is closest to the center of the vehicle's rear axle in the predicted lane line.

[0186] The following example illustrates the lane prediction process.

[0187] See Figure 2 This is a schematic diagram of lane lines provided in an embodiment of this application. The x-direction represents the longitudinal direction of the vehicle's movement, i.e., the X-axis direction of the vehicle's rear axle center coordinate system, and the y-direction represents the lateral left direction of the vehicle, i.e., the Y-axis direction of the vehicle's rear axle center coordinate system. Lane line -1 and lane line 1 are the two lane lines closest to the vehicle, determined through image detection. When predicting lane lines, lane lines to the left of lane line -1 can be predicted based on lane line -1, and lane lines to the right of lane line 1 can be predicted based on lane line 1. Taking lane lines -1 and lane line 1 as actual lane lines, lane lines -2, -3, etc., can be predicted based on lane line -1, and lane lines 2, 3, etc., can be predicted based on lane line 1. The specific number of lane lines to predict can be determined according to the actual situation. Here, -3, -2, -1, 1, 2, and 3 represent lane line IDs.

[0188] Taking ID=-1 as the actual lane line as an example, when predicting the lane line with ID=i (i<-1) based on this actual lane line, the prediction process is as follows:

[0189] Typically, lane width prediction can be calculated using the following formula:

[0190]

[0191] In the formula, w pre This represents the predicted lane width. This represents the Y-axis coordinate of the lane line point in lane line -1 that is closest to the center of the vehicle's rear axle. The Y-axis coordinate of the lane line point closest to the center of the rear axle of the vehicle.

[0192] Based on the above, if only the left lane line -1 is detected near both sides of the vehicle, the predicted lane width can be calculated using the following formula:

[0193]

[0194] If only the right lane line 1 is detected near both sides of the vehicle, the lane width prediction can be calculated using the following formula:

[0195]

[0196] The coordinates of the point in lane i that is closest to the center of the vehicle axle (hereinafter referred to as the second closest lane line point) can be determined according to the following formula:

[0197]

[0198]

[0199] In the formula, This represents the X-axis coordinate of the second nearest lane line point. This represents the Y-axis coordinate of the second nearest lane line point.

[0200] The coordinates of the point in lane i that is farthest from the center of the vehicle axle (hereinafter referred to as the second farthest lane line point) can be determined according to the following formula:

[0201]

[0202]

[0203] In the formula, This represents the X-axis coordinate of the second nearest lane line point. This represents the Y-axis coordinate of the second nearest lane line point.

[0204] The inclination angle corresponding to the predicted lane line i is determined according to the following formula:

[0205] θ i =θ -1

[0206] In the formula, θ i θ represents the inclination angle corresponding to lane line i in the prediction. -1 This indicates the angle of inclination corresponding to the actual lane line -1.

[0207] For the fitted equation of the predicted lane line i, the equation of the predicted lane line is regarded as a first-order fit, that is... in,

[0208] The width of the predicted lane line i is determined according to the following formula:

[0209] width i =width -1

[0210] In the formula, width i This represents the predicted width of lane line i. -1 This indicates the actual lane line width of lane -1.

[0211] When predicting lane lines with ID=i (i>1) based on the actual lane lines with ID=1, the prediction process is the same as described above, and will not be repeated here.

[0212] If only the lane line closest to the vehicle is detected based on image detection, then the lane line on the other side closest to the vehicle can be predicted using that lane line. For example, if only the lane line with ID=-1 is detected, then the lane line with ID=1 can be predicted using the lane line with ID=-1, and vice versa. If no lane lines are detected on either side closest to the vehicle, then no lane line prediction is performed.

[0213] Once the predicted lane lines are obtained, step S14 can be executed.

[0214] In some embodiments, for each lane line, considering the low detection accuracy of points at the far end of the lane line, the following steps can be performed when discretizing the lane line to improve the accuracy of the final lane line point cloud:

[0215] Obtain the preset first ratio value.

[0216] Based on the first ratio value, the X-axis coordinates of the lane line point closest to the center of the vehicle's rear axle, and the X-axis coordinates of the lane line point farthest from the center of the vehicle's rear axle, the upper limit of the X-axis coordinates of the discrete points corresponding to the lane line is calculated according to a preset discrete range calculation formula.

[0217] The X-axis coordinate of the lane line point closest to the center of the vehicle's rear axle is used as the lower limit of the X-axis coordinate of the discrete point corresponding to the lane line.

[0218] The interval formed by the upper and lower limits of the X-axis coordinates is used as the range of X-axis coordinate values ​​for the discrete points corresponding to the lane lines.

[0219] Based on the preset fitting equations for the discretization step length and lane lines, the lines within the X-axis coordinate range of the lane lines are discretized to obtain the discrete points corresponding to the lane lines.

[0220] The formula for calculating the discrete range is as follows:

[0221]

[0222] In the formula, This represents the upper limit of the X-axis coordinate of the discrete point corresponding to lane line i. This represents the X-axis coordinate of the point on lane line i that is closest to the center of the vehicle's rear axle. The X-axis coordinate of lane line i represents the point on lane line i that is farthest from the center of the rear axle of the vehicle, and α represents the first proportional value.

[0223] The first ratio is the optimal ratio set according to the actual situation. The value of this ratio ranges from 0 to 1. Based on this ratio, the range of X-axis coordinate values ​​of the discrete points corresponding to the lane line is determined, which can reduce the problem of low accuracy caused by the low detection accuracy of the far end of the lane line.

[0224] The Z-axis coordinate value of each discrete point can be set to 0.

[0225] The lane width expansion coefficient Φ can be set according to the actual situation, and its value can be greater than 1. The height of the two-dimensional preview frame can be set to infinity. The rotation matrix of the two-dimensional preview frame generated based on the lane line inclination angle is shown below:

[0226]

[0227] After assigning corresponding two-dimensional preview boxes to all discrete points, coarse extraction of the original point cloud or the second point cloud can be performed based on the two-dimensional preview boxes.

[0228] In some embodiments, when coarsely extracting the second point cloud from the two-dimensional preview box to obtain the third point cloud, the following steps may be performed:

[0229] For each point in the second point cloud, based on the center coordinates of the two-dimensional preview bounding box and the rotation matrix, the point's position coordinates are transformed to the center coordinate system of the two-dimensional preview bounding box.

[0230] The third coordinate condition is determined based on the length and width of the two-dimensional preview frame.

[0231] Extract the points from the second point cloud whose transformed coordinates meet the third coordinate condition.

[0232] The point cloud composed of the extracted points is used as the third point cloud.

[0233] The third coordinate conditions include:

[0234]

[0235]

[0236] In the formula, L2 represents the length of the two-dimensional preview box, and W2 represents the width of the two-dimensional preview box. This represents the X-axis coordinate of the second point cloud in the center coordinate system of the two-dimensional preview box. This represents the Y-axis coordinate value of the second point cloud in the center coordinate system of the two-dimensional preview frame.

[0237] The transformation of the point's position coordinates to the center coordinate system of the two-dimensional preview frame, based on the center coordinates and rotation matrix, can include the following transformation relationship:

[0238]

[0239] In the formula, R represents the position coordinates of a point in the center coordinate system of the two-dimensional preview box. pred_bbox Let t represent the rotation matrix of the two-dimensional preview box. pred_bbox This represents the center coordinates of the two-dimensional preview frame.

[0240] Point cloud extraction based on the third coordinate condition ensures that the extracted point cloud is located within the two-dimensional preview box.

[0241] After coarse extraction based on two-dimensional preview boxes to obtain the third point cloud, fine extraction is then performed on the third point cloud.

[0242] In some embodiments, when extracting the third point cloud based on the peak features of the point cloud reflection intensity, S17 can first cluster the third point cloud by pitch angle, then arrange each cluster of point clouds in ascending order of horizontal azimuth angle, and finally extract the real lane line point cloud based on the reflection intensity features.

[0243] This may include the following steps:

[0244] For each point in the third point cloud, determine the point's pitch angle, horizontal azimuth angle, and reflection intensity characterization value.

[0245] Based on the pitch angle of the points, the third point cloud is divided into multiple point cloud clusters. Within the same point cloud cluster, the difference in pitch angle between any two points is less than a preset difference threshold.

[0246] For each point cloud cluster, arrange the points contained in the cluster in ascending order of horizontal azimuth angle.

[0247] Based on the reflection intensity characterization value of each point in the point cloud cluster, the matching rising and falling edges in the point cloud cluster are determined. The rising edge consists of multiple points whose reflection intensity characterization values ​​show a continuous upward trend, and the falling edge consists of multiple points whose reflection intensity characterization values ​​show a continuous downward trend.

[0248] If the matching rising and falling edges meet the preset lane line conditions, extract the points that meet the preset extraction conditions from the matching rising and falling edges as lane line points.

[0249] The point cloud composed of lane line points extracted from the point cloud cluster is used as the lane line point cloud.

[0250] Lane line conditions include:

[0251] The difference between the Y-axis coordinate value at the end of the rising edge and the Y-axis coordinate value at the end of the falling edge is less than the preset lane line width threshold, and the difference between the reflection intensity characterization values ​​at the start and end of the falling edge is greater than the preset lane line reflection intensity difference threshold.

[0252] Extraction criteria include:

[0253] The reflection intensity characterization value is the highest, and the height value is less than the preset height threshold.

[0254] The pitch angle β at each point can be calculated using the following formula:

[0255]

[0256] The horizontal azimuth angle α of each point can be calculated using the following formula:

[0257]

[0258] Since the reflection intensity value of a single point is somewhat unstable, for each point, the average reflection intensity of points in its neighborhood can be used to characterize the reflection intensity of that point, that is, the average reflection intensity is used as the characterization value of the reflection intensity of that point.

[0259] For each point cloud cluster, when determining the matching rising and falling edges of the cluster based on the reflection intensity characterization value of each point in the cluster, the following steps can be performed:

[0260] For each cloud cluster, if the average reflection intensity of M1 consecutive points continuously increases or decreases, then the start and end indices of these M1 points are recorded and marked as rising or falling edges. The value of M1 can be set according to the actual situation.

[0261] For each cloud cluster, the rising and falling edges are traversed. A rising edge and a falling edge are considered to match if the following condition is met:

[0262] Arrange the edges in ascending order of horizontal azimuth angle, generating the rising edge first, followed by the falling edge. The rising and falling edges must be generated adjacent to each other.

[0263] Using the above method, points that meet the characteristics of reflection intensity spikes in point cloud clusters can be extracted, and then the extracted points can be identified as lane line points.

[0264] Traverse all point cloud clusters until all points matching the reflection intensity spike feature are extracted, and combine all extracted points into a lane line point cloud.

[0265] In some embodiments, the reflection intensity characterization value for each point can be determined in the following manner:

[0266] After determining the points in the third point cloud and arranging them in ascending order of horizontal azimuth, the points within the neighborhood of each given point are...

[0267] Calculate the mean reflection intensity of the point and its neighborhood.

[0268] The mean value of the reflection intensity is determined as the characterization value of the reflection intensity at a point.

[0269] The points located within the neighborhood of the store can be N1 points found before and after the current point, where the value of N1 can be set according to the actual situation.

[0270] By using the average reflection intensity of points within a given area to characterize the reflection intensity of a point, the adverse effects caused by the instability of the reflection intensity value of a single point can be eliminated. This application proposes a lane line point cloud map construction method, which constructs a lane line point cloud map based on image lane lines and point cloud reflection intensity. Compared to lane line detection methods using a single image or LiDAR, the detection accuracy is greatly improved, and it includes accurate elevation information (i.e., Z-axis information). Furthermore, by using an image lane line prediction method, the problem of relatively narrow image field of view is avoided, expanding the lateral detection range while ensuring detection accuracy.

[0271] Meanwhile, the lane line point cloud map construction method provided in this embodiment expands the detection range from a single frame to a multi-frame map concept, greatly increasing the detection range of lane lines. This can effectively serve tasks such as path planning and control in intelligent driving, and high-precision map production.

[0272] See Figure 3 This is a flowchart illustrating an optimal embodiment of the lane line point cloud map construction method provided in this application, which mainly includes:

[0273] After synchronizing the original point cloud, 3D target detection results, image lane line detection results, and localization information in time, the original point cloud is segmented and transformed. Dynamic point clouds are removed by combining the 3D target detection results. Then, lane lines are predicted using the image lane line detection results. Based on the prediction results, lane line point clouds are coarsely extracted. Then, lane line point clouds are finely extracted using reflection intensity peak features. Finally, lane line point cloud maps are constructed from multiple frames of lane line point clouds based on localization information.

[0274] Based on the lane line point cloud map construction method provided in the above embodiments, this application also provides specific implementation methods of the lane line point cloud map construction device. Please refer to the following embodiments.

[0275] See Figure 4 The lane line point cloud map construction device provided in this application embodiment includes the following modules:

[0276] The acquisition module 401 is used to acquire positioning information, at least one frame of original point cloud data collected by the LiDAR, and lane line detection results from the image.

[0277] Synchronization module 402 is used to synchronize the positioning information, at least one frame of original point cloud, and the lane line detection results of the image in time.

[0278] The lane prediction module 403 is used to predict lane lines for each frame of the original point cloud, provided that time synchronization is completed and the corresponding image lane detection result contains lane line information of at least one actual lane line.

[0279] Discretization module 404 is used to discretize each lane line in both the actual and predicted lane lines to obtain discrete points corresponding to the lane lines.

[0280] The preview bounding box determination module 405 is used to assign a two-dimensional preview bounding box to each discrete point.

[0281] The coarse extraction module 406 is used to extract the point cloud located within the two-dimensional preview box in the original point cloud to obtain the third point cloud.

[0282] The fine extraction module 407 is used to extract point clouds from the third point cloud based on the peak features of the point cloud reflection intensity, thereby obtaining the lane line point cloud from the original point cloud.

[0283] The map generation module 408 is used to generate a corresponding lane line point cloud map based on the positioning information and the lane line point cloud in at least one frame of the original point cloud.

[0284] The lane line point cloud map construction apparatus of this application embodiment acquires positioning information, at least one frame of original point cloud, and image lane line detection results. After time synchronization of the acquired information, and determining that the image lane line detection results corresponding to the original point cloud contain lane line information of the actual lane lines, lane line prediction is performed based on the lane line information to obtain predicted lane lines. The actual lane lines and predicted lane lines are discretized to obtain discrete points, and two-dimensional preview boxes corresponding to the discrete points are generated. Based on the two-dimensional preview boxes, coarse extraction is performed on the original point cloud to obtain a third point cloud. Based on the peak features of the point cloud reflection intensity, fine extraction is performed on the third point cloud to obtain the lane line point cloud in the original point cloud. A lane line point cloud map is generated based on the lane line point cloud of at least one frame of original point cloud and positioning information. According to this embodiment, the lane line detection range is expanded by lane line prediction, and the accuracy of the detected lane line point cloud is improved by combining coarse and fine extraction.

[0285] In some embodiments, the apparatus may further include: a point cloud processing module, which may include:

[0286] The point cloud segmentation submodule is used to extract points located within a 2D preview box from the original point cloud. Before obtaining the third point cloud, the original point cloud is segmented, and points whose corresponding position coordinates do not meet the preset first coordinate condition are removed, thus obtaining the first point cloud.

[0287] The 3D detection result acquisition submodule is used to acquire the 3D target detection results obtained from the original point cloud.

[0288] The dynamic point cloud removal submodule is used to remove dynamic points from the first point cloud based on the 3D object detection results, thus obtaining the second point cloud.

[0289] The first coordinate condition includes:

[0290] x min ≤x0≤x max

[0291] y min ≤y0≤y max

[0292] z min ≤z0≤z max

[0293] In the formula, x min This represents the minimum coordinate value of the X-axis of the set lidar center coordinate system. max This represents the maximum coordinate value of the X-axis of the set lidar center coordinate system, y min This represents the minimum Y-axis coordinate value of the set lidar center coordinate system. max This represents the maximum Y-axis coordinate value of the set lidar center coordinate system, z.min This represents the minimum Z-axis coordinate value of the set lidar center coordinate system. max This represents the maximum coordinate value of the Z-axis in the central coordinate system of the lidar, x0 represents the coordinate value of the point cloud on the X-axis in the central coordinate system of the lidar, y0 represents the coordinate value of the point cloud on the Y-axis in the central coordinate system of the lidar, and z0 represents the coordinate value of the point cloud on the Z-axis in the central coordinate system of the lidar.

[0294] The coarse extraction module 406 can be used for:

[0295] Extract the point cloud located within the two-dimensional preview box from the second point cloud to obtain the third point cloud.

[0296] In some embodiments, the dynamic point cloud culling submodule is used for:

[0297] The 3D target detection result includes 3D bounding box information, which includes the 3D coordinates, pose information, and size information of the center point of the 3D bounding box. Based on the 3D coordinates and pose information of the center point, the position coordinates of the first point cloud are transformed to the coordinate system of the center point of the 3D bounding box.

[0298] Determine the second coordinate condition based on the dimensional information.

[0299] In the first point cloud, those points whose position coordinates in the coordinate system of the center point of the 3D detection box satisfy the second coordinate condition are removed to obtain the second point cloud.

[0300] The second coordinate condition is as follows:

[0301]

[0302]

[0303]

[0304] Where L3 represents the length of the 3D detection bounding box included in the size information, W3 represents the width of the 3D detection bounding box included in the size information, and H3 represents the height of the 3D detection bounding box included in the size information. This represents the X-axis coordinate value in the coordinate system of the center point of the 3D detection box. This represents the Y-axis coordinate value in the coordinate system of the center point of the 3D detection box. This represents the Z-axis coordinate value in the coordinate system of the center point of the 3D detection frame.

[0305] In some embodiments, the lane prediction module 403 may include:

[0306] The lane width prediction submodule is used to predict the lane width based on the Y-axis coordinate value of the actual lane line in the vehicle's rear axle center coordinate system, as contained in the lane line information, thus obtaining the predicted lane width value.

[0307] The lane line endpoint prediction submodule is used to determine the position coordinates of the second closest lane line point and the second farthest lane line point in the predicted lane line, based on the lane width prediction value and the position coordinates of the first closest lane line point and the first farthest lane line point in the actual lane line information, which are closest to the center of the vehicle's rear axle.

[0308] The lane line inclination prediction submodule is used to determine the inclination angle of the predicted lane line from the inclination angle of the actual lane line contained in the lane line information.

[0309] The lane line width prediction submodule is used to determine the predicted lane line width from the actual lane line width contained in the lane line information.

[0310] The fitting equation prediction submodule is used to determine the fitting equation for the predicted lane line based on the lane line inclination angle.

[0311] The predicted lane line drawing submodule is used to draw lines based on the position coordinates of the second nearest lane line point, the position coordinates of the second farthest lane line point, the inclination angle of the predicted lane line, the line width of the predicted lane line, and the fitting equation of the predicted lane line, so as to obtain the predicted lane line.

[0312] In some embodiments, the lane line endpoint prediction submodule can be used for:

[0313] The X-axis coordinate value of the first nearest lane line point is used as the X-axis coordinate value of the second nearest lane line point.

[0314] Based on the Y-axis coordinates of the first nearest lane line point and the predicted lane width, determine the Y-axis coordinates of the second nearest lane line point that is closest to the center of the vehicle's rear axle within the predicted lane line.

[0315] The X-axis coordinate value of the first farthest lane line point is used as the X-axis coordinate value of the second farthest lane line point.

[0316] Based on the Y-axis coordinates of the first farthest lane line point and the predicted lane width, determine the Y-axis coordinates of the second farthest lane line point that is closest to the center of the vehicle's rear axle in the predicted lane line.

[0317] In some embodiments, the discrete module 404 can be used for:

[0318] Obtain the preset first ratio value.

[0319] Based on the first ratio value, the X-axis coordinates of the lane line point closest to the center of the vehicle's rear axle, and the X-axis coordinates of the lane line point farthest from the center of the vehicle's rear axle, the upper limit of the X-axis coordinates of the discrete points corresponding to the lane line is calculated according to a preset discrete range calculation formula.

[0320] The X-axis coordinate of the lane line point closest to the center of the vehicle's rear axle is used as the lower limit of the X-axis coordinate of the discrete point corresponding to the lane line.

[0321] The interval formed by the upper and lower limits of the X-axis coordinates is used as the range of X-axis coordinate values ​​for the discrete points corresponding to the lane lines.

[0322] Based on the preset fitting equations for the discretization step length and lane lines, the lines within the X-axis coordinate range of the lane lines are discretized to obtain the discrete points corresponding to the lane lines.

[0323] The formula for calculating the discrete range is as follows:

[0324]

[0325] In the formula, This represents the upper limit of the X-axis coordinate of the discrete point corresponding to lane line i. This represents the X-axis coordinate of the point on lane line i that is closest to the center of the vehicle's rear axle. The X-axis coordinate of lane line i represents the point on lane line i that is farthest from the center of the rear axle of the vehicle, and α represents the first proportional value.

[0326] In some embodiments, the pre-aiming frame determination module 405 may be used for:

[0327] The position coordinates of the discrete points are used as the center coordinates of their corresponding two-dimensional preview boxes.

[0328] Use the distance from the step size as the length of the two-dimensional preview box.

[0329] The width of the two-dimensional preview box is obtained by multiplying the preset lane width expansion coefficient by the width of the lane line to which the discrete point belongs.

[0330] Generate a rotation matrix for a two-dimensional preview frame based on the inclination angle of the lane line to which the discrete point belongs.

[0331] Assign a two-dimensional preview frame to a discrete point, along with its center coordinates, length, width, and rotation matrix.

[0332] In some embodiments, the coarse extraction module 406 can be used for:

[0333] For each point in the second point cloud, based on the center coordinates of the two-dimensional preview bounding box and the rotation matrix, the point's position coordinates are transformed to the center coordinate system of the two-dimensional preview bounding box.

[0334] The third coordinate condition is determined based on the length and width of the two-dimensional preview frame.

[0335] Extract the points from the second point cloud whose transformed coordinates meet the third coordinate condition.

[0336] The point cloud composed of the extracted points is used as the third point cloud.

[0337] The third coordinate conditions include:

[0338]

[0339]

[0340] In the formula, L2 represents the length of the two-dimensional preview box, and W2 represents the width of the two-dimensional preview box. This represents the X-axis coordinate of the second point cloud in the center coordinate system of the two-dimensional preview box. This represents the Y-axis coordinate value of the second point cloud in the center coordinate system of the two-dimensional preview frame.

[0341] In some embodiments, the fine extraction module 407 may include:

[0342] The point information determination submodule is used to determine the pitch angle, horizontal azimuth angle, and reflection intensity characterization value for each point in the third point cloud.

[0343] The point cloud segmentation submodule is used to divide the third point cloud into multiple point cloud clusters based on the pitch angle of the points. Within the same point cloud cluster, the difference in pitch angle between any two points is less than a preset difference threshold.

[0344] The sorting submodule is used to arrange the points contained in each point cloud cluster in ascending order of horizontal azimuth angle.

[0345] The edge detection submodule is used to determine the matching rising and falling edges in the point cloud cluster based on the reflection intensity characterization value of each point in the point cloud cluster. The rising edge consists of multiple points whose reflection intensity characterization values ​​show a continuous upward trend, and the falling edge consists of multiple points whose reflection intensity characterization values ​​show a continuous downward trend.

[0346] The lane line point extraction submodule is used to extract points that meet preset extraction conditions from the matching rising and falling edges as lane line points, provided that the matching rising and falling edges meet preset lane line conditions.

[0347] The lane line point cloud determination submodule is used to extract lane line points from point cloud clusters to form a point cloud as the lane line point cloud in the original point cloud.

[0348] Lane line conditions include:

[0349] The difference between the Y-axis coordinate value at the end of the rising edge and the Y-axis coordinate value at the end of the falling edge is less than a preset lane width threshold, and the difference between the reflection intensity values ​​at the start and end of the falling edge is greater than a preset lane reflection intensity difference threshold.

[0350] Extraction criteria include:

[0351] The reflection intensity characterization value is the highest, and the height value is less than the preset height threshold.

[0352] In some embodiments, the point information determination submodule can be used for:

[0353] After determining the points in the third point cloud and arranging them in ascending order of horizontal azimuth, the points located within the neighborhood of each point are...

[0354] Calculate the mean reflection intensity of the point and its neighborhood.

[0355] The mean value of the reflection intensity is determined as the characterization value of the reflection intensity at a point.

[0356] The lane line point cloud map construction device provided in this application embodiment can implement the various processes described in the above method embodiments. To avoid repetition, it will not be described again here.

[0357] Figure 5 A schematic diagram of the hardware structure of the electronic device provided in an embodiment of this application is shown.

[0358] The electronic device may include a processor 501 and a memory 502 storing computer program instructions.

[0359] Specifically, the processor 501 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0360] Memory 502 may include mass storage for data or instructions. For example, and not limitingly, memory 502 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 502 may include removable or non-removable (or fixed) media. Where appropriate, memory 502 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 502 is non-volatile solid-state memory.

[0361] Memory 502 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, typically, memory 502 includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it can perform the operations described in any of the lane line point cloud map construction methods in the above embodiments.

[0362] The processor 501 reads and executes computer program instructions stored in the memory 502 to implement any of the lane line point cloud map construction methods in the above embodiments.

[0363] In one example, the electronic device may also include a communication interface 503 and a bus 510. For example, Figure 5 As shown, the processor 501, memory 502, and communication interface 503 are connected through bus 510 and complete communication with each other.

[0364] The communication interface 503 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0365] Bus 510 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 510 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0366] Furthermore, in conjunction with the lane line point cloud map construction method in the above embodiments, this application embodiment can provide a computer storage medium for implementation. This computer storage medium stores computer program instructions, which, when executed by a processor, implement any of the lane line point cloud map construction methods in the above embodiments.

[0367] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0368] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0369] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0370] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0371] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for constructing a lane line point cloud map, characterized in that, include: Obtain positioning information, at least one frame of raw point cloud data acquired by LiDAR, and lane line detection results from the image. The positioning information, at least one frame of the original point cloud, and the lane line detection results of the image are synchronized in time. For each frame of the original point cloud, assuming time synchronization is complete and the corresponding image lane detection result contains lane information of at least one actual lane line, lane prediction is performed based on the lane information of the actual lane line to obtain the predicted lane line. For each lane line in the actual lane lines and the predicted lane lines, the lane line is discretized to obtain the discrete points corresponding to the lane line. For each of the discrete points, a two-dimensional preview bounding box is assigned to the discrete point. Extract the points within the two-dimensional preview box from the original point cloud to obtain the third point cloud. Point cloud extraction is performed on the third point cloud based on the peak characteristics of point cloud reflection intensity to obtain the lane line point cloud from the original point cloud. A corresponding lane line point cloud map is generated based on the positioning information and the lane line point cloud in at least one frame of the original point cloud. The step of predicting lane lines based on the lane line information of the actual lane lines to obtain predicted lane lines includes: The lane width is predicted based on the Y-axis coordinate value of the actual lane line in the vehicle's rear axle center coordinate system, as contained in the lane line information. Based on the predicted lane width value and the coordinates of the first closest lane line point closest to the vehicle's rear axle center and the first farthest lane line point farthest from the vehicle's rear axle center in the actual lane lines contained in the lane line information, the coordinates of the second closest lane line point closest to the vehicle's rear axle center and the second farthest lane line point farthest from the vehicle's rear axle center in the predicted lane lines are determined. The inclination angle of the actual lane line contained in the lane line information is determined as the inclination angle of the predicted lane line. The actual lane width contained in the lane line information is determined as the predicted lane width. The fitting equation for the predicted lane line is determined based on the inclination angle of the predicted lane line. The predicted lane line is obtained by drawing a line based on the position coordinates of the second nearest lane line point, the position coordinates of the second farthest lane line point, the inclination angle of the predicted lane line, the line width of the predicted lane line, and the fitting equation of the predicted lane line.

2. The method according to claim 1, characterized in that, Before extracting the point cloud located within the two-dimensional preview box from the original point cloud to obtain the third point cloud, the method further includes: The original point cloud is segmented, and point clouds whose corresponding position coordinates do not meet the preset first coordinate condition are removed to obtain the first point cloud. Obtain the 3D target detection results obtained by performing 3D target detection based on the original point cloud. Based on the three-dimensional target detection results, dynamic point clouds in the first point cloud are removed to obtain the second point cloud. The step of extracting the point cloud located within the two-dimensional preview box from the original point cloud to obtain the third point cloud includes: Extract the point cloud within the two-dimensional preview frame from the second point cloud to obtain the third point cloud. The first coordinate condition includes: In the formula, This represents the minimum coordinate value of the X-axis of the set lidar center coordinate system. This indicates the maximum X-axis coordinate value of the set lidar center coordinate system. This represents the minimum Y-axis coordinate value of the set lidar center coordinate system. This indicates the maximum Y-axis coordinate value of the set lidar center coordinate system. This represents the minimum coordinate value of the Z-axis in the set lidar center coordinate system. This indicates the maximum Z-axis coordinate value of the set lidar center coordinate system. This represents the coordinate value of the point cloud on the X-axis of the lidar center coordinate system. This represents the Y-axis coordinate value of the point cloud in the central coordinate system of the lidar. This represents the coordinate value of the point cloud on the Z-axis of the lidar center coordinate system.

3. The method according to claim 2, characterized in that, The step of removing dynamic point clouds from the first point cloud based on the three-dimensional target detection results to obtain a second point cloud includes: The 3D target detection result includes 3D detection box information, which includes the 3D coordinates of the center point of the 3D detection box, its pose information, and its size information. Based on the 3D coordinates of the center point and the pose information, the position coordinates of the first point cloud are transformed to the coordinate system of the center point of the 3D detection box. The second coordinate condition is determined based on the stated dimension information. The point cloud containing points whose position coordinates in the coordinate system of the center point of the 3D detection box satisfy the second coordinate condition is removed to obtain the second point cloud. The second coordinate condition is as follows: in, This indicates the length of the 3D detection frame included in the size information. This indicates the width of the 3D detection frame included in the size information. This indicates the height of the 3D detection frame included in the size information. This represents the X-axis coordinate value in the coordinate system of the center point of the three-dimensional detection frame. This represents the Y-axis coordinate value in the coordinate system of the center point of the three-dimensional detection frame. This represents the Z-axis coordinate value in the coordinate system of the center point of the three-dimensional detection frame.

4. The method according to claim 1, characterized in that, The step of determining the coordinates of the second closest lane line point and the second farthest lane line point in the predicted lane line based on the lane width prediction value and the coordinates of the first closest lane line point and the first farthest lane line point in the actual lane line that are closest to the rear axle center of the vehicle, as contained in the lane line information, includes: The X-axis coordinate value of the first nearest lane line point is determined as the X-axis coordinate value of the second nearest lane line point. Based on the Y-axis coordinate value of the first nearest lane line point and the predicted lane width value, determine the Y-axis coordinate value of the second nearest lane line point that is closest to the center of the vehicle's rear axle in the predicted lane line. The X-axis coordinate value of the first farthest lane line point is determined as the X-axis coordinate value of the second farthest lane line point. Based on the Y-axis coordinate value of the first farthest lane line point and the lane width prediction value, determine the Y-axis coordinate value of the second farthest lane line point that is closest to the center of the vehicle's rear axle in the predicted lane line.

5. The method according to claim 1, characterized in that, The discretization process of the lane lines to obtain the discrete points corresponding to the lane lines includes: Obtain the preset first ratio value. Based on the first ratio value, the X-axis coordinate value of the lane line point closest to the center of the vehicle's rear axle, and the X-axis coordinate value of the lane line point farthest from the center of the vehicle's rear axle, the upper limit value of the X-axis coordinate of the discrete point corresponding to the lane line is calculated according to a preset discrete range calculation formula. The X-axis coordinate value of the lane line point closest to the center of the vehicle's rear axle is used as the lower limit of the X-axis coordinate value of the discrete point corresponding to the lane line. The interval formed by the upper limit and lower limit of the X-axis coordinate is taken as the range of X-axis coordinate values ​​for the discrete points corresponding to the lane line. Based on the preset discretization step length and the fitting equation of the lane line, the lines within the X-axis coordinate range of the lane line are discretized to obtain the discrete points corresponding to the lane line. The formula for calculating the discrete range is as follows: In the formula, Lane lines The upper limit of the X-axis coordinate of the corresponding discrete point. Lane lines The X-axis coordinate of the lane line point closest to the center of the vehicle's rear axle. Lane lines The X-axis coordinate of the lane line point furthest from the center of the vehicle's rear axle. This represents the first proportional value.

6. The method according to claim 5, characterized in that, Assigning a two-dimensional preview bounding box to the discrete points includes: The position coordinates of the discrete points are used as the center coordinates of their corresponding two-dimensional preview boxes. The distance step size is used as the length of the two-dimensional preview frame. The width of the two-dimensional preview frame is obtained by multiplying the preset lane width expansion coefficient by the width of the lane line to which the discrete point belongs. The rotation matrix of the two-dimensional preview frame is generated based on the inclination angle of the lane line to which the discrete point belongs. Assign a two-dimensional preview frame to the discrete point, corresponding to the center coordinates, the length of the two-dimensional preview frame, the width of the two-dimensional preview frame, and the rotation matrix of the two-dimensional preview frame.

7. The method according to claim 2, characterized in that, The step of extracting the third point cloud from the second point cloud based on the two-dimensional preview bounding box includes: For each point in the second point cloud, based on the center coordinates and rotation matrix of the two-dimensional preview frame, the position coordinates of the point are transformed to the center coordinate system of the two-dimensional preview frame. The third coordinate condition is determined based on the length and width of the two-dimensional preview frame. Extract the points from the second point cloud whose transformed position coordinates meet the third coordinate condition. The point cloud composed of the extracted points is used as the third point cloud. The third coordinate condition includes: In the formula, This indicates the length of the two-dimensional preview frame. This represents the width of the two-dimensional preview frame. This represents the X-axis coordinate value of the second point cloud in the center coordinate system of the two-dimensional preview frame. This represents the Y-axis coordinate value of the second point cloud in the center coordinate system of the two-dimensional preview frame.

8. The method according to claim 7, characterized in that, The step of extracting the lane line point cloud from the third point cloud based on the peak characteristics of the point cloud reflection intensity to obtain the lane line point cloud in the original point cloud includes: For each point in the third point cloud, determine the pitch angle, horizontal azimuth angle, and reflection intensity characterization value of that point. Based on the pitch angle of the points, the third point cloud is divided into multiple point cloud clusters, wherein the difference in pitch angle between any two points in the same point cloud cluster is less than a preset difference threshold. For each point cloud cluster, the points contained in the point cloud cluster are arranged in ascending order of horizontal azimuth angle. Based on the reflection intensity characterization value of each point in the point cloud cluster, a matching rising edge and falling edge are determined in the point cloud cluster. The rising edge consists of multiple points whose reflection intensity characterization values ​​show a continuous upward trend, and the falling edge consists of multiple points whose reflection intensity characterization values ​​show a continuous downward trend. If the matching rising and falling edges meet preset lane line conditions, points that meet preset extraction conditions are extracted from the matching rising and falling edges as lane line points. The point cloud composed of lane line points extracted from the point cloud cluster is used as the lane line point cloud in the original point cloud. The lane line conditions include: The difference between the Y-axis coordinate value at the end of the rising edge and the Y-axis coordinate value at the end of the falling edge is less than a preset lane width threshold, and the difference between the reflection intensity values ​​at the start and end of the falling edge is greater than a preset lane reflection intensity difference threshold. The extraction conditions include: The reflection intensity characterization value is the highest, and the height value is less than the preset height threshold.

9. The method according to claim 8, characterized in that, The determination of the reflection intensity characterization value of the point includes: After determining that the points in the third point cloud are arranged in ascending order of horizontal azimuth angle, the points located within the neighborhood of the given points are... Calculate the average reflection intensity of the point and the points within its neighborhood. The average reflection intensity is determined as the reflection intensity characterization value of the point.

10. A lane line point cloud map construction device, characterized in that, include: The acquisition module is used to acquire positioning information, at least one frame of raw point cloud data collected by the LiDAR, and lane line detection results from the image. The synchronization module is used to synchronize the positioning information, at least one frame of the original point cloud, and the image lane line detection results in time. The lane prediction module is used to predict lane lines for each frame of the original point cloud, provided that time synchronization is completed and the corresponding image lane detection result contains lane line information of at least one actual lane line. The discretization module is used to discretize each lane line, both the actual lane line and the predicted lane line, to obtain discrete points corresponding to the lane line. The preview bounding box determination module is used to assign a two-dimensional preview bounding box to each of the discrete points. The coarse extraction module is used to extract the point cloud located within the two-dimensional preview box from the original point cloud to obtain the third point cloud. The fine extraction module is used to extract points from the third point cloud based on the peak features of the point cloud reflection intensity, thereby obtaining the lane line point cloud from the original point cloud. The map generation module is used to generate a corresponding lane line point cloud map based on the positioning information and the lane line point cloud in at least one frame of the original point cloud. The lane prediction module includes: The lane width prediction submodule is used to predict the lane width based on the Y-axis coordinate value of the actual lane line in the vehicle's rear axle center coordinate system, as contained in the lane line information, thus obtaining the predicted lane width value. The lane line endpoint prediction submodule is used to determine the position coordinates of the second closest lane line point and the second farthest lane line point in the predicted lane line, based on the lane width prediction value and the position coordinates of the first closest lane line point and the first farthest lane line point in the actual lane line information, which are closest to the center of the vehicle's rear axle. The lane line inclination prediction submodule is used to determine the inclination angle of the predicted lane line from the inclination angle of the actual lane line contained in the lane line information. The lane line width prediction submodule is used to determine the predicted lane line width from the actual lane line width contained in the lane line information. The fitting equation prediction submodule is used to determine the fitting equation for the predicted lane line based on the lane line inclination angle. The predicted lane line drawing submodule is used to draw lines based on the position coordinates of the second nearest lane line point, the position coordinates of the second farthest lane line point, the inclination angle of the predicted lane line, the line width of the predicted lane line, and the fitting equation of the predicted lane line, so as to obtain the predicted lane line.

11. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions. When the processor executes the computer program instructions, it implements the lane line point cloud map construction method as described in any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the lane line point cloud map construction method as described in any one of claims 1-9.

13. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device performs the lane line point cloud map construction method as described in any one of claims 1-9.

Images