Lane line detection error determination method, device, equipment and readable storage medium

Abstract

The application discloses a lane line detection error determination method, device and equipment and a readable storage medium. The method comprises the following steps: converting a lane line artificial labeling set in an image by using an inverse perspective transformation matrix to obtain a first set; filtering a first lane line of a target lane from the first set according to a high-precision map, and obtaining a corresponding second lane line from the artificial labeling set; each first lane line of the same target lane is a straight line; optimizing the inverse perspective transformation matrix until each third lane line of the same lane obtained by converting the second lane line by using the optimized matrix is parallel to each other; converting the artificial labeling set and a model detection set by using the optimized matrix; and determining a corresponding lane line detection error by using the converted set. According to the technical scheme disclosed by the application, the optimal inverse perspective transformation matrix is found by using the high-precision map, the lane line detection error is determined based on the conversion of the set by using the matrix, and therefore the accuracy of the lane line detection error determination is improved.

Description

Technical Field

[0001] This application relates to the field of lane line detection technology, and in particular to a method, apparatus, device and readable storage medium for determining lane line detection error. Background Technology

[0002] Lane detection plays a crucial role in advanced driver assistance systems (ADAS) and autonomous driving. Using neural networks to detect lane lines has become a popular technology in recent years. To improve the accuracy of lane detection using neural networks, it is necessary to determine the lane detection error.

[0003] Currently, the process for determining 2D lane detection error involves: acquiring lane lines from manually annotated 2D images (representing the ground truth), and obtaining lane line points detected and annotated in the 2D images using AI (Artificial Intelligence) methods. Then, the average deviation between the AI-detected lane line points and the manually annotated ground truth points in the same image is calculated; that is, the deviation between the manually annotated ground truth lane line points and the model detection results in the image coordinate system is calculated. Because the camera imaging model in the image coordinate system exhibits near-object size reduction, calculating lane line error using the above method leads to significant variations in statistical error, failing to accurately reflect the lane detection error situation.

[0004] In conclusion, improving the accuracy of lane line detection error determination is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a method, apparatus, device and readable storage medium for determining lane line detection errors, so as to improve the accuracy of lane line detection error determination.

[0006] To achieve the above objectives, this application provides the following technical solution:

[0007] A method for determining lane line detection error includes: acquiring a set of manually labeled lane lines and a set of model detection lines in an image; transforming the manually labeled set to the BEV plane using an inverse perspective transformation matrix to obtain a first set; selecting a first lane line of a target lane from the first set based on a high-precision map, and obtaining a second lane line corresponding to the first lane line from the manually labeled set; each first lane line of the same target lane is a straight line; optimizing the inverse perspective transformation matrix until each third lane line of the same lane obtained after transforming the second lane line using the optimized inverse perspective transformation matrix is ​​parallel to each other; transforming the manually labeled set and the model detection set using the optimized inverse perspective transformation matrix to obtain corresponding second and third sets; and determining the detection error of the corresponding lane line using the second and third sets.

[0008] Optionally, selecting the first lane line of the target lane from the first set based on the high-precision map includes: obtaining the vehicle coordinates at the time of image acquisition; obtaining map tiles from the high-precision map within a target distance range from the vehicle coordinates based on the vehicle coordinates; converting the map tiles to the coordinate system of the first set; obtaining a map lane line set from the converted map tiles; the map lane line set includes the fourth lane line of the target map lane, and the fourth lane lines of the same target map lane are parallel to each other and are straight lines; selecting the first lane line of the target lane from the first set based on the map lane line set.

[0009] Optionally, selecting the first lane line of the target lane from the first set based on the map lane line set includes: determining whether there exists a fourth lane line in the map lane line set whose distance to each marker point of the current lane line in the first set is less than a first distance threshold; if not, determining the next lane line in the first set as the current lane line, and returning to execute the step of determining whether there exists a fourth lane line in the map lane line set whose distance to each marker point of the current lane line in the first set is less than the first distance threshold, until the last lane line in the first set is determined as the current lane line and the determination is completed; if yes, determining the current lane line as a candidate lane line, and returning to execute the step of determining the next lane line in the first set as the current lane line; and determining the candidate lane lines belonging to the same lane among the candidate lane lines as the first lane line of the target lane.

[0010] Optionally, determining the detection error of the corresponding lane line using the second set and the third set includes: determining the matching relationship between lane lines in the second set and lane lines in the third set; calculating the distance between each detection point on each lane line in the third set and the matching lane line in the second set; and obtaining the detection error of the corresponding lane line based on the distance between each detection point on each lane line in the third set and the matching lane line in the second set.

[0011] Optionally, determining the matching relationship between lane lines in the second set and lane lines in the third set includes: calculating the distance between multiple detection points on each lane line in the third set and each lane line in the second set; and determining lane lines in the second set whose distances to multiple detection points on lane lines in the third set are all less than a second distance threshold as lane lines that match the corresponding lane lines in the third set.

[0012] Optionally, optimizing the inverse perspective transformation matrix includes: optimizing the inverse perspective transformation matrix using the Gauss-Newton method.

[0013] Optionally, it also includes: outputting the detection error of the corresponding lane line.

[0014] A lane line detection error determination device includes: an acquisition module for acquiring a set of manually labeled lane lines and a set of model detection lines in an image; a first conversion module for converting the manually labeled set to a BEV plane using an inverse perspective transformation matrix to obtain a first set; a filtering module for filtering first lane lines of a target lane from the first set based on a high-precision map, and obtaining second lane lines corresponding to the first lane lines from the manually labeled set; each first lane line of the same target lane is a straight line; an optimization module for optimizing the inverse perspective transformation matrix until each third lane line of the same lane obtained after converting the second lane lines using the optimized inverse perspective transformation matrix is ​​parallel to each other; a second conversion module for converting the manually labeled set and the model detection set using the optimized inverse perspective transformation matrix to obtain corresponding second and third sets; and a determination module for determining the detection error of the corresponding lane lines using the second and third sets.

[0015] A lane line detection error determination device includes: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the lane line detection error determination method as described above.

[0016] A readable storage medium storing a computer program that, when executed by a processor, implements the steps of the lane detection error determination method as described in any of the preceding claims.

[0017] This application provides a method, apparatus, device, and readable storage medium for determining lane line detection errors. The method includes: acquiring a set of manually labeled lane lines and a set of model-detected lane lines in an image; transforming the manually labeled set to the BEV plane using an inverse perspective transformation matrix to obtain a first set; selecting the first lane line of the target lane from the first set based on a high-precision map, and obtaining the second lane line corresponding to the first lane line from the manually labeled set; ensuring that each first lane line of the same target lane is a straight line; optimizing the inverse perspective transformation matrix until the third lane lines of the same lane obtained after transforming the second lane lines using the optimized inverse perspective transformation matrix are parallel to each other; transforming the manually labeled set and the model-detected lane line using the optimized inverse perspective transformation matrix to obtain corresponding second and third sets; and determining the detection error of the corresponding lane line using the second and third sets.

[0018] The technical solution disclosed in this application utilizes an inverse perspective transformation matrix to transform the manually labeled set of lane lines in an image to the BEV plane, obtaining a first set. Then, based on a high-precision map, the first lane line of the target lane is selected from the first set, wherein the first lane lines of the same target lane are all straight lines. The second lane line corresponding to the first lane line is then selected from the manually labeled set. Next, the inverse perspective transformation matrix is ​​optimized until the third lane lines of the same lane obtained after transforming the second lane lines using the optimized inverse perspective transformation matrix are parallel to each other. This optimization and correction of the inverse perspective transformation matrix is ​​based on a high-precision map, finding the optimal inverse perspective transformation matrix. Finally, the optimized inverse perspective transformation matrix is ​​used to transform the manually labeled set and the model detection set of lane lines in the image. The resulting second and third sets are used to determine the lane line detection error. By first finding the optimal inverse perspective transformation matrix using high-precision maps and manually labeled lane line data, and then transforming the manually labeled lane line set and model detection set in the image based on the optimal inverse perspective transformation matrix, and finally determining the detection error of the corresponding lane line based on the transformed result, the problem of inaccurate view transformation caused by introducing a planar assumption when using only inverse perspective transformation technology can be avoided, thereby improving the accuracy of lane line detection error determination.

[0019] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0020] Figure 1 This is a schematic diagram illustrating manual annotation and AI model detection in 2D images;

[0021] Figure 2 A flowchart illustrating a method for determining lane line detection error provided in an embodiment of this application;

[0022] Figure 3 A flowchart of another method for determining lane line detection error is provided for embodiments of this application;

[0023] Figure 4 This is a schematic diagram of the structure of a lane line detection error determination device provided in an embodiment of this application;

[0024] Figure 5 This is a schematic diagram of a lane line detection error determination device provided in an embodiment of this application. Detailed Implementation

[0025] To improve the accuracy of lane line detection by neural networks, it is necessary to detect and determine the error between manually marked lane lines and lane lines detected by neural networks. This allows for adjustments to the neural network parameters based on the error, thereby improving the accuracy of lane line detection by the application network.

[0026] Currently, the method for determining 2D lane line detection error is as follows: acquire 2D images, obtain lane lines from manually annotated 2D images, and obtain lane line points detected in the aforementioned 2D images using AI methods, specifically as follows: Figure 1 The diagram illustrates manual annotation and AI model detection in a 2D image. Dark dots at the ends of the lane lines represent manual annotations, while light-colored squares on the sides represent AI detection results. The distance between lane line points detected by AI and those annotated manually in the 2D image is then calculated, and the lane line detection error is determined based on this distance. In other words, the current calculation focuses on the deviation between manually annotated and model-detected lane line points in the pixel coordinate system. However, as... Figure 1 As shown, since the camera imaging model in the image coordinate system is that objects appear larger when they are closer and smaller when they are farther away, calculating lane line errors using the above method will result in large variations in statistical errors, which cannot reflect the error situation in the real world.

[0027] Therefore, this application provides a method, apparatus, device, and readable storage medium for determining lane line detection errors, in order to improve the accuracy of lane line detection error determination.

[0028] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0029] See Figure 2 The diagram illustrates a flowchart of a lane line detection error determination method provided in an embodiment of this application. The lane line detection error determination method provided in this application may include:

[0030] S11: Obtain the set of manually annotated lane lines and the set of model detection in the image.

[0031] Cameras mounted on the vehicle can capture images of the road surface. Lane lines in the images can be manually labeled, resulting in a set of manually labeled images. To improve the accuracy of lane line detection error determination, the manual labeling error cannot be too large; that is, the accuracy of the manual labeling needs to be guaranteed. Alternatively, a model can be used to detect lane lines in the images, resulting in a set of model-detected images.

[0032] Accordingly, processing devices such as processors with data processing capabilities can acquire a set of manually labeled lane lines in an image, and a set of model-detected lane lines in the same image, obtained by manually annotating lane lines. The lane lines in the image include solid lane lines and / or dashed lane lines. Solid lane lines primarily constrain the longitudinal direction and the left-right direction of the vehicle body, while dashed lane lines primarily constrain the lateral direction and the front-rear direction of the vehicle body.

[0033] The set of manually labeled points contains the set of manually labeled points corresponding to each lane line in the image. Each set contains the coordinates of multiple labeled points (i.e., manually labeled points) for the corresponding lane line in the image coordinate system. For a solid lane line, its corresponding set of manually labeled points can be represented as: Line2d = {annotation_line_0, annotation_line_1, ..., annotation_line_m}, where annotation_line_m represents the coordinates of the m-th manually labeled point on that solid lane line in the image coordinate system. For a virtual lane line, its corresponding set of manually annotated points can be represented as: Dash2d={(annotation_dash_start0,annotation_dash_end0),(annotation_dash_start1,annotation_dash_end1),...,(annotation_dash_start_t,annotation_dash_end_t)}, where annotation_dash_start_t represents the coordinates of the starting point of the manually annotated segment t of the virtual lane line in the image coordinate system, and annotation_dash_end_t represents the coordinates of the ending point of the manually annotated segment t of the virtual lane line in the image coordinate system.

[0034] The model detection set contains the model detection subsets corresponding to each lane line in the image, and the model detection point set contains the coordinates of multiple detection points (i.e. model detection points) of the corresponding lane line in the image coordinate system.

[0035] It should be noted that for the manually labeled dataset, it is known which lane lines belong to the same lane. Similarly, for the model-detected dataset, it is also known which lane lines belong to the same lane. Furthermore, the processing device may include a hard drive or other storage device to store relevant data from both the manually labeled and model-detected datasets, facilitating the determination of lane line detection errors based on the stored data.

[0036] S12: Use the inverse perspective transformation matrix to transform the manually labeled set to the BEV plane to obtain the first set.

[0037] After obtaining the manually labeled set of lane lines in the image, an inverse perspective transformation technique can be used to transform the manually labeled set to the BEV (Bird's Eye View) plane to obtain the corresponding first set. The first set in the BEV plane can be a set in a coordinate system centered on the vehicle (the vehicle from which the image was captured), or it can be a set in the world coordinate system. In other words, the inverse perspective transformation technique can be used to transform the lane line annotations in the manually labeled set from the image coordinate system to the lane line annotations in a coordinate system centered on the vehicle or in the world coordinate system. Among them, the inverse perspective transformation technology is based on the imaging projection model and the plane assumption to transform the 2D coordinates in the image to the ground plane coordinates. This technology mainly transforms the coordinates from the image coordinates to the coordinates in the real world by pre-calibrating the extrinsic parameters of the camera (the camera that acquires the above image) and the plane assumption (assuming that each pixel in the image is a projection of the position on the ground onto the pixel plane. Once the mapping relationship between the two planes is determined, each pixel can be mapped to the plane and has three-dimensional coordinates). The coordinates of the points on this plane have a characteristic: the z-axis is the same. It is assumed that the ground plane is flat.

[0038] Specifically, an inverse perspective transformation matrix Mraw can be generated based on the camera mounting parameters on the vehicle body. This matrix includes the camera's height above the ground, the vehicle's roll angle, and its pitch angle, and may also include its yaw angle. Then, the inverse perspective transformation matrix Mraw can be used to transform the coordinates of each annotation point in the manually labeled set from the image coordinate system to the BEV plane, thus obtaining the corresponding first set. During this transformation, the conversion correspondence between the lane lines in the manually labeled set and the first lane line in the first set can be simultaneously determined.

[0039] Specifically, for the set of manual annotations Line2d corresponding to the solid lane lines in the manual annotation set, after being transformed to the BEV plane by the inverse perspective transformation matrix Mraw, it can be represented as Line2d_ground_1={annotation_line_0_bev,annotation_line_1_bev,...,annotation_line_m_bev}; for the set of manual annotations Dash2d corresponding to the dashed lane lines in the manual annotation set, after being transformed to the BEV plane by the inverse perspective transformation matrix Mraw, it can be represented as Dash2d_ground_1={(annotation_dash_start0_bev,annotation_dash_end0_bev),(annotation_dash_start1_bev,annotation_dash_end1_bev),...,(annotation_dash_start_t_bev,annotation_dash_end_t_bev)}.

[0040] The above process transforms the manually labeled set of lane lines in the image to the first set of lane lines under normal viewing angle, thereby converting the coordinate system of the pixel plane into world coordinates or a coordinate system centered on the vehicle body, thus eliminating the camera scale problem.

[0041] S13: Based on the high-precision map, select the first lane line of the target lane from the first set, and obtain the second lane line corresponding to the first lane line from the manually labeled set; each first lane line of the same target lane is a straight line.

[0042] Considering that using only inverse perspective transformation technology introduces the problem of planar assumptions, it will create new issues for non-parallel lanes such as those with slopes and merging points. Therefore, this application optimizes the inverse perspective transformation matrix based on high-precision maps to solve the problems existing in using only inverse perspective transformation technology, thereby improving the accuracy of lane line detection error determination. High-precision maps are thematic maps serving autonomous driving systems, as opposed to ordinary navigation electronic maps. High-precision maps can provide lane-level map data, accurately and comprehensively representing road features.

[0043] Specifically, after executing step S12 to obtain the first set, the first lane lines of the target lane can be filtered from the first set based on the high-precision map. Each first lane line of the same target lane is a straight line; that is, the target lane is the lane whose corresponding first lane line is a straight line. The first lane lines of the target lane refer to the corresponding lane lines in the first set obtained by transforming the manually labeled set using inverse perspective transformation technology. The first lane lines of the target lanes filtered from the first set based on the high-precision map can be represented as Line2d_ground_1_filtered and Dash2d_ground_1_filtered.

[0044] Subsequently, based on the conversion correspondence between lane lines in the manually labeled set and the first lane lines in the first set, a second lane line corresponding to the first lane line of the target lane selected from the first set can be obtained from the manually labeled set. In other words, the second lane line can be obtained from the manually labeled set, and there is a conversion correspondence between the obtained second lane line and the first lane line of the target lane selected from the first set. Specifically, the second lane line corresponding to the first lane line obtained from the manually labeled set can be represented as Line2d_filtered and Dash2d_filtered, and there is a conversion correspondence between Line2d_filtered and Line2d_ground_1_filtered, and between Dash2d_filtered and Dash2d_ground_1_filtered.

[0045] S14: Optimize the inverse perspective transformation matrix until the third lane lines of the same lane obtained by transforming the second lane line using the optimized inverse perspective transformation matrix are parallel to each other.

[0046] After steps S11-S13, optimization techniques can be used to optimize the inverse perspective transformation matrix Mraw used in step S11. Specifically, optimization techniques can be used to adjust the parameters contained in the inverse perspective transformation matrix Mraw (such as the camera's height above the ground, the vehicle's roll angle, the vehicle's pitch angle, and the vehicle's yaw angle) to optimize the inverse perspective transformation matrix Mraw. Each optimization of the inverse perspective transformation matrix Mraw yields an optimized inverse perspective transformation matrix Mraw_i, where i represents the i-th optimization, i = 1, 2, 3, ..., and Mraw_i represents the inverse perspective transformation matrix obtained from the i-th optimization.

[0047] After obtaining the optimized inverse perspective transformation matrix Mraw_i, the second lane line can be transformed into the BEV plane to obtain the corresponding third lane line, and it can be determined whether the third lane lines of the same lane are parallel to each other.

[0048] If the third lane lines in the same lane are not parallel, the parameters in the optimized inverse perspective transformation matrix Mraw_i are adjusted to optimize it. The optimized inverse perspective transformation matrix is ​​then used to perform inverse perspective transformation on the second lane lines, and the parallelism of the third lane lines in the same lane is checked. This process continues until the third lane lines in the same lane are parallel. At this point, the final optimized inverse perspective transformation matrix Mopt is obtained. The final optimized inverse perspective transformation matrix Mopt is the optimal inverse perspective transformation matrix.

[0049] If the third lanes in the same lane are parallel to each other, then the final optimized inverse perspective transformation matrix Mopt can be obtained.

[0050] The third lane lines obtained by transforming the second lane lines Line2d_filtered and Dash2d_filtered using the optimized inverse perspective transformation matrix Mraw_i can be represented as Line2d_ground_2_filtered and Dash2d_ground_2_filtered. The third lane lines Line2d_ground_2_filtered and Dash2d_ground_2_filtered obtained after transforming the second lane lines Line2d_filtered and Dash2d_filtered using the final optimized inverse perspective transformation matrix Mopt satisfy the constraint that the third lane lines within the same lane are parallel to each other (meaning that the width of the lane is the same in different places when multiple lane lines within the same lane are parallel). In other words, the third lane lines Line2d_ground_2_filtered and Dash2d_ground_2_filtered within the same lane satisfy the constraint that they are parallel to each other and that the width of the lane is the same in different places.

[0051] The above implementation optimizes the inverse perspective transformation matrix based on the second lane lines obtained from the manually labeled set until the optimized inverse perspective transformation matrix transforms the second lane lines so that the resulting third lane lines of the same lane are parallel to each other. In other words, this process optimizes the inverse perspective transformation matrix using manually labeled 2D lane line data, finding the optimal inverse perspective transformation matrix that satisfies the constraint that multiple lane lines of the same lane are parallel (and in the case of multiple parallel lane lines of the same lane, the width of different parts of the lane is also the same). This solves the problem caused by the planar assumption introduced by using only inverse perspective transformation technology, thus improving the accuracy of lane line detection error determination. Furthermore, since manually labeled lane line data has higher precision, optimizing the inverse perspective transformation matrix based on manually labeled lane line data will improve the accuracy of the final optimized inverse perspective transformation matrix, thereby improving the effect of inverse perspective transformation and further improving the accuracy of lane line detection error determination.

[0052] S15: Use the optimized inverse perspective transformation matrix to transform the manually labeled set and the model detection set to obtain the corresponding second set and third set.

[0053] Based on step S14, the optimized inverse perspective transformation matrix Mopt can be used to transform the manually labeled set to the BEV plane, resulting in the corresponding second set. Furthermore, the optimized inverse perspective transformation matrix Mopt can be used to transform the model detection set to the BEV plane, resulting in the corresponding third set.

[0054] S16: Use the second and third sets to determine the detection error of the corresponding lane lines.

[0055] Then, the detection error of the corresponding lane line is determined by using the second set obtained by converting the manually labeled set and the third set obtained by converting the model detection model.

[0056] The above implementation utilizes the optimized inverse perspective transformation matrix Mopt to perform inverse perspective transformation on the manually labeled set and the model detection set, in order to determine the lane detection error in the BEV plane. This solves both the problem that determining the lane detection error in the image coordinate system cannot reflect the error situation in the real world, and the problem of inaccurate view transformation caused by the plane assumption introduced by only using inverse perspective transformation technology, thereby improving the accuracy of lane detection error determination.

[0057] The technical solution disclosed in this application uses an inverse perspective transformation matrix to transform the manually labeled set of lane lines in an image to the BEV plane to obtain a first set. Then, based on a high-precision map, the first lane lines of the target lane are selected from the first set, where each first lane line of the same target lane is a straight line. The second lane lines corresponding to the first lane lines are then selected from the manually labeled set. Next, the inverse perspective transformation matrix is ​​optimized until the third lane lines of the same lane obtained after transforming the second lane lines using the optimized inverse perspective transformation matrix are parallel to each other. This optimization and correction of the inverse perspective transformation matrix is ​​based on a high-precision map, finding the optimal inverse perspective transformation matrix. Then, the optimized inverse perspective transformation matrix is ​​used to transform the manually labeled set and the model detection set of lane lines in the image. The resulting second and third sets are used to determine the lane line detection error. By first finding the optimal inverse perspective transformation matrix using high-precision maps and manually labeled lane line data, and then using the optimal inverse perspective transformation matrix to transform the manually labeled set and model detection set of lane lines in the image, and then determining the detection error of the corresponding lane lines based on the transformation result, the problem of inaccurate view transformation caused by introducing a planar assumption when using only inverse perspective transformation technology can be avoided, thereby improving the accuracy of lane line detection error determination.

[0058] See Figure 3 The diagram illustrates a flowchart of another lane detection error determination method provided by an embodiment of this application. The lane detection error determination method provided by this application, which filters the first lane line of the target lane from a first set based on a high-precision map, may include:

[0059] Obtain the vehicle coordinates at the moment of image acquisition, and retrieve map tiles within the target distance range from the high-precision map based on the vehicle coordinates;

[0060] Transform the map tiles to the coordinate system of the first set, and obtain the map lane line set from the transformed map tiles; the map lane line set contains the fourth lane line of the target map lane, and the fourth lane lines of the same target map lane are parallel to each other and are straight lines;

[0061] The first lane line of the target lane is selected from the first set of lane lines based on the map lane line set.

[0062] In this embodiment of the application, the process of selecting the first lane line of the target lane from the first set based on the high-precision map can specifically be as follows:

[0063] Step S131: Obtain the vehicle coordinates at the time of image acquisition (the vehicle coordinates are specifically latitude and longitude orientation information). Then, based on the vehicle coordinates at the time of image acquisition, map tiles within the target distance range of the vehicle coordinates can be obtained from the high-precision map. That is, the map tiles closest to the vehicle coordinates can be obtained from the high-precision map. The size of the target distance range can be obtained based on experience or in advance through calibration, etc.

[0064] Step S132: Transform the acquired map tiles to the coordinate system of the first set, so that the first set and the map tiles are in the same coordinate system. For example, if the first set is in a coordinate system centered on the vehicle, then the acquired map tiles are transformed to the coordinate system centered on the vehicle; if the first set is in the world coordinate system, then the acquired map tiles are transformed to the world coordinate system. For ease of data processing, the first set can be in a coordinate system centered on the vehicle. In this case, the map tiles can be transformed to the coordinate system centered on the vehicle. After transforming the map tiles, a series of solid line point sets (Dash2d_map_raw, i.e., the point sets corresponding to each solid lane line in the transformed map tiles) and a series of dashed line point sets (Line2d_map_raw, i.e., the point sets corresponding to each dashed lane line in the transformed map tiles) are obtained.

[0065] Step S133: Remove curves, polylines or other irregularly shaped line segments from the solid line point set Dash2d_map_raw and the dashed line point set Line2d_map_raw corresponding to the converted map tile, and retain the fourth lane line of the target map lane (the map lane is the lane in the map tile), thus forming a map lane line set, wherein the fourth lane lines of the same target map lane are straight lines that are parallel to each other.

[0066] The map lane line set contains the point sets of each fourth lane line. Each fourth lane line can correspond to a separate point set, which can be represented as `Dash2d_map_filtered`, `Line2d_map_filtered`. In `Dash2d_map_filtered`, multiple lane lines within the same lane are parallel straight lines. For the map lane line set, it is known which fourth lane lines belong to the same target map lane.

[0067] Step S134: After obtaining the map lane line set from the converted map tiles, the first lane line of the target map lane can be filtered out from the first set based on the fourth lane line of the target map lane contained in the map lane line set.

[0068] First, map tiles near the vehicle are obtained from a high-precision map. Then, coordinate transformation is performed on the map tiles near the vehicle, and a set of lane lines consisting of parallel straight lines is obtained from the transformed map tiles. Then, the first lane line of the target lane is selected from the first set based on the lane lines contained in this set. This method can facilitate the accurate selection of the first lane line of the target lane, improve the optimization quality of the inverse perspective transformation matrix, and thus improve the accuracy of lane line detection error determination.

[0069] This application provides a method for determining lane line detection error, which filters the first lane line of the target lane from a first set of map lane lines, and may include:

[0070] Determine whether there exists a fourth lane line in the map lane line set whose distance to each marker point of the current lane line in the first set is less than a first distance threshold;

[0071] If not, the next lane line in the first set is determined as the current lane line, and the process returns to determine whether there is a fourth lane line in the map lane line set whose distance from each marker point of the current lane line in the first set is less than the first distance threshold, until the last lane line in the first set is determined as the current lane line and the judgment is completed.

[0072] If so, the current lane line is determined as the candidate lane line, and the process returns to the step of determining the next lane line in the first set as the current lane line;

[0073] The lane lines belonging to the same lane among the candidate lane lines are determined as the first lane line of the target lane.

[0074] In this embodiment of the application, the specific process of selecting the first lane line of the target lane from the first set based on the map lane line set can be as follows:

[0075] Step S1341: Take the first lane line in the first set as the current lane line.

[0076] Step S1342: Determine whether there exists a fourth lane line in the map lane line set whose distance to each marker point of the current lane line in the first set is less than a first distance threshold. The first distance threshold can be determined in advance through experimentation. To improve efficiency, the solid lane lines in the first set can be compared with the solid lane lines in the map lane line set, and the dashed lane lines in the first set can be compared with the dashed lane lines in the map lane line set.

[0077] If there is no fourth lane line in the map lane line set whose distance to each marker point of the current lane line in the first set is less than the first distance threshold, it indicates that there is no fourth lane line in the map lane line set that matches the current lane line in the first set, which means that the current lane line in the first set is not a straight line. Therefore, step S1343 can be executed. If there is a fourth lane line in the map lane line set whose distance to each marker point of the current lane line in the first set is less than the first distance threshold, it indicates that there is a fourth lane line in the map lane line set that matches the current lane line in the first set, which means that the current lane line in the first set is a straight line. Therefore, step S1344 can be executed.

[0078] Step S1343: Determine the next lane line in the first set as the current lane line, and return to execute step S1342, that is, return to execute the steps of determining whether there is a fourth lane line in the map lane line set whose distance from each marker point of the current lane line in the first set is less than the first distance threshold, until the last lane line in the first set is determined as the current lane line and the judgment is completed, that is, until each candidate lane line is selected from the first set.

[0079] Step S1344: Determine the current lane line in the first set as the candidate lane line and retain the candidate lane line. Then, return to execute step S1343, that is, return to execute the next lane line in the first set as the current lane line and related steps, until the last lane line in the first set is determined as the current lane line and the judgment is completed, that is, until each candidate lane line is selected from the first set.

[0080] For steps S1342-S1344, specifically: For Line2d_ground_1 in the first set, each line segment in Line2d_ground_1 can be matched with each fourth lane line Line2d_map_filtered in the map lane line set by line-to-line distance matching. If there is a fourth lane line Line2d_map_filtered in the map lane line set that satisfies the condition that the distance from Line2d_ground_1 to the fourth lane line Line2d_map_filtered is less than the first distance threshold, then Line2d_ground_1 can be retained and determined as a candidate lane line; otherwise, Line2d_ground_1 is deleted. For Dash2d_ground_1 in the first set, each line segment in Dash2d_ground_1 can be matched with the endpoint-to-endpoint distance of each fourth lane line Dash2d_map_filtered in the map lane line set. If there is a fourth lane line Dash2d_map_filtered in the map lane line set that satisfies the condition that the distance from Dash2d_ground_1 to the fourth lane line Dash2d_map_filtered is less than the first distance threshold, then Dash2d_ground_1 can be retained and determined as a candidate lane line; otherwise, Dash2d_ground_1 is deleted.

[0081] Step S1345: After selecting each candidate lane line from the first set, the candidate lane lines belonging to the same lane line are determined as the first lane line of the target lane. The lane corresponding to the candidate lane lines belonging to the same lane line is the target lane, and the multiple candidate lane lines of the target lane are the first lane lines of the target lane.

[0082] By using the distance judgment and traversal method between the fourth lane line in the map lane lines and each lane line in the first set, the first lane line of the target lane can be accurately selected from the first set, so as to improve the quality of the final inverse perspective transformation matrix Mopt, thereby improving the accuracy of lane line detection error determination.

[0083] This application provides a method for determining lane line detection error, which uses a second set and a third set to determine the detection error of the corresponding lane line, and may include:

[0084] Determine the matching relationship between lane lines in the second set and lane lines in the third set;

[0085] Calculate the distance between each detection point on each lane line in the third set and the matching lane line in the second set;

[0086] The detection error of the corresponding lane line is obtained by calculating the distance between each detection point on each lane line in the third set and the matching lane line in the second set.

[0087] In this embodiment of the application, the process of determining the detection error of the corresponding lane line using the second set obtained by transforming the manually labeled set and the third set obtained by transforming the model detection set can be specifically as follows:

[0088] Step S161: Determine the matching relationship between lane lines in the second set and lane lines in the third set.

[0089] Step S162: For each lane line in the third set, the distance between each detection point on that lane line in the third set and the matching lane line in the second set can be calculated.

[0090] Specifically, the point sets corresponding to each lane line in the second set obtained by transforming the manually labeled set to the BEV plane using the final optimized inverse perspective transformation matrix Mopt can all be represented as annotation_ground. Similarly, the point sets of each lane line in the third set obtained by transforming the model detection set to the BEV plane using the final optimized inverse perspective transformation matrix Mopt can all be represented as detect_ground. For solid lane lines, the distance from each detection point (detect_ground) on that lane line in the third set to the matching solid lane line (annotation_ground) in the second set can be calculated; that is, the point-to-line distance can be calculated for solid lane lines. For virtual lane lines, the distance from each detection point (detect_ground) on that lane line in the third set to the matching annotation point (annotation_ground) on the matching solid lane line in the second set that matches the aforementioned detection point can be calculated; that is, the point-to-point distance can be calculated for virtual lane lines.

[0091] Step S163: For each lane line in the third set, after calculating the distance between each detection point on that lane line in the third set and the matching lane line in the second set, the average distance between each detection point on that lane line in the third set and the matching lane line in the second set can be calculated to obtain the detection error of the corresponding lane line. That is, for real lane lines, the lane line detection error can be calculated based on the distance from point (detect_ground) to line (annotation_ground), and for virtual lane lines, the lane line detection error can be calculated based on the distance from point (detect_ground) to point (annotation_ground).

[0092] Of course, steps S162 and S163 can also involve calculating the distance between each marker point on each lane line in the second set and the matching lane line in the third set, and obtaining the detection error of the corresponding lane line based on the distance between each marker point on each lane line in the second set and the matching lane line in the third set.

[0093] The detection error of the corresponding lane line can be accurately calculated using the above method, thereby improving the accuracy of lane line detection error determination.

[0094] This application provides a method for determining lane line detection error, which determines the matching relationship between lane lines in a second set and lane lines in a third set, and may include:

[0095] Calculate the distances between multiple detection points on each lane line in the third set and each lane line in the second set;

[0096] Lane lines in the second set whose distances to multiple detection points on lane lines in both the second and third sets are less than a second distance threshold are identified as lane lines that match the corresponding lane lines in the third set.

[0097] In this embodiment, the matching relationship between lane lines in the second set and lane lines in the third set can be determined by distance calculation. Specifically, firstly, the distance between multiple detection points on each lane line in the third set and each lane line in the second set can be calculated. Then, lane lines whose distances to multiple detection points on both lane lines in the second and third sets are less than a second distance threshold are identified as lane lines that match the corresponding lane lines in the third set. The multiple detection points on the lane lines in the third set can specifically be all detection points on the lane lines in the third set, or a preset number of detection points. The preset number can be no less than the threshold, and the threshold value can be determined based on the number of detection points contained in the lane line. For example, the threshold can be 1 / 3, 1 / 2, etc., of the total number of detection points contained in the lane line. For virtual lane lines, in determining the matching relationship between lane lines in the second set and the third set, for each virtual lane line in the third set, the distance between the detection point of the virtual lane line in the third set and the marking points on the matching virtual lane lines in the second set can be calculated, and the marking point on the matching virtual lane lines in the second set that is closest to the detection point is determined as the marking point that matches the aforementioned detection point.

[0098] For example, the lane lines in the second set can be denoted as lane line A1, lane line A2, lane line A3, etc., and the lane lines in the third set can be denoted as lane line B1, lane line B2, lane line B3, etc. The distances between multiple detection points of lane line B1 in the third set and lane lines A1, A2, A3, etc. in the second set are calculated. If there exists a lane line An in the second set that satisfies the condition that the distances from multiple detection points of lane line B1 in the third set to lane line An in the second set are all less than a second distance threshold, then lane line An in the second set can be determined as a matching lane line of lane line B1 in the third set, that is, a matching relationship is determined between lane line B1 in the third set and An in the second set. The same method described above can be used to determine the matching relationship between the remaining lane lines in the third set and the lane lines in the second set.

[0099] This application provides a method for determining lane line detection error, which optimizes the inverse perspective transformation matrix, and may include:

[0100] The inverse perspective transformation matrix is ​​optimized using the Gauss-Newton method.

[0101] In this embodiment, the Gauss-Newton method can be used to optimize the inverse perspective transformation matrix. The Gauss-Newton method is one of the simplest optimization algorithms, possessing advantages such as fast convergence speed and strong applicability to high-dimensional and large-scale data.

[0102] Of course, other optimization algorithms can also be used to optimize the inverse perspective transformation matrix.

[0103] The lane line detection error determination method provided in this application embodiment may further include:

[0104] Output the detection error of the corresponding lane line.

[0105] In this embodiment of the application, after determining the detection error of the corresponding lane line using the second set and the third set, the detection error of the corresponding lane line can be output so that relevant personnel can obtain the detection error of the corresponding lane line in a timely manner, thereby facilitating adaptive adjustments to the parameters of the model used for detecting lane lines.

[0106] This application also provides a lane line detection error determination device, see [link to relevant documentation]. Figure 4 It shows a schematic diagram of the structure of a lane line detection error determination device provided in an embodiment of this application, which may include:

[0107] The acquisition module 41 is used to acquire the set of manually annotated lane lines and the set of model detection in the image;

[0108] The first conversion module 42 is used to convert the manually labeled set to the BEV plane using the inverse perspective transformation matrix to obtain the first set;

[0109] The filtering module 43 is used to filter out the first lane line of the target lane from the first set according to the high-precision map, and obtain the second lane line corresponding to the first lane line from the manually labeled set; each first lane line of the same target lane is a straight line.

[0110] Optimization module 44 is used to optimize the inverse perspective transformation matrix until the third lane lines of the same lane obtained by transforming the second lane line using the optimized inverse perspective transformation matrix are parallel to each other.

[0111] The second conversion module 45 is used to convert the manually labeled set and the model detection set using the optimized inverse perspective transformation matrix to obtain the corresponding second set and third set;

[0112] The determination module 46 is used to determine the detection error of the corresponding lane line using the second set and the third set.

[0113] This application provides a lane line detection error determination device, wherein the screening module 43 may include:

[0114] The first acquisition unit is used to acquire the vehicle coordinates at the time of image acquisition, and to acquire map tiles from the high-precision map within the target distance range relative to the vehicle coordinates based on the vehicle coordinates.

[0115] The second acquisition unit is used to convert the map tiles to the coordinate system of the first set and obtain the map lane line set from the converted map tiles; the map lane line set contains the fourth lane line of the target map lane, and the fourth lane lines of the same target map lane are parallel to each other and are straight lines;

[0116] The filtering unit is used to filter the first lane line of the target lane from the first set based on the map lane line set.

[0117] This application provides a lane line detection error determination device, wherein the screening unit may include:

[0118] The judgment sub-unit is used to determine whether there is a fourth lane line in the map lane line set whose distance to each marker point of the current lane line in the first set is less than the first distance threshold.

[0119] The first determining subunit is used to determine the next lane line in the first set as the current lane line if there is a fourth lane line in the map lane line set whose distance to each marker point of the current lane line in the first set is less than the first distance threshold, and then return to determine whether there is a fourth lane line in the map lane line set whose distance to each marker point of the current lane line in the first set is less than the first distance threshold, until the last lane line in the first set is determined as the current lane line and the determination is completed.

[0120] The second determining subunit is used to determine the current lane line as a candidate lane line if there is no fourth lane line in the map lane line set whose distance to each marker point of the current lane line in the first set is less than the first distance threshold, and then return to execute the step of determining the next lane line in the first set as the current lane line.

[0121] The third determining sub-unit is used to determine the candidate lane lines belonging to the same lane among the candidate lane lines as the first lane line of the target lane.

[0122] This application provides a lane line detection error determination device, wherein the determination module 46 may include:

[0123] The first determining unit is used to determine the matching relationship between lane lines in the second set and lane lines in the third set;

[0124] The calculation unit is used to calculate the distance between each detection point on each lane line in the third set and the matching lane line in the second set;

[0125] A detection error unit is obtained, which is used to obtain the detection error of the corresponding lane line based on the distance between each detection point on each lane line in the third set and the matching lane line in the second set.

[0126] This application provides a lane line detection error determination device, wherein the first determination unit may include:

[0127] The calculation subunit is used to calculate the distance between multiple detection points on each lane line in the third set and each lane line in the second set;

[0128] The fourth determining subunit is used to determine the lane lines in the second set whose distances to multiple detection points on the lane lines in the third set are all less than the second distance threshold as the lane lines that match the corresponding lane lines in the third set.

[0129] This application provides a lane line detection error determination device, wherein the optimization module 44 may include:

[0130] The optimization unit is used to optimize the inverse perspective transformation matrix using the Gauss-Newton method.

[0131] The lane line detection error determination device provided in this application embodiment may further include:

[0132] The output module is used to output the detection error of the corresponding lane line.

[0133] This application also provides a lane line detection error determination device, see [link to relevant documentation]. Figure 5 It shows a schematic diagram of the structure of a lane line detection error determination device provided in an embodiment of this application, which may include:

[0134] Memory 51 is used to store computer programs;

[0135] When processor 52 executes a computer program stored in memory 51, it can perform the following steps:

[0136] Obtain the manually labeled set and the model detection set of lane lines in the image; transform the manually labeled set to the BEV plane using an inverse perspective transformation matrix to obtain the first set; select the first lane line of the target lane from the first set based on the high-precision map, and obtain the second lane line corresponding to the first lane line from the manually labeled set; all first lane lines of the same target lane are straight lines; optimize the inverse perspective transformation matrix until the third lane lines of the same lane obtained by transforming the second lane line using the optimized inverse perspective transformation matrix are parallel to each other; transform the manually labeled set and the model detection set using the optimized inverse perspective transformation matrix to obtain the corresponding second set and third set; determine the detection error of the corresponding lane line using the second set and the third set.

[0137] This application embodiment also provides a readable storage medium storing a computer program, which, when executed by a processor, can perform the following steps:

[0138] Obtain the manually labeled set and the model detection set of lane lines in the image; transform the manually labeled set to the BEV plane using an inverse perspective transformation matrix to obtain the first set; select the first lane line of the target lane from the first set based on the high-precision map, and obtain the second lane line corresponding to the first lane line from the manually labeled set; all first lane lines of the same target lane are straight lines; optimize the inverse perspective transformation matrix until the third lane lines of the same lane obtained by transforming the second lane line using the optimized inverse perspective transformation matrix are parallel to each other; transform the manually labeled set and the model detection set using the optimized inverse perspective transformation matrix to obtain the corresponding second set and third set; determine the detection error of the corresponding lane line using the second set and the third set.

[0139] For a description of the relevant parts of the lane line detection error determination device, equipment and readable storage medium provided in the embodiments of this application, please refer to the detailed description of the corresponding parts of the lane line detection error determination method provided in the embodiments of this application, and will not be repeated here.

[0140] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any readable storage medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "readable storage medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the readable storage medium could even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0141] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0142] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0143] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0144] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "joining," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0145] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A lane line detection error determination method characterized by, The method comprises the following steps: obtaining a manual annotation set and a model detection set of lane lines in an image; converting the manual annotation set to a BEV plane by using an inverse perspective transformation matrix to obtain a first set; selecting first lane lines of a target lane from the first set according to a high-precision map, and obtaining second lane lines corresponding to the first lane lines from the manual annotation set; each first lane line of the target lane is a straight line; optimizing the inverse perspective transformation matrix until each third lane line of the same lane obtained by converting the second lane lines by using the optimized inverse perspective transformation matrix is parallel to each other; converting the manual annotation set and the model detection set by using the optimized inverse perspective transformation matrix to obtain corresponding second and third sets; determining a detection error of a corresponding lane line by using the second and third sets; selecting first lane lines of a target lane from the first set according to a high-precision map, comprising: obtaining a vehicle coordinate at an image collection time, and obtaining a map block within a target distance range from the vehicle coordinate from the high-precision map; converting the map block to a coordinate system in which the first set is located, and obtaining a map lane line set from the converted map block; the map lane line set contains fourth lane lines of a target map lane, and each fourth lane line of the target map lane is parallel to each other and a straight line; selecting the first lane lines of the target lane from the first set according to the map lane line set; selecting the first lane lines of the target lane from the first set according to the map lane line set, comprising: determining whether there is a fourth lane line in the map lane line set, the distance between each annotation point of the current lane line in the first set and the fourth lane line being less than a first distance threshold; if not, determining the next lane line in the first set as the current lane line, and returning to execute the step of determining whether there is a fourth lane line in the map lane line set, the distance between each annotation point of the current lane line in the first set and the fourth lane line being less than a first distance threshold, until the last lane line in the first set is determined as the current lane line and the determination is completed; if yes, determining the current lane line as a candidate lane line, and returning to execute the step of determining the next lane line in the first set as the current lane line; determining candidate lane lines belonging to the same lane in each candidate lane line as the first lane lines of the target lane.

2. The lane line detection error determination method of claim 1, wherein, determining a detection error of a corresponding lane line by using the second and third sets, comprising: determining a matching relationship between lane lines in the second set and lane lines in the third set; calculating the distance between each detection point on each lane line in the third set and the matching lane line in the second set; obtaining a detection error of a corresponding lane line according to the distance between each detection point on each lane line in the third set and the matching lane line in the second set.

3. The lane line detection error determination method of claim 2, wherein, determining a matching relationship between lane lines in the second set and lane lines in the third set, comprising: calculate distances between each lane line in the third set and each lane line in the second set; determine, as a lane line matched with a lane line in the third set, a lane line in the second set having a distance to each of the plurality of detection points on the lane line in the third set less than a second distance threshold.

4. The lane line detection error determination method of claim 1, wherein, optimizing the inverse perspective transformation matrix, comprising: optimizing the inverse perspective transformation matrix using a Gauss-Newton method.

5. The lane line detection error determination method of claim 1, wherein, further comprising: outputting a detection error of the corresponding lane line.

6. A lane line detection error determination apparatus characterized by comprising: comprising: an acquisition module, configured to acquire a manually labeled set and a model detection set of lane lines in an image; a first conversion module, configured to convert the manually labeled set to a BEV plane using an inverse perspective transformation matrix to obtain a first set; a screening module, configured to screen a first lane line of a target lane from the first set according to a high-definition map, and acquire a second lane line corresponding to the first lane line from the manually labeled set; each first lane line of the same target lane is a straight line; an optimization module, configured to optimize the inverse perspective transformation matrix until each third lane line of the same lane obtained by converting the second lane line using the optimized inverse perspective transformation matrix is parallel to each other; a second conversion module, configured to convert the manually labeled set and the model detection set using the optimized inverse perspective transformation matrix to obtain corresponding second and third sets; a determination module, configured to determine a detection error of the corresponding lane line using the second and third sets; the screening module comprises: a first acquisition unit, configured to acquire a vehicle body coordinate at an image acquisition time, and acquire a map block within a target distance range from the vehicle body coordinate from the high-definition map; a second acquisition unit, configured to convert the map block to a coordinate system in which the first set is located, and acquire a map lane line set from the converted map block; the map lane line set includes a fourth lane line of a target map lane, and each fourth lane line of the same target map lane is parallel to each other and is a straight line; a screening unit, configured to screen a first lane line of the target lane from the first set according to the map lane line set; the screening unit comprises: a judgment subunit, configured to judge whether there is a fourth lane line in the map lane line set having a distance to each annotation point of a current lane line in the first set less than a first distance threshold; a first determination subunit, configured to, if there is no fourth lane line in the map lane line set having a distance to each annotation point of the current lane line in the first set less than the first distance threshold, determine a next lane line in the first set as the current lane line, and return to execute the judgment of whether there is a fourth lane line in the map lane line set having a distance to each annotation point of the current lane line in the first set less than the first distance threshold, until the last lane line in the first set is determined as the current lane line and the judgment is completed. a second determining sub-unit, configured to determine a current lane line in the first set as a candidate lane line if there is a fourth lane line in the map lane line set which is closer to each of the marking points of the current lane line than a first distance threshold, and return to the step of determining a next lane line in the first set as the current lane line; a third determining sub-unit, configured to determine candidate lane lines belonging to the same lane as the first lane line of the target lane.

7. A lane line detection error determination apparatus characterized by comprising: comprising: a memory, configured to store a computer program; a processor, configured to implement the steps of the lane line detection error determination method according to any one of claims 1 to 5 when the computer program is executed.

8. A readable storage medium, characterized by, The readable storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the lane line detection error determination method according to any one of claims 1 to 5.

Images