Lane line detection method and device and storage medium

Abstract

The invention discloses a lane line detection method and device and a storage medium, and belongs to the technical field of vehicle control. The method comprises: acquiring an image of a road where a vehicle is; a key point set of a lane line of the road where the vehicle is located is recognized from the image, the key point set comprises a certain number of key points, and the key points are points representing features of the lane line of the road where the vehicle is located; an embedding vector corresponding to each key point is obtained; clustering the key points based on an embedding vector, and identifying an initial lane line of a road where the vehicle is located; obtaining the position of a Root point of each initial lane line; and screening outliers in the points contained in the initial lane line based on the position of the Root point, and determining a final lane line. Fine non-lane lines are prevented from being recognized as lane lines, the accuracy and stability of lane line detection are improved, and lane line detection is made to be suitable for various road conditions.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of vehicle control, and particularly to a method, device and storage medium for lane line detection. Background Art

[0002] The regression of lane lines refers to detecting and predicting the geometric parameters or positions of lane lines through a regression algorithm, which is one of the key algorithms for autonomous driving. In the related art, the coefficient regression of lane lines is performed using anchors, that is, a reference position or area is specified in the image, and then linear regression is performed based on the reference position or area through a deep learning model to identify and locate the lane lines.

[0003] In the related art, with the development of autonomous driving and the increase in driving scenarios, it is difficult to detect and predict anchors in scenarios such as sharp curves. Therefore, how to make lane line detection applicable to various road conditions and improve the stability of lane line prediction is a problem that needs to be solved. Summary of the Invention

[0004] The embodiments of the present application provide a method, device and storage medium for lane line detection, which can be used to make lane line detection applicable to various road conditions and improve the stability of lane line prediction. The technical solutions are as follows:

[0005] On the one hand, the embodiments of the present application provide a method for lane line detection, the method comprising:

[0006] Obtaining an image of the road where the vehicle is located;

[0007] Identifying a key point set of the lane lines of the road where the vehicle is located from the image, the key point set comprising a certain number of key points, and the key points being points representing the characteristics of the lane lines of the road where the vehicle is located;

[0008] Obtaining an embedding vector corresponding to each key point;

[0009] Clustering the key points based on the embedding vectors to identify the initial lane lines of the road where the vehicle is located;

[0010] Obtaining the position of the Root point of each initial lane line;

[0011] Filtering outlier points among the points included in the initial lane lines based on the position of the Root point to determine the final lane lines.

[0012] On the other hand, a device for lane line detection is provided, the device comprising:

[0013] A first obtaining module, configured to obtain an image of the road where the vehicle is located;

[0014] A first recognition module, configured to recognize a key point set of lane lines of a road where a vehicle is located from the image, the key point set includes a certain number of key points, and the key points are points representing characteristics of the lane lines of the road where the vehicle is located;

[0015] A second acquisition module, configured to acquire an embedding vector corresponding to each key point;

[0016] A second recognition module, configured to cluster the key points based on the embedding vectors to recognize an initial lane line of the road where the vehicle is located;

[0017] A third acquisition module, configured to acquire the position of the Root point of each initial lane line;

[0018] A screening module, configured to screen outlier points among the points included in the initial lane line based on the position of the Root point to determine a final lane line.

[0019] On the other hand, a non-transitory computer-readable storage medium is further provided, which is characterized in that a computer program is stored in the computer-readable storage medium, and the computer program is loaded and executed by a processor to implement the lane line detection method described in any one of the above.

[0020] On the other hand, a computer program product is further provided, the computer program product includes computer instructions, and when the computer instructions are executed by a processor, the steps of the lane line detection method described in any one of the above are implemented.

[0021] The technical solution provided by this application at least brings the following beneficial effects:

[0022] In this application, by acquiring an image of the road where the vehicle is located, a key point set of the lane lines of the road where the vehicle is located is recognized; then, an embedding vector corresponding to each key point is acquired, so as to cluster the key points based on the embedding vectors to recognize an initial lane line of the road where the vehicle is located; the position of the Root point of each initial lane line is acquired, and then outlier points among the points included in the initial lane line are screened and deleted according to the position of the Root point to determine a final lane line. By combining the embedding vector corresponding to the key point with the Root point to recognize the lane line, it is avoided to recognize fine non-lane line segments as lane lines, improving the accuracy and stability of lane line detection, and making lane line detection applicable to various road conditions. Description of the Drawings

[0023] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0024] Figure 1 It is a schematic diagram of an implementation environment provided by an embodiment of the present application;

[0025] Figure 2 It is a flowchart of a lane line detection method provided by an embodiment of the present application;

[0026] Figure 3 It is a logic flowchart of a lane line detection provided by an embodiment of the present application;

[0027] Figure 4 It is a schematic structural diagram of a lane line detection device provided by an embodiment of the present application. Detailed implementation manners

[0028] To make the objectives, technical solutions, and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the drawings.

[0029] An embodiment of the present application provides a lane line detection method. Please refer to Figure 1 , which shows a schematic diagram of the method implementation environment provided by the embodiment of the present application. The implementation environment may include: HCU (Hardware Control Unit, vehicle controller) 11, image recognition device 12, and in-vehicle camera 13.

[0030] Optionally, the in-vehicle camera 13 is installed at the bottom in front of the vehicle and is used to collect the front road surface image of the vehicle's location and send it to the image recognition device 12. The image recognition device 12 is installed on the vehicle and is used to receive the front road surface image of the vehicle's location collected by the in-vehicle camera 13, and detect the image of the vehicle's location road, identify key points of a certain number of lane lines in the image of the vehicle's location road, and send them to the HCU 11.

[0031] Among them, the HCU 11, the image recognition device 12, and the in-vehicle camera 13 establish a communication connection through a wired or wireless network.

[0032] Based on the above Figure 1 shown implementation environment, an embodiment of the present application provides a lane line detection method as Figure 2 shown. Taking this method applied to the HCU as an example, the method includes step 201-step 206.

[0033] In step 201, the HCU acquires an image of the road where the vehicle is located.

[0034] In a possible implementation, the way for the HCU to acquire an image of the road where the vehicle is located includes, but is not limited to: the HCU collects the front road surface image of the road where the vehicle is located through an on-vehicle camera, where the on-vehicle camera can be installed at the bottom in front of the vehicle.

[0035] In step 202, the HCU identifies a key point set of the lane lines of the road where the vehicle is located from the image. The key point set includes a certain number of key points, and the key points are points representing the characteristics of the lane lines of the road where the vehicle is located.

[0036] Exemplarily, after acquiring the image of the road where the vehicle is located, the HCU identifies a key point set of the lane lines of the road where the vehicle is located from the image. The key point set includes a certain number of key points, and the key points are points representing the characteristics of the lane lines of the road where the vehicle is located.

[0037] In a possible implementation, the HCU can detect the image of the road where the vehicle is located through an image recognition device, and identify a certain number of key points of the lane lines in the image of the road where the vehicle is located. The number of key points can be set according to experience, and the key points include, but are not limited to, the edge points, corner points and bifurcation points of the lane lines.

[0038] Exemplarily, the image recognition device is installed on the vehicle, can identify the key points of the lane lines of the road where the vehicle is located from the image of the road where the vehicle is located, and can further determine the coordinates of the key points in the coordinate system. The coordinate system can use the vehicle as the origin, the direction parallel to the lane line as the longitudinal axis, and the direction perpendicular to the lane line as the transverse axis.

[0039] In step 203, the HCU acquires the embedding vector corresponding to each key point.

[0040] In a possible implementation, after identifying the key point set of the lane lines of the road where the vehicle is located, the HCU acquires the embedding vector corresponding to each key point, including: the HCU acquires the embedding vector of the pixel corresponding to each key point through an embedding generation model.

[0041] Optionally, the embedding vector is a numerical vector representing the characteristics of the lane line. By encoding features such as shape, color, and position into a point in a multi-dimensional space to capture the essence of the lane line, similar lane lines can be made closer, thus facilitating the clustering and recognition of lane lines.

[0042] Exemplarily, the embedding generation model can be pre-trained, and the training process includes but is not limited to: selecting a neural network architecture as the initial recognition model. For example, a convolutional neural network can be selected; then, through a certain number of road images of lane lines with each key point pre-annotated with an embedding vector, the initial recognition model is trained to obtain the final recognition model.

[0043] In a possible implementation, the input of the final recognition model is the road image of the lane line where the key point is located. The output of the first specified connection layer of the neural network is the embedding vector of the key point, and the output of the second specified connection layer of the neural network is the confidence of the key point, where the confidence indicates the probability that the key point belongs to the point on the lane line.

[0044] Optionally, after obtaining the final recognition model, the HCU inputs the road images of the lane lines where each key point is located into the final recognition model, and then obtains the embedding vector corresponding to each key point from the output of the first specified connection layer of the neural network of the final recognition model.

[0045] In step 204, the HCU clusters the key points based on the embedding vectors to identify the initial lane lines of the road where the vehicle is located.

[0046] Exemplarily, after obtaining the embedding vector corresponding to each key point, before the HCU clusters the key points based on the embedding vector, the HCU performs normalization processing on the embedding vector corresponding to each key point, including: scaling the embedding vector to the same scale to eliminate the influence of the size difference between different embedding vectors on the clustering effect.

[0047] Optionally, after completing the normalization processing of the embedding vector, the HCU clusters the key points based on the embedding vector to identify the initial lane lines of the road where the vehicle is located, including: obtaining the confidence of the embedding vector corresponding to each key point; using the key points with a confidence greater than the confidence threshold as clustering seeds to cluster the remaining key points to determine the lane lines where each key point is located; deleting the lane lines with the number of key points less than the point threshold, and using the remaining lane lines as the initial lane lines.

[0048] Exemplarily, the HCU can obtain the confidence of the embedding vector corresponding to each key point from the output of the second specified connection layer of the neural network of the final recognition model. The key points with confidence greater than the confidence threshold are used as clustering seeds to cluster the remaining key points, and the lane line where each key point is located is determined, including: sorting each key point according to the confidence; in response to the confidence of the key point being greater than the confidence threshold, using the key point as the seed of the lane line cluster to cluster the remaining key points, and the seed of the lane line cluster is used to cluster the remaining key points around the key point into the lane line cluster.

[0049] In a possible implementation manner, after obtaining the confidence of the embedding vector corresponding to each key point, each key point is sorted according to the confidence level, and then the confidence is compared with the confidence threshold. If the confidence of the key point is greater than the confidence threshold, the key point is used as the seed of the lane line cluster to cluster the remaining key points, and the lane line where each key point is located is determined in turn. Among them, the seed of the lane line cluster is used to cluster the remaining key points around the key point into the lane line cluster, and the clustering can be completed by clustering algorithms such as K-means (mean).

[0050] Exemplarily, after determining the lane line where each key point is located, the HCU counts the key points included in each lane line, deletes the lane lines with the number of key points included less than the point threshold, and uses the remaining lane lines as the initial lane lines. Optionally, the confidence threshold and the point threshold can be set according to experience.

[0051] In step 205, the HCU obtains the position of the Root point of each initial lane line.

[0052] Optionally, obtaining the position of the Root point of each initial lane line includes: predicting the trend of each initial lane line based on the image of the road where the vehicle is located in the coordinate system with the vehicle as the origin, predicting the intersection point of the lane lines based on the trend of each initial lane line, and using the intersection point of the lane lines as the position of the Root point of the initial lane line.

[0053] In a possible implementation manner, the HCU can predict the trend of the initial lane line and the intersection point of the lane lines in the distance through the vanishing point estimation algorithm, including: calculating the slope of each initial lane line, selecting any two initial lane lines with close slopes, predicting the intersection point of the two initial lane lines in the distance, and using the intersection point as the position of the Root point of the two initial lane lines. Optionally, when the slope difference is less than the slope threshold, it indicates that the slopes of the two initial lane lines are close, and the slope threshold can be set according to experience.

[0054] In step 206, the HCU filters out the outliers among the points included in the initial lane line based on the position of the Root point, and determines the final lane line.

[0055] Optionally, after obtaining the positions of the Root points of each initial lane line, the HCU filters out the outliers among the points included in the initial lane line based on the positions of the Root points to determine the final lane line, including: calculating the average coordinate position of the root points of each initial lane line; identifying the outliers whose distances from the average coordinate position of the root points exceed the distance threshold; deleting the outliers, and taking the remaining initial lane lines as the final lane lines.

[0056] Exemplarily, the average coordinate position of the root point includes the average abscissa position and the average ordinate position. Calculating the average coordinate position of the root points of each initial lane line includes: calculating the sum of the abscissas and the sum of the ordinates of all the root points of each initial lane line; calculating the first calculation result of dividing the sum of the abscissas by the number of root points, and the second calculation result of dividing the sum of the ordinates by the number of root points; taking the first calculation result as the average abscissa position and the second calculation result as the average ordinate position.

[0057] In a possible implementation manner, the HCU counts the number of root points of each initial lane line, the sum of the abscissas of all the root points of each initial lane line, and the sum of the ordinates of all the root points of each initial lane line, and then calculates the first calculation result of dividing the sum of the abscissas of all the root points of each initial lane line by the number of root points of this initial lane line, and the second calculation result of dividing the sum of the ordinates of all the root points of each initial lane line by the number of root points of this initial lane line.

[0058] Optionally, after determining the first calculation result and the second calculation result, the HCU then takes the first calculation result as the average abscissa position, takes the second calculation result as the average ordinate position, and combines the average abscissa position of each initial lane line with the corresponding average ordinate position one by one to obtain the average coordinate position of the root points of each initial lane line.

[0059] Exemplarily, after determining the average coordinate position of the root points of each initial lane line, the HCU calculates the distance between the points on the lane line and the average coordinate position of the root points of this initial lane line through the coordinate distance calculation formula, and then compares the calculated distance with the distance threshold. If the calculated distance is greater than or equal to the distance threshold, then this point is determined as an outlier. Optionally, after determining the outlier, the HCU deletes the outlier, and takes the remaining initial lane lines after removing the outlier as the final lane lines.

[0060] To sum up, Figure 3Taking the logic flow chart of a lane line detection provided by the embodiment of the present application shown as an example for illustration. Among them, the execution subject can be the HCU. Step 301, obtain an image of the road where the vehicle is located, and identify a key point set of the lane lines on the road where the vehicle is located from the image. Step 302, obtain the embedding vector corresponding to each key point. Step 303, perform clustering on the key points based on the embedding vector, that is, judge whether the key points belong to the same lane line based on the embedding vector. After the judgment of the key points is completed, enter step 304.

[0061] Step 304, identify the initial lane lines on the road where the vehicle is located. Step 305, obtain the position of the Root point of each initial lane line. Step 306, judge whether there are outliers among the points included in the initial lane line. If there are outliers, enter step 307; if there are no outliers, enter step 308. Step 307, delete the outliers. Step 308, use the remaining initial lane lines as the final lane lines.

[0062] In the embodiment of the present application, by obtaining an image of the road where the vehicle is located, identifying a key point set of the lane lines on the road where the vehicle is located; then obtaining the embedding vector corresponding to each key point, so as to perform clustering on the key points based on the embedding vector and identify the initial lane lines on the road where the vehicle is located; obtaining the position of the Root point of each initial lane line, and then screening and deleting the outliers among the points included in the initial lane line according to the position of the Root point to determine the final lane lines. By combining the embedding vector corresponding to the key point with the Root point to identify the lane lines, it is avoided to identify fine non-lane line segments as lane lines, improving the accuracy and stability of lane line detection and making lane line detection applicable to various road conditions.

[0063] See Figure 4 , the embodiment of the present application provides a lane line detection device, and the device includes:

[0064] The first acquisition module 401 is used to acquire an image of the road where the vehicle is located;

[0065] The first recognition module 402 is used to identify a key point set of the lane lines on the road where the vehicle is located from the image. The key point set includes a certain number of key points, and the key points are the points representing the characteristics of the lane lines on the road where the vehicle is located;

[0066] The second acquisition module 403 is used to acquire the embedding vector corresponding to each key point;

[0067] The second recognition module 404 is used to perform clustering on the key points based on the embedding vector and identify the initial lane lines on the road where the vehicle is located;

[0068] A third acquisition module 405, configured to acquire the positions of the Root points of each initial lane line;

[0069] A screening module 406, configured to screen outlier points among the points included in the initial lane line based on the positions of the Root points, and determine the final lane line.

[0070] In a possible implementation manner, a second recognition module 404 is configured to acquire the confidence of the embedding vector corresponding to each key point; use the key points with a confidence greater than the confidence threshold as clustering seeds to cluster the remaining key points, and determine the lane line where each key point is located; delete the lane lines with the number of included key points less than the point threshold, and use the remaining lane lines as the initial lane lines.

[0071] In a possible implementation manner, a second recognition module 404 is configured to sort each key point according to the confidence; in response to the confidence of the key point being greater than the confidence threshold, use the key point as the seed of the lane line cluster to cluster the remaining key points, and the seed of the lane line cluster is used to cluster the remaining key points around the key point into the lane line cluster.

[0072] In a possible implementation manner, a third acquisition module 405 is configured to predict the trend of each initial lane line according to the image of the road where the vehicle is located; predict the intersection points of the lane lines based on the trend of each initial lane line, and use the intersection points of the lane lines as the positions of the Root points of the initial lane lines.

[0073] In a possible implementation manner, a screening module 406 is configured to calculate the average coordinate position of the root points of each initial lane line; identify outlier points whose distance from the average coordinate position of the root points exceeds the distance threshold; delete the initial lane lines containing outlier points, and use the remaining initial lane lines as the final lane lines.

[0074] In a possible implementation manner, the average coordinate position of the root point includes the average abscissa position and the average ordinate position. The screening module 406 is configured to calculate the sum of the abscissas and the sum of the ordinates of all the root points of each initial lane line; calculate the first calculation result of the sum of the abscissas divided by the number of root points, and the second calculation result of the sum of the ordinates divided by the number of root points; use the first calculation result as the average abscissa position and the second calculation result as the average ordinate position.

[0075] This device obtains an image of the road where the vehicle is located, and identifies the key point set of the lane lines on the road where the vehicle is located; then obtains the embedding vector corresponding to each key point, so as to cluster the key points based on the embedding vector and identify the initial lane lines on the road where the vehicle is located; obtains the position of the Root point of each initial lane line, and then filters and deletes the outliers among the points included in the initial lane line according to the position of the Root point to determine the final lane line. By combining the embedding vector corresponding to the key point with the Root point to identify the lane line, it avoids identifying fragmented non-lane line segments as lane lines, improves the accuracy and stability of lane line detection, and makes lane line detection applicable to various road conditions.

[0076] It should be noted that when the device provided in the above embodiment realizes its functions, only the above division of each functional module is used for illustration. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device provided in the above embodiment and the method embodiment belong to the same concept, and the specific implementation process can be seen in the method embodiment, which will not be elaborated here.

[0077] In an exemplary embodiment, a computer-readable storage medium is also provided. At least one computer program is stored in the computer-readable storage medium, and the at least one computer program is loaded and executed by a processor of a computer device so that the computer implements any one of the above lane line detection methods.

[0078] In a possible implementation manner, the above computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, etc.

[0079] In an exemplary embodiment, a computer program product or a computer program is also provided. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions so that the computer device executes any one of the above lane line detection methods.

[0080] It should be noted that the information involved in this application (including but not limited to user device information, user personal information, etc.), data (including but not limited to data for analysis, stored data, displayed data, etc.), and signals are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions. For example, the images of the road where the vehicle is located, the key point set of the lane lines of the road where the vehicle is located, the initial lane line and the final lane line of the road where the vehicle is located involved in this application are obtained under the condition of full authorization.

[0081] It should be understood that the term "a plurality of" mentioned herein refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0082] It should be noted that the terms "first", "second", etc. (if any) in the description and claims of this application are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances so that the embodiments of this application described here can be implemented in an order other than those illustrated or described here. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. On the contrary, they are only examples of devices and methods consistent with some aspects of this application as detailed in the appended claims.

[0083] The above are only exemplary embodiments of this application and are not intended to limit this application. Any modifications, equivalent replacements, improvements, etc. made within the principles of this application shall be included in the protection scope of this application.

Claims

1. A method for lane line detection, characterized in that, The method includes: Obtaining an image of the road where the vehicle is located; Identifying a key point set of the lane lines of the road where the vehicle is located from the image, the key point set including a certain number of key points, and the key points being points representing the characteristics of the lane lines of the road where the vehicle is located; Obtaining an embedding vector corresponding to each key point; Clustering the key points based on the embedding vectors to identify the initial lane lines of the road where the vehicle is located; Obtaining the position of the Root point of each initial lane line; Filtering outlier points among the points included in the initial lane lines based on the position of the Root point to determine the final lane lines.

2. The method according to claim 1, wherein The clustering the key points based on the embedding vectors to identify the initial lane lines of the road where the vehicle is located includes: Obtaining the confidence of the embedding vector corresponding to each key point; Using the key points with confidence greater than the confidence threshold as clustering seeds to cluster the remaining key points to determine the lane lines where each key point is located; Deleting the lane lines with the number of included key points less than the point threshold, and using the remaining lane lines as the initial lane lines.

3. The method according to claim 2, characterized in that, The using the key points with confidence greater than the confidence threshold as clustering seeds to cluster the remaining key points to determine the lane lines where each key point is located includes: Sorting each key point according to the confidence; In response to the confidence of the key point being greater than the confidence threshold, using the key point as the seed of the lane line cluster to cluster the remaining key points, and the seed of the lane line cluster is used to cluster the remaining key points around the key point into the lane line cluster.

4. The method according to claim 1, wherein The obtaining the position of the Root point of each initial lane line includes: Predicting the trend of each initial lane line according to the image of the road where the vehicle is located; Predicting the intersection points of the lane lines based on the trend of each initial lane line, and using the intersection points of the lane lines as the position of the Root point of the initial lane line.

5. The method according to claim 1, wherein The filtering outlier points among the points included in the initial lane lines based on the position of the Root point to determine the final lane lines includes: Calculating the average coordinate position of the root points of each initial lane line; Identifying outlier points whose distance from the average coordinate position of the root point exceeds the distance threshold; Deleting the outlier points, and using the remaining initial lane lines as the final lane lines.

6. The method according to claim 5, characterized in that, The average coordinate position of the root point includes the average abscissa position and the average ordinate position, and the calculating the average coordinate position of the root points of each initial lane line includes: Calculating the sum of the abscissas and the sum of the ordinates of all root points of each initial lane line; Calculating a first calculation result of dividing the sum of the abscissas by the number of root points, and a second calculation result of dividing the sum of the ordinates by the number of root points; Using the first calculation result as the average abscissa position and the second calculation result as the average ordinate position.

7. A device for lane line detection, characterized in that, The device includes: A first obtaining module, configured to obtain an image of the road where the vehicle is located; A first recognition module, configured to recognize a key point set of lane lines of a road where a vehicle is located from the image, the key point set includes a certain number of key points, and the key points are points representing the characteristics of the lane lines of the road where the vehicle is located; A second acquisition module, configured to acquire an embedding vector corresponding to each key point; A second recognition module, configured to cluster the key points based on the embedding vectors to recognize an initial lane line of the road where the vehicle is located; A third acquisition module, configured to acquire the position of the Root point of each initial lane line; A screening module, configured to screen outlier points among the points included in the initial lane line based on the position of the Root point to determine the final lane line.

8. The device according to claim 7, characterized in that, The second recognition module is configured to acquire the confidence of the embedding vector corresponding to each key point; use the key points with the confidence greater than the confidence threshold as clustering seeds to cluster the remaining key points to determine the lane line where each key point is located; Delete the lane lines with the number of included key points less than the point threshold, and use the remaining lane lines as the initial lane lines.

9. A computer program product, the computer program product includes computer instructions, and when the computer instructions are executed by a processor, the steps of the lane line detection method according to any one of claims 1 to 6 are implemented.

10. A non-transitory computer-readable storage medium, characterized in that, A computer program is stored in the computer-readable storage medium, and the computer program is loaded and executed by a processor to implement the lane line detection method according to any one of claims 1 to 6.

Images