Sample picture generation method, lane line generation method and related equipment

Abstract

The invention discloses a sample picture generation method, a lane line generation method and related equipment. The method comprises the following steps: acquiring a multi-view lane line sample, wherein the multi-view lane line sample is a picture obtained by shooting a lane line from different angles; processing the multi-view lane line sample to obtain a multi-view lane line-free sample; wherein the multi-view lane-line-free sample is used for training a lane line generation model of the lane line. The sensing ability of the automatic driving system is enhanced, and the driving stability of the vehicle is improved.

Description

Technical Field

[0001] The embodiments disclosed in the present application relate to the field of vehicle technologies, and more specifically, to a method for generating sample pictures, a method for generating lane lines, and related devices. Background Art

[0002] In the scenario of autonomous driving, the recognition of lane lines is very important. However, lane lines may become unclear or completely disappear due to various reasons. For example, in scenarios such as construction and bad weather, lane lines will become unclear, or in complex traffic intersection scenarios, there are too many lane lines to recognize the correct lane lines, etc., which greatly affects the stability of vehicle driving. Summary of the Invention

[0003] According to the embodiments of the present application, the present application provides a method for generating sample pictures, a method for generating lane lines, and related devices to solve the above problems.

[0004] The first aspect of the present application discloses a method for generating sample pictures, including: obtaining multi-perspective lane line samples, where the multi-perspective lane line samples are pictures taken of lane lines from different angles; processing the multi-perspective lane line samples to obtain multi-perspective lane-line-free samples; where the multi-perspective lane-line-free samples are used for training a lane line generation model of the lane lines.

[0005] In some embodiments, processing the multi-perspective lane line samples to obtain multi-perspective lane-line-free samples includes: segmenting the multi-perspective lane line samples to extract multi-view lane line regions corresponding to the lane lines in the multi-perspective lane line samples; performing lane line detection on the multi-view lane line regions, and erasing the lane lines in the multi-view lane line regions to obtain the multi-perspective lane-line-free samples.

[0006] The second aspect of the present application discloses a method for generating lane lines, including: obtaining multi-perspective lane-line-free pictures; processing the multi-perspective lane-line-free pictures to obtain multi-view lane-line-free regions in the multi-perspective lane-line-free pictures; inputting the multi-view lane-line-free regions into a lane line generation model to obtain multi-perspective virtual lane line pictures; where the lane line generation model is trained using the multi-perspective lane-line-free samples obtained by the method for generating sample pictures as described in the first aspect.

[0007] In some embodiments, the lane line generation model includes a diffusion model and a control model; the multi-view lane line-free area is input into the lane line generation model to obtain a multi-view virtual lane line image, including: inputting the multi-view lane line-free area into the control model to generate control conditions for the lane line generation model; based on preset noise information and the control conditions, using the diffusion model to generate the multi-view virtual lane line image.

[0008] In some embodiments, the lane line generation model includes a diffusion model and a control model;

[0009] The step of inputting the multi-view lane line-free area into the lane line generation model to obtain a multi-view virtual lane line image includes: inputting the multi-view lane line-free area into the control model to generate control conditions for the lane line generation model; and generating the multi-view virtual lane line image using a diffusion model based on preset noise information, prompt word information and the control conditions.

[0010] In some embodiments, the prompt word information includes at least: parameters of a camera used to obtain the multi-view image without lane lines; lane line attributes; and road information.

[0011] In some embodiments, the multi-view image without lane lines is processed to obtain a multi-view area without lane lines in the multi-view image without lane lines, including: segmenting the multi-view image without lane lines to extract a multi-view road surface area in the multi-view image without lane lines; and splicing the multi-view road surface area to generate the multi-view area without lane lines.

[0012] In some embodiments, obtaining a multi-perspective image without lane lines includes: obtaining a multi-perspective image with lane lines; segmenting the multi-perspective image with lane lines to extract a multi-view lane line area corresponding to the lane lines in the multi-perspective image with lane lines; performing lane line detection on the multi-view lane line area and erasing the lane lines in the multi-view lane line area to obtain the multi-perspective image with no lane lines.

[0013] The third aspect of the present application discloses an electronic device, comprising a memory and a processor coupled to each other, wherein the processor is used to execute program instructions stored in the memory to implement the sample image generation method described in the first aspect, or to implement the lane line generation method described in the second aspect.

[0014] The fourth aspect of the present application discloses a non-volatile computer-readable storage medium having program instructions stored thereon. When the program instructions are executed by a processor, the sample image generation method described in the first aspect is implemented, or the lane line generation method described in the second aspect is implemented.

[0015] The beneficial effects of the present application include: obtaining multi-perspective lane line samples, which are pictures obtained by taking lane lines from different angles, processing the multi-perspective lane line samples to obtain multi-perspective lane-free samples, and using the multi-perspective lane-free samples for training a lane line generation model for lane lines, thereby enabling the generation of virtual lane lines when there are no lane lines or the lane lines are unclear, thereby enhancing the perception capability of the autonomous driving system and improving the stability of vehicle driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present application will be further described below in conjunction with the accompanying drawings and implementation methods, in which:

[0017] Figure 1 is a flow chart of a sample image generation method according to an embodiment of the present application;

[0018] Figure 2 is a flowchart of a lane line generation method according to an embodiment of the present application;

[0019] Figure 3 is a schematic diagram of a lane line generation model according to an embodiment of the present application;

[0020] Figure 4 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application;

[0021] Figure 5 It is a schematic diagram of the structure of the non-volatile computer-readable storage medium of an embodiment of the present application. DETAILED DESCRIPTION

[0022] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0023] The term "and / or" in this application is only a description of the association relationship of associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the objects associated before and after are in an "or" relationship. In addition, "many" in this article means two or more than two. In addition, the term "at least one" in this article means any combination of at least two of any one or more of a plurality of, for example, including at least one of A, B, and C, and can mean including any one or more elements selected from the set consisting of A, B, and C. In addition, the terms "first", "second", and "third" in this application are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features.

[0024] In order to enable those skilled in the art to better understand the technical solution of the present application, the technical solution of the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0025] See also Figure 1 , Figure 1 The method can be executed by an electronic device with computing functions, such as a microcomputer, a server, a notebook computer, a tablet computer, or other mobile devices.

[0026] It should be noted that if there are substantially the same results, the method of the present application is not limited to Figure 1 The process sequence shown is limited.

[0027] In some possible implementations, the method may be implemented by a processor calling computer-readable instructions stored in a memory, such as Figure 1 As shown, the method may include the following steps:

[0028] S11: Obtain multi-view lane line samples, where the multi-view lane line samples are pictures of lane lines taken from different angles.

[0029] The multi-view lane line samples are pictures taken from different angles of the lane line. For example, road pictures taken from different angles at the same time are obtained from pictures / videos taken by multiple cameras as multi-view lane line samples, and the corresponding road pictures contain clear lane lines. Among them, multiple cameras can be installed at different positions of the vehicle. For example, the multi-view lane line samples can include pictures obtained by the front camera on the vehicle, pictures obtained by the left camera on the vehicle, pictures obtained by the right camera on the vehicle, etc.

[0030] S12: Process the multi-view lane line samples to obtain multi-view lane line-free samples.

[0031] By processing the multi-view lane line samples, that is, pre-processing the road pictures containing lane lines at different viewpoints, the road pictures without lane lines at different viewpoints are generated as the multi-view lane line-free samples.

[0032] Among them, the multi-view lane-line-free samples are used for training the lane line generation model of the lane lines. For example, the lane line generation model is obtained by using the obtained multi-view lane-line-free samples for model training. The lane line generation model can generate corresponding virtual lane lines in the multi-view lane-line-free images. The virtual lane lines can be used for lane tracking and path planning of autonomous driving vehicles.

[0033] In this embodiment, multi-perspective lane line samples are obtained, which are pictures obtained by photographing lane lines from different angles. The multi-perspective lane line samples are processed to obtain multi-perspective lane-free samples, which are used to train a lane line generation model for lane lines. This allows direct generation of virtual lane lines when there are no lane lines or the lane lines are unclear, thereby enhancing the perception capability of the autonomous driving system and improving the stability of vehicle driving.

[0034] In some embodiments, multi-perspective lane line samples are processed to obtain multi-perspective lane line-free samples, including: segmenting the multi-perspective lane line samples to extract multi-view lane line areas corresponding to the lane lines in the multi-perspective lane line samples; performing lane line detection on the multi-view lane line areas, and erasing the lane lines in the multi-view lane line areas to obtain multi-perspective lane line-free samples.

[0035] The multi-view lane line samples are segmented to extract the multi-view lane line regions corresponding to the lane lines in the multi-view lane line samples. For example, the multi-view lane line images are segmented using a road surface segmentation model to identify the road surface in different view images, thereby obtaining images that only retain road surface information. The road surface segmentation model can be a deep learning model, such as a pre-trained semantic segmentation network model, i.e., a SETR model, a TransUNet model, a SegFormer model, a MaskFormer model, etc. Lane line detection is performed on the multi-view lane line region using a lane line detection model, i.e., lane line recognition is performed on the image that only retains road surface information, and lane line information is extracted therefrom. The lane line detection model can also be a deep learning model, such as a U-Net model, a ResNet model, etc. Further, the lane lines are erased in the multi-view lane line region to obtain multi-view lane line-free samples. For example, a preset erasing method (e.g., an inpaint method) is used to erase the lane lines in the image that only retains road surface information, thereby generating a road image without lane lines as a multi-view lane line-free sample. Among them, the road pictures without lane lines and the lane lines after erasure can also be reviewed to eliminate incorrectly labeled data and improve the accuracy of the data.

[0036] See also Figure 2 , Figure 2 is a flow chart of a lane line generation method according to an embodiment of the present application. The method can be applied to an on-board device having functions such as calculation. It should be noted that if there are substantially the same results, the method of the present application is not limited to the above. Figure 2 The process sequence shown is limited.

[0037] In some possible implementations, the method may be implemented by a processor calling computer-readable instructions stored in a memory, such as Figure 2 As shown, the method may include the following steps:

[0038] S21: Obtain a multi-view lane-free image.

[0039] Acquire multi-perspective lane-line-free images, for example, road images / videos taken at the same time by multiple cameras installed on the vehicle while the vehicle is driving, from which road images with different perspectives are acquired, and the lane lines in the corresponding road images are unclear or do not exist.

[0040] S22: Process the multi-view lane-line-free image to obtain a multi-view lane-line-free region in the multi-view lane-line-free image.

[0041] The multi-view lane-free image is processed to obtain multi-view lane-free areas in the multi-view lane-free image. For example, the multi-view lane-free image is preset to identify road surfaces in road images at different perspectives from the multi-view lane-free image, thereby obtaining corresponding data features.

[0042] S23: Inputting the multi-view lane line-free area into the lane line generation model to obtain a multi-view virtual lane line image.

[0043] The multi-view lane-line-free area is input into the lane line generation model, that is, the trained lane line generation model is used for prediction to generate virtual lane line images under different perspectives. The virtual lane line images can be understood as virtual lane lines generated from the acquired multi-view lane-line-free images. The virtual lane lines can be used for lane tracking and path planning of autonomous driving vehicles.

[0044] The lane line generation model is obtained by training the multi-view lane line-free samples obtained by the above-mentioned sample image generation method. The multi-view lane line-free samples are obtained by the above-mentioned sample image generation method, and further, the lane line generation model is obtained by model training using the multi-view lane line samples and their corresponding multi-view lane line-free samples.

[0045] In this embodiment, a multi-view image with no lane lines is obtained, and the multi-view image with no lane lines is processed to obtain a multi-view area with no lane lines in the multi-view image with no lane lines, and the multi-view area with no lane lines is input into a lane line generation model to obtain a multi-view image with virtual lane lines. The lane line generation model trained based on multi-view samples with no lane lines is used to directly generate virtual lane lines when there are no lane lines or the lane lines are unclear, thereby enhancing the perception capability of the autonomous driving system and improving the stability of vehicle driving.

[0046] In some embodiments, Figure 3 As shown, Figure 3 : is a schematic diagram of a lane line generation model of an embodiment of the present application, and the lane line generation model includes at least a diffusion model and a control model. Among them, the control model can be a neural network model (for example, ControlNet), which can be used to guide the diffusion model to generate image content to achieve more accurate image generation. The diffusion model can be an image generation model based on deep learning (for example, a Stable Diffusion model), which can be used to generate pictures with virtual lane lines. And, the diffusion model can also include a feature encoder, which can be used to process the acquired multi-view lane-free pictures.

[0047] In some embodiments, a multi-perspective lane-free image is processed to obtain a multi-view lane-free area in the multi-perspective lane-free image, including: segmenting the multi-perspective lane-free image to extract a multi-view road surface area in the multi-perspective lane-free image; and splicing the multi-view road surface area to generate a multi-view lane-free area.

[0048] Segment the multi-view lane-free images to extract the multi-view road surface areas in the multi-view lane-free images. For example, obtain 7 lane-free images taken by 7 cameras with different view angles at the same time, use the road surface segmentation model to segment the road area of ​​the 7 lane-free images, and extract the road surface area from the 7 lane-free images, that is, obtain 7 images that only retain the road surface information, such as Figure 3 The multi-view road area is stitched to generate a multi-view lane-free area, for example, using a visual feature encoder (ie, Figure 3 The feature encoder shown in ) performs feature splicing on the 7 images that only retain the road surface information to generate a multi-view lane-free area, that is, to obtain data features of corresponding dimensions. For example, feature splicing is performed on the 7 images that only retain the road surface information, where the dimension of each image is 3, and then splicing can obtain 21-dimensional data features.

[0049] In some embodiments, a multi-view lane line-free area is input into a lane line generation model to obtain a multi-view virtual lane line image, including: inputting a multi-view lane line-free area into a control model to generate control conditions for the lane line generation model; based on preset noise information and control conditions, using a diffusion model to generate a multi-view virtual lane line image.

[0050] In some examples, a multi-view lane-free area is input into a control model to generate control conditions for a lane-generating model. For example, n lane-free images are processed, and data features obtained by feature splicing of n images that only retain road surface information are input into the control model. The control model generates corresponding control conditions and inputs them into the diffusion model. Furthermore, the diffusion model generates multi-view virtual lane line images based on preset noise information and control conditions. For example, the diffusion model performs steps such as gradually adding noise and removing noise based on the guidance of the input control conditions and according to preset noise information and preset time steps, thereby obtaining multi-view virtual lane line images. For example, virtual lane lines are generated in n lane-free images, such as Figure 3 W2 shown in .

[0051] The preset noise information is random noise, and the random noise can be determined based on multi-view lane line-free samples during the training process of the lane line generation model. For example, when training the lane line generation model, the initial random noise is obtained by adding noise to the multi-view lane line samples, and the initial random noise is restored with reference to the corresponding multi-view lane line-free samples to generate a picture with lane lines. In this process, the target random noise is gradually determined, and the target random noise is used as the preset noise information for the diffusion model to generate an image, so as to generate a virtual lane line in the lane line-free picture. In addition, the latitude of the random noise is consistent with the multi-view lane line-free area.

[0052] In an embodiment of the present application, a multi-view lane-line-free area is input into a control model to generate control conditions for a lane line generation model, and then a multi-view virtual lane line image is generated using a diffusion model based on preset noise information and control conditions. This enables direct generation of virtual lane lines when there are no lane lines or the lane lines are unclear, thereby enhancing the perception capability of the autonomous driving system and improving the stability of vehicle driving.

[0053] In some embodiments, a multi-view lane line-free area is input into a lane line generation model to obtain a multi-perspective virtual lane line image, including: inputting a multi-view lane line-free area into a control model to generate control conditions for the lane line generation model; based on preset noise information, prompt word information and control conditions, using a diffusion model to generate a multi-perspective virtual lane line image.

[0054] In some examples, a multi-view lane-free area is input into a control model to generate control conditions for a lane-free generation model. For example, n lane-free images are processed, and data features obtained by feature splicing of n images that only retain road surface information are input into the control model. The control model generates corresponding control conditions and inputs them into the diffusion model. Furthermore, the diffusion model generates multi-view virtual lane line images based on preset noise information, prompt word information, and control conditions. For example, the diffusion model performs steps such as gradually adding noise and removing noise based on the guidance of the input control conditions and according to preset noise information, prompt word information, and preset time steps, thereby obtaining multi-view virtual lane line images. For example, virtual lane lines are generated in the n acquired lane-free images, such as Figure 3 W2 shown in .

[0055] Among them, the preset noise information is random noise, and the random noise can be determined based on multi-view lane line-free samples during the training process of the lane line generation model. For example, when training the lane line generation model, the initial random noise is obtained by adding noise to the multi-view lane line samples, and the initial random noise is restored to generate a picture with lane lines with reference to the corresponding multi-view lane line-free samples. In this process, the target random noise is gradually determined, and the target random noise is used as the preset noise information for the diffusion model to generate an image to generate virtual lane lines in the lane line-free picture. The latitude size of the random noise is consistent with the multi-view lane line-free area. In addition, the prompt word information can be used to clarify the generation target of the diffusion model. For example, the prompt word can be information about the subject, resolution, color, etc. in the multi-view virtual lane line picture.

[0056] In an embodiment of the present application, a multi-view lane-line-free area is input into a control model to generate control conditions for a lane line generation model, and then a multi-view virtual lane line image is generated using a diffusion model based on preset noise information, prompt word information, and control conditions. This enables direct generation of virtual lane lines when there are no lane lines or the lane lines are unclear, thereby enhancing the perception capability of the autonomous driving system and improving the stability of vehicle driving.

[0057] In some embodiments, the prompt word information includes at least: camera parameters, the camera is used to obtain multi-view images without lane lines; lane line attributes; road information.

[0058] The prompt word information may include camera parameters, for example, obtaining the internal and external parameters of the camera of the multi-view lane line-free image, encoding the internal and external parameters of the camera to obtain the corresponding prompt words, and inputting them into the diffusion model, so that the multi-view virtual lane line images generated by the diffusion model can correspond to the imaging angles of different cameras, etc. The prompt word information may include lane line attributes, such as the type of lane line (for example, solid lane line, dashed lane line, etc.), color, etc., encoding the lane line attributes to obtain the corresponding prompt words, and inputting them into the diffusion model, so that the virtual lane lines in the multi-view virtual lane line images generated by the diffusion model are more in line with reality, etc. The prompt word information may also include road information, such as curvature information and slope information of the road, encoding the road information to obtain the corresponding prompt words, and inputting them into the diffusion model, so that the virtual lane lines in the multi-view virtual lane line images generated by the diffusion model are more in line with reality, which can improve the adaptability of the autonomous driving system.

[0059] In some embodiments, obtaining a multi-perspective image without lane lines includes: obtaining a multi-perspective image with lane lines; segmenting the multi-perspective image with lane lines to extract a multi-view lane line area corresponding to the lane lines in the multi-perspective image with lane lines; performing lane line detection on the multi-view lane line area, and erasing the lane lines in the multi-view lane line area to obtain a multi-perspective image with no lane lines.

[0060] In some examples, a multi-view lane line image is obtained, for example, road images of different viewpoints at the same time are obtained from images / videos taken by multiple cameras, and the corresponding road images contain complex lane lines, so that the autonomous driving system cannot accurately identify them. At this time, the multi-view lane line image is segmented to extract the multi-view lane line area corresponding to the lane line in the multi-view lane line image, for example, the multi-view lane line image is segmented using a road surface segmentation model, and the road surface area of ​​the image under different viewpoints is identified, thereby obtaining an image that only retains road surface information. Lane line detection is performed on the multi-view lane line area, that is, lane line recognition is performed on the image that only retains road surface information, and lane line information is extracted therefrom. Further, the lane lines are erased in the multi-view lane line area to obtain a multi-view lane line-free sample, for example, the lane lines in the image that only retains road surface information are erased using a preset erasing method (for example, inpaint method), thereby generating a multi-view lane line-free image for generating a virtual lane.

[0061] Those skilled in the art will appreciate that, in the above method of specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.

[0062] See also Figure 4 , Figure 4 4 is a schematic diagram of the structure of an electronic device of an embodiment of the present application. The electronic device 40 includes a memory 41 and a processor 42 coupled to each other, and the processor 42 is used to execute program instructions stored in the memory 41 to implement the steps of the above-mentioned sample image generation method embodiment, or to implement the steps of the above-mentioned lane line generation method embodiment. In a specific implementation scenario, the electronic device 40 may include but is not limited to: a microcomputer, a server, which is not limited here.

[0063] Specifically, the processor 42 is used to control itself and the memory 41 to implement the steps of the above-mentioned sample image generation method embodiment, or to implement the steps of the above-mentioned lane line generation method embodiment. The processor 42 can also be called a CPU (Central Processing Unit), and the processor 42 may be an integrated circuit chip with signal processing capabilities. The processor 42 can also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field-programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. In addition, the processor 42 can be implemented by an integrated circuit chip.

[0064] See also Figure 5 , Figure 5 Schematic diagram of the structure of a non-volatile computer-readable storage medium according to an embodiment of the present application. The non-volatile computer-readable storage medium 50 is used to store a computer program 501. When the computer program 501 is executed by a processor, for example, Figure 4 When the processor 42 in the embodiment is executed, it is used to implement the steps of the above-mentioned sample image generation method embodiment, or to implement the steps of the above-mentioned lane line generation method embodiment.

[0065] The above description of various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced to each other, and for the sake of brevity, they will not be repeated herein.

[0066] In the several embodiments provided in the present application, it should be understood that the disclosed methods and related devices can be implemented in other ways. For example, the above-described related device implementations are only schematic, for example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation, such as units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication disconnection shown or discussed can be through some interfaces, indirect coupling or communication disconnection of devices or units, which can be electrical, mechanical or other forms.

[0067] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0068] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of each implementation method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

[0069] It is easy for a person skilled in the art to know that many modifications and changes can be made to the device and method while maintaining the teaching content of the present application. Therefore, the above disclosure should be considered as limited only by the scope of the appended claims.

Claims

1. A sample image generation method, characterized in that: include: Acquire multi-view lane line samples, where the multi-view lane line samples are pictures obtained by photographing the lane lines from different angles; Processing the multi-view lane line samples to obtain multi-view lane line-free samples; The multi-view lane line-free samples are used for training a lane line generation model of the lane lines.

2. The method according to claim 1, characterized in that The multi-view lane line samples are processed to obtain multi-view lane line-free samples, including: Segmenting the multi-view lane line samples to extract a multi-view lane line region corresponding to the lane line in the multi-view lane line samples; Lane line detection is performed on the multi-view lane line area, and the lane lines are erased in the multi-view lane line area to obtain the multi-view lane line-free samples.

3. A lane line generation method, characterized in that: include: Get multi-view lane-free images; Processing the multi-view lane-line-free image to obtain a multi-view lane-line-free region in the multi-view lane-line-free image; Inputting the multi-view lane line-free area into a lane line generation model to obtain a multi-view virtual lane line image; The lane line generation model is obtained by training with multi-view lane line-free samples obtained by the sample image generation method according to any one of claims 1 to 2.

4. The method according to claim 3, characterized in that The lane line generation model includes a diffusion model and a control model; The step of inputting the multi-view lane line free area into a lane line generation model to obtain a multi-view virtual lane line image includes: Inputting the multi-view lane line-free area into the control model to generate control conditions for the lane line generation model; Based on the preset noise information and the control conditions, the multi-view virtual lane line image is generated using a diffusion model.

5. The method according to claim 3, characterized in that: The lane line generation model includes a diffusion model and a control model; The step of inputting the multi-view lane line free area into a lane line generation model to obtain a multi-view virtual lane line image includes: Inputting the multi-view lane line-free area into the control model to generate control conditions for the lane line generation model; Based on the preset noise information, the prompt word information and the control conditions, the multi-view virtual lane line image is generated using a diffusion model.

6. The method according to claim 5, characterized in that The prompt word information at least includes: Parameters of a camera, the camera being used to obtain the multi-view lane-free image; Lane line attributes; Road information.

7. The method according to claim 3, characterized in that Processing the multi-view lane-line-free image to obtain a multi-view lane-line-free region in the multi-view lane-line-free image includes: Segmenting the multi-view image without lane lines to extract multi-view road surface areas in the multi-view image without lane lines; The multi-view road surface areas are spliced ​​to generate the multi-view lane-free area.

8. The method according to claim 3, characterized in that The obtaining of a multi-view lane-free image includes: Get multi-view lane line images; Segmenting the multi-view lane line image to extract a multi-view lane line area corresponding to the lane line in the multi-view lane line image; Lane line detection is performed on the multi-view lane line area, and the lane lines are erased in the multi-view lane line area to obtain the multi-view lane line-free image.

9. An electronic device, characterized in that: It includes a memory and a processor coupled to each other, and the processor is used to execute program instructions stored in the memory to implement the sample image generation method described in any one of claims 1 to 2, or to implement the lane line generation method described in any one of claims 3 to 8.

10. A non-volatile computer-readable storage medium having program instructions stored thereon, characterized in that: When the program instructions are executed by the processor, the sample image generation method described in any one of claims 1 to 2 is implemented, or the lane line generation method described in any one of claims 3 to 8 is implemented.

Images