A lane line detection method and device, a vehicle, and a storage medium

Abstract

Embodiments of the present application provide a lane line detection method and device, a vehicle and a storage medium. The method comprises: acquiring at least one area image captured relative to a current road area and position coding of a camera corresponding to each area image; determining target feature information of the current road area in a vehicle body coordinate system according to each area image and the corresponding position coding; and determining a lane line detection result in the current road area according to the target feature information and in combination with a set lane line detection model. With this method, area images captured by multiple cameras relative to the current road area can be fused, and the area images can be uniformly mapped to the same vehicle body coordinate system based on the corresponding position coding, so that the lane line detection is converted from two-dimensional image information to three-dimensional feature information, without the need for rear-end ranging, splicing, fusion and deduplication, thereby reducing detection errors and improving lane line detection effect.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a method, device, vehicle, and storage medium for detecting lane lines. Background Technology

[0002] In the field of autonomous driving perception, lane lines serve as a crucial basis for vehicle movement, and accurate lane line perception is a necessary prerequisite for safe driving. Lane line detection refers to obtaining lane line information around the vehicle by analyzing and calculating data collected in real time by various sensors on the vehicle. Cameras are a very important sensor, as their images possess rich color and texture information, and neural networks can be used to detect lane lines around the vehicle.

[0003] To comprehensively detect lane markings around a vehicle, autonomous vehicles are often equipped with multiple cameras to cover a 360° detection range. Currently, most lane detection methods involve multiple cameras detecting lanes individually, measuring distances, and then stitching the data together to obtain the final lane list. This method has several drawbacks: each camera has a limited field of view, meaning only a portion of the lane lines can be seen within a single camera, leading to poor detection results; and the detected targets contain errors after distance measurement, resulting in complex parameter tuning and unsatisfactory results during the backend stitching and fusion process. Summary of the Invention

[0004] This invention provides a lane line detection method, device, vehicle, and storage medium that eliminates the need for back-end ranging, stitching, fusion, and deduplication, thereby reducing detection errors and improving lane line detection performance.

[0005] Firstly, this embodiment provides a method for detecting lane lines, including:

[0006] Acquire at least one region image captured relative to the current road area and the position code of the camera corresponding to each region image;

[0007] Based on the images of each region and their corresponding location codes, the target feature information of the current road region in the vehicle coordinate system is determined;

[0008] Based on the target feature information and the established lane detection model, the lane detection results in the current road area are determined.

[0009] Secondly, this embodiment provides a lane line detection device, including:

[0010] The acquisition module is used to acquire at least one area image captured relative to the current road area and the position code of the camera corresponding to each area image;

[0011] The feature determination module is used to determine the target feature information of the current road area in the vehicle coordinate system based on the images of each area and the corresponding position codes;

[0012] The result determination module is used to determine the lane detection result in the current road area based on the target feature information and the set lane detection model.

[0013] Thirdly, this embodiment provides a vehicle, the vehicle comprising:

[0014] The vehicles include:

[0015] Vehicle body;

[0016] A camera is mounted on the vehicle body;

[0017] A controller, communicatively connected to the camera, the controller comprising:

[0018] At least one processor; and

[0019] A memory communicatively connected to the at least one processor; wherein,

[0020] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the lane line detection method described in the first aspect embodiment.

[0021] Fourthly, this embodiment provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the lane line detection method described in any one of the first aspect embodiments.

[0022] This invention provides a lane line detection method, device, vehicle, and storage medium. The method includes: acquiring at least one region image captured relative to a current road area and the position code of the camera corresponding to each region image; determining target feature information of the current road area in a vehicle coordinate system based on each region image and the corresponding position code; and determining the lane line detection result in the current road area based on the target feature information and a set lane line detection model. This technical solution can fuse region images captured by multiple cameras on the current road area and uniformly map the region images to the same vehicle coordinate system based on the corresponding position codes, converting two-dimensional image information into three-dimensional feature information for lane line detection. Compared to the prior art where multiple cameras detect and measure distances separately, then stitch the images together to obtain a final lane line list, resulting in poor detection performance, this technical solution eliminates the need for backend distance measurement, stitching, fusion, and deduplication, reducing detection errors and improving lane line detection performance.

[0023] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart illustrating a lane line detection method provided in Embodiment 1 of the present invention;

[0026] Figure 1a This is an example diagram of the camera deployment on a vehicle in a lane line detection method provided in Embodiment 1 of the present invention;

[0027] Figure 2 This is a flowchart illustrating a lane line detection method provided in Embodiment 2 of the present invention;

[0028] Figure 2a This is an example diagram of the feature information splicing process in a lane line detection method provided in Embodiment 2 of the present invention;

[0029] Figure 2b This is a schematic diagram of lane line sampling points detected in a lane line detection method provided in Embodiment 2 of the present invention;

[0030] Figure 2c This is an example diagram of the measurement quantity in a lane line detection method provided in Embodiment 2 of the present invention;

[0031] Figure 2d This is an example diagram of the lane line processing result in a lane line detection method provided in Embodiment 2 of the present invention;

[0032] Figure 3 This is a schematic diagram of the structure of a lane line detection device provided in Embodiment 3 of the present invention;

[0033] Figure 4 This is a structural schematic diagram of a vehicle provided in Embodiment 4 of the present invention. Detailed Implementation

[0034] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0035] It should be noted that the terms "original," "target," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0036] Example 1

[0037] Figure 1 This is a flowchart illustrating a lane line detection method provided in Embodiment 1 of the present invention. This method is applicable to situations where lane lines are detected. The method can be executed by a lane line detection device, which can be implemented in hardware and / or software and can be configured in a vehicle.

[0038] like Figure 1 As shown, the lane line detection method provided in this embodiment may specifically include the following steps:

[0039] S110. Obtain at least one area image captured relative to the current road area and the position code of the camera corresponding to each area image.

[0040] The lane detection method provided in this embodiment can be used to detect lane lines on a road to obtain specific lane line information as a basis for vehicle driving. Specifically, the current road area can be understood as the surrounding road area of ​​the vehicle's current position. During vehicle travel, it is necessary to continuously detect lane line information contained within the road area to assist vehicle driving. To determine the lane line information of the current road area, several cameras are deployed on the vehicle to capture regional images of the current road area.

[0041] It should be noted that this embodiment does not impose specific limitations on the number and location of cameras deployed on the vehicle. The number and location of cameras can be reasonably deployed based on the detection range of the cameras themselves. Preferably, after the cameras are deployed on the vehicle, it is necessary to ensure that the area images captured by all cameras can completely cover the surrounding road area where the vehicle is located. For example, Figure 1a This is an example diagram illustrating the camera deployment on a vehicle in a lane line detection method provided in Embodiment 1 of the present invention. Figure 1a As shown, the vehicle is equipped with six cameras, cameras 1 to 6, which are located at the front, rear, and sides of the vehicle, respectively. Each camera has a corresponding detection range, and it can be seen that the detection range of the six cameras basically covers the road area around the vehicle.

[0042] The camera's position encoding represents the transformation relationship from the region image to the vehicle body coordinate system. The encoding information can be represented in matrix form, specifically including pixel coordinates and depth values ​​in the region image. Each camera corresponds to a position code, which is predetermined and fixed for each camera. The camera's position code can be determined through its intrinsic parameters. It is understood that the size of the captured image from each camera is fixed. Given the length and width of the captured image, it can be sampled according to a set image size unit. Sampling the image width yields a series of horizontal coordinates, and sampling the image length yields a series of vertical coordinates. By sampling the width and length of the image, it can be divided into multiple pixels, each corresponding to a coordinate. The captured image, through sampling and partitioning, generates a pixel coordinate matrix.

[0043] In addition, each camera has a corresponding detection distance. The detection distance can be sampled according to a set sampling distance to obtain a series of distance values, forming a distance matrix. The pixel coordinate matrix obtained in the above steps is then concatenated with the distance matrix to obtain the concatenated result. By applying the transformation relationship between the image coordinate system and the camera coordinate system to the concatenated result, the camera's position code can be obtained.

[0044] It should be noted that when multiple cameras deployed on the vehicle capture area images in real time, the time at which each camera captures the image may differ. To obtain area images at the same time as possible, it is necessary to synchronize the area images captured by each camera. For example, the timestamp of the previous camera is used as the reference, and the area image from other cameras that is closest to that time is selected.

[0045] S120. Based on the images of each region and their corresponding location codes, determine the target feature information of the current road area in the vehicle coordinate system.

[0046] Specifically, the target feature information can be understood as converting the region image into relevant information in the vehicle body coordinate system. In this embodiment, through position encoding and network learning, the mapping relationship between the region image in the image coordinate system and the coordinates in the vehicle body coordinate system can be determined. After determining the mapping relationship through network learning, it can be determined from which point in the region image the feature information needs to be obtained for the coordinates of a certain point in the vehicle body coordinate system. For example, assuming the coordinates of ten meters ahead in the vehicle body coordinate system, the corresponding image feature information can be obtained according to the mapping relationship.

[0047] In this embodiment, after acquiring a region image relative to the current road area, the region image can be processed using a deep learning method to obtain the corresponding feature map. For example, the deep learning method can be based on a deep residual network (ResNet) to obtain multi-scale images, and then fuse these multi-scale images to obtain the feature map. Before obtaining the feature map, each region image can be sampled to obtain images that better meet the requirements of deep learning input. For example, the original region images captured by each camera are six images taken from different angles, each with a resolution of 1920×1080. First, a 2x downsampling through bilinear interpolation is used to obtain a 960×540 image, which is then input into the deep learning model.

[0048] In this embodiment, after acquiring a region image relative to the current road area, image feature information is extracted using deep learning methods to determine a feature map. After acquiring the feature map, the corresponding location encoding is combined with the network model to learn and determine the mapping relationship between the feature map and the vehicle coordinate system. To obtain the mapping relationship between the feature map and the vehicle coordinate system, the image features can be transformed from image space to camera coordinate system using the feature map and location encoding, thus determining the feature information of the feature map in the corresponding camera coordinate system.

[0049] It is known that the detection range of each camera is fixed. Based on the set resolution, the detection range is divided into several grid units, each with corresponding coordinates in the camera coordinate system. For example, assuming the detection range of the forward-looking camera on a vehicle is 30 meters to the left and right of the vehicle and 50 meters in front, and the detection range is set as x∈[-30, 30], y∈[0, 50], where x represents the detection range to the left and right of the vehicle, y represents the detection range in front of the vehicle, and the resolution is 1 (in meters), the detection range contains [(30-(-30)) / 1]×[(50-0) / 1]=3000 grid units. If the feature dimension is set to 128, then the size of the feature information is [128, 3000], with an initial value of 0. The feature dimension can be understood as the length of the feature information; a larger value indicates a longer length and more features. In this embodiment, based on historical experience, a feature dimension of 128 is preferred. It is clear that the camera's detection range in this step is consistent with the camera's detection range when the position code is determined.

[0050] In existing technologies, each camera detects and measures distances individually before stitching together the data to obtain the final lane list. This method has several drawbacks: each camera has a limited field of view, meaning only a portion of the lane lines can be seen from a single camera, leading to poor detection results; and the detected targets contain errors after distance measurement, resulting in complex parameter tuning and unsatisfactory results during the backend stitching and fusion process. In this embodiment, regional image information from multiple cameras is fused and unified into the same vehicle coordinate system. Three-dimensional voxel features are generated from two-dimensional image features, eliminating the need for backend distance measurement, stitching, fusion, and deduplication.

[0051] Following the above description, after obtaining the feature information in each camera coordinate system, the feature information is transformed to the vehicle coordinate system using a conversion formula from the camera coordinate system to the vehicle coordinate system. Based on the camera's detection range and resolution, the detection range is divided into several grid cells, each corresponding to coordinates in the camera coordinate system. Each grid cell's coordinates in the camera coordinate system correspond to feature information. Using camera extrinsic parameters, the coordinates in the camera coordinate system are converted to coordinates in the vehicle coordinate system. The feature information corresponding to each grid cell is then filled into the coordinates in the vehicle coordinate system, completing the feature information stitching and determining the target feature information.

[0052] S130. Based on the target feature information and the established lane detection model, determine the lane detection results in the current road area.

[0053] The lane detection network model can be understood as a neural network model capable of detecting lane lines. The lane detection model includes multiple detection network heads, through which lane line categories, confidence levels, offsets, lateral direction vectors, and longitudinal direction vectors can be obtained. It can be understood that before detecting lane-related information from target features, further extraction of target features can be performed to obtain more distinct feature information. In this step, each grid cell is used as the detection object, and the lane-related information corresponding to each grid cell is determined separately.

[0054] In this embodiment, after determining the lane line information, a preliminary screening of the lane lines is required to determine whether each grid cell contains lane lines or not. The preliminary screening can be achieved by comparing the confidence level of the lane line category corresponding to each grid cell with a pre-set confidence threshold. If the confidence level of the lane line category corresponding to the grid cell is higher than the confidence threshold, then the grid cell is determined to contain lane lines; if the confidence level of the lane line category corresponding to the grid cell is equal to or lower than the confidence threshold, then the grid cell is determined to be background.

[0055] After determining whether each grid cell contains lane lines, the lane line situation of each grid cell can be initially depicted based on the judgment result combined with the lane line category, offset, lateral direction vector, and longitudinal direction vector. Further, based on the lane line category, a set of grid cells with lane lines of the same category is determined. Clustering is then performed based on the distance between points and the distance of the direction vectors, resulting in all sampling points containing a lane line. These sampling points are then fitted to determine the lane lines. Additionally, based on the coordinates of the grid cells and the offset, the coordinates of the lane line center point can be determined. Finally, the lane line detection result for the current road area can be the lane line category, the coordinates of the lane line center point, and the fitted curve.

[0056] This invention provides a lane line detection method, which includes: acquiring at least one region image captured relative to the current road area and the position code of the camera corresponding to each region image; determining the target feature information of the current road area in the vehicle coordinate system based on each region image and the corresponding position code; and determining the lane line detection result in the current road area based on the target feature information and a set lane line detection model. Using this method, region images captured by multiple cameras on the current road area can be fused and mapped uniformly to the same vehicle coordinate system based on the corresponding position codes, converting two-dimensional image information into three-dimensional feature information for lane line detection. Compared to the existing technology where multiple cameras detect and measure distances separately, then stitch the images together to obtain the final lane line list, resulting in poor detection performance, this technical solution eliminates the need for backend distance measurement, stitching, fusion, and deduplication, reducing detection errors and improving lane line detection performance.

[0057] As an optional embodiment of the present invention, based on the above embodiments, the step of determining the camera position code corresponding to each region image includes:

[0058] a1. Divide the length and width of the images captured by each camera according to the first sampling distance to determine the pixel coordinate matrix corresponding to the images captured by each camera.

[0059] The first sampling distance can be set according to the actual situation. In this step, the length and width of the captured image of each camera are divided according to the first sampling distance to generate grid sampling points. Each grid sampling point can be understood as a pixel, and each pixel corresponds to a coordinate. The coordinates of each pixel point constitute a pixel coordinate matrix. In this step, the meshgrid function is preferably used to generate grid sampling points to perform grid sampling on the captured image. Specifically, let the width of the captured image be w, and the first sampling distance be 1 pixel, to obtain a series of horizontal coordinates W = [0,1,…,w-2,w-1], where pixel is the image size unit. Let the length of the captured image be h, and the sampling distance be 1 pixel, to obtain a series of vertical coordinates H = [0,1,…,h-2,h-1].

[0060] b1. Divide the detection distance of each camera according to the second sampling distance to determine the distance matrix corresponding to each camera.

[0061] The second sampling distance can be determined based on the actual situation. In this step, the detection distance of each camera is fixed. The detection distance is divided according to the second sampling distance. Based on each distance sampling point, the distance matrix corresponding to each camera can be determined. Specifically, let the detection distance of a single camera be 1m-50m, and the second sampling distance be 0.5m, to obtain a series of distance values ​​Z = [1m, 1.5m, ... 49.5m, 50m].

[0062] c1. Based on the pixel coordinate matrix and distance matrix, and combined with the set array construction algorithm, determine the position code of each camera.

[0063] The array construction algorithm specified here refers to the array concatenation function `cat`. In this step, the pixel coordinate matrix and the distance matrix are concatenated using the array concatenation function `cat` to construct a multidimensional matrix. The multidimensional matrix is ​​then transformed using camera intrinsic parameters to obtain the position encoding information.

[0064] Furthermore, based on the pixel coordinate matrix and distance matrix, and combined with the established array construction algorithm, the position code of each camera is determined, including:

[0065] c11. Determine the unit matrix based on the length and width of the image captured by the camera.

[0066] Specifically, the identity matrix is ​​determined based on the length and width of the captured image. Assuming the width of the captured image is w and the length is h, the size of the identity matrix is ​​[w, h, 1].

[0067] c12. Input the pixel coordinate matrix, distance matrix, and identity matrix into the set array construction algorithm to obtain the construction result.

[0068] Specifically, assume the constructed result is represented as PI, where PI is an empty array list, and ones is a two-dimensional identity matrix of size w*h. Assume the one-dimensional identity matrix is ​​represented as Z, and iterate through the one-dimensional identity matrix Z, P... i =cat(P×Z[i], ones×Z[i], 2), where cat is the array concatenation function, parameter 2 is the concatenation dimension index, and P i The dimensions are [w, h, 3]. P I For all generated P i The splicing result at latitude 2, i.e., the constructed result, has a data length of Z of (50-1) / 0.5 = 98, therefore P I The final dimensions are [w, h, 98, 3].

[0069] c13. Based on the construction results and the transformation formula between the image coordinate system and the camera coordinate system, determine the camera's position code.

[0070] The transformation formula from the image coordinate system to the camera coordinate system is as follows: Where u is the x-coordinate of the image, v is the y-coordinate of the image, and X... C Y C Z C Here are the coordinates in the camera coordinate system, and K is the camera intrinsic parameter.

[0071] Each camera intrinsic parameter can be represented as: i∈[1,6],f x f xThe camera focal length is used; in this embodiment, the camera intrinsic parameters can be used as known parameters.

[0072] Specifically, the camera's location code can be represented as: P C =K -1 ×P I The meanings of the parameters in this step can be found in the description above, and will not be elaborated further here.

[0073] For example, the pseudocode logic implementation of this step can be represented as follows:

[0074]

[0075] This optional embodiment specifies the steps for determining the camera's position code. The camera's position code can be obtained through the camera's intrinsic parameters. The camera's position code is used to map image information to the vehicle coordinate system. It can be used to transform the camera's two-dimensional image features to the same vehicle coordinate system, generate three-dimensional features from the two-dimensional features, and improve the detection effect of lane lines.

[0076] Example 2

[0077] Figure 2 This is a flowchart illustrating a lane line detection method provided in Embodiment 2 of the present invention. This embodiment is a further optimization of the above embodiment. In this embodiment, the definition of "determining the target feature information of the current road area in the vehicle coordinate system based on the image of each region and the corresponding position code" is further optimized, and the definition of "determining the lane line detection result in the current road area based on the target feature information and the set lane line detection model" is also optimized.

[0078] like Figure 2 As shown in the figure, this embodiment 2 provides a lane line detection method, which specifically includes the following steps:

[0079] S210. Obtain at least one area image captured relative to the current road area and the position code of the camera corresponding to each area image.

[0080] S220. Based on the images of each region and their corresponding location codes, determine the feature information of the current road area in each camera coordinate system.

[0081] In this embodiment, the region image is used as input, and image information is extracted based on a deep learning method. This image information is a feature map in the image coordinate system. For example, the deep learning method in this step can be a neural network model based on a ResNet-34 (Deep Residual Network). In this embodiment, to meet the input requirements of the neural network model, the captured original image needs to be bilinearly interpolated to obtain an image that meets the resolution requirements. For example, assuming six images taken from different angles, each with a resolution of 1920×1080, the image resolution is first obtained by downsampling by a factor of 2 using bilinear interpolation to 960×540. Then, the image with the converted resolution is used as input to the neural network to extract image features. After obtaining multi-scale image information, the multi-scale image information is fused based on Feature Pyramid Networks (FPN) to obtain a feature map in the image coordinate system.

[0082] Furthermore, it is necessary to convert the feature maps from the image feature information into feature information in the camera coordinate system. Having obtained the feature maps in the image coordinate system corresponding to each camera, and combining them with the pre-determined camera position encoding, a set neural network model can transform the image features from the image space to the camera coordinate system, thus obtaining the feature information in the camera coordinate system. For example, the Transformer model can be preferred as the neural network model.

[0083] Furthermore, the step of determining the feature information of the current road area in each camera coordinate system based on the images of each region and their corresponding location codes can be described as follows:

[0084] a2. Input the images of each region into the set first network model to determine the feature map of the current road region in each image coordinate system.

[0085] The first network model can be understood as primarily using a deep residual neural network, ResNet-34. For example, assuming the resolution of the region image is 1920×1080, to better meet the resolution requirements of the first network model, the region image needs to be downsampled by a factor of 2 using bilinear interpolation to obtain an image with a resolution of 960×540. A size of 6×3×960×540 (6 being the number of images, 3 being the number of image channels, and 960×540 being the image's width and height) is used as the input to the first network model. The output sizes of the last three layers are 6×128×120×68, 6×256×60×34, and 6×512×30×17, respectively. Based on this, an FPN layer is added, with an output size of 6×64×120×68, representing the feature map in the image coordinate system. This completes the extraction of image information.

[0086] b2. Based on each feature map and the corresponding position code, and combined with the established second network model, determine the feature information of the current road area in the camera coordinate system.

[0087] The second network model is set as a Transformer model. Let Q... i The feature information obtained by transforming the image features corresponding to the i-th camera is initially set as follows: Assume the detection range of the forward-looking camera on the vehicle is 30 meters to the left and right of the vehicle and 50 meters in front. Set the detection range x∈[-30,30],∈[0,50], the resolution is 1, and the unit is m. There are a total of ((30-(-30)) / 1)×((50-0) / 1)=3000 feature grid units. Let the feature dimension be 128, then the size of Q is [128,3000], and the initial value is 0.

[0088] Specifically, the feature maps and their corresponding location codes are input into the second network model, which outputs the feature information of the current road area in the camera coordinate system. For example, assume F... i For the feature map, P C i Q is the location code corresponding to the i-th camera. i This represents the feature information obtained for the i-th camera in the camera coordinate system.

[0089] S230. Perform splicing processing on each feature information to determine the target feature information of the current road area in the vehicle coordinate system.

[0090] In this step, the feature information from each camera needs to be fused into the same vehicle coordinate system, and the feature information is then stitched together. The stitched feature information in the vehicle coordinate system is denoted as the target feature information. For example, assuming the combined detection range of each camera is 50 meters in front and behind the vehicle, and 30 meters to the left and right, with a resolution of 0.5m × 0.5m and a total of 120 × 200 grids, and assuming the feature information in the vehicle coordinate system is F... Veh The latitude is 128 and the dimensions are [128, 120, 200]. The feature information Q corresponding to each camera was obtained through the above steps. i The Q coordinate system can be transformed from the camera to the vehicle body coordinate system using the transformation matrix. i Mapping to F Veh From this, we can know that Q i Each grid cell has its corresponding coordinates [xyz] in the camera coordinate system. c via camera extrinsic C2V i Its coordinates [xyz] in the vehicle body coordinate system were calculated. VehThe feature vector with latitude 128 at that point, i.e., the feature information, is then filled into F. Veh In the process, after all Q-mappings are completed, the final target feature information F is obtained. Veh .

[0091] Furthermore, the steps of stitching together the various feature information to determine the target feature information of the current road area in the vehicle coordinate system can be described as follows:

[0092] a3. Rearrange the feature information to obtain the three-dimensional arrays.

[0093] Specifically, the feature information Q corresponding to each camera i The matrix is ​​rearranged to obtain a three-dimensional array with dimensions of 128*60*50. The rearrangement method uses the `reshape` operation. The `reshape` function in MATLAB transforms a specified matrix into a matrix of a specific dimension while keeping the number of elements in the matrix unchanged. The `reshape` function can readjust the number of rows, columns, and dimensions of a matrix.

[0094] b3. Based on the detection range and resolution of each camera, determine the camera coordinates of each grid cell within the detection range in the camera coordinate system.

[0095] For example, suppose the detection range is 30 meters to the left and right of the vehicle, and 50 meters in front of the vehicle, with x∈[-30, 30], y∈[0, 50], and a resolution of 1 (in meters). There are a total of ((30-(-30)) / 1)×((50-0) / 1)=3000 feature grid cells. Each grid cell corresponds to camera coordinates in the camera coordinate system. For example, the coordinates of the center point of the grid cell can be used as the grid cell's coordinates.

[0096] c3. Based on the coordinates of each camera and the camera's extrinsic parameters, determine the target coordinate matrix of each mesh unit in the vehicle body coordinate system.

[0097] Among these, the camera extrinsic parameters can be directly obtained as known parameters. The camera coordinates in each camera coordinate system are transformed to the vehicle body coordinate system, and the transformed coordinates in the vehicle body coordinate system can be represented as [x, y, z]. Veh The conversion formula is: Where, [x, y, z] c The coordinates are those of the camera. In this embodiment, the coordinate matrix of each grid cell in the vehicle body coordinate system is denoted as the target coordinate matrix.

[0098] d3. Fill the corresponding positions in the target coordinate matrix with the feature information in each three-dimensional array as matrix elements to determine the target feature information of the current road area in the vehicle coordinate system.

[0099] For example, according to [xyz]Veh Q i The eigenvectors of each 128-dimensional grid are filled into F. Veh middle.

[0100] For example, Figure 2a This is an example diagram of the feature information stitching process in a lane line detection method provided in Embodiment 2 of the present invention, as shown below. Figure 2a As shown, the feature information Q corresponding to each camera is... i Q i-1 Rearranging the array yields a three-dimensional array Q with dimensions 128*60*50. i Q i-1 Each grid cell has its corresponding camera coordinates [x, y, z]. c According to the formula Transform it to the vehicle body coordinate system [x, y, z] Veh According to [xyz] Veh Q i The eigenvectors of each 128-dimensional grid are filled into F. Veh middle.

[0101] S240. Input the target feature information into the set lane line detection model to obtain the lane line information corresponding to each grid cell.

[0102] The lane detection model is a neural network structure consisting of an 8-layer residual network module and 4 detection network heads. The target feature information is input into the 8-layer residual network module for further feature extraction, resulting in more accurate feature information. For example, assume that step S230 obtains target feature information F with a size of 128×120×200. VehThe final feature information of 256×120×200 is obtained after passing through 8 layers of residual network modules. Then, 4 detection network heads are used to obtain the class head, offset head, horizontal direction vector, and vertical direction vector for each grid. The class head outputs the confidence score of the grid at the corresponding position belonging to a certain class, with an output size of n×120×200, where n is the number of lane line classes, defined as a Class Map. The offset head outputs the positional deviation of the center point at the corresponding position, with an output size of 2×120×200, defined as an Offset Map. The horizontal direction head and vertical direction head output the horizontal and vertical direction vectors of the lane line center at the corresponding position, respectively, with an output size of 1×120×200, defined as HDir Map and VDir Map.

[0103] S250: Process the lane line information to determine the lane line detection results in the current road area.

[0104] The lane line information includes the lane line category and confidence level at the corresponding location, the center point offset, and the lateral and longitudinal direction vectors. In this step, lane lines are initially filtered based on their confidence level, identifying locations containing lane lines and those not containing them; the locations not containing lane lines are designated as background. Further, clustering is performed based on the center point offset and the lateral and longitudinal direction vectors in the lane line information to obtain sampling points for each lane line, which are then fitted to obtain the final state of the lane lines.

[0105] Further, the lane line information is processed to determine the lane line detection results in the current road area, including:

[0106] a4. Extract the confidence level of the lane line category from the lane line information of each lane.

[0107] Specifically, lane line information includes the lane line category and confidence level at the corresponding location, the center point offset, and the horizontal and vertical direction vectors. The confidence level of the lane line category is extracted from the lane line information. In essence, each grid cell, i.e., each location, corresponds to a lane line information.

[0108] b4. Grid cells with a confidence level greater than the set threshold are identified as target grid cells with lane lines.

[0109] The threshold is determined based on a large amount of historical data. Specifically, the confidence level of the lane line category of each grid cell is compared with the set threshold. When the confidence level of the lane line category corresponding to the grid cell is greater than the set threshold, the grid cell is determined to have lane lines, which is denoted as the target grid cell in this embodiment.

[0110] c4. Otherwise, determine the grid cell as a background without lane lines.

[0111] Specifically, if the confidence level of the lane line category corresponding to a grid cell is less than or equal to a set threshold, then the grid cell is determined to be a background without lane lines.

[0112] For example, Figure 2b This is a schematic diagram of lane line sampling points detected in a lane line detection method provided in Embodiment 2 of the present invention. Figure 2b As shown, the grid with arrows represents lane line sampling points that have been detected. Dark arrows represent the lateral and longitudinal direction vectors, while light arrows represent the center point offset.

[0113] d4. Based on the center point offset and the horizontal and vertical direction vectors in the target grid cells and lane line information, determine the lane line detection results in the current road area.

[0114] Specifically, once the target grid cell is determined, the location of lane lines in the current road area is determined. Further, based on the center point offset and the horizontal and vertical direction vectors in the lane line information corresponding to each target grid cell, the lane lines can be clustered and fitted to determine the lane line detection results in the current road area.

[0115] Further, determining the lane line detection result in the current road area based on the target mesh cell and the center point offset, and the lateral and longitudinal direction vectors in the lane line information includes:

[0116] b41. Classify each target grid cell according to the lane line category and determine the target grid cell corresponding to each lane line category.

[0117] Specifically, each target grid cell is classified according to lane line type, and each lane line type corresponds to a target grid cell of that type.

[0118] b42. Cluster the target grid cells corresponding to each lane line category according to the horizontal and vertical direction vectors to obtain the sampling points contained in each lane line.

[0119] In this step, the sampling points corresponding to the target grid cell for each lane line category are clustered to obtain the sampling points contained in each lane line. An exemplary clustering method could be the density-based clustering algorithm DBSCAN, using the distance between points and the distance between their direction vectors as metrics. For example, Figure 2c This is an example diagram of the measurement quantity in a lane line detection method provided in Embodiment 2 of the present invention, as shown below. Figure 2c As shown, the distance between points is the Euclidean distance D1, and the direction vector distance is the angle between the direction vectors of the two points D2.

[0120] b43. Perform fitting processing on the sampling points contained in each lane line to determine the fitting curve of each lane line.

[0121] Specifically, the clustered lane lines contain all the sampling points of that lane line. Further fitting processing is performed on the sampling points contained in each lane line. For example, a cubic curve fitting algorithm is used to calculate the cubic fitting equation of the lane line, which can be expressed as: y = k1 × x 3 +k2×x 2 +k3×x+c, where k1, k2, k3, and c are constants.

[0122] b44. Determine the coordinates of the lane line center point based on the center point offset in the lane information.

[0123] Specifically, the coordinates of each grid cell can be determined. For example, by determining the coordinates of the center point of the grid cell and combining them with the offset of the center point of the lane line, the coordinates of the center point of the lane line can be calculated.

[0124] b45. Use the lane line category, lane line center point coordinates, and fitted curve as the lane line detection results in the current road area.

[0125] Specifically, the lane line category, lane line center point coordinates, and fitted curve are used as the lane line detection results in the current road area. The output result for each lane line is:

[0126] Lane i =[class points k1 k2 k3 c]

[0127] Where class represents the lane category, points represents a series of lane center point coordinates, and [k1 k2 k3 c] represents the cubic fitted curve equation of the lane.

[0128] For example, Figure 2d This is an example diagram of the lane line processing result in a lane line detection method provided in Embodiment 2 of the present invention, as shown in the figure. Figure 2dAs shown, the lane line detection method provided in this embodiment can determine the lane line type, the coordinates of the lane line center point, and the curve shape of the lane line.

[0129] This embodiment captures regional images of the current road area using multiple cameras. First, a neural network extracts image information and obtains a positional code mapping the image information to the vehicle coordinate system. The image information and positional code are then learned by the neural network model to obtain feature information in the vehicle coordinate system. Finally, a lane detection network is used to obtain lane lines in the vehicle coordinate system. This technical solution can fuse image information from multiple cameras and unify it into the same vehicle coordinate system. Three-dimensional voxel features are generated from two-dimensional image features. Further high-dimensional features are extracted based on the three-dimensional voxel features to detect lane lines in three-dimensional space. This eliminates the need for backend ranging, stitching, fusion, and deduplication, directly obtaining a list of lane line-related information in the vehicle coordinate system, reducing detection errors and improving lane line detection performance.

[0130] Example 3

[0131] Figure 3 This is a schematic diagram of a lane line detection device provided in Embodiment 3 of the present invention. It is applicable to situations involving lane line detection. The device can be implemented in hardware and / or software and is generally integrated into a vehicle. Figure 3 As shown, the device includes: an acquisition module 31, a feature determination module 32, and a result determination module 33, wherein,

[0132] The acquisition module 31 is used to acquire at least one area image captured relative to the current road area and the position code of the camera corresponding to each area image;

[0133] The feature determination module 32 is used to determine the target feature information of the current road area in the vehicle coordinate system based on the images of each area and the corresponding position codes;

[0134] The result determination module 33 is used to determine the lane detection results in the current road area based on the target feature information and the set lane detection model.

[0135] Optionally, the device further includes an encoding determination module, which may include:

[0136] The first matrix determination unit is used to divide the length and width of the images captured by each camera according to the first sampling distance, and determine the pixel coordinate matrix corresponding to the images captured by each camera.

[0137] The second matrix determination unit is used to divide the detection distance of each camera according to the second sampling distance and determine the distance matrix corresponding to each camera.

[0138] The encoding determination unit is used to determine the position encoding of each camera based on the pixel coordinate matrix and the distance matrix, combined with a set array construction algorithm.

[0139] Optionally, the encoding determination unit is specifically used for:

[0140] Determine the identity matrix based on the first sampling distance and the second sampling distance;

[0141] The pixel coordinate matrix, distance matrix, and identity matrix are input into the set array construction algorithm to obtain the construction result;

[0142] Based on the construction results, and combined with the transformation formula between the image coordinate system and the camera coordinate system, the camera position code is determined.

[0143] Optionally, the feature determination module 32 includes:

[0144] The information determination unit is used to determine the feature information of the current road area in each camera coordinate system based on the images of each area and the corresponding location codes;

[0145] The information stitching unit is used to stitch together various feature information to determine the target feature information of the current road area in the vehicle coordinate system.

[0146] Optionally, the information determining unit is specifically used for:

[0147] The images of each region are input into the first network model to determine the feature map of the current road region in each image coordinate system;

[0148] Based on each feature map and its corresponding location encoding, and combined with the established second network model, the feature information of the current road area in the camera coordinate system is determined.

[0149] Optionally, the information splicing unit is specifically used for:

[0150] The feature information is rearranged to obtain the three-dimensional arrays;

[0151] Based on the detection range and resolution of each camera, determine the camera coordinates of each grid cell within the detection range in the camera coordinate system;

[0152] Based on the coordinates of each camera and the camera's extrinsic parameters, determine the target coordinate matrix of each mesh unit in the vehicle body coordinate system;

[0153] The feature information in each three-dimensional array is used as matrix elements to fill the corresponding positions in the target coordinate matrix, thereby determining the target feature information of the current road area in the vehicle coordinate system.

[0154] Optionally, the result determination module 33 may include:

[0155] The lane line information determination unit is used to input target feature information into the set lane line detection model to obtain lane line information corresponding to each grid cell.

[0156] The result determination unit is used to process the lane line information of each lane and determine the lane line detection results in the current road area.

[0157] Optionally, the result determination unit is specifically used for:

[0158] Extract the confidence level of lane line category from the lane line information of each lane;

[0159] Grid cells with a confidence level greater than a set threshold are identified as target grid cells with lane lines;

[0160] Otherwise, determine the grid cell as a background without lane lines;

[0161] Based on the center point offset and the lateral and longitudinal direction vectors in the target grid cells and lane line information, the lane line detection results in the current road area are determined.

[0162] Optionally, the result determination unit is used to determine the lane detection result in the current road area based on the center point offset and the lateral and longitudinal direction vectors in the target mesh cell and lane line information, including the following steps:

[0163] Classify each target grid cell according to lane line type and determine the target grid cell corresponding to each lane line type;

[0164] Cluster the target grid cells corresponding to each lane line category based on the horizontal and vertical direction vectors to obtain the sampling points contained in each lane line;

[0165] The sampling points contained in each lane line are fitted to determine the fitting curve of each lane line.

[0166] Determine the coordinates of the lane line center point based on the center point offset in the lane information;

[0167] The lane line category, lane line center point coordinates, and fitted curve are used as the lane line detection results in the current road area.

[0168] The lane line detection device provided in this embodiment of the invention can execute the lane line detection method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0169] Example 4

[0170] Figure 4 This is a structural schematic diagram of a vehicle provided in Embodiment 4 of the present invention. Figure 4As shown, the vehicle provided in Embodiment 4 of the present invention includes: a vehicle body (not shown in the figure); a camera 2, mounted on the vehicle body; and a controller 4, which is communicatively connected to the camera 2.

[0171] Controller 4 includes: one or more processors 41 and a storage device 42; the processors 41 in controller 4 may be one or more. Figure 4 Taking a processor 41 as an example; storage device 42 is used to store one or more programs; the one or more programs are executed by the one or more processors 41, so that the one or more processors 41 implement the lane line detection method as described in any one embodiment of the present invention.

[0172] The processor 41 and storage device 42 in controller 4 can be connected via a bus or other means. Figure 4 Taking the example of a connection between China and Israel via a bus.

[0173] The storage device 42 in the controller 4 serves as a computer-readable storage medium, which can be used to store one or more programs. These programs can be software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the lane line detection method provided in Embodiment 1 or 2 of this invention (e.g., attached...). Figure 3 The lane line detection device shown includes modules such as an acquisition module 31, a feature determination module 32, and a result determination module 33. The processor 41 executes various functional applications and data processing of the controller 4 by running software programs, instructions, and modules stored in the storage device 42, thereby implementing the lane line detection method in the above method embodiment.

[0174] Storage device 42 may include a stored program area and a stored data area, wherein the stored program area may store the operating system and applications required for at least one function; the stored data area may store data created based on the use of controller 4, etc. Furthermore, storage device 42 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, storage device 42 may further include memory remotely located relative to processor 41, and this remote memory may be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0175] Input device 43 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of controller 4. Output device 44 may include display devices such as a display screen.

[0176] Furthermore, when one or more programs included in the aforementioned controller 4 are executed by one or more processors 41, the programs perform the following operations:

[0177] Acquire at least one region image captured relative to the current road area and the position code of the camera corresponding to each region image;

[0178] Based on the images of each region and their corresponding location codes, the target feature information of the current road region in the vehicle coordinate system is determined;

[0179] Based on the target feature information and the established lane detection model, the lane detection results in the current road area are determined.

[0180] Example 5

[0181] Embodiment 5 of the present invention provides a computer-readable storage medium having a computer program stored thereon. When executed by a processor, the program is used to perform a lane line detection method, the method comprising:

[0182] Acquire at least one region image captured relative to the current road area and the position code of the camera corresponding to each region image;

[0183] Based on the images of each region and their corresponding location codes, the target feature information of the current road region in the vehicle coordinate system is determined;

[0184] Based on the target feature information and the established lane detection model, the lane detection results in the current road area are determined.

[0185] Optionally, when executed by a processor, the program can also be used to execute the lane line detection method provided in any embodiment of the present invention.

[0186] The computer storage medium of this invention can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable CD-ROM, optical storage device, magnetic storage device, or any suitable combination thereof. The computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0187] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit programs for use by or in connection with an instruction execution system, apparatus, or device.

[0188] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, radio frequency (RF), etc., or any suitable combination thereof.

[0189] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0190] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A method for detecting lane lines, characterized in that, include: Acquire at least one region image captured relative to the current road area and the position code of the camera corresponding to each region image; Based on the images of each region and their corresponding location codes, the target feature information of the current road region in the vehicle coordinate system is determined; Based on the target feature information and the established lane detection model, the lane detection results in the current road area are determined. The step of determining the position encoding of the camera corresponding to each of the aforementioned regions includes: The length and width of the images captured by each camera are divided according to the first sampling distance to determine the pixel coordinate matrix corresponding to the images captured by each camera; The detection distance of each camera is divided according to the second sampling distance to determine the distance matrix corresponding to each camera; Based on the pixel coordinate matrix and the distance matrix, and combined with the set array construction algorithm, the position code of each camera is determined; The step of determining the position code of each camera based on the pixel coordinate matrix and the distance matrix, combined with a set array construction algorithm, includes: Determine the unit matrix based on the length and width of the images captured by each camera; The pixel coordinate matrix, the distance matrix, and the identity matrix are input into a set array construction algorithm to obtain the construction result; Based on the constructed results, and combined with the transformation formula between the image coordinate system and the camera coordinate system, the position code of the camera is determined.

2. The method according to claim 1, characterized in that, The step of determining the target feature information of the current road area in the vehicle coordinate system based on each of the region images and corresponding location codes includes: Based on the images of each region and their corresponding location codes, the feature information of the current road region in each camera coordinate system is determined; The aforementioned feature information is spliced ​​together to determine the target feature information of the current road area in the vehicle coordinate system.

3. The method according to claim 2, characterized in that, The step of determining the feature information of the current road area in each camera coordinate system based on the images of each area and their corresponding location codes includes: Each of the aforementioned regional images is input into a set first network model to determine the feature map of the current road region in each image coordinate system; Based on the feature maps and corresponding location codes, and in conjunction with the established second network model, the feature information of the current road area in the camera coordinate system is determined.

4. The method according to claim 2, characterized in that, The step of concatenating the aforementioned feature information to determine the target feature information of the current road area in the vehicle coordinate system includes: The feature information is rearranged to obtain the three-dimensional arrays; Based on the detection range and resolution of each camera, determine the camera coordinates of each grid cell included in the detection range in the camera coordinate system; Based on the coordinates of each camera and the extrinsic parameters of the camera, determine the target coordinate matrix of each mesh unit in the vehicle body coordinate system; The feature information in each of the three-dimensional arrays is used as matrix elements to fill the corresponding positions in the target coordinate matrix, thereby determining the target feature information of the current road area in the vehicle coordinate system.

5. The method according to claim 1, characterized in that, The step of determining the lane detection result in the current road area based on the target feature information and the established lane detection model includes: The target feature information is input into the set lane line detection model to obtain the lane line information corresponding to each grid cell. The lane line information is processed to determine the lane line detection results in the current road area.

6. The method according to claim 5, characterized in that, The process of processing each lane line information to determine the lane line detection result in the current road area includes: Extract the confidence level of the lane line category from each of the lane line information; Grid cells with a confidence level greater than a set threshold are identified as target grid cells with lane lines. Otherwise, the grid cell is determined to be a background without lane lines; Based on the target grid cell and the center point offset, and the lateral and longitudinal direction vectors in the lane line information, the lane line detection result in the current road area is determined.

7. The method according to claim 6, characterized in that, The step of determining the lane line detection result in the current road area based on the target grid cell and the center point offset and the lateral and longitudinal direction vectors in the lane line information includes: The target grid cells are classified according to the lane line category to determine the target grid cells corresponding to each lane line category; Cluster the target grid cells corresponding to each lane line category according to the horizontal and vertical direction vectors to obtain the sampling points contained in each lane line; The sampling points contained in each lane line are fitted to determine the fitting curve of each lane line; The coordinates of the lane line center point are determined based on the center point offset in the lane information. The lane line category, lane line center point coordinates, and fitted curve are used as the lane line detection results in the current road area.

8. A lane line detection device, characterized in that, include: The acquisition module is used to acquire at least one area image captured relative to the current road area and the position code of the camera corresponding to each area image; The feature determination module is used to determine the target feature information of the current road area in the vehicle coordinate system based on the images of each area and the corresponding position codes; The result determination module is used to determine the lane line detection result in the current road area based on the target feature information and the set lane line detection model. The device further includes an encoding determination module, comprising: The first matrix determination unit is used to divide the length and width of the images captured by each camera according to the first sampling distance, and determine the pixel coordinate matrix corresponding to each image captured by the camera. The second matrix determination unit is used to divide the detection distance of each camera according to the second sampling distance and determine the distance matrix corresponding to each camera. The encoding determination unit is used to determine the position encoding of each camera based on the pixel coordinate matrix and the distance matrix, combined with a set array construction algorithm; The encoding determination unit is specifically used for: Determine the unit matrix based on the length and width of the images captured by each camera; The pixel coordinate matrix, the distance matrix, and the identity matrix are input into a set array construction algorithm to obtain the construction result; Based on the constructed results, and combined with the transformation formula between the image coordinate system and the camera coordinate system, the position code of the camera is determined.

9. A vehicle, characterized in that, The vehicles include: Vehicle body; A camera is mounted on the vehicle body; A controller, communicatively connected to the camera, the controller comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the lane line detection method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the lane line detection method according to any one of claims 1-7.

Images