Method and apparatus for lane line detection post-processing

Abstract

The application relates to the technical field of intelligent driving, in particular to a lane line detection post-processing method and device, wherein the method comprises the following steps: acquiring single-frame image data in front of a vehicle; extracting key point position information with a label attribute on a lane line according to the single-frame image data; performing inverse perspective transformation on the key point position information by using an inverse perspective transformation matrix, projecting to a bird's-eye view angle, and removing noise points based on the coordinate relationship between adjacent lanes and adjacent rows to obtain a plurality of lane line key points; and combining an actual distance and key point confidence to perform local weighted linear regression fitting on lane line parameters to obtain a final lane line curve equation. Thus, the technical problem that in the related art, output information needs to be post-processed to fit a lane line, the fitting effect is poor, the accuracy of post-processing is reduced, the accuracy of lane line detection is reduced, the automation level of the vehicle is low, and the driving and riding requirements of a user cannot be met is solved.

Description

Technical Field

[0001] This application relates to the field of intelligent driving technology, and in particular to a method and apparatus for lane line detection post-processing. Background Technology

[0002] In related technologies, lane line detection methods such as traditional image processing, deep learning semantic segmentation, or deep learning key point detection can be used to perform corresponding post-processing after outputting information to fit lane lines, thereby enabling a clear description of lane line information.

[0003] However, the relevant technologies require post-processing after outputting information to fit lane lines, resulting in poor fitting effects, reduced accuracy of post-processing, and decreased accuracy of lane line detection. This leads to a low level of vehicle automation, failing to meet users' driving needs and urgently requiring a solution. Summary of the Invention

[0004] This application provides a method and apparatus for lane line detection post-processing to solve the technical problems in related technologies where post-processing is required after outputting information to fit lane lines, resulting in poor fitting effects, reduced accuracy of post-processing, and reduced accuracy of lane line detection. This leads to a low level of vehicle automation and fails to meet the driving needs of users.

[0005] The first aspect of this application provides a method for post-processing lane line detection, comprising the following steps: acquiring single-frame image data in front of a vehicle; extracting key point location information with label attributes on the lane line based on the single-frame image data; performing inverse perspective transformation on the key point location information using an inverse perspective transformation matrix, projecting it onto a bird's-eye view, and removing noise based on the coordinate relationship between adjacent lanes and adjacent rows to obtain multiple lane line key points; and performing local weighted linear regression to fit lane line parameters by combining actual distance and key point confidence to obtain the final lane line curve equation.

[0006] Optionally, in one embodiment of this application, the description formula for the key point location information is:

[0007]

[0008] Where N represents the total number of lane lines, and K represents that each lane line is described by K key points. i It is the descriptor of the i-th lane line. This represents the j-th keypoint on the i-th lane, which is represented by pixels in the image. The coordinates are used to represent the confidence score of the current key point.

[0009] Optionally, in one embodiment of this application, the removal of noise based on the coordinate relationship between adjacent lanes and adjacent rows includes: filtering out abnormal points based on the consistency criteria of multiple key points adjacent to the current lane line and the distance constraint relationship between key points between adjacent lanes.

[0010] Optionally, in one embodiment of this application, the step of combining the actual distance and the confidence of key points to perform local weighted linear regression to fit the lane line parameters and obtain the final lane line curve equation includes: calculating a first longitudinal distance based on the pixel key points, and calculating a second longitudinal distance from the farthest point on the lane line; and obtaining the weight of the pixel key points based on the first longitudinal distance and the second longitudinal distance.

[0011] Optionally, in one embodiment of this application, the formula for calculating the weight is:

[0012]

[0013] in, Indicates based on pixel keypoints (x) j ,y j The calculated longitudinal distance, This represents the point (x) furthest away on the lane line. K ,y K The longitudinal distance is calculated.

[0014] A second aspect of this application provides a lane line detection post-processing apparatus, comprising: an acquisition module for acquiring single-frame image data in front of a vehicle; an extraction module for extracting key point location information with label attributes on the lane line based on the single-frame image data; a projection module for performing inverse perspective transformation on the key point location information using an inverse perspective transformation matrix, projecting it onto a bird's-eye view, and removing noise based on the coordinate relationship between adjacent lanes and adjacent rows to obtain multiple lane line key points; and a fitting module for performing local weighted linear regression to fit lane line parameters by combining actual distance and key point confidence, to obtain the final lane line curve equation.

[0015] Optionally, in one embodiment of this application, the description formula for the key point location information is:

[0016]

[0017] Where N represents the total number of lane lines, and K represents that each lane line is described by K key points. i It is the descriptor of the i-th lane line. This represents the j-th keypoint on the i-th lane, which is represented by pixels in the image. The coordinates are used to represent the confidence score of the current key point.

[0018] Optionally, in one embodiment of this application, the projection module includes: a filtering unit, used to filter out abnormal points based on the consistency criteria of multiple key points adjacent to the current lane line and the distance constraint relationship between key points between adjacent lanes.

[0019] Optionally, in one embodiment of this application, the fitting module includes: a calculation unit, configured to calculate a first longitudinal distance based on pixel key points and a second longitudinal distance based on the farthest point on the lane line; and an acquisition unit, configured to obtain the weights of the pixel key points based on the first longitudinal distance and the second longitudinal distance.

[0020] Optionally, in one embodiment of this application, the formula for calculating the weight is:

[0021]

[0022] in, Indicates based on pixel keypoints (x) j ,y j The calculated longitudinal distance, This represents the point (x) furthest away on the lane line. K ,y K The longitudinal distance is calculated.

[0023] A third aspect of this application provides a vehicle, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the lane line detection post-processing method as described in the above embodiments.

[0024] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the lane line detection post-processing method described above.

[0025] This application embodiment can extract key point location information with labeled attributes on the lane line based on single-frame image data acquired from the vehicle ahead. It then uses an inverse perspective transformation matrix to perform an inverse perspective transformation on the key point location information, projects it onto a bird's-eye view, and removes noise based on the coordinate relationship between adjacent lanes and rows, obtaining multiple lane line key points. Finally, it combines the actual distance and key point confidence levels to perform a locally weighted linear regression to fit the lane line parameters, obtaining the final lane line curve equation. This improves the accuracy of lane line detection, enhances the vehicle's intelligence level, and meets the user's driving needs. Therefore, it solves the technical problem in related technologies where post-processing is required after outputting information to fit the lane line, resulting in poor fitting effects, reduced post-processing accuracy, and lower lane line detection accuracy, leading to a lower level of vehicle automation and an inability to meet the user's driving needs.

[0026] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0027] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0028] Figure 1 This is a flowchart of a lane line detection post-processing method provided according to an embodiment of this application;

[0029] Figure 2 This is a schematic diagram of key point data with IDs on each lane line output by a deep learning algorithm, according to a specific embodiment of this application.

[0030] Figure 3 This is a schematic diagram of the reverse perspective transformation of a specific embodiment of this application;

[0031] Figure 4 This is a schematic diagram illustrating the consistency of three adjacent key points of the current lane line in a specific embodiment of this application;

[0032] Figure 5 This is a schematic diagram illustrating the distance constraint relationship between key points between adjacent lanes in a specific embodiment of this application;

[0033] Figure 6 A flowchart illustrating a lane line detection post-processing method according to a specific embodiment of this application;

[0034] Figure 7 This is a schematic diagram of the structure of the lane line detection post-processing device provided according to an embodiment of this application;

[0035] Figure 8This is a structural schematic diagram of a vehicle provided according to an embodiment of this application. Detailed Implementation

[0036] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0037] The method and apparatus for lane line detection post-processing according to embodiments of this application are described below with reference to the accompanying drawings. Addressing the issues mentioned in the background art where post-processing is required after outputting information to fit lane lines, resulting in poor fitting effects, reduced post-processing accuracy, and decreased lane line detection accuracy, leading to lower vehicle automation levels and failure to meet user driving needs, this application provides a method for lane line detection post-processing. In this method, key point location information with labeled attributes on the lane line is extracted from single-frame image data acquired from the vehicle ahead. The key point location information is then subjected to inverse perspective transformation using an inverse perspective transformation matrix, projected onto a bird's-eye view, and noise is removed based on the coordinate relationships between adjacent lanes and adjacent rows, resulting in multiple lane line key points. These are then combined with actual distance and key point confidence levels to perform local weighted linear regression fitting of lane line parameters, yielding the final lane line curve equation. This improves lane line detection accuracy, enhances vehicle intelligence, and meets user driving needs. This solves the technical problem in related technologies where post-processing is required after outputting information to fit lane lines, resulting in poor fitting effects, reduced accuracy of post-processing, and decreased accuracy of lane line detection. Consequently, the level of vehicle automation is low, failing to meet the driving needs of users.

[0038] Specifically, Figure 1 This is a schematic flowchart illustrating a lane line detection post-processing method provided in an embodiment of this application.

[0039] like Figure 1 As shown, the method for post-processing lane line detection includes the following steps:

[0040] In step S101, a single frame image of the front of the vehicle is acquired.

[0041] It is understood that the embodiments of this application can acquire single-frame image data in front of the vehicle. For example, single-frame image data in front of the vehicle can be acquired by a forward-view camera installed on the windshield, thereby effectively improving the feasibility of lane line detection.

[0042] In step S102, the key point location information with label attributes on the lane line is extracted based on the single frame image data.

[0043] It is understood that embodiments of this application can extract key point location information with label attributes on lane lines based on single-frame image data, for example, such as Figure 2 As shown, the acquired single-frame image data in front of the vehicle, i.e. the original image, can be input into the vehicle's lane line key point detection function. The deep convolutional network model outputs labeled key point data on each lane line. For example, if the original image is input at a resolution of 1280*720, the trained lane line key point detection function will output the corresponding labeled key point location information in the following steps, thereby effectively improving the accuracy of vehicle detection.

[0044] In one embodiment of this application, the description formula for the key point location information is as follows:

[0045]

[0046] Where N represents the total number of lane lines, and K represents that each lane line is described by K key points. i It is the descriptor of the i-th lane line. This represents the j-th keypoint on the i-th lane, which is represented by pixels in the image. The coordinates are used to represent the confidence score of the current key point.

[0047] Here, score represents the confidence score of the current key point, with a value between 0 and 1.

[0048] In step S103, the key point location information is transformed by inverse perspective transformation matrix and projected onto the bird's-eye view. Then, noise is removed based on the coordinate relationship between adjacent lanes and adjacent rows to obtain multiple lane line key points.

[0049] It is understood that the embodiments of this application can utilize an inverse perspective transformation matrix to perform inverse perspective transformation on the key point location information. For example, such as Figure 3 As shown, due to the perspective effect in the image captured by the forward-looking camera, the lane lines are not parallel. Therefore, the perspective effect can be eliminated by using IPM (Inverse Perspective Mapping) and projected onto the bird's-eye view, so that the lane lines retain their original characteristics. The IPM transformation matrix is ​​generated by the camera calibration parameters, i.e.:

[0050]

[0051] Where M represents the corresponding IPM inverse perspective transformation matrix (3*3 size), (x s ,y s(x) represents the original pixel coordinates, (x) d ,y d ) represents the new pixel coordinates obtained after IPM transformation.

[0052] In addition, the M matrix can be 3x3 in size, and it is generated from the camera calibration parameters.

[0053] Next, the embodiments of this application can remove noise points based on the coordinate relationship between adjacent lanes and adjacent rows in the following steps to obtain multiple lane line key points, thereby effectively improving the accuracy and reliability of lane line detection and enhancing the intelligence level of the vehicle.

[0054] In one embodiment of this application, removing noise based on the coordinate relationship between adjacent lanes and adjacent rows includes: filtering out abnormal points based on the consistency criteria of multiple key points adjacent to the current lane line and the distance constraint relationship between key points between adjacent lanes.

[0055] For example, such as Figure 4 The embodiment of this application, as shown, can be based on the consistency criterion of three adjacent key points of the current lane line. Regardless of whether it is a straight line or a curve, the transformation trend between adjacent key points is consistent from the BEV (Birds Eye View) perspective. When the consistency is not met, there is a high probability of key point deviation. The slope between the coordinates of the current key point p1(x1,y1) and the adjacent key points p0(x0,y0) and p2(x2,y2) can be calculated as an important basis for judgment. The criterion for judging whether the key point p1 is abnormal is as follows:

[0056]

[0057]

[0058] Here, Flag1 being 1 indicates that the current key point has a certain probability of being an anomaly point, and is used as a candidate anomaly point. The distance constraint relationship between key points in adjacent lanes in the following steps is used for judgment.

[0059] Next, as Figure 5 As shown, in this embodiment of the application, the distance constraint relationship criterion between key points of adjacent lanes can be used to verify the candidate anomaly point after the consistency criterion of three key points adjacent to the current lane line is used. The verification selects key points of adjacent lane lines.

[0060] Finally, this embodiment of the application can determine whether point p1 on lane line L1 is abnormal. Using the consistency criterion of three adjacent key points on the current lane line, if there are errors at curve bends, a distance constraint can be added. For the key point information on the corresponding rows of the two adjacent lane lines L1 and L2, the distance in the X direction is calculated:

[0061]

[0062] And calculate the mean and variance:

[0063] meanD 12 ,δD 12 ,

[0064] but:

[0065]

[0066] Here, Flag2 being 1 indicates that the j-th key point is an anomaly and needs to be removed, such as... Figure 5 If the left and right boundary points of the ellipse in the diagram exceed the boundary points, they are directly judged as abnormal points, and those with too large deviations are filtered out.

[0067] In summary, the outlier filtering function in this application embodiment mainly processes key points on each lane line after transformation from the BEV perspective, filtering out some points with large deviations, thereby improving the accuracy of lane line detection post-processing and enhancing vehicle safety.

[0068] In step S104, the lane line parameters are fitted by local weighted linear regression, which combines the actual distance and the confidence level of the key points, to obtain the final lane line curve equation.

[0069] It is understood that the embodiments of this application can combine the actual distance and the confidence of key points in the following steps to perform local weighted linear regression to fit the lane line parameters, thereby obtaining the final lane line curve equation, which effectively improves the accuracy of the fitted data, makes the lane line description information more accurate and reliable, improves the vehicle's intelligence level, and effectively meets the user's driving needs.

[0070] For example, the lane line fitting function in this embodiment can receive the filtered lane line key points from the above steps, and can use local weighted regression to fit the parameters. The minimized objective function is then:

[0071]

[0072]

[0073] Among them, y j The y-direction value of the j-th key point on the lane line. w represents the predicted value in the y-direction of the j-th key point on the lane line. j This represents the weight of the corresponding key point.

[0074] In addition, the curve fitting is second-order, w j The values ​​represent the weights of the corresponding key points. The lane line fitting function can select the optimal parameter combination (a, b, c) to minimize the loss.

[0075] In some embodiments, the innovation of the lane line fitting function mainly lies in improving the weight values ​​w of key points. j The lane line key point detection function can output corresponding key points and their corresponding scores, i.e., the confidence scores of the current key points. The scores range from 0 to 1, with higher scores indicating higher accuracy. This score is used as one of the components of the key point weight. In addition, the importance of key points at different distances will vary during the lane line fitting process. Key points at greater distances are more important for fitting higher-order parameters. Therefore, the longitudinal distance between the key point and the vehicle can be considered as another important factor, which can effectively improve the accuracy of the lane line fitting parameters and enhance the robustness of the lane line detection post-processing.

[0076] Optionally, in one embodiment of this application, the lane line parameters are fitted by a locally weighted linear regression combining the actual distance and the confidence level of the key points to obtain the final lane line curve equation, including: calculating a first longitudinal distance based on the pixel key points, and calculating a second longitudinal distance from the farthest point on the lane line; and obtaining the weights of the pixel key points based on the first longitudinal distance and the second longitudinal distance.

[0077] For example, embodiments of this application can use pixel key points (x j ,y j Substitute (u, v, 1) into the equation to calculate the longitudinal distance. For the farthest point on the lane line (x K ,y K ), calculate longitudinal distance To obtain the final fitted (x) j ,y j The weight w j This can improve the accuracy of pixel key point weights and enhance the reliability of intelligent driving in vehicles.

[0078] In one embodiment of this application, the formula for calculating the weight is:

[0079]

[0080] in, Indicates based on pixel keypoints (x) j ,y jThe calculated longitudinal distance, This represents the point (x) furthest away on the lane line. K ,y K The longitudinal distance is calculated.

[0081] In addition, θ1 takes the value 0.6 and θ2 takes the value 0.4.

[0082] For example, such as Figure 6 As shown, the working principle of the embodiments of this application will be described in detail below with a specific example.

[0083] Step S601: Front view image data.

[0084] In other words, the embodiments of this application can acquire single-frame image data in front of the vehicle to improve the feasibility of lane line detection.

[0085] Step S602: Lane line key point detection function.

[0086] In other words, the lane detection function of this application can output a large number of key point location information on the lane line, thereby improving the accuracy of vehicle detection.

[0087] Step S603: IPM inverse perspective transformation.

[0088] In other words, the embodiments of this application can use an inverse perspective transformation matrix to transform image data and map key point data, thereby improving the intelligence level of the vehicle.

[0089] Step S604: Filter out outliers.

[0090] In other words, the embodiments of this application provide a method for filtering out abnormal key points from a BEV perspective, which improves the accuracy of lane line detection post-processing and enhances vehicle safety.

[0091] Step S605: Lane line fitting.

[0092] In other words, the embodiments of this application can improve the local weighted regression fitting of lane lines, enhance the intelligence level of the vehicle, and effectively meet the driving needs of users.

[0093] The lane line detection post-processing method proposed in this application extracts key point location information with labeled attributes from single-frame image data acquired from vehicles ahead. It then performs an inverse perspective transformation on the key point location information using an inverse perspective transformation matrix, projects it onto a bird's-eye view, and removes noise based on the coordinate relationships between adjacent lanes and rows, resulting in multiple lane line key points. Finally, it combines actual distance and key point confidence levels to perform local weighted linear regression to fit lane line parameters, obtaining the final lane line curve equation. This improves the accuracy of lane line detection, enhances vehicle intelligence, and meets user driving needs. Therefore, it solves the technical problems in related technologies where poor fitting results reduce post-processing accuracy and lane line detection accuracy, failing to meet user driving needs.

[0094] Next, the apparatus for lane line detection post-processing according to an embodiment of this application is described with reference to the accompanying drawings.

[0095] Figure 7 This is a block diagram of a lane line detection post-processing device according to an embodiment of this application.

[0096] like Figure 7 As shown, the lane line detection post-processing device 10 includes: an acquisition module 100, an extraction module 200, a projection module 300, and a fitting module 400.

[0097] Specifically, the acquisition module 100 is used to acquire single-frame image data of the front of the vehicle.

[0098] The extraction module 200 is used to extract key point location information with label attributes on the lane line based on single frame image data.

[0099] The projection module 300 is used to perform inverse perspective transformation on the key point position information using the inverse perspective transformation matrix, project it onto the bird's-eye view, and remove noise based on the coordinate relationship between adjacent lanes and adjacent rows to obtain multiple lane line key points.

[0100] The fitting module 400 is used to combine the actual distance and the confidence level of key points to perform local weighted linear regression to fit the lane line parameters and obtain the final lane line curve equation.

[0101] Optionally, in one embodiment of this application, the description formula for the key point location information is:

[0102]

[0103] Where N represents the total number of lane lines, and K represents that each lane line is described by K key points. i It is the descriptor of the i-th lane line. This represents the j-th keypoint on the i-th lane, which is represented by pixels in the image. The coordinates are used to represent the confidence score of the current key point.

[0104] Optionally, in one embodiment of this application, the projection module includes a filtering unit.

[0105] The filtering unit is used to filter out outliers based on the consistency criteria of multiple key points adjacent to the current lane line and the distance constraint relationship between key points between adjacent lanes.

[0106] Optionally, in one embodiment of this application, the fitting module includes a calculation unit and an acquisition unit.

[0107] The calculation unit is used to calculate a first longitudinal distance based on pixel key points and a second longitudinal distance based on the farthest point on the lane line.

[0108] The acquisition unit is used to obtain the weights of pixel key points based on the first vertical distance and the second vertical distance.

[0109] Optionally, in one embodiment of this application, the formula for calculating the weight is:

[0110]

[0111] in, Indicates based on pixel keypoints (x) j ,y j The calculated longitudinal distance, This represents the point (x) furthest away on the lane line. K ,y K The longitudinal distance is calculated.

[0112] It should be noted that the explanation of the above-described method embodiment for lane line detection post-processing also applies to the device for lane line detection post-processing in this embodiment, and will not be repeated here.

[0113] The lane line detection post-processing apparatus proposed in this application can extract key point location information with labeled attributes on the lane line based on single-frame image data acquired from the vehicle ahead. It then performs an inverse perspective transformation on the key point location information using an inverse perspective transformation matrix, projects it onto a bird's-eye view, and removes noise based on the coordinate relationship between adjacent lanes and rows, obtaining multiple lane line key points. Finally, it combines the actual distance and key point confidence levels to perform a locally weighted linear regression to fit the lane line parameters, obtaining the final lane line curve equation. This improves the accuracy of lane line detection, enhances the vehicle's intelligence level, and meets the user's driving needs. Therefore, it solves the technical problems in related technologies where poor fitting results reduce the accuracy of post-processing and lane line detection, failing to meet the user's driving needs.

[0114] Figure 8 A schematic diagram of the structure of a vehicle provided in an embodiment of this application. The vehicle may include:

[0115] The memory 801, the processor 802, and the computer program stored on the memory 801 and capable of running on the processor 802.

[0116] When the processor 802 executes the program, it implements the lane line detection post-processing method provided in the above embodiments.

[0117] Furthermore, the vehicle also includes:

[0118] Communication interface 803 is used for communication between memory 801 and processor 802.

[0119] The memory 801 is used to store computer programs that can run on the processor 802.

[0120] The memory 801 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0121] If the memory 801, processor 802, and communication interface 803 are implemented independently, then the communication interface 803, memory 801, and processor 802 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be divided into address buses, data buses, control buses, etc. For ease of representation, Figure 8The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0122] Optionally, in a specific implementation, if the memory 801, processor 802, and communication interface 803 are integrated on a single chip, then the memory 801, processor 802, and communication interface 803 can communicate with each other through an internal interface.

[0123] The processor 802 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0124] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described lane line detection post-processing method.

[0125] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0126] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0127] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0128] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0129] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0130] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it includes one or a combination of the steps of the method embodiments.

[0131] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0132] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for post-processing lane line detection, characterized in that, Includes the following steps: Acquire single-frame image data of the front of the vehicle; Extract the location information of key points with tagged attributes on the lane line from the single frame image data; The key point location information is transformed by inverse perspective transformation matrix and projected onto the bird's-eye view. Noise is removed based on the coordinate relationship between adjacent lanes and adjacent rows to obtain multiple lane line key points. as well as By combining the actual distance and the confidence level of key points, a locally weighted linear regression is performed to fit the lane line parameters, and the final lane line curve equation is obtained. The formula for describing the location information of the key points is: ， Where N represents the total number of lane lines, and K represents the number of key points describing each lane line. It is the first i The description of the lane lines, Representing the i The first lane on the line j A key point, which is composed of pixels on the image. Represented by coordinates, The confidence score represents the current key point; The process of combining actual distance and key point confidence levels to perform local weighted linear regression fitting of lane line parameters yields the final lane line curve equation, including: The first longitudinal distance is calculated based on the pixel key points, and the second longitudinal distance is calculated based on the farthest point on the lane line; The weights of the pixel key points are obtained based on the first vertical distance and the second vertical distance; The formula for calculating the weight is: ， in, Indicates based on pixel key points Calculated longitudinal distance, This indicates the furthest point on the lane line. Calculate the longitudinal distance.

2. The method according to claim 1, characterized in that, The method of removing noise based on the coordinate relationship between adjacent lanes and adjacent rows includes: Outlier filtering is performed based on the consistency criteria of multiple key points adjacent to the current lane line and the distance constraint relationship between key points between adjacent lanes.

3. A device for post-processing lane line detection, characterized in that, include: The acquisition module is used to acquire single-frame image data of the front of the vehicle; The extraction module is used to extract key point location information with label attributes on the lane line based on the single frame image data; The projection module is used to perform inverse perspective transformation on the key point position information using an inverse perspective transformation matrix, project it onto the bird's-eye view, and remove noise based on the coordinate relationship between adjacent lanes and adjacent rows to obtain multiple lane line key points. as well as The fitting module is used to combine the actual distance and the confidence level of key points to perform local weighted linear regression to fit the lane line parameters and obtain the final lane line curve equation. The formula for describing the key point location information is: ， Where N represents the total number of lane lines, and K represents the number of key points describing each lane line. It is the first i The description of the lane lines, Representing the i The first lane on the line j A key point, which is composed of pixels on the image. Represented by coordinates, The confidence score represents the current key point; The fitting module includes: a calculation unit, used to calculate a first longitudinal distance based on pixel key points, and a second longitudinal distance based on the farthest point on the lane line; and an acquisition unit, used to obtain the weights of the pixel key points based on the first longitudinal distance and the second longitudinal distance. The formula for calculating the weight is: ， in, Indicates based on pixel key points Calculated longitudinal distance, This indicates the furthest point on the lane line. Calculate the longitudinal distance.

4. The apparatus according to claim 3, characterized in that, The projection module includes: The filtering unit is used to filter out outliers based on the consistency criteria of multiple key points adjacent to the current lane line and the distance constraint relationship between key points between adjacent lanes.

5. A vehicle, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the lane line detection post-processing method as described in any one of claims 1-2.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the lane line detection post-processing method as described in any one of claims 1-2.

Images