Method for constructing lane model

Abstract

The invention relates to a method for constructing a lane model, and the method comprises the steps: carrying out the processing of a detected lane line, and carrying out the parallelization processing of the lane line; adaptively determining a lane center line coefficient by using the processed lane line according to the identified lane scene; and sampling points on the lane center line according to the determined lane center line coefficient under the condition of keeping a safe distance from the road edge so as to construct the lane model. The application also relates to a computer program product and an electronic device.

Description

Technical Field

[0001] The present application relates to the field of lane model construction, and more specifically, to a method, a computer program product, and an electronic device for constructing a lane model. Background Art

[0002] With the rapid development of autonomous driving technology, lane modeling, as a key component of autonomous driving systems, is crucial for achieving safe driving and intelligent decision-making. Lane models are the foundation for autonomous driving systems to perform functions such as path planning, lane keeping, and lane changing. Their accuracy and stability directly impact the overall performance of autonomous driving systems.

[0003] However, existing lane modeling methods still have some flaws. For example, they fail to fully account for lane line non-parallelism caused by camera calibration errors or detection errors. In practical applications, directly using the original detected lane lines can lead to deviations in the width of the lanes at the beginning and end, affecting the accuracy and reliability of the lane model. Summary of the Invention

[0004] The inventors of this application recognized that a common prior art method for constructing a three-lane model is to extract four lane lines from an image: the left lane line, the left-left lane line, the right lane line, and the right-right lane line, and then perform a cubic polynomial fit on these lines. Based on the coefficients of each lane's left and right lane lines, the centerline points of that lane are sampled to construct a three-lane model. However, this prior art suffers from the following drawbacks: Firstly, it fails to account for lane line non-parallelism caused by camera calibration errors or detection errors, and directly using the original lane lines can lead to deviations in lane width at the beginning and end. Secondly, directly sampling the center points using the lane lines on both sides is only suitable for normal scenarios and is less suitable for special scenarios such as missing lane lines, occlusions, separations, and merging. This affects the accuracy and reliability of the road model, and consequently, the safety and performance of autonomous vehicles.

[0005] According to one aspect of the present application, a method for constructing a lane model is provided, the method comprising: processing detected lane lines, the processing comprising parallelizing the lane lines; adaptively determining a lane centerline coefficient using the processed lane lines based on the identified lane scene; and sampling points on the lane centerline based on the determined lane centerline coefficient while maintaining a safe distance from the roadside to construct the lane model.

[0006] As a supplement or alternative to the above scheme, the above method also includes: extracting lane line points from the image and converting the lane line points into a vehicle coordinate system; fitting the lane line points to obtain lane line coefficients; and determining lane line position attributes based on the lane line coefficients.

[0007] As a supplement or alternative to the above solution, the above method also includes: after determining the lane line position attribute, processing the historical lane line information to align it with the current lane line information.

[0008] As a supplement or alternative to the above-mentioned scheme, in the above-mentioned method, after determining the lane line position attributes, processing the historical lane line information to align it with the current lane line information includes: predicting the historical frame lane line to estimate the position and shape of the historical frame lane line at the current moment; and considering the lane change situation, aligning the position of the historical frame lane line with the position of the lane line at the current moment.

[0009] As a supplement or alternative to the above scheme, in the above method, the processing also includes: after the parallelization processing, performing stability and continuity processing on the lane line, and the stability and continuity processing includes multi-frame verification, missing compensation and / or fusion smoothing.

[0010] As a supplement or alternative to the above scheme, in the above method, parallelizing the lane lines includes: extracting multiple relatively parallel lane lines from the lane lines; and performing weighted averaging on the lane line coefficients of the multiple relatively parallel lane lines.

[0011] As a supplement or replacement for the above scheme, in the above method, the multi-frame verification includes: when a jump in the lane line is detected, the lane lines of multiple consecutive frames are detected, and if the lane lines in the multiple consecutive frames are stable, switching to the new lane line; otherwise, maintaining the lane line of the historical frame.

[0012] As a supplement or alternative to the above-mentioned scheme, in the above-mentioned method, the missing lane line compensation includes: when there is a missing lane line, compensating for the missing lane line by extending the lane line of the historical frame forward, using the curb on the same side, using the default lane width, or using the lane line of the adjacent lane.

[0013] As a supplement or replacement for the above solution, in the above method, the fusion smoothing includes: fusing the lane lines of the historical frames and the lane lines at the current moment by weighted averaging or Kalman filtering to eliminate the jitter of the lane lines.

[0014] As a supplement or alternative to the above-mentioned scheme, in the above-mentioned method, the lane centerline coefficient is adaptively determined using the processed lane line according to the identified lane scene, including: when the lane scene indicates a normal lane, the lane centerline coefficient is calculated based on the left and right lane lines of the lane by length weighting; when the lane scene indicates an extra-wide lane, the lane centerline coefficient is calculated based on the lane line closest to the vehicle; when the lane scene indicates a lane separation scenario, the lane centerline coefficient is calculated based on the lane line with a smaller absolute value of curvature; and when the lane scene indicates a lane merging scenario, the merging direction is determined and the lane centerline coefficient is calculated based on the lane line on the opposite side of the merging direction.

[0015] As a supplement or alternative to the above-mentioned scheme, in the above-mentioned method, adaptively determining the lane centerline coefficient using the processed lane line according to the identified lane scene also includes: fusing the lane centerline coefficient calculated for the lane at the current moment with the lane centerline coefficient calculated for the lane of the historical frame.

[0016] As a supplement or alternative to the above-mentioned scheme, in the above-mentioned method, while maintaining a safe distance from the roadside, sampling the points on the lane centerline according to the determined lane centerline coefficient to construct the lane model includes: sampling the points on the lane centerline into a series of point sets according to the centerline coefficient, wherein the roadside on both sides are searched during the sampling process, and if the distance between the centerline point and the roadside is less than a preset threshold, the sampled centerline point is offset to maintain a safe distance from the roadside; the sampled point set is smoothed; and the sampled centerline points of the left lane, center lane and right lane are combined with the lane lines and other lane markings to generate a three-lane model.

[0017] According to another aspect of the present application, a driving assistance control method is provided, which includes: constructing a three-lane model based on the method described above; updating lane line information according to the three-lane model; and performing a driving assistance control operation based on the updated lane line information.

[0018] As a supplement or replacement for the above solution, in the above method, the driving assistance control operation includes any one of path planning, vehicle following target identification / screening, target trajectory prediction, fusion positioning and human-computer interaction HMI display.

[0019] As a supplement or alternative to the above scheme, in the above method, the driving assistance control operation includes any one of centering control, lane keeping, departure warning, automatic correction, automatic lane change / lane change assistance, following control, collision risk warning, automatic emergency braking, cruise control, driver-commanded lane change, nudge and bypass functions.

[0020] As a supplement or alternative to the above solution, in the above method, the driving assistance control operation includes any one of an automatic cruise control ACC function and a traffic jam assistance TJA function.

[0021] According to another aspect of the present application, a computer program product is provided, comprising a computer program, which implements the above method when executed by a processor.

[0022] According to another aspect of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method described above.

[0023] As a supplement or alternative to the above solution, in the above electronic device, the electronic device is a domain controller, a camera or a radar.

[0024] The lane model construction scheme of the embodiment of the present application significantly improves the accuracy and stability of lane model construction in various scenarios as a whole. Specifically, by processing the detected lane lines, including parallelizing the lane lines, the impact of camera calibration errors or detection errors on the lane lines can be effectively reduced, making the lane lines more parallel and improving the accuracy of the lane model. In addition, the lane centerline coefficient is adaptively determined according to the identified lane scene, which is conducive to generating a lane centerline that is more in line with driving habits and improves the adaptability of the autonomous driving system in complex scenarios. Moreover, the curb constraint can ensure that the generated centerline point always maintains a certain safe distance from the curb, thereby improving driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above and other objects and advantages of the present application will become more fully apparent from the following detailed description taken in conjunction with the accompanying drawings, wherein the same or similar elements are denoted by the same reference numerals.

[0026] Figure 1 A schematic flow chart of a method for constructing a lane model according to an embodiment of the present application is shown;

[0027] Figure 2 A schematic diagram of a process for parallelizing lane lines according to an embodiment of the present application is shown;

[0028] Figure 3 A schematic diagram showing the effect of parallelizing lane lines according to an embodiment of the present application is shown;

[0029] Figure 4 A schematic diagram of a scenario for calculating a lane centerline coefficient in a lane separation scenario according to an embodiment of the present application is shown;

[0030] Figure 5 A schematic diagram of a scenario for calculating a lane centerline coefficient in a lane merging scenario according to an embodiment of the present application is shown;

[0031] Figure 6 A schematic diagram showing a method of maintaining a safe distance between a sampling point on a lane centerline and a roadside according to an embodiment of the present application is shown; and

[0032] Figure 7 A schematic structural diagram of an electronic device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0033] Hereinafter, a scheme for constructing a lane model according to various exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0034] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the embodiments of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in an appropriate manner in any one or more embodiments or examples. In addition, the terms "first", "second" and "third" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are mutually inconsistent.

[0035] Figure 1 FIG. 1 is a flow chart of a method 1000 for constructing a lane model according to an embodiment of the present application. Figure 1 As shown, the method 1000 includes the following steps:

[0036] In step S110, the detected lane lines are processed, and the processing includes parallelizing the lane lines;

[0037] In step S120, according to the identified lane scene, the lane centerline coefficient is adaptively determined using the processed lane lines; and

[0038] In step S130 , while maintaining a safe distance from the roadside, points on the lane centerline are sampled according to the determined lane centerline coefficients to construct the lane model.

[0039] In the context of this application, the term "lane model" is a concept used in autonomous / assisted driving vehicles to describe and represent road geometry. It helps the vehicle better understand the surrounding road environment, thereby achieving safer and more efficient autonomous / assisted driving behavior. In one or more embodiments, the lane model may include information such as lane lines, lane centerlines, lane width, and lane type. This information can be used for vehicle path planning, obstacle detection, lane keeping, and other functions.

[0040] In step S110, the detected lane markings are processed. Here, the term "lane marking" is a fundamental component of the lane model and refers to the markings on the road used to separate lanes, including solid lines, dashed lines, and double solid lines. For example, a solid line represents a lane boundary that vehicles cannot cross and is typically used to separate lanes or lane groups in different directions. For another example, a dashed line represents a lane boundary that vehicles can cross safely and is used to separate lanes in the same direction. For another example, a double solid line typically represents a lane boundary that vehicles cannot cross and is often used to separate opposing lanes or lane groups.

[0041] In one embodiment, lane lines can be obtained through image recognition. For example, before step S110, method 1000 may further include: extracting lane line points from the image and converting the lane line points into a vehicle coordinate system; fitting the lane line points to obtain lane line coefficients; and determining lane line position attributes based on the lane line coefficients.

[0042] Specifically, a camera can capture road images. For example, image data is typically in the form of a two-dimensional array containing RGB or grayscale values. After acquiring the image, it can be preprocessed, such as by denoising and enhancing the image, to improve the accuracy of subsequent lane detection. Lane points can then be extracted from the image using various methods. For example, edge detection can be performed on the preprocessed image using operators such as Canny and Sobel to identify edge feature points within the image. Then, by combining features such as the color and shape of the lane lines, methods such as Hough transforms and machine learning can be used to identify lane points from these edge feature points. In one embodiment, converting lane points to the vehicle coordinate system can further include determining camera intrinsic parameters (such as focal length and optical center) and extrinsic parameters (such as the camera's position and attitude relative to the vehicle coordinate system); and converting the lane points from the image coordinate system to the vehicle coordinate system based on the camera's intrinsic and extrinsic parameters.

[0043] In one embodiment, a suitable mathematical method is selected to fit the lane line points in order to obtain the lane line coefficients. Common fitting methods include polynomial fitting, spline fitting, etc. For example, polynomial fitting is suitable for situations where the lane line is relatively smooth and can be described by a polynomial function, while spline fitting is suitable for situations where the lane line curvature varies greatly and requires more flexible fitting. In addition, the order or number of fitting can be determined based on the complexity of the lane line and the fitting accuracy requirements. For example, for ordinary lane lines, quadratic or cubic polynomial fitting can usually provide better fitting results. The lane line points are fitted using the selected fitting method and parameters. The coefficients of the fitting curve are obtained by the least squares method or other optimization algorithms. These coefficients describe the shape and position of the lane line. For example, using the cubic polynomial y=C3*x 3 +C2*x 2 +C1*x+C0 to represent the lane line, where x and y are the coordinates in the vehicle coordinate system, and C3, C2, C1, and C0 are the lane line coefficients.

[0044] In the lane line equation above, C0 is the intercept, and C1, C2, and C3 are the coefficients of the first, second, and third terms of the polynomial, respectively. In one embodiment, the position attribute of each lane line (for example, left-left lane line, left lane line, right lane line, or right-right lane line) can be accurately determined based on the value of the intercept C0. Because C0 represents the intercept of the lane line on the y-axis, it reflects the lateral position of the lane line. Therefore, it can be understood that in the vehicle coordinate system, the C0 of the left-left lane line should be the largest positive value because it is farthest to the left; similarly, the C0 of the right-right lane line should be the smallest negative value because it is farthest to the right (Note: in the vehicle coordinate system, the left side is the positive direction of the y-axis (i.e., y is a positive value) and the right side is the negative direction of the y-axis (i.e., y is a negative value)).

[0045] It should be noted that the left-left lane marking refers to the lane marking on the left side of the left lane (of the current vehicle). The left lane marking refers to the lane marking on the left side of the vehicle's lane (or the lane marking on the right side of the left lane). Similarly, the right lane marking refers to the lane marking on the right side of the vehicle's lane (or the lane marking on the left side of the right lane). The right-right lane marking refers to the lane marking on the right side of the right lane (of the current vehicle). In other words, the left-left lane marking refers to the left lane marking of the other lane to the left of the current lane. However, if the vehicle's lane is the leftmost lane, then the left-left lane marking may not exist (i.e., there is no lane further to the left than the current lane). Similarly, the right-right lane marking refers to the right lane marking of the other lane to the right of the current lane. If the vehicle's lane is the rightmost lane, then the right-right lane marking may not exist.

[0046] although Figure 1Not shown in the figure, in one embodiment, method 1000 may further include, before step S110, processing historical lane line information to align it with current lane line information after determining the lane line position attribute.

[0047] In the context of this application, "lane position attributes" refer to the relative position and orientation of a lane line on the road. These attributes are typically determined using coefficients (such as C0, C1, and C2) derived from lane line fitting. Lane position attributes can help identify whether a lane line is left, right, left-left, or right-right.

[0048] "Historical lane marking information" refers to lane marking data detected and recorded at previous moments or frames. This data includes lane marking coefficients, location attributes, and lane marking shape. Historical lane marking information can help predict the current lane marking position and shape, especially during vehicle motion or lane changes. "Current lane marking information" refers to lane marking data detected at the current moment or frame. This data, also obtained through image processing and fitting, includes lane marking coefficients, location attributes, and shape. Current lane marking information reflects the actual lane conditions of the vehicle.

[0049] In one embodiment, after determining lane line position attributes, processing historical lane line information to align it with current lane line information includes: predicting lane lines in historical frames to estimate their current position and shape; and aligning the positions of lane lines in historical frames with those of the current moment, taking into account lane changes. In this embodiment, "alignment" refers to inter-frame alignment, which ensures a smooth transition between lane line positions in consecutive frames. This alignment ensures the continuity and consistency of lane line information, particularly when a vehicle changes lanes or lane lines change.

[0050] In this embodiment, "historical lane lines" are also referred to as "historical lane lines," representing lane lines detected in the past (i.e., before the current moment). "Current lane lines" are also referred to as "current lane lines" or "lane lines at the current moment," representing lane lines detected at the current moment, reflecting the actual conditions of the vehicle's current lane and adjacent lanes.

[0051] Due to issues such as frame rate and processing delay, the lane line information of the historical frame may be inconsistent in time with the lane line information at the current moment. Through prediction, the position and shape of the historical lane line at the current moment can be estimated so that it is aligned in time with the lane line at the current moment, thereby improving the accuracy and reliability of the fusion. In one embodiment, predicting the lane line of the historical frame may include: using a time series model (such as a Kalman filter) to predict the position and shape of the historical lane line at the current moment. For example, the Kalman filter can dynamically predict the position and shape of the lane line based on the movement trend of the historical lane line and the driving status of the vehicle.

[0052] In addition, when the vehicle changes lanes, the position of the lane line will change. In order to ensure that the historical lane line information accurately corresponds to the current lane line information, it is necessary to consider the lane change situation and align the position of the historical lane line with the position of the current lane line. In one embodiment, considering the lane change situation, aligning the position of the lane line in the historical frame with the position of the lane line at the current moment can include: by analyzing the vehicle's lane change intention (such as turn signal, lateral acceleration of the vehicle, etc.) and the changing trend of the lane line, determining whether the vehicle has changed lanes; if a lane change is detected, adjusting the position of the historical lane line to match the position of the current lane line. For example, if the vehicle changes from the left lane to the right lane, the position of the historical lane line needs to be moved to the right by the width of one lane.

[0053] In step S110, the processing performed on the detected lane lines includes parallelization. In one embodiment, the parallelization process includes: extracting multiple relatively parallel lane lines from the detected lane lines; and performing a weighted average of the lane line coefficients of the multiple relatively parallel lane lines. For example, the angle or curvature between the lane lines can be calculated to identify relatively parallel lane lines. When the angle between the lane lines is small or the curvature is close, they can be considered to be relatively parallel. Then, based on the integrity and confidence of the lane lines, the clearest and most stable lane line is selected as the reference lane line. Then, based on the reference lane line, lane lines that are relatively parallel to it are extracted. Usually, lane lines with an angle with the reference lane line less than a certain threshold (such as 5 degrees) are selected as relatively parallel lane lines. Subsequently, the weight of each lane line is determined based on factors such as the length of the lane line and the detection confidence. Longer lane lines with higher detection confidence are usually given larger weights. Finally, the coefficients of the extracted relatively parallel lane lines are weighted averaged.

[0054] For example, for the coefficients C1, C2, and C3 of multiple lane lines, the weighted average value can be calculated as follows:

[0055] C 1_avg =(w1*C 1_1 +w2*C 1_2+...+wn*C 1_n ) / (w1+w2+...+wn);

[0056] C 2_avg =(w1*C 2_1 +w2*C 2_2 +...+wn*C 2_n ) / (w1+w2+...+wn);

[0057] C 3_avg =(w1*C 3_1 +w2*C 3_2 +...+wn*C 3_n ) / (w1+w2+...+wn).

[0058] In the above formula, w1 is the weight for lane line 1, w2 is the weight for lane line 2, and so on. 1_1 represents the C1 parameter of lane line 1, C 2_1 represents the C2 parameter of lane line 1, C 3_1 Indicates the C3 parameter of lane line 1, C 1_2 Indicates the C1 parameter of lane 2, and so on. 1_avg 、C 2_avg and C 3_avg Represents the weighted average of each coefficient.

[0059] Figure 2 FIG. 1 shows a flow chart of parallelizing lane lines according to an embodiment of the present application. Figure 2 As shown, in step S210, the entire process begins; then, in step S220, it is determined whether a reference path exists; if so, step S232 is executed, otherwise step S242 is executed. In step S232, multiple lane lines that are relatively parallel to the reference path are extracted; then, in step S234, the multiple lane lines are parallelized to calculate new lane line coefficients; then, in step S236, the processed lane lines are verified for deviation from the original lane lines, that is, the deviation between the processed lane lines and the original lane lines is verified to ensure that it is within the allowable error range. In step S238, it is determined whether the lane lines pass the (deviation) verification; if so, the process proceeds to step S250 and the entire process ends. If not, the process jumps to step S242.

[0060] In step S242, the parallelism of each lane line relative to the other lane lines is calculated. Next, in step S244, the lane line with the lowest parallelism is eliminated, and the remaining lane lines are then parallelized to calculate new lane line coefficients. Finally, the process proceeds to step S250 to terminate.

[0061] The above example demonstrates the complete process of lane parallelization, ensuring lane parallelism and improving the accuracy and stability of the lane model. By extracting relatively parallel lane lines, performing parallelization, calculating new lane coefficients, and verifying deviations, the problems caused by non-parallel lane lines can be effectively reduced, providing a more reliable driving reference for autonomous and assisted driving vehicles.

[0062] Figure 3 FIG. 1 shows a schematic diagram of the effect of parallelizing lane lines according to an embodiment of the present application. Figure 3 As shown, lane lines 310, 330 and 350 represent original lane lines, i.e., lane lines detected without parallelization processing. Figure 3 As can be seen in the figure, lane lines 310 and 330 are located on the left and right sides of vehicle 300, and the width of the lane defined by lane lines 310 and 330 deviates from the width of the lane at the front and rear ends (for example, due to camera calibration or detection errors). By parallelizing lane lines 310, 330, and 350, processed lane lines 320, 340, and 360 are obtained. Compared to the original lane lines, the parallelism between the processed lane lines 320, 340, and 360 is significantly improved, effectively reducing the problems caused by lane line misparallelism and improving the accuracy and stability of the lane model.

[0063] In one embodiment, in addition to parallelization, the processing of detected lane lines in step S110 may also include stability and continuity processing of the lane lines after parallelization, including multi-frame verification, loss compensation, and / or fusion smoothing. This stability and continuity processing ensures that the lane line information is stable and continuous over time, thereby providing a reliable road model for autonomous / assisted driving systems.

[0064] "Multi-frame verification" is used to ensure the consistency of lane line detection and avoid false detection or missed detection due to accidental factors. Specifically, when there is an obvious jump or mutation in the lane line detection, the "multi-frame verification" mechanism is triggered. For example, the lane line detection suddenly has a large change in position, shape, or the appearance of a new lane line. After the "multi-frame verification" mechanism is triggered, the lane line detection results of the subsequent multiple frames are checked continuously. If the lane line detection results of multiple consecutive frames are stable and meet expectations, the new lane line detection results are considered reliable and can be switched to the new lane line. However, if the lane line detection results of multiple consecutive frames are unstable or do not meet expectations, the new lane line detection results are considered unreliable, and the lane line information of the historical frame is maintained until the detection results are stable. In one embodiment, the multi-frame verification includes: when a jump in the lane line is detected, the lane lines of multiple consecutive frames are detected, wherein if the lane lines in the multiple consecutive frames are stable, the lane lines are switched to the new lane lines; otherwise, the lane lines of the historical frames are maintained.

[0065] "Missing lane compensation" addresses lane line gaps caused by obstructions, lighting changes, or other reasons. In one embodiment, missing lane line compensation includes compensating for missing lane lines by extending lane lines from historical frames forward, using the same-side curb, using the default lane width, or using lane lines from adjacent lanes. "Extending historical lane lines forward" can, for example, use a mathematical model (such as polynomial fitting) to extend lane lines detected stably in historical frames forward to fill in the missing lane lines in the current frame. "Using the same-side curb" can, for example, use the detected curb as a reference for lane boundaries and infer the likely position of the lane line using the same-side curb when the lane line is missing. "Using the default lane width" can, for example, set a default lane width based on road type and regional standards and use this default value to estimate the lane line position when a lane line is partially or completely missing. "Using the adjacent lane lines" can, for example, use the lane line position and shape of the adjacent lanes (such as the left or right lane) to estimate the position of the missing lane line through interpolation or mapping.

[0066] Fusion smoothing is used to reduce jitter and unevenness in lane detection, improving the smoothness and continuity of the lane model. In one embodiment, fusion smoothing involves fusing lane lines from historical frames with those at the current moment through weighted averaging or Kalman filtering to eliminate jitter.

[0067] In step S120, the lane centerline coefficient is adaptively determined using the processed lane line based on the identified lane scene. In the context of this application, "lane centerline" refers to a virtual line located in the middle of the lane, which is used to indicate the ideal driving path of the vehicle within the lane and provide a reference for the driving path of the vehicle. It is usually located between the left and right lane lines and matches the geometry of the road. The term "lane centerline coefficient" is a mathematical parameter used to describe the position and shape of the lane centerline. For example, the lane centerline coefficient can be obtained by fitting a set of points of the lane centerline, which are extracted from the image and converted into the vehicle coordinate system. The coefficient defines the geometry of the lane centerline so that the vehicle can accurately identify and track the lane centerline. In one or more embodiments, the lane centerline can generally be represented by a polynomial equation, such as:

[0068] y=C3*x 3 +C2*x 2 +C1*x+C0,

[0069] In the above formula, x and y are the coordinates in the vehicle coordinate system, and C3, C2, C1, and C0 are the lane centerline coefficients. Specifically, C0 represents the intercept of the lane centerline on the y-axis; C1 represents the linear slope of the lane centerline; and C3 and C2 represent the curvature of the lane centerline, with C2 affecting the quadratic curvature and C3 affecting the cubic curvature.

[0070] In one embodiment, step S120 includes: when the lane scenario indicates a normal lane, calculating the lane centerline coefficient based on the left and right lane lines of the lane by length weighting; when the lane scenario indicates an extra-wide lane, calculating the lane centerline coefficient based on the lane line closest to the vehicle; when the lane scenario indicates a lane separation scenario, calculating the lane centerline coefficient based on the lane line with a smaller absolute value of curvature; and when the lane scenario indicates a lane merging scenario, determining the merging direction and calculating the lane centerline coefficient based on the lane line on the opposite side of the merging direction.

[0071] In one or more embodiments, lane scenarios may be divided into the following types: normal lanes, extra-wide lanes, lane separation scenarios, and lane merging scenarios, etc.

[0072] Specifically, a normal lane is the most common lane type, typically consisting of two or more parallel lane lines, with vehicles traveling in straight lines. Normal lanes have standard lane widths, clear lane lines, and a clear direction of vehicle travel. In one embodiment, in a normal lane scenario, the lane centerline coefficient is calculated using length weighting based on the left and right lane lines. Here, "length weighting" means that when calculating the centerline coefficient, the coefficients of the left and right lane lines are weighted according to their lengths. The core idea is that the longer the lane line distribution length on the road, the greater its impact on the lane centerline.

[0073] An extra-wide lane is a lane that is wider than a standard lane and may exist due to road design or for special purposes (such as an emergency lane). In one embodiment, in an extra-wide lane scenario, the lane centerline coefficient is calculated based on the lane line closest to the vehicle to ensure that the vehicle remains centered in the extra-wide lane.

[0074] Lane separation occurs when a lane splits into multiple directions ahead, typically occurring at the entrance or exit of a highway. In one embodiment, in a lane separation scenario, a lane centerline coefficient is calculated based on the lane line with the smallest absolute curvature to guide the vehicle to the correct lane.

[0075] See also Figure 4 , which shows a schematic diagram of a scenario for calculating lane centerline coefficients in a lane separation scenario according to an embodiment of the present application. Figure 4 In the figure, lane lines 410 and 420 form the first lane, with lane line 415 representing the centerline of the first lane. Lane lines 420 and 430 form the second lane, with lane line 425 representing the centerline of the second lane. The first lane formed by lane lines 410 and 420 separates into multiple directions at the far end, a phenomenon known as "lane separation." Therefore, the coefficient for lane centerline 415 can be generated based on the lane line with the smaller absolute curvature value (e.g., lane line 420).

[0076] A lane merge occurs when multiple lanes merge into a single lane ahead, typically occurring at highway entrances or where the road narrows. In one embodiment, in a lane merge scenario, the merging direction is first determined. Then, the lane centerline coefficient is calculated based on the lane markings on the opposite side of the merging direction to help vehicles merge smoothly. For example, if merging to the left, the right lane marking will serve as the primary reference; if merging to the right, the left lane marking will serve as the reference.

[0077] See also Figure 5 , which shows a schematic diagram of a scenario for calculating the lane centerline coefficient in a lane merging scenario according to an embodiment of the present application. Figure 5In FIG, lane lines 510 and 520 form a first lane, and 515 represents the center line of the first lane; lane lines 520 and 530 form a second lane, and 525 represents the center line of the second lane. Figure 5 As can be seen, the second lane will merge to the left ahead, ie, it will merge with the first lane into the same lane. In this case, when calculating the centerline 525 of the second lane, the lane line 530 on the right side will be used as the primary reference.

[0078] In one or more embodiments, step S120 may further include: fusing the lane centerline coefficient calculated for the lane at the current moment with the lane centerline coefficient calculated for the lane in the historical frame (e.g., using weighted averaging, Kalman filtering, or other filtering algorithms). Fusion of the lane centerline coefficients of the current moment and the historical frame may have at least the following effects: (1) improving stability: reducing lane centerline fluctuations due to temporary detection errors or environmental changes; (2) enhancing continuity: ensuring that changes in the lane centerline are smooth and continuous during vehicle driving; and (3) adapting to dynamic changes: responding adaptively to dynamic changes in road conditions (e.g., wear of lane lines or temporary obstruction of lane markings).

[0079] In step S130, points on the lane centerline are sampled based on the determined lane centerline coefficients while maintaining a safe distance from the curb to construct the lane model. Specifically, after the lane centerline coefficients have been calculated based on the lane scene and lane line information in step S120, points on the lane centerline may be sampled (e.g., a series of points are sampled at equal intervals along the lane direction in the vehicle coordinate system based on the lane centerline coefficients). These points represent positions on the lane centerline for constructing the lane model. Next, during the sampling process, the positions of the curbs on both sides of the vehicle are simultaneously detected (e.g., using sensor data or map information). If the distance between the sampled centerline points and the curb is less than a preset safety threshold, the positions of these points are adjusted (typically, these points are offset toward the lane center to ensure a safe distance from the curb). The sampled point set is then smoothed (e.g., using a moving average, spline interpolation, or other methods) to eliminate broken lines or discontinuities that may result from adjusting the safety distance. Finally, the smoothed centerline points are combined with the lane lines and other lane markings to generate a complete lane model.

[0080] In one embodiment, step S130 can be used to generate a three-lane model, specifically including: sampling the points on the lane centerline into a series of point sets based on the centerline coefficient, searching the curbs on both sides during the sampling process, and if the distance between the centerline point and the curb is less than a preset threshold, offsetting the sampled centerline point to maintain a safe distance from the curb; smoothing the sampled point set; and combining the sampled centerline points of the left lane, center lane, and right lane with the lane lines and other lane markings to generate a three-lane model.

[0081] Figure 6 FIG1 shows a schematic diagram of maintaining a safe distance between a sampling point on a lane centerline and a roadside according to an embodiment of the present application. Figure 6 As shown, when sampling lane centerline 615 of the first lane formed by lane lines 610 and 620, if a point on lane centerline 615 is found to be too close to the curb (i.e., one side 642 of the curb), the sampled centerline point is offset to maintain a safe distance from curb side 642. Similarly, when sampling lane centerline 625 of the second lane formed by lane lines 620 and 630, if a point on lane centerline 625 is found to be too close to the curb (i.e., one side 644 of the curb), the sampled centerline point is offset to maintain a safe distance from curb side 644.

[0082] This reduces the risk of collisions by ensuring that vehicles maintain a safe distance from the curb. Furthermore, the lane modeling method described above is adaptable to varying road conditions and obstacle distribution, ensuring a reliable lane model can be constructed in all situations.

[0083] Another aspect of the present application provides a driving assistance control method, which includes: constructing a three-lane model based on the method 1000 for constructing a lane model as described above; updating lane line information according to the three-lane model; and performing a driving assistance control operation based on the updated lane line information.

[0084] In one embodiment, from the perspective of the technology stack, the driving assistance control operation may include any one of path planning, vehicle following target identification / screening, target trajectory prediction, fusion positioning, and human-computer interaction HMI display. For example, in the example of target trajectory prediction, the target vehicle's driving trajectory can be predicted based on the lane centerline information in the three-lane model and the driving status of the vehicle in front, providing a basis for the vehicle's path planning and decision-making. In the example of fusion positioning, the positioning information from multiple sensors (such as GPS, cameras, radars, etc.) can be fused with the lane line information in the three-lane model to improve the accuracy and reliability of vehicle positioning. In the example of HMI display, the three-lane model, lane centerline, target trajectory and other information can be displayed on the vehicle's human-computer interaction interface (HMI), so that the driver can intuitively understand the road conditions around the vehicle and the operating status of the auxiliary system.

[0085] In one embodiment, from a single functional dimension, the driving assistance control operation includes any one of centering control, lane keeping, departure warning, automatic deviation correction, automatic lane change / lane change assist, following control, collision risk warning, automatic emergency braking, cruise control, driver-commanded lane change, nudge, and bypass functions. For example, in the driver-commanded lane change example, after the driver issues a lane change command via the steering wheel paddles or other control devices, the system assists the vehicle in completing the lane change, improving the smoothness and safety of lane changes, especially in highway driving or heavy traffic conditions. In the nudge example, when the vehicle slightly deviates from the lane centerline, the system automatically makes slight steering adjustments to quickly return the vehicle to the lane center, enhancing driving stability and comfort. In the bypass example, when encountering a slower vehicle, obstacle, or other situation ahead that requires avoidance, the system assists the vehicle in safely circumventing the obstacle, plans a temporary avoidance path, and returns to the original lane when conditions permit, ensuring continuous and safe driving.

[0086] In one embodiment, from an integrated functional perspective, the driving assistance control operation includes any one of the following: Adaptive Cruise Control (ACC) and Traffic Jam Assist (TJA). For example, in the ACC example, the system automatically adjusts the vehicle speed to maintain a safe distance from the vehicle ahead, combining vehicle speed and distance information in a three-lane model. In the Traffic Jam Assist (TJA) example, in congested low-speed traffic conditions, the system simultaneously controls vehicle acceleration, deceleration, and steering to keep the vehicle in the centerline of the lane and maintain a safe distance from the vehicle ahead.

[0087] It is easy for those skilled in the art to understand that the method 1000 for constructing a lane model or the driving assistance control method provided in the above one or more embodiments of the present application can be implemented by a computer program. For example, the computer program is contained in a computer program product, and when the computer program is executed by a processor, the method 1000 for constructing a lane model or the driving assistance control method of one or more embodiments of the present application is implemented. For another example, when a computer-readable storage medium (such as a USB flash drive) storing the computer program is connected to a computer, running the computer program can execute the method 1000 for constructing a lane model or the driving assistance control method of one or more embodiments of the present application.

[0088] refer to Figure 7 , which shows a schematic structural diagram of an electronic device 7000 according to an embodiment of the present application. Figure 7 As shown, electronic device 7000 includes memory 710 and processor 720, with a computer program stored on memory 710. In one embodiment, processor 720 executes the computer program to implement the method for constructing a lane model or the driving assistance control method of one or more embodiments of the present application. In one or more embodiments, the electronic device can be a domain controller, a camera, or a radar.

[0089] In summary, the lane model construction scheme of the embodiment of the present application significantly improves the accuracy and stability of lane model construction in various scenarios as a whole. Specifically, by processing the detected lane lines, including parallelizing the lane lines, the impact of camera calibration errors or detection errors on the lane lines can be effectively reduced, making the lane lines more parallel and improving the accuracy of the lane model. In addition, the lane centerline coefficient is adaptively determined according to the identified lane scene, which is conducive to generating a lane centerline that is more in line with driving habits and improves the adaptability of the autonomous driving system in complex scenarios. Moreover, the curb constraint can ensure that the generated centerline point always maintains a certain safe distance from the curb, thereby improving driving safety.

[0090] The above examples primarily illustrate the lane modeling scheme of an embodiment of the present application. Although only some of the embodiments of the present application have been described, those skilled in the art will appreciate that the present application may be implemented in many other forms without departing from its spirit and scope. Therefore, the examples and embodiments presented are to be considered illustrative rather than restrictive, and the present application may encompass various modifications and substitutions without departing from the spirit and scope of the present application as defined in the claims.

Claims

1. A method for constructing a lane model, characterized in that: The method comprises: Processing the detected lane lines, wherein the processing includes parallelizing the lane lines; Adaptively determining a lane centerline coefficient using the processed lane lines according to the identified lane scene; and While keeping a safe distance from the roadside, points on the lane centerline are sampled according to the determined lane centerline coefficients to construct the lane model.

2. The method of claim 1, further comprising: Extracting lane line points from the image and converting the lane line points into a vehicle coordinate system; Fitting the lane line points to obtain lane line coefficients; as well as The lane line position attribute is determined according to the lane line coefficient.

3. The method of claim 2, further comprising: After determining the lane line position attributes, the historical lane line information is processed to align it with the current lane line information.

4. The method according to claim 3, wherein: After determining the lane line position attributes, processing the historical lane line information to align it with the current lane line information includes: Predicting lane lines in historical frames to estimate the position and shape of lane lines in historical frames at a current moment; and Considering the lane change situation, the position of the lane line in the historical frame is aligned with the position of the lane line at the current moment.

5. The method according to claim 1, wherein The processing further includes: after the parallelization processing, performing stability and continuity processing on the lane line, wherein the stability and continuity processing includes multi-frame verification, missing compensation and / or fusion smoothing.

6. The method of claim 1, wherein: Parallelizing the lane lines includes: extracting a plurality of relatively parallel lane lines from the lane lines; and A weighted average is performed on the lane line coefficients of the plurality of relatively parallel lane lines.

7. The method according to claim 5, wherein: The multi-frame verification includes: When a lane line jump is detected, the lane lines of multiple consecutive frames are detected. If the lane lines in the multiple consecutive frames are stable, the new lane line is switched; otherwise, the lane line of the historical frame is maintained.

8. The method of claim 5, wherein: The missing compensation includes: In the case of missing lane lines, the missing lane lines are compensated by extending the lane lines of the historical frame forward, using the same-side curb, using the default lane width, or using the lane lines of the adjacent lane.

9. The method of claim 5, wherein: The fusion smoothing includes: The lane lines in the historical frames and the lane lines at the current moment are fused through weighted averaging or Kalman filtering to eliminate the jitter of the lane lines.

10. The method of claim 1, wherein: Based on the identified lane scene, adaptively determining the lane centerline coefficient using the processed lane lines includes: When the lane scene indicates a normal lane, the lane centerline coefficient is calculated based on the left and right lane lines of the lane by length weighting; When the lane scene indicates an extra-wide lane, calculating the lane centerline coefficient based on the lane line closest to the ego vehicle; When the lane scenario indicates a lane separation scenario, calculating the lane centerline coefficient based on a lane line having a smaller absolute value of curvature; and When the lane scenario indicates a lane merging scenario, a merging direction is determined and the lane centerline coefficient is calculated based on a lane line on an opposite side of the merging direction.

11. The method according to claim 10, wherein: Adaptively determining a lane centerline coefficient using the processed lane lines according to the identified lane scene also includes: The lane centerline coefficient calculated for the lane at the current moment is fused with the lane centerline coefficient calculated for the lane in the historical frame.

12. The method of claim 1, wherein: While maintaining a safe distance from the roadside, sampling points on the lane centerline according to the determined lane centerline coefficient to construct the lane model includes: The points on the lane centerline are sampled into a series of point sets according to the centerline coefficient, wherein the curbs on both sides are searched during the sampling process. If the distance between the centerline point and the curb is less than a preset threshold, the sampled centerline point is offset to keep it at a safe distance from the curb; Smoothing the sampled point set; and The sampled centerline points of the left lane, center lane, and right lane are combined with the lane lines and other lane markings to generate a three-lane model.

13. A driving auxiliary control method, characterized in that: The method comprises: Constructing a three-lane model based on the method according to any one of claims 1 to 12; Updating lane line information according to the three-lane model; and The driving assistance control operation is performed based on the updated lane marking information.

14. The method of claim 13, wherein: The driving auxiliary control operation includes any one of path planning, vehicle following target identification / screening, target trajectory prediction, fusion positioning and human-computer interaction HMI display.

15. The method of claim 13, wherein: The driving assistance control operation includes any one of centering control, lane keeping, departure warning, automatic correction, automatic lane change / lane change assistance, follow-up control, collision risk warning, automatic emergency braking, cruise control, driver-commanded lane change, slight adjustment and bypassing functions.

16. The method of claim 13, wherein: The driving assistance control operation includes any one of an automatic cruise control ACC function and a traffic jam assist TJA function.

17. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 16 is implemented.

18. An electronic device comprising a memory, a processor, and a computer program stored in the memory, wherein: The processor executes the computer program to implement the method according to any one of claims 1 to 16.

19. The electronic device according to claim 18, wherein The electronic device is a domain controller, a camera or a radar.

Images