A training method of a lane line detection model

Abstract

This invention provides a training method for a lane detection model, relating to the field of neural network model training technology. The invention first acquires road sample images labeled with real lane lines. Then, the road sample images are input into a pre-established initial model. Anchor points are selected on the horizontal axis, using the bottom edge of the road sample images as the horizontal axis. Multiple straight anchor lines with different preset angles are generated from these anchor points, and curved anchor lines tangent to the straight anchor lines are also generated. Finally, the parameters of the initial model are adjusted based on the distances between the real lane lines and each straight anchor line, as well as the distances between the real lane lines and each curved anchor line, to train the initial model. This technical solution generates not only straight anchor lines but also curved anchor lines during the initial model training, thus addressing both lane line detection in normal scenarios and lane line detection in curved scenarios, improving the accuracy of lane line detection in curved scenarios.

Description

Technical Field

[0001] This invention relates to the field of neural network model training technology, and in particular to a training method for a lane detection model. Background Technology

[0002] Currently, lane detection methods using deep learning are divided into those with anchors (anchor lines or anchor boxes) and those without anchors. In lane detection methods with anchors, the role of anchors is to provide positional references for lane line prediction by pre-setting some fixed anchor boxes or anchor lines. During the training of the convolutional neural network model, the distance between the labeled real lane lines and the pre-set anchor boxes or anchor lines is calculated to allow the neural network model to learn parameters, and then the trained model is used to predict lane lines. The setting and generation method of anchors can greatly affect the effect and efficiency of lane line detection.

[0003] In existing technologies, lane detection methods can achieve an accuracy of 92.14% in ordinary scenarios, but their accuracy in curved scenarios is only 67.72%. One reason for this poor performance in curve detection is that the lane detection methods use straight anchor lines, which is not conducive to the model learning the characteristics of curves during training. Summary of the Invention

[0004] The purpose of this invention is to provide a training method for a lane detection model, thereby solving the technical problem of low accuracy in lane detection in curved scenarios in the prior art.

[0005] According to the purpose of this invention, a method for training a lane detection model is provided, comprising the following steps:

[0006] Acquire a road sample image, wherein the road sample image is marked with real lane lines;

[0007] The road sample images are input into a pre-established initial model;

[0008] Using the bottom edge of the road sample image as the horizontal axis, select anchor points on the horizontal axis; generate multiple straight anchor lines with different preset angles starting from the anchor points, and generate curved anchor lines tangent to the straight anchor lines.

[0009] The parameters of the initial model are adjusted based on the distances between the actual lane lines and each straight anchor line, and the distances between the actual lane lines and each curved anchor line, in order to train the initial model.

[0010] Optionally, in the step of generating multiple straight anchor lines with different preset angles starting from the anchor point and generating curved anchor lines tangent to the straight anchor lines, the curvature of all the curved anchor lines is a preset curvature, and except for the straight anchor lines perpendicular to the horizontal axis, any straight anchor line generates only one curved anchor line, and the straight anchor lines perpendicular to the horizontal axis generate two curved anchor lines.

[0011] Optionally, the ordinate of the endpoint of the curved anchor line is less than or equal to the ordinate of the endpoint of the corresponding straight anchor line;

[0012] The central angle of the curved anchor line is greater than 0° and less than 90°.

[0013] Optionally, the step of inputting the road sample image into a pre-established initial model includes the following steps:

[0014] Using the bottom edge of the road sample image as the horizontal axis, select multiple anchor points arranged at intervals on the horizontal axis; generate multiple straight anchor lines with different preset angles starting from the multiple anchor points, and generate curved anchor lines tangent to the straight anchor lines.

[0015] Optionally, after generating multiple straight anchor lines with different preset angles starting from the anchor point, and generating curved anchor lines tangent to the straight anchor lines, the method further includes the following steps:

[0016] The function expression of the curved anchor line is calculated based on the first included angle of the straight anchor line and the radius of the curved anchor line, wherein the preset angle is the first included angle between the straight anchor line and the horizontal axis.

[0017] Optionally, the step of calculating the functional expression of the curved anchor line based on the first included angle of the straight anchor line and the radius of the curved anchor line specifically includes the following steps:

[0018] Calculate the second angle between the starting line of the curved anchor line and the horizontal axis based on the first included angle of the straight anchor line. The starting line of the curved anchor line is the straight line between the center of the curved anchor line and the anchor point.

[0019] Calculate the center coordinates of the curved anchor line based on the second included angle and the radius of the curved anchor line;

[0020] The function expression of the curved anchor line is determined based on the center coordinates of the curved anchor line.

[0021] Optionally, after the step of calculating the functional expression of the curved anchor line based on the center coordinates of the curved anchor line, the following steps are further included:

[0022] Select multiple dividing lines, wherein the multiple dividing lines are straight lines parallel to the horizontal axis and whose vertical coordinates are not zero;

[0023] Based on the ordinates of the multiple dividing lines, the function expression of the straight anchor line, and the function expression of the curved anchor line, the multiple abscissas of the intersection points of the multiple dividing lines with the straight anchor line and the curved anchor line are calculated respectively. The function expression of the straight anchor line is calculated based on the first included angle and the coordinates of the anchor point.

[0024] Optionally, the step of adjusting the parameters of the initial model based on the distances between the actual lane lines and each straight anchor line and the distances between the actual lane lines and each curved anchor line to train the initial model specifically includes the following steps:

[0025] The parameters of the initial model are adjusted based on the x-coordinates of the intersection points of the multiple dividing lines with the straight anchor line and the curved anchor line, and the lateral distance between the lane lines in the lane line training image, thereby training the initial model.

[0026] Optionally, the center coordinates of the curved anchor line are calculated according to the following formula:

[0027] x1 = x0 + R*cos(π / 2 - θ);

[0028] y1 = -R*sin(π / 2 - θ);

[0029] Where x1 is the x-coordinate of the center of the circle and y1 is the y-coordinate of the center of the circle;

[0030] x s R is the x-coordinate of the anchor point, R is the radius of the curved anchor line, θ is the first included angle, and π / 2-θ is the second included angle.

[0031] Optionally, the functional expression of the curved anchor line is: (x – x0) 2 +(y-y0) 2 =0.

[0032] This invention first acquires road sample images labeled with real lane lines. These images are then input into a pre-established initial model. Anchor points are selected on the horizontal axis, using the bottom edge of the road sample image as the starting point. Multiple straight anchor lines at different preset angles are generated from these anchor points, along with curved anchor lines tangent to the straight anchor lines. Finally, the parameters of the initial model are adjusted based on the distances between the real lane lines and each straight anchor line, as well as the distances between the real lane lines and each curved anchor line, to train the initial model. This technical solution generates not only straight anchor lines but also curved anchor lines during initial model training, thus addressing both lane line detection in normal scenarios and curved scenarios, improving the accuracy of lane line detection in curved scenarios.

[0033] The above and other objects, advantages and features of the present invention will become more apparent to those skilled in the art from the following detailed description of specific embodiments of the invention in conjunction with the accompanying drawings. Attached Figure Description

[0034] The following sections will describe some specific embodiments of the invention in a detailed manner by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or portions. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0035] Figure 1 This is a schematic diagram of the anchor line in the existing lane line detection method;

[0036] Figure 2 This is a schematic diagram of an anchor line and a real lane line in existing lane line detection methods;

[0037] Figure 3 This is a schematic diagram of another anchor line and a real lane line in the existing lane line detection method;

[0038] Figure 4 This is a schematic flowchart of a training method for a lane detection model according to an embodiment of the present invention;

[0039] Figure 5 This is a schematic diagram of the anchor line in the training method of the lane line detection model according to an embodiment of the present invention;

[0040] Figure 6 This is a schematic diagram of anchor lines and real lane lines in a training method for a lane line detection model according to an embodiment of the present invention;

[0041] Figure 7 This is a schematic diagram of straight anchor lines and curved anchor lines in a training method for a lane line detection model according to an embodiment of the present invention. Detailed Implementation

[0042] Embodiments of the present invention are described in detail below, examples of which are illustrated 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 the present invention, and should not be construed as limiting the present invention.

[0043] Figure 1 This is a schematic diagram of the anchor line in existing lane line detection methods. Figure 2 This is a schematic diagram of an anchor line and a real lane line in existing lane line detection methods. Figure 3 This is a schematic diagram of another anchor line and a real lane line in existing lane line detection methods. For example... Figure 1 , Figure 2 and Figure 3 As shown, in the existing anchor line generation method for lane line detection, the anchor line generation process is as follows: An RGB image is input, and a certain number of points are equally selected on its left, right, and bottom edges as the starting points of the anchor lines, along with a certain number of angles θ. Straight lines with an angle θ to the horizontal line are generated from these starting points as the straight anchor lines 10'. Specifically, the number of starting points on the left is 72, and the values ​​of θ are {θ1, θ2, ..., θ6} = {72°, 60°, 49°, 39°, 30°, 22°}. Each starting point generates 6 anchor lines with angles θ1, θ2, ..., θ6. The number of starting points on the right is also 72, and the values ​​of θ are {θ1, θ2, ..., θ6} = {108°, 120°, 131°, 141°, 150°, 158°}. The generation method for the straight anchor lines 10' is the same as on the left. The number of starting points on the bottom edge is 128, and the values ​​of θ are {θ1, θ2, ..., θ15} = {165°, 150°, 141°, 131°, 120°, 108°, 100°, 90°, 80°, 72°, 60°, 49°, 39°, 30°, 15°}. The generation method of the straight anchor lines 10' is the same as that of the left and right sides, generating 15 lines for each starting point, corresponding to 15 different θ angles. For an RGB image, the total number of generated straight anchor lines 10' is 2*72*6 + 128*15 = 2784. The existing lane line detection method can achieve an average accuracy of 76.68% in all scenarios.

[0044] The principle of this algorithm is as follows: During the training phase of the convolutional neural network model, an RGB image with labeled lane line values ​​is input. First, straight anchor lines 10' are generated according to the method described above. (See...) Figure 2 and Figure 3The image is then vertically divided into 72 equal parts, with each dividing line parallel to the bottom edge. These dividing lines intersect the straight anchor line 10' and the actual lane lines at several cutoff points. The number of cutoff points is determined by the actual length of the actual lane lines. Parts of the image above or below the actual lane lines are not segmented. Since the vertical coordinate of each cutoff point is fixed, the difference in horizontal coordinate between the cutoff points of the straight anchor line 10' and the actual lane lines on the same horizontal line represents the distance between the two lines at that horizontal section. Predicted lane lines are then generated based on the position of the straight anchor line 10'. The network learns and iteratively updates its parameters, shortening the distance between the predicted values ​​and the actual lane lines at the cutoff points. Finally, a pre-trained convolutional neural network model is used for lane line detection.

[0045] Taking a starting point 'a' on the bottom edge as an example, the generated straight anchor line 10' is as follows: Figure 1 As shown. Figure 1 The center is a set of 15 straight anchor lines 10' at different angles generated from point a.

[0046] like Figure 2 and Figure 3 As shown, the straight line with the arrow is the straight anchor line 10' generated by the lane detection algorithm, and the curve is the actual lane line that has been marked in the curve scenario. Figure 2 and Figure 3 These represent two possible scenarios between the anchor line and the label.

[0047] Because the curvature of the lane lines on curves is large, in both cases, the distance between the straight anchor line 10' and the actual lane line increases at the cut-off point, and the distance at the end is large. Therefore, the straight anchor line 10' cannot fit the actual lane line well, which increases the difficulty of model parameter learning during the training of the convolutional neural network and makes the loss function converge very slowly. Therefore, in the case of curves, the lane line detection method using straight lines as anchor lines has low detection accuracy.

[0048] Figure 4 This is a schematic flowchart illustrating a training method for a lane detection model according to an embodiment of the present invention. Figure 4 As shown, the training method for the lane detection model includes the following steps:

[0049] Step S100: Obtain a road sample image, which is marked with real lane lines;

[0050] Step S200: Input the road sample image into the pre-established initial model;

[0051] Step S300: Using the bottom edge of the road sample image as the horizontal axis, select anchor point 50 on the horizontal axis;

[0052] Step S400: Generate multiple straight anchor lines 10 with different preset angles starting from anchor point 50, and generate curved anchor lines 20 that are tangent to the straight anchor lines 10.

[0053] Step S500: Adjust the parameters of the initial model based on the distance between the actual lane lines and each straight anchor line 10 and the distance between the actual lane lines and each curved anchor line 20 to train the initial model.

[0054] This embodiment generates not only straight anchor lines 10 but also curved anchor lines 20 during the initial model training, thus addressing both lane line detection in normal scenarios and curved scenarios, thereby improving the accuracy of lane line detection in curved scenarios. In other words, this embodiment can improve the problem of lane line detection models struggling to identify lane lines in curved scenarios.

[0055] Figure 5 This is a schematic diagram of the anchor line in the training method of the lane line detection model according to an embodiment of the present invention. Figure 6 This is a schematic diagram of the anchor line and the real lane line 30 in a lane line detection model training method according to an embodiment of the present invention. Figure 5 and Figure 6 As shown, the anchor line generation method proposed in this embodiment generates both straight lines with different preset angles as anchor lines and curves with a fixed curvature that are tangent to the straight lines as anchor lines. During model training, this allows the curve anchor line 20 to better and faster fit the labeled real curve lane lines. Figure 6 The curve with arrows on the right is the fixed-curvature curve anchor line 20, while the left side shows the actual labeled curve lane lines. It can be seen that because the curve anchor line 20 has a certain curvature, it better fits the real curve lane lines. At each cutoff point, the distance between the curve anchor line 20 and the real lane line 30 is small, allowing for a better fit to the real lane line 30 in a curve scenario. This embodiment incorporates curves during the anchor line generation process, enabling the lane line detection model to better learn curve characteristics and improving the accuracy and efficiency of the lane line detection model in curve scenarios.

[0056] In this embodiment, in step S400, the curvature of all curved anchor lines 20 is a preset curvature, and except for the straight anchor lines 10 perpendicular to the horizontal axis, each of the other straight anchor lines 10 generates only one curved anchor line 20, while the straight anchor lines 10 perpendicular to the horizontal axis generate two curved anchor lines 20. See details. Figure 5 The preset curvature here is also called the fixed curvature, which is set according to specific design requirements.

[0057] In this embodiment, the ordinate of the endpoint of the curved anchor line 20 is less than or equal to the ordinate of the endpoint of the corresponding straight anchor line 10. This can be understood as the starting point of the curved anchor line 20 being the starting point of the previously selected straight anchor line 10, the curved anchor line 20 extending in the direction of the arrow, and its endpoint not exceeding the point of tangency with the straight anchor line 10 above it. Furthermore, the central angle of the curved anchor line 20 is greater than 0° and less than 90°. Therefore, except for the straight anchor lines 10 perpendicular to the horizontal axis, each of the remaining straight anchor lines 10 generates only one corresponding curved anchor line 20.

[0058] In this embodiment, the following steps are included after S200:

[0059] Step 1: Using the bottom edge of the road sample image as the horizontal axis, select multiple anchor points 50 spaced apart on the horizontal axis;

[0060] Step 2: Generate multiple straight anchor lines 10 with different preset angles starting from multiple anchor points 50, and generate curved anchor lines 20 that are tangent to the straight anchor lines 10.

[0061] Specifically, for an RGB image input with width w and height h, 128 points are taken at equal intervals along its bottom edge as the starting point of the anchor line. Each point has 9 θ values, namely θ1, θ2, ..., θ9, generating 9 straight anchor lines 10. Then, taking each straight anchor line 10 as a tangent, 10 arcs with curvature h are generated as curved anchor lines 20. Among them, an arc is generated on each side of the straight anchor line 10 perpendicular to the horizontal line.

[0062] Dividing line 40 is still divided into 72 equal parts on the vertical axis, forming 72 horizontal lines parallel to the horizontal axis. These lines intersect the straight anchor line 10, the curved anchor line 20, and the actual lane lines at several cutoff points. The straight line segments parallel to the horizontal lines represent the horizontal distances between each straight anchor line 10, curved anchor line 20, and the actual lane line. It can be seen that curved anchor line 20 can better fit the curved lane line; see [reference needed]. Figure 6 Here, the number and coordinates of the anchor line starting points, the number and magnitude of θ, the curvature of the curved anchor line 20, and the ordinate of the dividing line 40 are all selected based on experience and can be adjusted according to the situation in practical applications. Selecting different curvatures or θ also falls within the scope of this invention. In this embodiment, the curvature of all curved anchor lines 20 is a single fixed value. Multiple different curvatures can also be selected when generating the curved anchor lines 20 to generate various curved anchor lines 20 with different curvatures. This invention only selects anchor line starting points at the bottom edge to generate anchor lines, without selecting points on the left and right sides. A certain number of points can also be selected on the left and right sides as anchor line starting points.

[0063] Figure 7 This is a schematic diagram of straight anchor line 10 and curved anchor line 20 in a lane line detection model training method according to an embodiment of the present invention. Figure 7 As shown, in this embodiment, after step S400, the following steps are also included:

[0064] Step 3: Calculate the function expression of the curved anchor line 20 based on the first included angle θ of the straight anchor line 10 and the radius R of the curved anchor line 20. The preset angle is the first included angle θ between the straight anchor line 10 and the horizontal axis.

[0065] Specifically, step three includes the following steps:

[0066] The first step is to calculate the second angle between the starting line of the curved anchor line 20 and the horizontal axis based on the first included angle θ of the straight anchor line 10. The starting line of the curved anchor line 20 is the straight line between the center of the curved anchor line 20 and the anchor point 50.

[0067] The second step is based on the second included angle. Calculate the center coordinates of the curved anchor line 20 using the radius R of the curved anchor line 20.

[0068] The third step is to determine the function expression of the curve anchor line 20 based on the center coordinates of the curve anchor line 20.

[0069] In this embodiment, after the third step, the following steps are also included:

[0070] The fourth step is to select multiple dividing lines 40, which are straight lines parallel to the horizontal axis and whose vertical coordinates are not zero.

[0071] Fifth, based on the ordinates of the multiple dividing lines 40, the function expression of the straight anchor line 10, and the function expression of the curved anchor line 20, calculate the multiple abscissas of the intersection points of the multiple dividing lines 40 with the straight anchor line 10 and the curved anchor line 20, respectively. The function expression of the straight anchor line 10 is calculated based on the first included angle and the coordinates of the anchor point 50. Here, the distances between the multiple dividing lines 40 are equal, which can be understood as the distances between the ordinates of the multiple dividing lines 40 being equal.

[0072] In this embodiment, step S500 specifically includes the following steps:

[0073] The parameters of the initial model are adjusted by using the x-coordinates of the intersection points of multiple segmentation lines 40 with straight anchor lines 10 and curved anchor lines 20 and the lateral distance between lane lines in the lane line training image, thereby training the initial model.

[0074] Specifically, the function for the straight anchor line 10 is: y = (xx s )*tanθ;

[0075] Where, x s Let θ be the x-coordinate of anchor point 50, and θ be the first included angle.

[0076] Specifically, the coordinates of the center of the curve anchor line 20 are calculated according to the following formula:

[0077] x1 = x0 + R*cos(π / 2 - θ);

[0078] y1 = -R*sin(π / 2 - θ);

[0079] Where x1 is the x-coordinate of the center of the circle and y1 is the y-coordinate of the center of the circle;

[0080] x s Let be the x-coordinate of anchor point 50, R be the radius of the curved anchor line 20, θ be the first included angle, and π / 2 - θ be the second included angle. The second included angle It is calculated based on the properties of circles and tangents.

[0081] According to mathematical properties, the coordinates of the center of the circle corresponding to the anchor line 20 are (x0-y0) = [(x s +R*cos(π / 2–θ), -R*sin(π / 2–θ)].

[0082] Therefore, the functional expression for curve anchor line 20 is: (x – x0) 2 +(y-y0) 2 =0.

[0083] Assume (x) l y l ), (x c y c The coordinates of the intersection points of the dividing line 40 with the straight anchor line 10 and the curved anchor line 20 are respectively, where y l =y c , are the ordinates of the two intersection points, x and y respectively. l With x c Let y be the x-coordinates of the two intersection points. l and y c Since this is known, the x-coordinates of the straight anchor line 10 and the curved anchor line 20 at their intersection point can be obtained respectively based on the functions of the straight anchor line 10 and the curved anchor line 20. l and x c Therefore, the x-coordinate can be calculated. l x c The distance between the model and the actual lane line 40 is used to adjust the parameters of the initial model.

[0084] Therefore, those skilled in the art should recognize that although numerous exemplary embodiments of the present invention have been shown and described in detail herein, many other variations or modifications conforming to the principles of the present invention can be directly determined or derived from the disclosure of the present invention without departing from the spirit and scope of the invention. Thus, the scope of the present invention should be understood and construed as covering all such other variations or modifications.

Claims

1. A method for training a lane line detection model, characterized in that, The method comprises the following steps: obtaining a road sample image, wherein the road sample image is labeled with a real lane line; inputting the road sample image into a pre-established initial model; selecting an anchor point on a horizontal axis of a bottom edge of the road sample image; generating a plurality of straight anchor lines with different preset angles and a curve anchor line tangent to the straight anchor lines, with the anchor point as a starting point; adjusting parameters of the initial model according to distances between the real lane line and each straight anchor line and distances between the real lane line and each curve anchor line, to train the initial model; wherein, after the step of generating a plurality of straight anchor lines with different preset angles and a curve anchor line tangent to the straight anchor lines, with the anchor point as a starting point, the method further comprises the following steps: calculating a function expression of the curve anchor line according to a first included angle of the straight anchor line and a radius of the curve anchor line, wherein the preset angle is the first included angle between the straight anchor line and the horizontal axis; and, the step of calculating the function expression of the curve anchor line according to the first included angle of the straight anchor line and the radius of the curve anchor line specifically comprises the following steps: calculating a second included angle between a starting line of the curve anchor line and the horizontal axis according to the first included angle of the straight anchor line, wherein the starting line of the curve anchor line is a straight line between a center of the curve anchor line and the anchor point; calculating a center coordinate of the curve anchor line according to the second included angle and the radius of the curve anchor line; determining the function expression of the curve anchor line according to the center coordinate of the curve anchor line.

2. The training method of claim 1, wherein, In the step of generating a plurality of straight anchor lines with different preset angles and a curve anchor line tangent to the straight anchor lines, with the anchor point as a starting point, curvatures of all the curve anchor lines are preset curvatures, and only one curve anchor line is generated for each straight anchor line except the straight anchor line perpendicular to the horizontal axis, and two curve anchor lines are generated for the straight anchor line perpendicular to the horizontal axis.

3. The training method of claim 2, wherein, A vertical coordinate of an end point of the curve anchor line is less than or equal to a vertical coordinate of an end point of the corresponding straight anchor line. A central angle of the curve anchor line is greater than 0° and less than 90°.

4. The training method according to any one of claims 1-3, characterized in that, After the step of inputting the road sample image into the pre-established initial model, the method further comprises the following steps: selecting a plurality of anchor points arranged at intervals on the horizontal axis of the bottom edge of the road sample image; and generating a plurality of straight anchor lines with different preset angles and a curve anchor line tangent to the straight anchor lines, with the plurality of anchor points as starting points.

5. The training method of claim 1, wherein, After the step of calculating the function expression of the curve anchor line according to the center coordinate of the curve anchor line, the method further comprises the following steps: selecting a plurality of segmentation lines, wherein the plurality of segmentation lines are straight lines parallel to the horizontal axis and having non-zero vertical coordinates; calculating a plurality of horizontal coordinates of intersection points of the plurality of segmentation lines and the straight anchor line and the curve anchor line according to vertical coordinates of the plurality of segmentation lines, the function expression of the straight anchor line and the function expression of the curve anchor line, wherein the function expression of the straight anchor line is calculated according to the first included angle and a coordinate of the anchor point.

6. The training method of claim 5, wherein, The step of adjusting parameters of the initial model according to distances between the real lane line and each straight anchor line and distances between the real lane line and each curved anchor line comprises the following steps: The step of training the initial model according to the parameters of the initial model comprises the following steps:

7. The training method of claim 1, wherein, The step of training the initial model according to the parameters of the initial model comprises the following steps: x0 = x s + R*cos(pi / 2-theta); The center coordinates of the curved anchor line are calculated according to the following formula: y0=-R*sin(π / 2-θ); x s x is the abscissa of the anchor point, R is the radius of the curve anchor line, θ is the first angle, and π / 2 - θ is the value of the second angle.

8. The training method of claim 1, wherein, Wherein, x0 and y0 are the horizontal and vertical center coordinates of the circle corresponding to the curved anchor line. (x - x0) 2 + (y - y0) 2 = R 2 ; The function expression of the curved anchor line is: Wherein, x and y are the horizontal and vertical coordinates of any point on the curved anchor line. x0 and y0 are the horizontal and vertical center coordinates of the circle corresponding to the curved anchor line. R is the radius of the curved anchor line.

Images