Positioning method and device for assisting vehicle in lane changing and transverse position

Abstract

The invention discloses a positioning method and device for assisting vehicle lane changing and transverse position, and belongs to the technical field of computer application, and the positioning method for assisting vehicle lane changing and transverse position comprises the following steps: S1, obtaining vehicle position information and vehicle speed from a navigation module; s2, performing camera internal reference calibration and camera external reference calibration; s3, acquiring image data from a camera and preprocessing the image data; s4, extracting a vehicle drivable area contour; s5, acquiring related information of the vehicle; and S6, determining a vehicle lane changing signal and a vehicle transverse position. According to the invention, through a lightweight model, real-time image processing and camera calibration technology, vehicle transverse position positioning and lane changing judgment can be accurately carried out in real time on a low-performance vehicle machine.

Description

Technical Field

[0001] The present invention relates to the field of computer application technology, and in particular to a method and device for assisting a vehicle in changing lanes and locating a lateral position. Background Art

[0002] In recent years, with the improvement of traffic conditions and the increase in demand for driving safety, real-time lateral position information of vehicles has become particularly important in assisted driving and advanced driver assistance systems. This information can help vehicles optimize their driving routes based on the lane they are in and combined with map data. Especially in the non-entry and exit areas of highways, accurate lateral position information of vehicles can not only assist vehicles in changing lanes quickly, but also avoid missing highway exits.

[0003] At present, vehicle positioning in the scene of high-speed trunk roads is mainly determined by lane-level maps or high-precision map positioning, but ordinary maps still cannot obtain accurate vehicle location information. Among them, the patent application numbers are CN201710487137.8 and CN201811393431.3, which mainly detect the drivable area based on deep learning algorithms. For low-performance car computers, the complex network structure and huge parameter algorithms are difficult to run in real time and occupy a lot of memory resources. For this reason, a method and device for assisting vehicle lane change and lateral position positioning is proposed. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, one object of the present invention is to provide a method and device for assisting a vehicle in changing lanes and locating a lateral position.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for assisting a vehicle in changing lanes and locating a lateral position comprises the following steps:

[0007] S1. Obtain vehicle location information and speed from the navigation module;

[0008] S2. Perform camera internal parameter calibration and camera external parameter calibration;

[0009] S3. Preprocessing of image data obtained from the camera;

[0010] S4. extracting the contour of the vehicle drivable area;

[0011] S5. Obtain relevant information of the vehicle;

[0012] S6. Determine the vehicle lane change signal and the vehicle lateral position.

[0013] As a further optimization scheme of the present invention, in step S2, the camera internal parameter calibration is performed by using the kalibr calibration tool or opencv, and the focal length f in the x and y directions in the image coordinate system is obtained using the internal parameter calibration. x ,f y ;

[0014] The optical center coordinate is C x ,C y Thus, the camera internal parameter matrix M1 = [f x ,0,C x ; 0,f y ,C y ; 0,0,1], at the same time, the five parameters D = (k1, k2, P1, P2, k3) of the distortion coefficient distortion mathematical model are obtained through camera internal calibration;

[0015] For camera extrinsic calibration, the camera position is selected as the origin on the ground projection point. According to the right-hand coordinate system, the horizontal right direction is the positive direction of the X axis, the vehicle driving direction is the positive direction of the Y axis, and the Z axis is vertical to the ground and upwards is the positive direction. The camera extrinsic parameter matrix M2 is calculated through the vanishing point. At the same time, the lateral distance S between the camera and the left wheel is calibrated. l The lateral distance S2 from the camera to the right wheel and the height H from the camera to the ground.

[0016] As a further optimization scheme of the present invention, in step S4, for the modification of the Decoder stage of the SegFormer model, firstly, the Features of different layers are respectively passed through the linear layer to ensure that the channel dimensions are the same, then they are all deconvolved to the original resolution and weighted averaged, and then fused with a linear layer, and a 1x1 convolution kernel is used for convolution to reduce the convolution kernel parameters and simplify the model, and finally a linear layer predicts the result;

[0017]

[0018]

[0019]

[0020] F conv =Conv 1×1 (F),

[0021] M=Linear(C,N cls )(F conv )

[0022] As a further optimization scheme of the present invention, in step S4, the modification of the loss function of the SegFormer model is expressed as:

[0023] Ljoint_seg =L ce +L energy

[0024]

[0025] L energy =S*E(i,j),

[0026] in:

[0027] in, is the normalized weight, S is the weight hyperparameter, D represents the spatial pixel position of image I, and G is essentially a Gaussian kernel filter.

[0028] As a further optimization scheme of the present invention, in step S4, for the segmentation of the drivable area contour, the image is first preprocessed, the SegFormer model and weights are loaded, and according to the loss function and the Adam optimizer, the hyperparameters are adjusted during the training process to improve the model performance. Finally, the trained improved SegFormer model is applied to load the partially preprocessed image in step S3 to perform drivable area contour segmentation.

[0029] As a further optimization solution of the present invention, in step S4, the contour of the drivable area is post-processed, and the specific steps are as follows:

[0030] S401. Perform polygon fitting on the contour points of the drivable area based on the corresponding opencv function approxPolyDP in the Douglas-Peucker algorithm to obtain a polygon point set;

[0031] S402. Traverse the polygon point set and search for the first point that meets the conditions and re-store the polygon point set in a certain order;

[0032] S403. When the number of polygon point sets is less than the threshold, all polygon point sets are used as key points of the drivable area contour; otherwise, key point selection is performed based on the re-stored polygon point sets;

[0033] S404. For the selection of key points of the contour of the drivable area, three points are taken each time, namely the current point, the second point after the current point, and the third point, and the Euclidean distance, the lateral distance, and the longitudinal distance are calculated with the current point respectively, and the Euclidean distance, the lateral distance, and the longitudinal distance are compared with the set threshold value to determine the traversal step length of the key point selection and store the points that meet the conditions as key points;

[0034] S405. Based on the above key points, the slope, horizontal distance, vertical distance and Euclidean distance of two adjacent points and the horizontal coordinate of the key points are compared with the threshold to determine the left boundary point set and the right boundary point set of the drivable area contour;

[0035] S406. Based on the slope and bottom boundary intersection features, the boundary point set is further optimized to obtain the optimal boundary point set. The optimized point set includes the top point and the bottom end point of the boundary. The top point and the bottom end point are converted back to the original image coordinates, and the intrinsic parameter matrix and the extrinsic parameter matrix based on the camera calibration are converted into coordinates in the world coordinate system, and the feasibility is calculated.

[0036] As a further optimization scheme of the present invention, in step S5, SegFormer improved based on the semantic segmentation network extracts other moving objects around the vehicle, screens out the closest moving obstacles in front and behind the vehicle lane or in the adjacent left and right lanes, and converts the pixel coordinates of the obstacle's distance from the vehicle into the actual distance in the world coordinate system according to the camera's intrinsic parameter matrix M1, extrinsic parameter matrix M2 and distortion matrix D.

[0037] As a further optimization scheme of the present invention, in step S6, to determine the lateral position of the vehicle, firstly, according to the drivable area boundary obtained in step S4, the confidence of the left and right boundaries is compared with the set threshold value, and when the confidence is greater than the threshold value, the horizontal coordinate mean S of the corresponding boundary is calculated, and the lane width W of the road is calculated. d The estimated vehicle position is the floor (S / W) corresponding to the boundary of the drivable area. d )+ lanes, where floor is rounded down.

[0038] As a further optimization scheme of the present invention, in step S6, for determining the vehicle changeable lane signal, according to the drivable area boundary obtained in step S4, the confidence of the left and right boundaries are compared with the set threshold value respectively. When the confidence is greater than the threshold value, the horizontal coordinate mean S1 of the corresponding boundary is calculated, and according to the lane line parameter equation in step S5, several points are sampled respectively to calculate the corresponding horizontal coordinate mean S2;

[0039] The calculation formula is: d=abs(S1-S2), where abs() is an absolute value operation.

[0040] A device for assisting a vehicle in changing lanes and positioning its lateral position, comprising:

[0041] Memory for storing computer programs;

[0042] The processor is used to implement the steps of the method for assisting vehicle lane change and lateral position positioning when executing the computer program.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] The present invention obtains the current position information and speed of the vehicle through the navigation and GPS modules. When the vehicle is detected in the non-high-speed entrance and exit area and the speed is greater than the threshold, it enters the drivable area detection function, obtains the contour of the drivable area and finally obtains the left and right boundaries of the drivable area. Based on the camera external parameter calibration information, the lateral pixels of the vehicle from the left and right boundaries of the drivable area are converted into distances. On the one hand, the lateral position of the vehicle can be determined according to the distance from the vehicle to the left and right boundaries of the drivable area; on the other hand, the lane line information of the vehicle is obtained and compared with the contour boundary of the drivable area and whether the lane can be changed based on the surrounding vehicle information. Through lightweight models, real-time image processing and camera calibration technology, real-time and accurate vehicle lateral position positioning and lane change judgment on low-performance vehicle computers are achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A flow chart of the method for assisting a vehicle in changing lanes and lateral positioning according to the present invention;

[0046] Figure 2 This is a decoder flow chart of step S4 in the present invention;

[0047] Figure 3 The flowchart of the drivable area contour segmentation in step S4 of the present invention;

[0048] Figure 4 This is a flowchart of post-processing of the drivable area detection in step S4 of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] See also Figure 1-Figure 4 , the present invention provides a technical solution:

[0051] A method for assisting a vehicle in changing lanes and locating a lateral position comprises the following steps:

[0052] S1. Obtain vehicle location information and speed from the navigation module;

[0053] Specifically, the current vehicle location information is obtained from the navigation map data module to determine that the vehicle location is not in the high-speed entrance and exit area;

[0054] The acquisition standard is that the distance from a standard highway entrance to exit is less than or equal to 500m, and the distance from a structured urban road to exit is less than or equal to 30m; after the vehicle speed is acquired by the navigation data module, when the vehicle speed meets the minimum threshold (such as 50km / h), the functions of steps S2-S6 of the present invention are started;

[0055] Since the lane lines within a certain distance range of the entrance and exit are all solid lines, they are not within the scope of application of the present invention.

[0056] S2. Perform camera internal parameter calibration and camera external parameter calibration;

[0057] Specifically, for camera intrinsic calibration, the camera intrinsic calibration is performed through the kalibr calibration tool or the calibration method in opencv, and the focal length f in the x-direction and y-direction in the image coordinate system is obtained through the intrinsic calibration. x ,f y ;

[0058] The optical center coordinate is C x ,C y Thus, the camera internal parameter matrix M1 = [f x ,0,C x ; 0,f y ,C y ; 0,0,1], at the same time, the five parameters D = (k1, k2, P1, P2, k3) of the distortion coefficient distortion mathematical model are obtained through camera internal calibration;

[0059] For camera extrinsic calibration, the camera position is selected as the origin on the ground projection point. According to the right-hand coordinate system, the horizontal right direction is the positive direction of the X axis, the vehicle driving direction is the positive direction of the Y axis, and the Z axis is vertical to the ground and upwards is the positive direction. The camera extrinsic parameter matrix M2 is calculated through the vanishing point. At the same time, the lateral distance S between the camera and the left wheel is calibrated. l The lateral distance S2 from the camera to the right wheel and the height H from the camera to the ground.

[0060] S3. Preprocessing of image data obtained from the camera;

[0061] Specifically, the camera selection should choose a FOV of 60°-100°, which can ensure a large enough lateral field of view without affecting the accuracy of the driving direction. The data format uses the YUV (YV12, NV21) format, and the resolution is selected according to its own needs, such as 720p. Data is collected at a regular period and converted into RGB format.

[0062] S4. extracting the contour of the vehicle drivable area;

[0063] The drivable area in the present invention refers to the area between two solid lines on the road, and the boundary of the drivable area mainly refers to the detected left and right solid line boundaries or road edges.

[0064] Specifically, for the modification of the Decoder stage of the SegFormer model, first, the features of different layers are passed through the linear layer to ensure that the channel dimensions are the same, then they are all deconvolved to the original resolution and weighted averaged, and then fused with a linear layer, and convolution with a 1x1 convolution kernel is applied to reduce the convolution kernel parameters and simplify the model, and finally a linear layer predicts the results.

[0065]

[0066]

[0067]

[0068] F conv =Conv 1×1 (F),

[0069] M=Linear(C,N cls )(F conv )

[0070] The modification of the loss function of the SegFormer model is expressed as:

[0071] L joint_seg =L ce +L energy

[0072] Among them, the first part is the cross entropy loss and the second part is the energy loss.

[0073]

[0074] L energy =S*E(i,j),

[0075] in:

[0076] in, is the normalized weight, S is the weight hyperparameter, D represents the spatial pixel position of image I, and G is essentially a Gaussian kernel filter.

[0077] For drivable area contour segmentation, the image is preprocessed. First, 10,000 images from the BDD 100K dataset are used to adjust the SegFormer model. Load the image and mask. Convert the image to RGB format and the mask to YUV format. Then convert the mask to binary format, where non-zero pixels are considered part of the lane. Resize the mask to match the image dimensions and then convert it to a tensor. Finally, threshold the mask back to binary values ​​and convert it to a LongTensor.

[0078] Load the SegFormer model and weights, and adjust the hyperparameters during the training process to improve the model performance according to the loss function and the Adam optimizer. Finally, use the trained improved SegFormer model to load some of the preprocessed images in step S3 to perform drivable area contour segmentation.

[0079] For post-processing of the drivable area contour, the specific steps are:

[0080] S401. Based on the corresponding function of the Douglas-Peucker algorithm (iterative endpoint fitting algorithm), the opencv function approxPolyDP performs polygon fitting on the contour points of the drivable area to obtain a polygon point set;

[0081] The first parameter of the pproxPolyDP function is the input function which is the image contour point set, the first parameter is the output which is the polygon contour point set, the third parameter is the output accuracy which is the maximum distance between two points, and the fourth parameter is whether the output parameter is closed.

[0082] S402. Traverse the polygon point set and search for the first point that meets the conditions and re-store the polygon point set in a certain order, wherein the condition of the first point is that the distance between the current point and the next point is greater than the set threshold for the first time, set the current point as the first point, and re-store the polygon point set;

[0083] S403. When the number of polygon point sets is less than the threshold, all polygon point sets are used as key points of the drivable area contour; otherwise, key point selection is performed based on the re-stored polygon point sets;

[0084] S404. For the selection of key points of the drivable area contour, three points are taken each time, namely the current point, the second point after the current point, and the third point, and the Euclidean distance, the lateral distance, and the longitudinal distance are calculated with the current point respectively, and the Euclidean distance, the lateral distance, and the longitudinal distance are compared with the set threshold value to determine the traversal step length of the key point selection and store the points that meet the conditions as key points;

[0085] S405. Based on the above key points, the slope, horizontal distance, vertical distance and Euclidean distance of two adjacent points and the horizontal coordinate of the key points are compared with the threshold to determine the left boundary point set and the right boundary point set of the drivable area contour;

[0086] S405. The boundary point set is further optimized based on the slope and bottom boundary intersection features to obtain the optimal boundary point set. The optimized point set includes the top point and the bottom end point of the boundary. The top point and the bottom end point are converted back to the original image coordinates, and the intrinsic parameter matrix and the extrinsic parameter matrix based on the camera calibration are converted into coordinates in the world coordinate system, and the feasibility is calculated.

[0087] S5. Obtain relevant information of the vehicle;

[0088] Specifically, to obtain the lane line information of the lane where the vehicle is located, firstly, the lane line is extracted by improving SegFormer through semantic segmentation, and the lane line is fitted based on the consistency of random sampling. According to the camera intrinsic parameter matrix M1, extrinsic parameter matrix M2 and distortion matrix D, the lane line parameter equation in the world coordinate system is finally obtained.

[0089] To obtain information about other moving objects around the ego vehicle, we first improve SegFormer based on the semantic segmentation network to extract other moving objects such as vehicles and motorcycles around the ego vehicle, and further screen out the nearest moving obstacles in front and behind the ego vehicle lane or in the adjacent left and right lanes. According to the camera intrinsic parameter matrix M1, extrinsic parameter matrix M2 and distortion matrix D, the pixel coordinates of the obstacle's distance from the ego vehicle are converted into the actual distance in the world coordinate system.

[0090] S6. Determine the vehicle lane change signal and the vehicle lateral position.

[0091] Specifically, to determine the lateral position of the vehicle, first, according to the drivable area boundary obtained in step S4, the confidence of the left and right boundaries is compared with the set threshold. When the confidence is greater than the threshold, it means that the corresponding drivable area boundary is valid. The horizontal coordinate mean S of the corresponding boundary is calculated, and the lane width W of the road is calculated. d The estimated vehicle position is the floor (S / W) corresponding to the boundary of the drivable area. d )+ lanes, where floor is rounded down.

[0092] For determining the vehicle lane change signal, according to the drivable area boundary obtained in step S4, the confidence of the left and right boundaries are compared with the set threshold value respectively. When the confidence is greater than the threshold value, it means that the corresponding drivable area boundary is valid, and the horizontal coordinate mean S1 of the corresponding boundary is calculated. According to the lane line parameter equation provided in step S5, several points are sampled respectively to calculate the corresponding horizontal coordinate mean S2;

[0093] The calculation formula is: d=abs(S1-S2), where abs() is an absolute value operation.

[0094] Assume the lane width of the road where the vehicle is located is W d , taking the right boundary of the drivable area as an example, when the right boundary is valid and d>0.5*W d When d<0.5*W, it indicates that the lane where the vehicle is located is not the right edge lane. If it is further confirmed in step S5 that the distance between other moving obstacles around the vehicle and the vehicle is greater than the safe distance, it is considered that the current vehicle can change lanes to the right. Otherwise, when d<0.5*W d This indicates that the vehicle is currently in the lane at the right edge of the drivable area, and further confirmation is made through the navigation line as to whether a lane change in the opposite direction is required.

[0095] A device for assisting a vehicle in changing lanes and positioning its lateral position, comprising:

[0096] Memory for storing computer programs;

[0097] The processor is used to implement the steps of the method for assisting vehicle lane change and lateral position positioning when executing the computer program.

[0098] In summary, this patent uses a lightweight semantic segmentation network to achieve road drivable area segmentation, lane line extraction and road moving target detection, and can provide the vehicle with the lateral position information of the road where the vehicle is located and a method and device for signaling whether the lane can be changed.

[0099] Parts not involved in the present invention are the same as the prior art or can be implemented by the prior art. Although the embodiments of the present invention have been shown and described, it is understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for assisting a vehicle in changing lanes and locating a lateral position, characterized in that: The following steps are involved: S1. Obtain vehicle location information and speed from the navigation module; S2. Perform camera internal parameter calibration and camera external parameter calibration; S3. Preprocessing of image data obtained from the camera; S4. extracting the contour of the vehicle drivable area; S5. Obtain relevant information of the vehicle; S6. Determine the vehicle lane change signal and the vehicle lateral position.

2. A method for assisting a vehicle in lane change and lateral position positioning according to claim 1, characterized in that: In step S2, the camera internal calibration is performed using the kalibr calibration tool or opencv, and the focal length f in the x and y directions in the image coordinate system is obtained using the internal calibration. x ,f y ; The optical center coordinate is C x ,C y Thus, the camera internal parameter matrix M1 = [f x ,0,C x ; 0,f y ,C y ; 0,0,1], at the same time, the five parameters D = (k1, k2, P1, P2, k3) of the distortion coefficient distortion mathematical model are obtained through camera internal calibration; For camera extrinsic calibration, the camera position is selected as the origin on the ground projection point. According to the right-hand coordinate system, the horizontal right direction is the positive direction of the X axis, the vehicle driving direction is the positive direction of the Y axis, and the Z axis is vertical to the ground and upwards is the positive direction. The camera extrinsic parameter matrix M2 is calculated through the vanishing point. At the same time, the lateral distance S between the camera and the left wheel is calibrated. l The lateral distance S2 from the camera to the right wheel and the height H from the camera to the ground.

3. A method for assisting a vehicle in lane change and lateral position positioning according to claim 1, characterized in that: In step S4, for the modification of the Decoder stage of the SegFormer model, first, the features of different layers are passed through the linear layer respectively to ensure that the channel dimensions are the same, then they are all deconvolved to the original resolution and weighted averaged, and then fused with a linear layer, and convolution with a 1x1 convolution kernel is applied to reduce the convolution kernel parameters and simplify the model, and finally a linear layer predicts the results.

4. A method for assisting a vehicle in lane change and lateral position positioning according to claim 1, characterized in that: In step S4, the modification of the SegFormer model loss function is expressed as: L joint_seg =L ce +L energy L energy =S*E(i,j), in: in, is the normalized weight, S is the weight hyperparameter, D represents the spatial pixel position of image I, and G is essentially a Gaussian kernel filter.

5. A method for assisting a vehicle in lane change and lateral position positioning according to claim 4, characterized in that: In step S4, for the drivable area contour segmentation, the image is first preprocessed, the SegFormer model and weights are loaded, and according to the loss function and the Adam optimizer, the hyperparameters are adjusted during the training process to improve the model performance. Finally, the trained improved SegFormer model is applied to load part of the preprocessed image in step S3 for drivable area contour segmentation.

6. A method for assisting a vehicle in lane change and lateral position positioning according to claim 1, characterized in that: In step S4, the drivable area contour is post-processed, and the specific steps are as follows: S401. Perform polygon fitting on the contour points of the drivable area based on the corresponding opencv function approxPolyDP in the Douglas-Peucker algorithm to obtain a polygon point set; S402. Traverse the polygon point set and search for the first point that meets the conditions and re-store the polygon point set in a certain order; S403. When the number of polygon point sets is less than the threshold, all polygon point sets are used as key points of the drivable area contour; otherwise, key point selection is performed based on the re-stored polygon point sets; S404. For the selection of key points of the contour of the drivable area, three points are taken each time, namely the current point, the second point after the current point, and the third point, and the Euclidean distance, the lateral distance, and the longitudinal distance are calculated with the current point respectively, and the Euclidean distance, the lateral distance, and the longitudinal distance are compared with the set threshold value to determine the traversal step length of the key point selection and store the points that meet the conditions as key points; S405. Based on the above key points, the slope, horizontal distance, vertical distance and Euclidean distance of two adjacent points and the horizontal coordinate of the key points are compared with the threshold to determine the left boundary point set and the right boundary point set of the drivable area contour; S406. Based on the slope and bottom boundary intersection features, the boundary point set is further optimized to obtain the optimal boundary point set. The optimized point set includes the top point and the bottom end point of the boundary. The top point and the bottom end point are converted back to the original image coordinates, and the intrinsic parameter matrix and the extrinsic parameter matrix based on the camera calibration are converted into coordinates in the world coordinate system, and the feasibility is calculated.

7. The method for assisting a vehicle in lane change and lateral position positioning according to claim 1, characterized in that: In step S5, SegFormer, which is improved based on the semantic segmentation network, extracts other moving objects around the ego vehicle, and screens out the closest moving obstacles in front and behind the ego vehicle lane or in the adjacent left and right lanes. According to the camera intrinsic parameter matrix M1, extrinsic parameter matrix M2 and distortion matrix D, the pixel coordinates of the obstacle's distance from the ego vehicle are converted into the actual distance in the world coordinate system.

8. The method for assisting a vehicle in lane change and lateral position positioning according to claim 1, characterized in that: In step S6, to determine the lateral position of the vehicle, firstly, according to the drivable area boundary obtained in step S4, the confidence of the left and right boundaries is compared with the set threshold. When the confidence is greater than the threshold, the horizontal coordinate mean S of the corresponding boundary is calculated, and the lane width W of the road is calculated. d The estimated vehicle position is the floor (S / W) corresponding to the boundary of the drivable area. d )+ lanes, where floor is rounded down.

9. The method for assisting a vehicle in lane change and lateral position positioning according to claim 1, characterized in that: In step S6, for determining the vehicle changeable lane signal, the confidence of the left and right boundaries are compared with the set thresholds according to the drivable area boundary obtained in step S4. When the confidence is greater than the threshold, the mean value S1 of the horizontal coordinate of the corresponding boundary is calculated. According to the lane line parameter equation in step S5, several points are sampled to calculate the mean value S2 of the corresponding horizontal coordinate. The calculation formula is: d=abs(S1-S2), where abs() is an absolute value operation.

10. A device for assisting a vehicle in changing lanes and positioning its lateral position, characterized in that: include: Memory for storing computer programs; A processor is used to implement the steps of the method for assisting vehicle lane change and lateral position positioning as described in any one of claims 1 to 9 when executing the computer program.

Images