A vision-based vehicle following state recognition method

Abstract

The application discloses a kind of vehicle car following state identification methods based on vision, this method is installed in car front camera and gathers driving environment information, and reads the vehicle speed information from vehicle CAN bus. Adopt target detection algorithm to identify various vehicle targets in traffic environment, obtain the category information and position information of vehicle target, and develop ranging algorithm to calculate the relative distance information of vehicle and vehicle target. Identification module first determines whether the vehicle is in driving state according to the vehicle speed, then determines whether there is a front vehicle driving in the same direction as the vehicle according to the vehicle target information and lane line position information. Finally, according to the longitudinal relative distance and lateral relative distance between the vehicle and the front vehicle, it is determined whether the vehicle is in car following state. It can accurately identify the vehicle car following state, provide important technical support for the development and application of driving assistance system, and provide a new idea for establishing a car following scene database from natural driving data.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent driving technology of automobile, and particularly relates to a vehicle car-following state recognition method. BACKGROUND

[0002] With the continuous development of intelligent driving technology and the gradual maturity of active safety technology, the application of advanced driving assistance system (ADAS) on intelligent driving vehicles is becoming more and more widespread. Some driving assistance systems such as adaptive cruise control (ACC) and forward collision warning system (FCW) all play a role when the vehicle is in a car-following state. Therefore, it is necessary to accurately determine whether the vehicle is in a car-following state to realize the automatic switching of intelligent driving mode and ensure the function of driving assistance system.

[0003] With the development of intelligent cameras, it has become very easy and low-cost to collect traffic environment information in the driver's attention area at the vehicle end. The application of intelligent driving key technologies based on collected video information, such as traffic participant recognition and detection, distance measurement, etc., will greatly reduce the cost of intelligent driving technology development and application. However, these information has not been effectively utilized, and there is a lack of "new method" to solve the problem of intelligent driving using "new data". SUMMARY

[0004] In view of the defects or deficiencies of the prior art, the purpose of the present application is to provide a vehicle car-following state recognition method based on vision. This method only uses visual information and vehicle motion speed as input to analyze the traffic environment and relative motion state, and then determines whether the vehicle is in a car-following state, thereby providing technical support for the development and application of driving assistance systems and intelligent driving technology.

[0005] In order to achieve the above-mentioned task, the present application adopts the following technical solutions:

[0006] A car-following state recognition method based on vision, characterized in that the method installs a front camera at the vehicle end to collect driving environment information, adopts a car-following state recognition system based on vision composed of a data collection module, a target detection module, a distance measurement module and a car-following state recognition module to realize the detection of different vehicle targets and the measurement of relative distance, and then realize the recognition of vehicle car-following state. Specifically, it includes:

[0007] 1) The data collection module collects information including traffic environment information and vehicle motion state information. The traffic environment information in the driving process of the vehicle is collected by the front camera installed at the vehicle end. The driving speed of the vehicle is obtained from the vehicle CAN bus.

[0008] 2) The target detection module uses a target detection algorithm to identify vehicle targets in the traffic environment. The module outputs information including category information and position information of the vehicle target in the image, wherein the position information is represented by the coordinates of the top-left corner and the bottom-right corner of the target detection box;

[0009] 3) The ranging module develops a ranging algorithm based on the pinhole imaging principle, calibrates the camera to obtain camera intrinsic parameters, and uses the vehicle target information output by the target detection module and the lane line position information obtained by manual calibration to calculate the relative distance between the host vehicle and the vehicle target.

[0010] 4) The car-following state recognition module determines whether the host vehicle is in a car-following state based on the obtained host vehicle speed information, vehicle target detection information, lane line position information, and calculated relative distance information.

[0011] According to the present application, the target detection module distinguishes vehicle target types and driving directions of vehicle targets relative to the host vehicle by classifying vehicle targets. Specifically, vehicle targets are classified into large vehicles and small vehicles according to vehicle volume, wherein large vehicles mainly include heavy commercial vehicles such as trucks and vans, and small vehicles mainly include small cars. Vehicle targets are divided into side, front, and rear according to the vehicle orientation observed by the driver. When the side is recognized, it indicates that the vehicle target is perpendicular to the driving direction of the host vehicle. When the front is recognized, it indicates that the vehicle target is moving towards the host vehicle. When the rear is recognized, it indicates that the vehicle target is in the same driving direction as the host vehicle. Finally, vehicle targets are divided into six categories, i.e., large vehicle side, large vehicle front, large vehicle rear, small vehicle side, small vehicle front, and small vehicle rear, according to the above two classification methods. This method can initially exclude many vehicles that do not have a front-back interaction with the host vehicle.

[0012] Specifically, the longitudinal relative distance and the lateral relative distance between the host vehicle and the preceding vehicle calculated by the ranging module are used to constrain the car-following state.

[0013] Further, the lane line position information used by the ranging module and the car-following state recognition module is obtained by manual annotation in the video image.

[0014] On the other hand, the car-following state judgment step is:

[0015] S1: In the collected video, an image of the host vehicle driving in the middle of the lane is intercepted. Two lane lines of the host lane are drawn from the intersection of the lane line and the host vehicle front, and the lane line position information is obtained by manual calibration to determine the trapezoidal area of the future driving of the host vehicle.

[0016] S2: read the vehicle speed information v from the vehicle CAN bus, and identify the vehicle target in the real-time collected video data by using the target detection algorithm to obtain the category information and position information of the vehicle target;

[0017] S3: make a preliminary judgment according to the driving speed of the vehicle, if v = 0, it is considered that the vehicle is not in the car following state, otherwise step S4 is executed;

[0018] S4: make a further judgment according to the vehicle target category information output by the target detection module, if no large vehicle tail or small vehicle tail is identified, it is considered that the vehicle is not in the car following state, otherwise step S5 is executed;

[0019] S5: calculate the position coordinates A1(a1, b1), A2(a2, b2), … of the center points of the bottom edges of the vehicle tail target detection boxes according to the vehicle target position information output by the target detection module, and judge whether these positioning points are located in the trapezoidal area of the future driving of the vehicle, if no positioning point is located in the trapezoidal area, it is considered that the vehicle is not in the car following state, otherwise step S6 is executed.

[0020] S6: calculate the longitudinal relative distance and transverse relative distance between the vehicle and the vehicle tail target by using the ranging algorithm. If there are multiple positioning points of the vehicle tail target located in the trapezoidal area of the future driving of the vehicle, the closest vehicle tail target to the vehicle is selected according to the longitudinal relative distance, and then step S7 is executed; if only one positioning point of the vehicle tail target is located in the trapezoidal area of the future driving of the vehicle, step S7 is directly executed.

[0021] S7: judge whether the vehicle is in the car following state according to the longitudinal relative distance and transverse relative distance between the vehicle and the vehicle tail target output by the ranging module, if the longitudinal relative distance and transverse relative distance are both within the threshold range, i.e. X max and Y max , it is considered that the vehicle is in the car following state, otherwise it is considered that the vehicle is not in the car following state.

[0022] Further, in step S1, the two lane lines of the driving lane of the vehicle are represented by a first function expression in an image pixel coordinate system, and the trapezoidal area of the future driving of the vehicle is represented by the above two lane lines and two horizontal straight lines, and the specific method is as follows:

[0023] Determine four coordinate points P1(x1, y1), P2(x2, y2), P3(x3, y3), P4(x4, y4) of the two lane lines, wherein y2 = y3, y1 = y4.

[0024] The trapezoidal area of the future driving of the vehicle is represented as follows:

[0025]

[0026] Further, in step S5, whether the rear target positioning point A1(a1, b1) is located in the trapezoidal region of the future driving of the vehicle is judged, and the specific method is as follows:

[0027] The coordinate origin of the image pixel coordinate system is the top left corner of the image, therefore, if x1

[0028] Further, in step S6, the ranging algorithm is used to calculate the longitudinal relative distance and the transverse relative distance between the vehicle and the rear target, and the specific method is as follows:

[0029] First, the pixel lane width L corresponding to the positioning point A1(a1, b1) is calculated according to the longitudinal coordinate of the positioning point, that is, y=b1 is substituted into the equation of the two lane lines to obtain x=d1 and d2, and then the lane width is calculated as L=d2-d1.

[0030] According to the above information, the transverse distance s of the positioning point to the longitudinal center line of the future driving region of the vehicle is calculated x , and the calculation is as follows: x s=a1-(d1+d2) / 2.

[0031] The longitudinal relative distance and the transverse relative distance are calculated as follows:

[0032]

[0033] Wherein, S L is the longitudinal relative distance, S H is the transverse relative distance, f x is the pixel length of the focal length in the x-axis direction after the camera is calibrated, and L real is the real lane width.

[0034] The technical innovation of the vehicle following state recognition method based on vision of the present application is as follows:

[0035] Using only visual information and vehicle CAN bus data as input, this invention achieves sophisticated vehicle target classification and detection, as well as relative distance measurement. This effectively improves the intelligent driving system's understanding of the traffic environment, thereby accurately identifying vehicle following status and providing crucial technical support for the perception and decision-making of intelligent driving vehicles. Simultaneously, this method significantly reduces the development and application costs of driver assistance systems, which is of great significance for the popularization and promotion of intelligent driving systems. Furthermore, this invention provides a new approach to the automatic identification of following scenarios during the establishment of a following scenario database, offering a possibility for automatically identifying and extracting following scenarios from natural driving data lacking radar data. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of a vision-based vehicle following status recognition system.

[0037] Figure 2 The images are vehicle target recognition maps based on the target detection algorithm. In (a) the images are bus_front, car_rear and bus_rear from left to right, and in (b) the images are car_side, bus_side and car_rear from left to right.

[0038] Figure 3 This is a flowchart of the vision-based vehicle following state recognition method of the present invention.

[0039] Figure 4 This is a schematic diagram of the trapezoidal area in which the vehicle will travel in the future, and a schematic diagram of the ranging algorithm principle, as shown in the embodiment.

[0040] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. Detailed Implementation

[0041] Part One:

[0042] like Figure 1 As shown, this embodiment provides a vision-based car-following scene recognition method. A front-facing camera is installed on the vehicle to record the driving environment. A vision-based car-following scene recognition system, consisting of a data acquisition module, a target detection module, a ranging module, and a car-following state recognition module, is used to detect different vehicle targets and measure their relative distances, thereby recognizing the vehicle's car-following state. Specifically, it includes:

[0043] The data acquisition module collects information including traffic environment information and vehicle motion status information. It records traffic environment information during vehicle operation using a front-facing camera installed on the vehicle. It also obtains the vehicle's speed from the vehicle's CAN bus.

[0044] like Figure 2As shown, the target detection module uses the target detection algorithm YOLOv5 to identify vehicle targets in the traffic environment. The (a) figure shows bus_front, car_rear, and bus_rear from left to right, and the (b) figure shows car_side, bus_side, and car_rear from left to right. The module outputs the class information of the vehicle target and the position information in the image, where the position information is represented by the coordinates of the top-left corner and the bottom-right corner of the target detection box.

[0045] The ranging module uses the pinhole imaging principle to develop a ranging algorithm, calibrates the camera to obtain the camera intrinsic parameters, and uses the vehicle target information output by the target detection module and the lane line position information manually calibrated to calculate the relative distance between the ego vehicle and the vehicle target.

[0046] The car-following state recognition module determines whether the ego vehicle is in a driving state according to the ego vehicle speed information, filters out the same-direction vehicles that may have a front-back interaction relationship with the ego vehicle, i.e., the car tail, according to the vehicle target class information, determines whether there is a same-direction leading vehicle in the lane according to the vehicle target position information and the lane line position information, and finally determines whether the ego vehicle is in a car-following state according to the relative distance information.

[0047] In this embodiment, the target detection module uses a vehicle target classification method to distinguish the vehicle target type and the driving direction of the vehicle target relative to the ego vehicle. Specifically, the vehicle target is classified into large vehicles and small vehicles according to the vehicle volume, where the large vehicles mainly include trucks, vans, and other heavy commercial vehicles, and the small vehicles are mainly small cars. The vehicle target is further classified into side, front, and rear according to the vehicle orientation observed by the ego vehicle driver. When the side is recognized, it indicates that the vehicle target is perpendicular to the driving direction of the ego vehicle, when the front is recognized, it indicates that the vehicle target is moving towards the ego vehicle, and when the rear is recognized, it indicates that the vehicle target is moving in the same direction as the ego vehicle. Finally, the vehicle target is classified into 6 categories, i.e., large vehicle side, large vehicle front, large vehicle rear, small vehicle side, small vehicle front, and small vehicle rear, according to the above two classification methods. Using this classification method can initially exclude many other vehicles that do not have a front-back interaction relationship with the ego vehicle.

[0048] The longitudinal relative distance and the lateral relative distance between the ego vehicle and the leading vehicle calculated by the ranging module are used to constrain the car-following state.

[0049] The lane line position information used by the ranging module and the car-following state recognition module is obtained by manual calibration in the video image.

[0050] Second part:

[0051] As Figure 3As shown in the embodiment, the vision-based vehicle following state recognition method includes the following steps for determining the following state:

[0052] S1: Extract an image of the vehicle driving in the middle of the lane from the captured video. Draw two lane lines for this lane starting from the intersection of the lane line and the front of the vehicle. Manually mark the position information of the lane lines to determine the trapezoidal area where the vehicle will drive in the future.

[0053] S2: Read vehicle speed information v from the vehicle CAN bus, use a target detection algorithm to identify vehicle targets in the real-time video data, and obtain the category and location information of the vehicle targets;

[0054] S3: Make a preliminary judgment based on the vehicle's speed. If v = 0, it is considered that the vehicle is not in a following state; otherwise, proceed to step S4.

[0055] S4: Further judgment is made based on the vehicle target category information output by the target detection module. If the rear of the vehicle is not detected, it is assumed that the vehicle is not in a following state; otherwise, step S5 is executed.

[0056] S5: Calculate the position coordinates A1(a1,b1), A2(a2,b2), ... of the center point of the bottom edge of the target detection box at the rear of the vehicle based on the vehicle target position information output by the target detection module, and determine whether these positioning points are located in the trapezoidal area where the vehicle will drive in the future. If no positioning point is located in the area, it is considered that the vehicle is not in a following state; otherwise, proceed to step S6.

[0057] S6: Calculate the longitudinal and lateral relative distances between the vehicle and the target behind it using the developed ranging algorithm. If multiple target locations are within the trapezoidal area where the vehicle will travel in the future, filter for the closest target based on the longitudinal relative distance, and then proceed to step S7. If only one target location is within the trapezoidal area where the vehicle will travel in the future, proceed directly to step S7.

[0058] S7: Determine whether the vehicle is in a following state based on the calculated longitudinal and lateral relative distances between the vehicle and the target behind it. If both the longitudinal and lateral relative distances are within the threshold range, i.e., X... max and Y max If the following conditions are met, the vehicle is considered to be in a following state; otherwise, it is considered not to be in a following state.

[0059] like Figure 4 As shown, in step S1, the two lane lines of the vehicle's driving lane are represented by a linear function expression in the image pixel coordinate system, and the trapezoidal region where the vehicle will drive in the future is represented by the above two lane lines and two horizontal straight lines. The specific method is as follows:

[0060] Four coordinate points of two lane lines are determined, P1(x1, y1), P2(x2, y2), P3(x3, y3), P4(x4, y4), wherein y2=y3, y1=y4.

[0061] The trapezoidal area of the future driving of the vehicle is represented as follows:

[0062]

[0063] As shown in step S5, the judgment of whether the rear target positioning point A1(a1, b1) is located in the trapezoidal area of the future driving of the vehicle is performed, and the specific method is as follows: Figure 4

[0064] The coordinate origin of the image pixel coordinate system is the top left corner of the image, therefore, if x1

[0065] As shown in step S6, the developed ranging algorithm is used to calculate the longitudinal relative distance and the transverse relative distance between the vehicle and the rear target, and the specific method is as follows: Figure 4

[0066] Firstly, the lane width corresponding to the rear target positioning point A1(a1, b1) is positioned and calculated according to the longitudinal coordinate of the positioning point, i.e. y=b1 is substituted into the equation of the two lane lines to obtain x=d1 and d2, and then the lane width is calculated as L=d2-d1.

[0067] The transverse distance s of the positioning point to the longitudinal center line of the future driving area of the vehicle is calculated according to the above information x , and the calculation is as follows: x s=a1-(d1+d2) / 2

[0068] The longitudinal relative distance and the transverse relative distance are calculated as follows:

[0069]

[0070] wherein S L is the longitudinal relative distance, S H is the transverse relative distance, f x is the pixel length of the focal length in the x-axis direction after the camera is calibrated, and L real is the real lane width.

[0071] Part III: Specific experimental examples

[0072] As shown in step S6, the developed ranging algorithm is used to calculate the longitudinal relative distance and the transverse relative distance between the vehicle and the rear target, and the specific method is as follows: Figure 4 ​​As shown, taking the current traffic environment as an example, a camera with a sensor size of 1 / 3" is used to identify whether the vehicle is in a following state.

[0073] In the collected video, the image of the vehicle driving in the middle of the lane is intercepted, and two lane lines of the vehicle lane are drawn from the intersection of the lane line and the vehicle head. The position information of the lane line is obtained by manual calibration to determine the trapezoidal area of the future driving of the vehicle:

[0074]

[0075] The data acquisition module reads the vehicle speed information from the CAN bus as 25km / h, and the target detection algorithm YOLOv5 identifies the small car tail car_rear in the current traffic environment, with a confidence of 1.

[0076] The preliminary judgment is that the vehicle speed is not 0, and the small car tail is identified. According to the vehicle target position information, the positioning point A1(960, 738) is calculated, i.e. a1=960, b1=738.

[0077] Further judgment, 180<a1<1550, then x=960, substitute the function expression of the two lane lines to get c1=599.32, c2=631.25, which satisfies b1>c1, b1>c2, and simultaneously satisfies y2<b1<y1, so the positioning point A1 is located in the trapezoidal area of the future driving of the vehicle.

[0078] According to the longitudinal coordinate of the tail target positioning point A1(960, 738), the pixel lane width L corresponding to the positioning point is calculated, and y=738 is substituted into the equation of the two lane lines to get d1=690.02, d2=1130.8, and the lane width is calculated as: L=440.78.

[0079] The transverse distance s of the positioning point to the longitudinal center line of the future driving area of the vehicle is calculated x =a1-(d1+d2) / 2=49.59.

[0080] The actual lane width is 3.5m, and the pixel length of the camera focal length in the x-axis direction is 1600. The longitudinal relative distance and the transverse relative distance are calculated as: S L =12.70m, S H =0.39m.

[0081] Preferably, X max =1.2m, Y max =120m. The longitudinal relative distance and the transverse relative distance are within the threshold range, and it is judged that the vehicle is in a following state.

[0082] The above examples are only the preferred examples of the present application, the present application is not limited to the above examples, any modification made on the basis of the technical solutions of the present application falls within the protection scope of the patent claims of the present application.

Claims

1. A vision-based car-following state recognition method, characterized in that, The method records the driving environment before the front camera is installed at the vehicle end, adopts a visual-based following state recognition system composed of a data acquisition module, a target detection module, a distance measurement module and a following state recognition module, realizes detection of different vehicle targets and measurement of relative distance, and further realizes recognition of the following state of the vehicle, specifically including: 1) The information collected by the data acquisition module is traffic environment information and self-vehicle motion state information, the traffic environment information in the driving process of the vehicle is collected by the front camera installed at the vehicle end, and the driving speed of the vehicle is obtained from the vehicle CAN bus; 2) The target detection module adopts a target detection algorithm to recognize vehicle targets in the traffic environment, the output information of the target detection module is category information and position information of the vehicle target in the image, wherein the position information is represented by the coordinates of the top left corner and the bottom right corner of the target detection frame; 3) The distance measurement module develops a distance measurement algorithm based on the pinhole imaging principle, calibrates the camera to obtain the camera intrinsic parameters, and calculates the relative distance between the vehicle and the vehicle target using the vehicle target position information output by the target detection module and the lane line position information artificially calibrated; 4) The following state recognition module is used to determine whether the vehicle is in the following state according to the obtained vehicle speed information, vehicle target detection information, lane line position information and calculated relative distance information; The judgment steps of the following state are: S1: In the collected video, the image of the vehicle driving in the middle of the lane is intercepted, two lane lines of the lane are drawn from the intersection of the lane line and the vehicle head, the position information of the lane line is artificially calibrated, and the trapezoidal area of the future driving of the vehicle is determined; S2: reading vehicle speed information from a vehicle CAN bus v a target detection algorithm is used to identify vehicle targets in the real-time collected video data, and category information and position information of the vehicle targets are obtained; S3: According to the driving speed of the host vehicle, make a preliminary judgment, if v= 0, it is considered that the host vehicle is not in the car-following state, otherwise step S4 is executed; S4: According to the vehicle target category information output by the target detection module, further judgment is made, if the tail of a large vehicle or the tail of a small vehicle is not recognized, it is considered that the vehicle is not in the following state, otherwise step S5 is executed; S5: Calculate the position coordinates A1 of the center point of the bottom edge of the rear target detection box based on the vehicle target position information output by the target detection module. a 1, b 1), A2 a 2, b 2), ..., and determine whether these positioning points are located in the trapezoidal area where the vehicle will drive in the future. If no positioning point is located in the area, it is considered that the vehicle is not in a following state; otherwise, proceed to step S6. S6: The distance measurement algorithm is used to calculate the longitudinal relative distance and the transverse relative distance between the vehicle and the tail target; if there are multiple tail target positioning points located in the trapezoidal area of the future driving of the vehicle, the tail target closest to the vehicle is selected according to the longitudinal relative distance, and then step S7 is executed; if there is only one tail target positioning point located in the trapezoidal area of the future driving of the vehicle, step S7 is directly executed; S7: According to the longitudinal relative distance and the lateral relative distance between the vehicle and the tail target output by the ranging module, it is judged whether the vehicle is in the car-following state. If the longitudinal relative distance and the lateral relative distance are both within the threshold range, that is, X max and Y am the vehicle is considered to be in the car-following state, otherwise the vehicle is considered not to be in the car-following state.

2. The method of claim 1, wherein, The target detection module distinguishes the vehicle target type and the driving direction of the vehicle target relative to the vehicle by classifying the vehicle target, specifically: According to the volume of the vehicle, the vehicle target is classified into large vehicles and small vehicles, wherein the large vehicles are heavy commercial vehicles, and the small vehicles are small cars; According to the vehicle position observed by the driver of the vehicle, the vehicle target is divided into side, front and rear; when the side is recognized, it indicates that the vehicle target is perpendicular to the driving direction of the vehicle, when the front is recognized, it indicates that the vehicle target is moving towards the vehicle, and when the rear is recognized, it indicates that the vehicle target is in the same driving direction as the vehicle; finally, the vehicle target is divided into 6 categories, i.e. large vehicle side, large vehicle front, large vehicle rear, small vehicle side, small vehicle front and small vehicle rear; using this classification method, many other vehicles which do not have a front-back relationship with the vehicle are preliminarily excluded.

3. The method of claim 1, wherein, The longitudinal relative distance and the lateral relative distance between the host vehicle and the front vehicle calculated by the distance measurement module are used to constrain the car following state.

4. The method of claim 1, wherein, The lane line position information used by the distance measurement module and the car following state recognition module is obtained by manual annotation in a video image.

5. The method of claim 1, wherein, In step S1, the two lane lines of the host vehicle's driving lane are represented by a first function expression in an image pixel coordinate system, and the trapezoidal area in which the host vehicle will travel in the future is represented by the above two lane lines and two horizontal straight lines, and the specific method is as follows: determining four coordinate points of two lane lines, P1 x 1, y 1), P2 x 2, y 2), P3 x 3, y 3), P4 x 4, y 4), wherein y 2= y 3, y 1= y 4; The trapezoidal area in which the host vehicle will travel in the future is represented as follows: 。 6. The method of claim 5, wherein, In step S5, it is judged whether the rear positioning point A1 a 1, b 1) is located in the trapezoidal region in which the host vehicle travels, specifically by The coordinate origin of the image pixel coordinate system is the top left corner of the image, thus, if x 1 <a 1 <x 4, let x = a 1, substituting the function expression of the two lane lines, we get y = c 1 and c 2, if b 1 c 1 b 1 c 2, and y 2 <b 1 y 1, then the positioning point A1 is located in the future driving area of the vehicle.

7. The method of claim 5, wherein, In step S6, the longitudinal relative distance and the lateral relative distance between the host vehicle and the rear target are calculated by using a distance measurement algorithm, and the specific method is as follows: First, according to the rear target positioning point A1 a 1, b 1) longitudinal coordinate positioning and calculating the pixel lane width corresponding to the positioning point L , that is, substituting y = b 1 into the equation of the two lane lines x = d 1 and d 2, the lane width is calculated as: L = d 2- d 1; The lateral distance of the positioning point to the longitudinal center line of the future travel area of the vehicle is calculated according to the above information s x , calculated as: s x = a 1-( d 1 +d 2) / 2; The longitudinal relative distance and the lateral relative distance are calculated as follows: wherein S L is a longitudinal relative distance, S H is a transverse relative distance, f x is a focal length in the x axis direction after calibrating the camera, L real is a real lane width.

Images