Road marking recognition device

Abstract

A road marking recognition device includes: an outside detecting unit mounted on a vehicle and detecting an outside situation of a periphery of the vehicle; a behavior detecting unit detecting a driving behavior of the vehicle; a road marking recognizing unit setting a coordinate system corresponding to a current position and a traveling direction of the vehicle based on the outside situation and recognizing a road marking of the periphery of the vehicle in the coordinate system; a movement amount calculating unit calculating a movement amount of the vehicle from a first time to a second time based on the driving behavior; and a road marking correcting unit correcting the recognized road marking. The road marking recognizing unit generates first road marking information indicating the recognized road marking at the first time and second road marking information indicating the recognized road marking at the second time, and the road marking correcting unit has: a predicting unit predicting the recognized road marking at the second time based on the first road marking information and the movement amount of the vehicle; and an updating unit updating the second road marking information based on a difference between the predicted road marking and the actually recognized road marking at the second time.

Description

Technical Field

[0001] This invention relates to a road marking recognition device for identifying road markings in the driving lanes of vehicles with autonomous driving or driver assistance functions. Background Technology

[0002] As such a device, there are known devices for recognizing road markings based on camera images (for example, see Patent Document 1). In the device described in Patent Document 1, road marking information in each frame is obtained in the form of a polynomial based on the detection results of the road markings by the camera. When the difference between the road marking parameters (coefficients of the polynomial) between adjacent frames is less than a predetermined value, the parameters of the current frame are subjected to Kalman filtering and smoothed.

[0003] The widespread adoption of vehicles with autonomous driving and driver assistance functions can improve the overall safety and convenience of the transportation system, enabling a sustainable transportation system. Furthermore, by improving the efficiency and smoothness of transportation, CO2 emissions can be reduced, lessening the environmental impact.

[0004] However, as described in Patent Document 1, Kalman filtering is only performed when the difference between the recognition results of adjacent frames is small, which makes it difficult to improve the recognition accuracy of road markings.

[0005] Existing technical documents

[0006] Patent documents

[0007] Patent Document 1: Japanese Patent Publication No. 2021-508901 (JP2021-508901A). Summary of the Invention

[0008] The road marking recognition device of this invention comprises: an external detection unit mounted on a vehicle to detect the external conditions surrounding the vehicle; a behavior detection unit to detect the vehicle's driving behavior; a road marking recognition unit that, based on the external conditions detected by the external detection unit, sets a coordinate system corresponding to the vehicle's current position and direction of travel, and recognizes road markings around the vehicle in the coordinate system; a movement calculation unit that, based on the driving behavior detected by the behavior detection unit, calculates the vehicle's movement from a first time moment to a second time moment; and a road marking correction unit that corrects the road markings recognized by the road marking recognition unit. The road marking recognition unit generates first road marking information representing the road markings recognized at the first time moment and second road marking information representing the road markings recognized at the second time moment. The road marking correction unit includes: a prediction unit that predicts the road markings identified by the road marking recognition unit at a second time based on first road marking information generated by the road marking recognition unit and the vehicle movement calculated by the movement calculation unit; and an update unit that updates the second road marking information generated by the road marking recognition unit based on the difference between the road markings predicted by the prediction unit and the road markings actually identified by the road marking recognition unit at the second time. Attached Figure Description

[0009] The objectives, features, and advantages of the present invention are further illustrated by the following description of embodiments in conjunction with the accompanying drawings.

[0010] Figure 1 This is a graph illustrating an example of the frequency distribution of road marking recognition errors.

[0011] Figure 2 This is a block diagram that schematically illustrates an example of the main components and processing flow of a road marking recognition device according to an embodiment of the present invention.

[0012] Figure 3 It is used for explanation Figure 2 The diagram shows the road marking recognition function.

[0013] Figure 4 It is used for explanation Figure 2 The diagram shows the correction of road markings by the road marking correction department. Detailed Implementation

[0014] The following is for reference Figures 1-4The embodiments of the present invention will be described. The road marking recognition device of the present invention is applied to a vehicle having a driving assistance function that controls driving actuators to assist the driver or enable automatic driving of the vehicle, and identifies road markings in the designated driving lanes. In this embodiment, "driving assistance" includes driving assistance that assists the driver's driving operations and automatic driving that enables the vehicle to drive automatically without the driver's driving operations, equivalent to Levels 1 to 4 of automatic driving as defined by SAE (Society of Automotive Engineers), with "automatic driving" equivalent to Level 5 automatic driving.

[0015] In driver assistance or autonomous driving systems, the vehicle's speed, direction of travel, and surrounding external conditions are detected at predetermined intervals. Based on the detection results, a target driving path is generated, and the vehicle is controlled to travel along this path. External sensors, such as cameras or LiDAR (Light Detection and Ranging) sensors, are mounted on the vehicle at predetermined positions and angles (or orientations) during manufacturing. For external targets, including road markings, their positions are estimated and identified based on the sensor's orientation information relative to the vehicle, but the identification results (position estimation results) based on these sensor values ​​for road markings, etc., contain a certain degree of error.

[0016] Figure 1 This is a graph illustrating an example of the frequency distribution of road marking recognition errors, showing the recognition error in the vehicle width direction. Road marking recognition results based on sensor values ​​include certain errors caused by factors such as the offset between the posture information from external sensors and the actual posture. For example... Figure 1 As shown, errors can follow a normal distribution or become non-normally distributed. If the variance of such errors is large, it is difficult to identify road markings with high accuracy. Therefore, in this embodiment, in order to improve the recognition accuracy of road markings by applying Kalman filtering to the recognition results of road markings based on sensor values, a road marking recognition device is configured as follows.

[0017] Figure 2 This is a block diagram that schematically illustrates an example of the main components and processing flow of a road marking recognition device (hereinafter referred to as the device) 100 according to an embodiment of the present invention. Figure 2 As shown, the device 100 is mainly composed of an electronic control unit (ECU) 10. The ECU 10 is configured as a computer having an arithmetic unit 11 such as a CPU (central processing unit), a storage unit 12 such as RAM (random access memory) and ROM (read-only memory), an I / O (input / output) interface, and other peripheral circuits. For example, the ECU 10 is configured as part of a group of multiple ECUs mounted on the vehicle 1 and controlling the operation of the vehicle 1. Figure 2 The process, for example, begins when vehicle 1 is started or activated and ECU 10 is activated, and is repeated at a predetermined cycle.

[0018] ECU10 is connected to external sensor 2, which is mounted on vehicle 1 and detects the external conditions in front of vehicle 1, behavior sensor 3, which detects the driving behavior of vehicle 1, and driving actuator 4.

[0019] External sensor 2 detects the external conditions in front of the vehicle 1, centered on the vehicle 1's direction of travel. External sensor 2 may be composed of, for example, a camera with an imaging element such as a CCD (charge-coupled device) or CMOS (complementary metal-oxide-semiconductor) sensor that captures images of the area in front of the vehicle. External sensor 2 may also be composed of a LiDAR (Light Detection and Ranging) sensor, which illuminates an object and determines the distance and direction of the object based on the time it takes for the light to return from the object, and also detects the reflected brightness at each measurement point.

[0020] Based on the external conditions detected by external sensor 2, the system can estimate the vehicle 1's own position (current position) according to the SLAM (Simultaneous Localization and Mapping) algorithm, and create a map that includes the road lines around the vehicle 1.

[0021] The behavior sensor 3 detects the driving behavior of vehicle 1, such as its speed and direction of travel. The behavior sensor 3 consists of an inertial measurement unit (IMU), which detects the angular velocity of the vehicle 1's center of gravity around three axes: the vertical direction, the direction of travel, and the width direction, as well as the acceleration in those three axes. The behavior sensor 3 may also include wheel speed sensors that detect the rotational speed of each wheel of vehicle 1, and a positioning unit that determines the absolute position (latitude and longitude) of vehicle 1 based on positioning signals from positioning satellites.

[0022] The driving actuator 4 includes a steering mechanism such as a steering gear that steers the vehicle 1, a driving mechanism such as an engine or electric motor that drives the vehicle 1, and a braking mechanism such as a brake that brakes the vehicle 1.

[0023] The ECU 10 has a functional structure comprising a road marking recognition unit 13, a movement calculation unit 14, a road marking correction unit 15, a prediction unit 16, an update unit 17, a driving control unit 18, and a map information update unit 19, which together form the computing unit. Specifically, the computing unit 11 of the ECU 10 functions as the road marking recognition unit 13, the movement calculation unit 14, the road marking correction unit 15, the prediction unit 16, the update unit 17, the driving control unit 18, and the map information update unit 19. The storage unit 12 stores map information and other data used for autonomous driving.

[0024] Figure 3 This diagram illustrates the recognition of road markings by the road marking recognition unit 13, showing the road marking L recognized by the road marking recognition unit 13 at time t. t One example. It should be noted that, for convenience, only the road marking L on the left side of this lane is shown in each diagram. t .

[0025] like Figure 3 As shown, the road marking recognition unit 13 identifies the road markings L that define the driving lane for the vehicle 1 based on the external conditions detected by the external sensor 2. t More specifically, the origin O is set to the current position of vehicle 1. t Using a coordinate system with the vehicle's direction of travel as the X-axis and the vehicle width as the Y-axis, estimate the road marking L in the established coordinate system. t The location coordinates.

[0026] Furthermore, the road marking recognition unit 13 uses curve fitting methods such as the least squares method to generate a curve fitting algorithm for the recognized road markings L. t The cubic function F is approximated by the following equation (i). t The information. More specifically, the road marking recognition unit 13 determines the function F. t Parameter C 3t C 2t C 1t C 0t And generate the determined parameter C. 3t C 2t C 1t C 0t Information as function F t Information.

[0027] Y = C 3t X 3 +C 2t X 2 +C 1t X+C 0t (i)

[0028] The shape of a typical curved road is designed using a spiral curve with curvature varying proportionally. Certain sections of the spiral curve corresponding to the road shape can be designed using functions such as cubic functions F. t To approximate the road shape, a cubic function is used. Compared to using a spiral curve to approximate the road shape, the computational load can be reduced. The recognition of road marking Lt and the generation of information about the function Ft by the road marking recognition unit 13 are performed, for example, in each control cycle of the ECU 10.

[0029] The movement calculation unit 14 calculates the movement of vehicle 1 from a first time t1 to a second time t2 based on the driving behavior detected by the behavior sensor 3. The first time t1 is, for example, the previous control cycle of ECU 10, and the second time t2 is, for example, the current control cycle. More specifically, by integrating the detection values ​​from the behavior sensor 3, the movement of vehicle 1 from the first time t1 to the second time t2 is calculated as follows: ΔX in the direction of travel, ΔY in the direction of vehicle width, and Δθ in the direction of rotation. In other words, the movement from the first time t1 to the second time t2 is calculated... Figure 2 The translational displacement (ΔX, ΔY) and rotational displacement Δθ of the coordinate system.

[0030] The road marking correction unit 15 applies Kalman filtering to the recognition result of the road marking Lt by the road marking recognition unit 13, thereby correcting the road marking Lt recognized by the road marking recognition unit 13. t For example, nonlinear Kalman filters such as the Extended Kalman Filter (EKF), which is applicable to large and nonlinear state-space models, the Unscented Kalman Filter (UKF), which estimates the covariance matrix in the form of a non-normal distribution, and the Particle Filter (PF) can be used. The road marking correction unit 15 has a prediction unit 16 responsible for the prediction step of the Kalman filter and an update unit 17 responsible for the update step of the Kalman filter.

[0031] Figure 4 This is used to explain the road marking correction unit 15 for road marking L. t The correction diagram shows the road marking L actually identified by the road marking recognition unit 13 at the first time t1, with solid lines. t1 and the road markings L identified at the second time t2 t2 One example. Additionally, an example of a road marking L(^) corrected by the road marking correction unit 15 is shown with dashed lines.

[0032] like Figure 4 As shown, the prediction unit 16 of the road marking correction unit 15, based on the translational movement (ΔX, ΔY) and rotational movement Δθ calculated by the movement calculation unit 14, calculates the first function F of the following formula (ii) determined by the road marking recognition unit 13 at the first time t1. t1 Perform parallel and rotational translations. First function F t1 This is equivalent to the road marking L actually identified by the road marking recognition unit 13 at the first moment t1. t1 An approximation of .

[0033] Y = C 3t1 X3 +C 2t1 X 2 +C 1t1 X+C 0t1 (ii)

[0034] The first function F, which is shifted parallel to ΔX in the X-axis direction and parallel to ΔY in the Y-axis direction by the prediction unit 16, is... t1 The first function F after parallel translation is represented by equation (iii). t1 The origin is the same as the current position of vehicle 1 at the second time t2 (the new origin O). t2 Here, the time interval from the first time t1 to the second time t2 is short enough (e.g., equivalent to a few milliseconds of the control cycle of ECU10), and the movement of vehicle 1 during this period, ΔX and ΔY, is small enough to be approximated as (ΔX). 2 =0, (△Y) 2 =0.

[0035] (Y+△Y)=C 3t1 (X+△X) 3 +C 2t1 (X+△X) 2

[0036] +C 1t1 (X+△X)+C 0t1

[0037] Y = C 3t1 X 3 +(3C 3t1 △X+C 2t1 )X 2

[0038] +(2C 2t1 △X+C 1t1 )X+C 1t1 △X-△Y+C 0t1 (iii)

[0039] Furthermore, the prediction department 16 uses a new origin O. t2 The first function F, rotated and shifted by Δθ around the center. t1 It is represented by the following equation (iv). Here, since the rotational movement Δθ of vehicle 1 from the first time t1 to the second time t2 is small enough, it can be approximated as cosΔθ=1, sinΔθ=Δθ, YΔθ=0.

[0040] (-Xsin△θ+Ycos△θ)=C 3t1 (Xcos△θ+Ysin△θ) 3

[0041] +(3C 3t1△X+C 2t1 (Xcos△θ+Ysin△θ) 2

[0042] +(2C 2t1 △X+C 1t1 (Xcos△θ+Ysin△θ)

[0043] +C 1t1 △X-△Y+C 0t1

[0044] Y = C 3t1 X 3 +(3C 3t1 △X+C 2t1 )X 2 +(2C 2t1 △X+△θ+C 1t1 )X

[0045] +C 1t1 △X-△Y+C 0t1 (iv)

[0046] The prediction unit 16 performs parallel and rotational translations and transforms the first function F into the form of equation (iv). t1 The road marking L is calculated based on sensor values ​​up to the first time t1 and identified by the road marking recognition unit 13 at the second time t2. t2 The pre-determined value.

[0047] The updating unit 17 of the road marking correction unit 15 updates the road marking L actually identified by the road marking recognition unit 13 at the second time t2 based on the pre-estimated value of the prediction unit 16. t2 The difference between (observed values) updates the second function F of the following equation (v) determined by the road marking recognition unit 13. t2 The second function F t2 This is equivalent to the road marking L actually identified by the road marking recognition unit 13 at the second time t2. t2 An approximation of .

[0048] Y = C 3t2 X 3 +C 2t2 X 2 +C 1t2 X+C 0t2 (v)

[0049] Thus, F, the first function after parallel and rotational shifts of the pre-estimated value in Kalman filtering, t1 Compared with the actual road markings L identified as observation values t2 The second function F t2The order is the same. Therefore, the difference between the pre-estimated value and the observed value can also be expressed as a function of the same order. It should be noted that in Kalman filtering, x(t) = [C 3t C 2t C 1t C 0t The predetermined values ​​are represented by the matrix of equation (vi). When using an extended Kalman filter, the errors of ΔX, ΔY, and Δθ can also be considered.

[0050]

[0051] Update unit 17 calculates the ex-post estimate F in Kalman filtering based on the difference between the pre-estimated value and the observed value and the Kalman gain. t2 (^), and the second function F determined by the road marking recognition unit 13 t2 Updated to the calculated ex post facto estimate F t2 (^). Post-hoc estimated value F t2 (^) is equivalent to the road marking L(^) corrected by the road marking correction unit 15. Therefore, the road marking recognition unit 13 corrects the road marking L. t The recognition results.

[0052] The driving control unit 18 controls the driving actuator 4 based on the road markings L(^) corrected by the road marking correction unit 15. For example, it generates a target driving path for the vehicle 1 by passing through the center of the left and right road markings L(^), and controls the driving actuator 4 to drive the vehicle 1 along the generated target driving path. In this way, the road markings L are corrected sequentially by online Kalman filtering. t The recognition results enable real-time improvement of road markings used in driver assistance and autonomous driving. t The accuracy of recognition.

[0053] Map information update unit 19 updates the road lines L(^) included in the autonomous driving map information stored in storage unit 12 based on the road lines L(^) corrected by road line correction unit 15. t Location information. In this case, by utilizing offline Kalman smoothing, road markings L within a specified period are retrospectively corrected. t The recognition results can further improve the road markings used in map information for autonomous driving. t The recognition accuracy. Such processing, for example, when obtaining... Figure 1 The frequency distribution of road marking recognition errors shown is then implemented.

[0054] The map information updating unit 19 updates the map information under the condition that the computational load of the driving control unit 18 is below a predetermined value, such as when the vehicle 1 has just started (started), when the engine is stopped, when the drive motor is stopped, when the vehicle is parked, or when manual control is not engaged in automatic driving. When the map information updating unit 19 updates the map information, it estimates the vehicle 1's own position (current position) based on the updated map information and the external conditions detected by the external sensor 2. In this case, the road markings L used in the map information are used to determine the position. t The improved recognition accuracy also improves the accuracy of estimating the position of vehicle 1.

[0055] The following effects can be achieved by adopting this implementation method.

[0056] (1) The device 100 includes: an external sensor 2 mounted on the vehicle 1 to detect the external conditions around the vehicle 1; a behavior sensor 3 to detect the driving behavior of the vehicle 1; and a road marking recognition unit 13, which sets a coordinate system corresponding to the current position and direction of travel of the vehicle 1 based on the external conditions detected by the external sensor 2, and recognizes the road markings L around the vehicle 1 in the coordinate system. t The movement calculation unit 14 calculates the translational movement of vehicle 1 from the first time t1 to the second time t2, ΔX, ΔY, and Δθ; and the road marking correction unit 15 corrects the road markings L identified by the road marking recognition unit 13. t Perform correction ( Figure 2 ).

[0057] The road marking recognition unit 13 generates a representation of the road marking L recognized at the first time t1. t1 The first function F t1 And represents the road marking L identified at the second time t2. t2 The second function F t2 The road marking correction unit 15 includes a prediction unit 16, which, based on a first function F generated by the road marking recognition unit 13, provides information. t1 Based on the information and the vehicle movement amounts △X, △Y, and △θ calculated by the movement amount calculation unit 14, the road marking L identified by the road marking recognition unit 13 at the second time t2 is predicted. t2 (Pre-determined value); and the updating unit 17, which updates the road marking L actually identified by the road marking recognition unit 13 at the second time t2 based on the pre-determined value of the prediction unit 16. t2 The difference updates the second function F generated by the road marking recognition unit 13. t2 Information ( Figure 2 ).

[0058] In this way, by utilizing the information on the movement of vehicle 1, it is possible to convert the road marking L identified at the first time t1 into a representation.t1 The first function F t1 Using the coordinate system, predict the road marking L identified at the second time t2. t2 (Kalman filter prediction steps). Additionally, by updating the road markings L actually identified at time t2... t2 The corresponding second function F t2 (The update steps of Kalman filtering) can improve the accuracy of road markings L. t The accuracy of recognition (position estimation accuracy).

[0059] (2) The road marking recognition unit 13 generates the road marking L recognized at the first time t1. t1 The first function F to be approximated t1 The information is used to generate the road markings L identified at the second time t2. t2 The second function F is approximated t2 Information such as cubic functions that can be used to approximate the shape of a general road using spiral curves.

[0060] (3) The movement calculation unit 14 calculates the translational movement (ΔX, ΔY) and rotational movement (Δθ) of vehicle 1 from the first time t1 to the second time t2. The prediction unit 16, based on the translational movement (ΔX, ΔY) calculated by the movement calculation unit 14, adjusts the first function F determined by the road marking recognition unit 13. t1 The system moves parallel to the target area and rotates it based on the rotational movement amount Δθ, thereby predicting the road marking L identified by the road marking recognition unit 13 at the second time t2. t2 That is, by utilizing the information of the movement of vehicle 1, ΔX, ΔY, and Δθ, the first function F is made... t1 Parallel translation and rotational translation can transform the first function F t1 Transformed into the second function F t2 Same coordinate system.

[0061] (4) The first function F after parallel and rotational movement of the prediction unit 16 t1 The number of times and the first function F determined by the road marking recognition unit 13 t1 Second function F t2 The order of the values ​​is the same. Therefore, the difference between the pre-estimated value and the observed value in Kalman filtering can also be expressed as a function of the same order.

[0062] (5) The device 100 also includes a driving control unit 18, which controls the driving actuator 4 based on the road markings L(^) corrected by the road marking correction unit 15. Figure 2 In this case, the road markings L are corrected sequentially by online Kalman filtering. tThe recognition results can improve the road markings used in driver assistance and autonomous driving in real time. t The accuracy of recognition.

[0063] (6) The device 100 further includes: a storage unit 12 that stores map information; and a map information updating unit 19 that updates the map information stored in the storage unit 12 based on the road lines L(^) corrected by the road line correction unit 15. In this case, by utilizing offline Kalman smoothing, the road lines L within a specified period are corrected retrospectively. t The recognition results can further improve the road markings used in map information for autonomous driving. t This improves the recognition accuracy and further enhances the accuracy of estimating the position of vehicle 1.

[0064] In the above embodiments, utilizing Figure 3 The road markings L, etc., indicate the identification of this lane. t Examples include those for identifying road markings around a vehicle, but the road marking recognition unit is not limited to this. For example, it can also identify road markings around a vehicle, such as those in adjacent lanes, either in place of the lane markings or together with the lane markings.

[0065] In the above embodiments, utilizing Figure 3 The description explains how to determine the road marking L identified by the road marking recognition unit 13. t Approximate the cubic function F t Examples are given, but the first and second functions used to approximate road markings are not limited to these. For example, they could also be quadratic functions, etc.

[0066] It is possible to combine one or more of the above-described embodiments and variations, and to combine the variations with each other.

[0067] Using this invention can improve the recognition accuracy of road markings.

[0068] The present invention has been described above in conjunction with preferred embodiments, but those skilled in the art should understand that various modifications and alterations can be made without departing from the scope of the following claims.

Claims

1. A road marking recognition device (100), characterized in that, have: An external detection unit (2) is mounted on the vehicle (1) to detect the external conditions around the vehicle (1); The behavior detection unit (3) detects the driving behavior of the vehicle (1); The road marking recognition unit (13) sets a coordinate system corresponding to the current position and direction of travel of the vehicle (1) based on the external conditions detected by the external detection unit (2), and identifies the road markings around the vehicle (1) in the coordinate system. The movement calculation unit (14) calculates the movement of the vehicle (1) from the first moment to the second moment based on the driving behavior detected by the behavior detection unit (3); as well as The road marking correction unit (15) corrects the road markings identified by the road marking recognition unit (13). The road marking recognition unit (13) generates first road marking information representing the road markings recognized at the first time and second road marking information representing the road markings recognized at the second time. The road marking correction unit (15) has: The prediction unit (16) predicts the road markings identified by the road marking recognition unit (13) at the second time moment based on the first road marking information generated by the road marking recognition unit (13) and the movement of the vehicle (1) calculated by the movement calculation unit (14); and The updating unit (17) updates the second road marking information generated by the road marking recognition unit (13) based on the difference between the road marking predicted by the prediction unit (16) and the road marking actually recognized by the road marking recognition unit (13) at the second time.

2. The road marking recognition device (100) according to claim 1, characterized in that, The road marking recognition unit (13) generates a first function that approximates the road markings identified at the first time as first road marking information, and generates a second function that approximates the road markings identified at the second time as second road marking information.

3. The road marking recognition device (100) according to claim 2, characterized in that, The movement calculation unit (14) calculates the translational movement and rotational movement of the vehicle (1) from the first moment to the second moment. The prediction unit (16) moves the first function generated by the road marking recognition unit (13) in parallel according to the translational movement amount calculated by the movement amount calculation unit (14), and rotates the first function according to the rotational movement amount, thereby predicting the road markings recognized by the road marking recognition unit (13) at the second moment.

4. The road marking recognition device (100) according to claim 3, characterized in that, The number of times the first function is moved in parallel and rotated by the prediction unit (16) is consistent with the number of times the first function and the second function are generated by the road marking recognition unit (13).

5. The road marking recognition device (100) according to any one of claims 1 to 4, characterized in that, It also includes a driving control unit (18) that controls the driving actuator (4) based on the road markings corrected by the road marking correction unit (15).

6. The road marking recognition device (100) according to any one of claims 1 to 4, characterized in that, It also has: Storage unit (12), which stores map information; as well as The map information update unit (19) updates the map information stored in the storage unit (12) based on the road lines corrected by the road line correction unit (15).

Images