Lane boundary line recognition method and lane boundary line recognition apparatus

Abstract

To improve recognition accuracy of a lane boundary line when a point group of feature points loses continuity, in a lane boundary line recognition method for recognizing the lane boundary line on the basis of a detection result of the feature points of a road surface marking.SOLUTION: A lane boundary line recognition method includes: calculating spatially distributed characteristics and discrete characteristics of feature points (S6 and S7); setting a plurality of coefficients for determining approximation functions for calculating approximation curves of point groups of the feature points on the basis of the spatially distributed characteristics and the discrete characteristics (S9); and estimating the approximate curves of the point groups of the feature points calculated by the approximate functions as a lane boundary line.SELECTED DRAWING: Figure 8

Description

[Technical Field]

[0001] The present invention relates to a lane boundary line recognition method and a lane boundary line recognition device. [Background technology]

[0002] The road shape recognition device described in Patent Document 1 below calculates an approximate curve that approximates the shape of the road boundary based on multiple feature points that indicate the road boundary on which the vehicle is traveling in a captured image, and recognizes the road shape based on the calculated approximate curve. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2018-005617 Summary of the Invention [Problem to be solved by the invention]

[0004] The lane shape recognition device described in Patent Document 1 estimates an approximate curve of the lane boundary shape based on a point cloud of observation data of the lane boundary. Therefore, if the continuity of the lane boundary data is lost due to a lack of observation data or other reasons, there is a risk that the estimation accuracy will decrease. The present invention aims to improve the accuracy of lane boundary line recognition in a lane boundary line recognition method that recognizes lane boundary lines based on the detection results of feature points of road markings when the point group of feature points loses continuity. [Means for solving the problem]

[0005] According to one aspect of the present invention, there is provided a lane boundary line recognition method that detects feature points on road markings around the vehicle using a sensor mounted on the vehicle and recognizes lane boundary lines based on the detected feature points. The lane boundary line recognition method calculates spatial distribution characteristics and discrete characteristics of the feature points, sets multiple coefficients based on the spatial distribution characteristics and discrete characteristics to define an approximation function that calculates an approximation curve of a point cloud of the feature points, and estimates the approximation curve of the point cloud of the feature points calculated by the approximation function as a lane boundary line. [Effects of the Invention]

[0006] According to the present invention, in a lane boundary line recognition method that recognizes lane boundary lines based on the detection results of feature points of road markings, it is possible to improve the recognition accuracy of lane boundary lines when the point group of feature points loses continuity. [Brief explanation of the drawings]

[0007] [Figure 1] 1 is a schematic configuration diagram of an example of a lane boundary line recognition device according to an embodiment; [Figure 2] FIG. 1 is a schematic diagram of a point cloud of feature points obtained by detecting road markings. [Figure 3] 10(a) and 10(b) are diagrams illustrating a method for setting a search range for feature points for recognizing each lane boundary line. [Figure 4] FIG. 10 is a schematic diagram illustrating the extraction result of a point cloud for each lane boundary line. [Figure 5] FIG. 10 is an explanatory diagram of a method for calculating an evaluation value of a discrete characteristic of a feature point. [Figure 6] FIG. 10 is an explanatory diagram of a method for setting a reference value for the spatial distribution of feature points. [Figure 7] FIG. 10 is an explanatory diagram of a method for calculating an evaluation value of the spatial distribution of feature points. [Figure 8] 1 is a flowchart illustrating an example of a lane boundary line recognition method according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] Hereinafter, embodiments of the present invention will be described with reference to the drawings. Note that the drawings are schematic and may differ from the actual product. Furthermore, the embodiments of the present invention shown below are examples of devices and methods for embodying the technical concept of the present invention, and the technical concept of the present invention does not limit the structure, arrangement, etc. of component parts to those described below. The technical concept of the present invention can be modified in various ways within the technical scope defined by the claims.

[0009] (First embodiment) (composition) 1 is a schematic diagram of an example of a lane boundary line recognition device according to an embodiment. A vehicle 1 includes a lane boundary line recognition device 10 that recognizes lane boundary lines around the vehicle 1, and an actuator 20 that generates vehicle behavior of the vehicle 1 by operating the steering device, accelerator opening, and braking device of the vehicle 1. The lane boundary line recognition device 10 includes an object sensor 11 and a controller 12 .

[0010] The object sensor 11 is a sensor that detects objects around the vehicle 1. The object sensor 11 detects the surrounding environment of the vehicle 1, such as the relative position between the vehicle 1 and an object present around the vehicle 1, the distance between the vehicle 1 and the object, and the direction in which the object is present. The object sensor 11 may include, for example, a camera that captures the surrounding environment of the vehicle 1. Furthermore, for example, the vehicle 1 may include a distance measuring device such as a laser range finder (LRF), radar, or LiDAR (Light Detection and Ranging) laser radar, or a sonar sensor that detects obstacles around the vehicle 1. The object sensor 11 outputs surrounding environment information, which is information on the detected surrounding environment of the vehicle 1, to the controller 12.

[0011] The controller 12 is an electronic control unit (ECU) that recognizes lane boundaries around the vehicle 1 based on the detection results of the object sensor 11 of the vehicle 1, and performs vehicle control of the vehicle 1 based on the recognized lane boundaries. The controller 12 includes a processor and peripheral components such as a storage device. The processor may be, for example, a CPU (Central Processing Unit) or an MPU (Micro-Processing Unit). The storage device may include a semiconductor storage device, a magnetic storage device, an optical storage device, etc. The storage device may include a register, a cache memory, and memories such as a ROM (Read Only Memory) and a RAM (Random Access Memory) used as a main memory device.

[0012] The functions of the controller 12 described below are realized, for example, by a processor executing a computer program stored in a storage device. The controller 12 may be formed of dedicated hardware for executing each information processing described below. For example, the controller 12 may include a functional logic circuit configured in a general-purpose semiconductor integrated circuit. For example, the controller 12 may include a programmable logic device (PLD) such as a field-programmable gate array (FPGA).

[0013] The actuator 20 operates the steering device, accelerator opening, and braking device of the vehicle 1 in response to control signals from the controller 12. The actuator 20 includes a steering actuator, an accelerator opening actuator, and a brake control actuator. The steering actuator controls the steering direction and steering amount of the steering device of the vehicle 1. The accelerator opening actuator controls the accelerator opening of the vehicle 1. The brake control actuator controls the braking operation of the braking device of the vehicle.

[0014] The controller 12 includes a driving image storage unit 31, a road marking recognition unit 32, a point cloud overhead image creation unit 33, a lane boundary recognition unit 34, and a vehicle control unit 35. The driving image storage unit 31 stores surrounding environment information detected by the object sensor 11. For example, the driving image storage unit 31 stores color image data captured by an on-board camera included in the object sensor 11. Note that the number of on-board cameras capturing the color image data stored in the driving image storage unit 31 is not limited to one. Each of the color image data captured by multiple on-board cameras may be stored in the driving image storage unit 31.

[0015] The road marking recognition unit 32 detects a point cloud of characteristic points of road markings from the surrounding environment information stored in the running image storage unit 31. Note that various methods can be used to detect road markings, and the method is not limited to a specific method. For example, if color image data captured by an in-vehicle camera is stored in the running image storage unit 31 as surrounding environment information, road markings may be detected using a deep learning method such as semantic segmentation, or a general pattern recognition method such as machine learning.

[0016] The point cloud overhead image creation unit 33 converts the coordinate system of the point cloud of characteristic points of the road markings detected by the road marking recognition unit 32 from the image coordinate system into point cloud information in a vehicle coordinate system with the vehicle 1 as its origin. Fig. 2 is a schematic diagram of the point cloud Gp of the characteristic points of the road markings after being converted into a point cloud on the vehicle coordinate system. The converted point cloud information of the road markings is input to the lane boundary line recognition unit 34. The lane boundary line recognition unit 34 extracts point cloud information that will be used to recognize lane boundary lines from the point cloud information of all input road markings, and recognizes lane boundary lines based on the extracted point cloud information. The lane boundary lines recognized by the lane boundary line recognition unit 34 may be, for example, white solid lines, white dashed lines, yellow solid lines, or light blue solid lines on the road surface.

[0017] First, the lane boundary line recognition unit 34 extracts candidate points that are candidates for the point cloud of each lane boundary line from the point cloud information of the road markings input from the point cloud overhead image creation unit 33. At this time, the lane boundary line recognition unit 34 sets a search range Rs for each lane boundary line in which to search for candidate points for each lane boundary line, and extracts a group of points included in the search range Rs as candidate points for one lane boundary line. 3(a) and 3(b) are diagrams illustrating an example of a method for setting a search range Rs for searching for candidate points of lane boundary lines.

[0018] 3(a) is a trapezoidal region with the current position of the vehicle 1 as the origin O and the center line as a straight line OA extending from the origin O in the traveling direction of the vehicle 1. In the following description, the traveling direction of the vehicle 1 is set as the X direction, and the direction perpendicular to the X direction on the horizontal plane is set as the Y direction. The length of the short side of the trapezoid region Rt is a predetermined value W. The predetermined value W may be set according to the width of the lane boundary line (e.g., 0.15 m) specified in laws and regulations that stipulate the installation style of lane boundary lines (e.g., the "Ordinance on Road Signs, Dividing Lines, and Road Markings"). For example, the predetermined value W may be 0.15 m (i.e., the width of the lane boundary line).

[0019] The length f(d) of the long side of the trapezoid region Rt may be set to increase as the distance d from the vehicle 1 (from the origin O) in the traveling direction (X direction) of the vehicle 1 increases. For example, the length f(d) of the long side may be f(d) = 0.15 + 0.01 × d. The height d of the trapezoidal region Rt may be set to the length of the detectable range of the object sensor 11 (for example, 70 m).

[0020] Referring to Fig. 3(b), the lane boundary line recognition unit 34 randomly selects a search start point Pa from all point cloud data from which candidate points for lane boundary lines are extracted. The lane boundary line recognition unit 34 moves the trapezoidal area Rt in a direction perpendicular to the direction of travel of the vehicle 1 (Y direction) so that the center line OA of the trapezoidal area Rt passes through the search start point Pa, and sets the moved trapezoidal area Rt as the search range Rs.

[0021] 4 is a schematic diagram of the extraction results of candidate points for point clouds for each lane boundary line. In the example of Fig. 4, point clouds Gp1, Gp2, Gp3, and Gp4 are extracted as candidate points for point clouds of four lane boundary lines. Please refer to Figure 5. The lane boundary line recognition unit 34 calculates the maximum separation distance between feature points included in each of the point groups Gp1 to Gp4, which are candidate points for lane boundary lines. In the example of Figure 5, the maximum separation distances d2 to d4 between the point groups Gp2 to Gp4 are shown. Since the point group Gp1 has no discontinuous parts, the maximum separation distance d1 is "0". In the following description, the maximum separation distances d1 to d4 may be collectively referred to as the "maximum separation distance d".

[0022] The lane boundary line recognition unit 34 determines for each lane boundary line whether the maximum separation distance d is greater than a predetermined determination threshold dth. For example, the determination threshold dth may be set according to the maximum distance (e.g., 9 m) between lines of a broken lane boundary line specified in laws and regulations that stipulate the installation style of lane boundary lines. For example, the determination threshold dth may be 9 m. If the maximum separation distance d is equal to or less than the predetermined threshold value dth, the lane boundary line recognition unit 34 determines that there is no missing observation data for the lane boundary line candidate point. In this case, the lane boundary line recognition unit 34 sets the weighting coefficients k1, k2, and k3 to "1."

[0023] If the maximum separation distance d is longer than a predetermined determination threshold dth, the lane boundary line recognition unit 34 determines that the observation data of the lane boundary line candidate point is missing. In this case, the lane boundary line recognition unit 34 estimates the approximate shape of the observation data of the lane boundary line candidate point and reflects the estimation result in the estimation of the approximate curve of the lane boundary line. Specifically, the lane boundary line recognition unit 34 calculates a spatial distribution evaluation value α that represents the spatial distribution characteristics of the candidate points (feature points) and a discrete characteristic evaluation value β of the candidate points as parameters that represent the approximate shape of the observation data of the candidate points of the lane boundary line. The lane boundary line recognition unit 34 reflects the spatial distribution evaluation value α and the discrete characteristic evaluation value β in estimating the approximate curve of the lane boundary line.

[0024] When multiple lane boundary lines are extracted from the point cloud information input from the point cloud overhead image creation unit 33, a common spatial distribution evaluation value α may be calculated for the multiple lane boundary lines, or a spatial distribution evaluation value α may be calculated individually for each lane boundary line. On the other hand, with regard to the discrete characteristic evaluation value β, the lane boundary line recognition unit 34 calculates the discrete characteristic evaluation value β individually for each lane boundary line. The spatial distribution characteristics of feature points are feature quantities that represent the distribution shape of feature points in space, and feature quantities such as variance values ​​in Euclidean space and eigenvalues ​​in eigenspace can be used.

[0025] (Calculation of spatial distribution evaluation value α) In order to calculate the spatial distribution evaluation value α, the reference value α of the spatial distribution of the candidate points is c For example, as shown in Fig. 6, a straight three-lane road is used as the evaluation reference model, and the data distribution range a in the X direction (travel direction of vehicle 1) of the points on the lane boundary of the evaluation reference model is set as c and the data distribution range b in the Y direction (lateral direction of vehicle 1). c Set.

[0026] For example, the data distribution range a in the X direction c may be set to the length of the detectable range of the object sensor 11 (for example, 70 m). Also, the data distribution range in the Y direction b c may be set to the width of the number of lanes in the evaluation reference model based on the laws and regulations that stipulate the installation style of lane boundary lines. For example, if the inner width of the lane boundary line of one lane is W1 (e.g., 3.5 m) and the width of the lane boundary line is W2 (e.g., 0.15 m), the data distribution range b for one lane c may be set to 3×W1+4×W2 (e.g., approximately 11 m). Reference value α c is the data distribution range a c Data distribution range b for c The ratio (α c =b c / a c )

[0027] The lane boundary line recognition unit 34 calculates an evaluation value α of the spatial distribution range of the observation data of the candidate points of the lane boundary lines. o The lane boundary line recognition unit 34 calculates the data distribution range a of the observation data Gp in the X direction as shown in FIG. o and the data distribution range b of the observation data Gp in the Y direction o The lane boundary line recognition unit 34 calculates the data distribution range a o Data distribution range b for o The ratio (α o =b o / a o ) is evaluated as α o When calculating the spatial distribution evaluation value α for each lane boundary individually, the distribution range a of the observation data extracted as a candidate point for one lane boundary is calculated as o , b o Based on the evaluation value α o may be calculated. The lane boundary line recognition unit 34 determines the reference value α c Evaluation value α for o The ratio (α = α o / α c ) is calculated as the spatial distribution evaluation value α.

[0028] (Calculation of discrete characteristic evaluation value β) The lane boundary recognition unit 34 calculates the ratio (β=d / dth) of the maximum separation distance d to the maximum distance dth (e.g., 9 m) between the lines of dashed lane boundary lines specified in laws and regulations that stipulate the installation style of lane boundary lines, as the discrete characteristic evaluation value β.

[0029] Next, the lane boundary line recognition unit 34 calculates weighting factors k1, k2, and k3 based on the spatial distribution evaluation value α and the discrete characteristic evaluation value β according to the following equation (1).

[0030]

number

[0031] The lane boundary line recognition unit 34 calculates a third-order polynomial y=C0+C1x+C2x of an approximate curve of the point group of candidate points by the least squares method based on the following (a) to (c):2 +C3x 3 Estimate. (a) The lane boundary line recognition unit 34 substitutes weighting coefficients k1, k2, and k3 into the evaluation function E of the following equation (2). The combination of coefficients C0, C1, C2, and C3 that minimizes this evaluation function E becomes the coefficients of each term of the polynomial that represents the approximate curve of the point group of candidate points.

[0032]

number

[0033] The first term in equation (2) is an estimation error term that represents the error between the position of the candidate point (feature point) and the estimated value of the lane boundary line. The second term in equation (2) is a curve shape evaluation term that evaluates the shape of the approximation curve defined by the polynomial coefficients C1, C2, and C3. The curve shape evaluation term includes a composite value of the coefficients C1, C2, and C3 weighted by the weighting coefficients k1, k2, and k3. For example, the composite value is the sum of squares of the products of the weighting coefficients k1, k2, and k3 and the coefficients C1, C2, and C3 ((k1C1) 2 +(k2C2) 2 +(k3C3) 2 ) The composite value is not limited to this example, and may be, for example, a weighted sum of the coefficients C1, C2, and C3 using the weighting coefficients k1, k2, and k3. By substituting equation (2) into a matrix form, the following equation (3) is obtained.

[0034]

number

[0035] In Equation (2) and Equation (3), the variable i is a subscript that identifies each candidate point in the point cloud of one lane boundary line, the natural number N is the total number of candidate points included in the point cloud of the lane boundary line, and x i is the x coordinate of the candidate point, and y i and t i is the y coordinate of the candidate point, and φ(x i )=(1,x1,x2 2 ,…x iD ) T and δ i is the measurement accuracy of the candidate point (the inverse of the variance), λ is the regularization coefficient, vector k is a vector whose elements are the weighting coefficients k0, k1, k2, and k3, and vector ω is a vector whose elements are the coefficients C0, C1, C2, and C3 of each term of the polynomial.

[0036] The lane boundary line recognition unit 34 calculates a combination of coefficients C0, C1, C2, and C3 that minimizes the evaluation function E by solving the partial differential equation of the following equation (4).

[0037]

number

[0038] The solution of equation (4) is given by the following equation (5): The lane boundary line recognition unit 34 uses equation (5) to calculate the polynomial y=C0+C1x+C2x of the approximate curve of the point group of candidate points. 2 +C3x 3 Estimate.

number

[0039] In equation (3), vector t is an N-dimensional vector whose elements are the y-coordinates of all candidate points, and Φ is the design matrix (φ(x1) T ,φ(x2) T ,…φ(x N ) T ) T , and Δ is the precision matrix (δ i is a square matrix with diagonal components. The lane boundary line recognition unit 34 estimates the approximate curve y=C0+C1x+C2x using the equation (5). 2 +C3x 3 are recognized as lane markings.

[0040] Referring again to the evaluation function E in equation (2), the evaluation function E includes a curve shape evaluation term (the second term). Therefore, if the point group of candidate points for lane boundary lines is close to a straight line and the weighting coefficient k1 is large, even a slight change in coefficient C1 will cause a large change in the evaluation function E (i.e., the absolute value of the partial differential ∂E / ∂w will become large). Therefore, when the partial differential ∂E / ∂w is converged to 0 (i.e., when the evaluation function E is minimized), the coefficient C1 is preferentially estimated to the optimal value. The remaining coefficients C2 and C3 are estimated after optimizing the coefficient C1.

[0041] Similarly, if the point cloud of candidate points for lane boundary lines is close to a curve and the weighting coefficients k2 and k3 are large, the coefficients C2 and C3 are optimized before the remaining coefficients are estimated. Therefore, the weighting coefficients k1, k2, and k3 play a role in amplifying the influence that the coefficients C1, C2, and C3 of the approximation function that approximates the shape of the point group of feature points have on the evaluation function E. In this way, the lane boundary line recognition unit 34 evaluates the approximate shape of the observed data of the candidate points of the lane boundary lines, and based on the evaluation results, estimates one of the coefficients C1, C2, and C3 with priority. 2 +C3x 3 This can improve the accuracy of approximating lane boundary lines.

[0042] For example, when the shape of the point group of candidate points is close to a straight line and an approximate curve having a shape close to a straight line is estimated as a lane boundary line, the lane boundary line having a shape close to a straight line can be accurately approximated by estimating coefficient C1, which determines the characteristics of the straight line, with priority over other coefficients. Similarly, when the shape of the point group of characteristic points is close to a curve and an approximate curve having a shape similar to a curve is estimated as a lane boundary line, the lane boundary line having a shape similar to a curve can be accurately approximated by estimating coefficients C2 and C3 that determine the characteristics of the curve with priority over other coefficients.

[0043] Thereafter, the lane boundary line recognition unit 34 removes the observation data of candidate points that can be placed on the estimated approximate curve (for example, candidate points located on or near the approximate curve) as inlier points from all point cloud information received from the point cloud overhead image creation unit 33. The lane boundary line recognition unit 34 calculates the number of data (point clouds) contained in the remaining point cloud information after the inlier points have been deleted, and determines whether or not it satisfies the minimum required number of points needed to estimate the approximate curve of the next lane boundary line.

[0044] If the remaining point cloud information satisfies the required number of points, the lane boundary line recognition unit 34 repeats the above process to estimate the next lane boundary line. If the remaining point cloud information does not satisfy the required number of points, the lane boundary line recognition unit 34 outputs the recognition result of the lane boundary line to the vehicle control unit 35. The vehicle control unit 35 executes driving assistance control to assist the vehicle 1 in driving based on the lane boundary lines recognized by the lane boundary line recognition unit 34.

[0045] For example, the driving assistance control by the vehicle control unit 35 may be autonomous driving control that causes the vehicle 1 to drive autonomously to a set destination, or driving assistance control that assists in driving the vehicle 1 by controlling at least the steering mechanism of the vehicle 1. The vehicle control unit 35 controls the actuator 20 to cause the vehicle 1 to drive autonomously and to control the steering mechanism, driving force, and braking force of the vehicle 1.

[0046] For example, the vehicle control unit 35 may execute a lane departure prevention function (lane keeping function) that controls the actuator 20 so that the vehicle 1 maintains a predetermined lateral position (e.g., near the center) within the lane based on the lane boundary lines recognized by the lane boundary line recognition unit 34, or a lane change assistance function.

[0047] (operation) In step S1, the lane boundary line recognition unit 34 receives the point cloud data (point cloud information) of the characteristic points of the road markings from the point cloud overhead image creation unit 33. In step S2, the lane boundary line recognition unit 34 extracts a point cloud (candidate points) that will be used to recognize one lane boundary line from the point cloud information of all input road markings. In step S3, the lane boundary line recognition unit 34 calculates the maximum separation distance d between the candidate points extracted in step S2.

[0048] In step S4, the lane boundary line recognition unit 34 determines whether the maximum separation distance d is greater than the determination threshold dth. If the maximum separation distance d is greater than the determination threshold dth (step S4: Y), the process proceeds to step S6. If the maximum separation distance d is equal to or less than the determination threshold dth (step S4: N), the process proceeds to step S5. In step S5, the lane boundary line recognition unit 34 sets the weighting factors k1, k2, and k3 to 1. Thereafter, the process proceeds to step S9.

[0049] In step S6, the lane boundary line recognition unit 34 calculates a spatial distribution evaluation value α of the point cloud of candidate points. In step S7, the lane boundary line recognition unit 34 calculates a discrete characteristic evaluation value β of the point cloud of candidate points. In step S8, the lane boundary line recognition unit 34 calculates weighting coefficients k1, k2, and k3 based on the spatial distribution evaluation value α and the discrete characteristic evaluation value β. The process then proceeds to step S9.

[0050] In step S9, the lane boundary line recognition unit 34 defines the evaluation function E of equation (2) using the weighting coefficients k1, k2, and k3, and calculates the polynomial y=C0+C1x+C2x of the approximate curve of the point group of candidate points based on the defined evaluation function E. 2 +C3x 3 The estimated approximate curve is recognized as the lane boundary line. In step S10, the lane boundary line recognition unit 34 deletes the observation data of candidate points that can be placed on the estimated approximate curve from the point cloud information received from the point cloud overhead image creation unit 33 as inlier points.

[0051] In step S11, the lane boundary line recognition unit 34 calculates the number of data points remaining in the point cloud information after the inlier points have been deleted, and determines whether the number of data points satisfies the minimum required number of points necessary to estimate the approximate curve of the next lane boundary line. If the remaining point cloud information satisfies the required number of points (step S11: Y), the process returns to step S2. If the remaining point cloud information does not satisfy the required number of points (step S11: N), the process proceeds to step S12. In step S12, the lane boundary line recognition unit 34 outputs the recognition result of the lane boundary line to the vehicle control unit 35. Thereafter, the process ends.

[0052] (Effects of the embodiment) (1) A lane boundary line recognition method detects feature points on road markings around the vehicle using a sensor mounted on the vehicle, and recognizes lane boundary lines based on the detected feature points. The lane boundary line recognition method calculates the spatial distribution characteristics and discrete characteristics of the feature points, sets multiple coefficients that define an approximation function that calculates an approximation curve of the point cloud of the feature points based on the spatial distribution characteristics and discrete characteristics, and estimates the approximation curve of the point cloud of the feature points calculated by the approximation function as the lane boundary line. This makes it possible to determine the approximate shape of the lane boundary line to be estimated (for example, whether it is a shape close to a straight line, a shape close to a monotonic curve such as a quadratic curve, or a shape close to a complex curve such as a cubic curve), and reflect the determination result in estimating the approximate curve of the lane boundary line.As a result, even if continuity is lost due to a gap in the observed data of the lane boundary line, it is possible to restore the correspondence between the feature points before and after the gap, and to estimate an approximate curve that can correctly represent the shape of the lane boundary line.

[0053] (2) Weighting coefficients for weighting the multiple coefficients of the approximation function may be set based on the spatial distribution characteristics and the discrete characteristics, and the coefficients of the approximation function may be set based on an evaluation function that includes the error between the position of the feature point and the estimated value of the lane boundary line and a composite value of the multiple coefficients weighted by the weighting coefficients. This makes it possible to limit the fluctuation range of each coefficient of the approximation function within a certain range when minimizing the estimation error of the evaluation function. As a result, extreme deformation of the approximation curve is avoided, and a correct approximation curve is obtained.

[0054] (3) The spatial distribution characteristics may be calculated based on the distribution range of the feature points in the traveling direction of the vehicle and the distribution range of the feature points in a direction perpendicular to the traveling direction. This allows defining a method for evaluating the spatial distribution characteristics of the feature points. (4) The ratio between the length of the sensor's detectable range in the traveling direction and a constant that defines the distance in the road width direction occupied by a predetermined number of lanes may be set as a reference value, and the ratio between the distribution range of feature points in the traveling direction and the distribution range of feature points in a direction perpendicular to the traveling direction may be calculated as an evaluation value of the distribution range of feature points, and the spatial distribution characteristics may be calculated based on the evaluation value and the reference value. This allows for defining a method for evaluating the spatial distribution characteristics of feature points.

[0055] (5) The discrete characteristic may be the distance between the feature points in the traveling direction of the vehicle, which allows defining a method for evaluating the discrete characteristic of the feature points. (6) Lane markings may be solid white lines, broken white lines, solid yellow lines, or solid light blue lines, allowing various types of lane markings to be recognized.

[0056] (7) The approximation function may include a third-order polynomial, which can approximate most lane boundaries. (8) A search range may be set such that the width in the direction perpendicular to the vehicle's direction of travel increases as the distance from the vehicle increases, and an approximate curve may be calculated using feature points included in the search range. This allows the search for feature points of lane boundary lines not only for straight lanes but also for curved lanes. [Explanation of symbols]

[0057] 1...vehicle, 10...lane boundary line recognition device, 11...object sensor, 12...controller, 20...actuator, 31...driving image storage unit, 32...road marking recognition unit, 33...point cloud overhead image creation unit, 34...lane boundary line recognition unit, 35...vehicle control unit

Claims

1. A lane boundary line recognition method for detecting characteristic points of road markings around a vehicle using a sensor mounted on the vehicle and recognizing lane boundary lines based on the detected characteristic points, comprising: Calculating the spatial distribution characteristics and discrete characteristics of the feature points; setting a plurality of coefficients that define an approximation function for calculating an approximation curve of the point group of the feature points based on the spatial distribution characteristics and the discrete characteristics; an approximation curve of the point group of the feature points calculated by the approximation function is estimated as the lane boundary line; A lane boundary line recognition method characterized by:

2. setting a weighting coefficient for weighting the plurality of coefficients of the approximation function based on the spatial distribution characteristic and the discrete characteristic; setting the coefficients of the approximation function based on an evaluation function including an error between the position of the feature point and the estimated value of the lane boundary line and a composite value of the plurality of coefficients weighted by the weighting coefficient; 2. The lane boundary line recognition method according to claim 1.

3. 2. The lane boundary line recognition method according to claim 1, wherein the spatial distribution characteristics are calculated based on a distribution range of the feature points in the traveling direction of the vehicle and a distribution range of the feature points in a direction perpendicular to the traveling direction.

4. setting a ratio between the length of the detectable range of the sensor in the traveling direction and a constant that defines a distance in the road width direction occupied by a predetermined number of lanes as a reference value; calculating a ratio between a distribution range of the feature points in the traveling direction and a distribution range of the feature points in a direction perpendicular to the traveling direction as an evaluation value of the distribution range of the feature points; calculating the spatial distribution characteristic based on the evaluation value and the reference value; 4. The lane boundary line recognition method according to claim 3.

5. 2. The lane boundary line recognition method according to claim 1, wherein the discrete characteristic is a distance between the point groups of the feature points in the traveling direction of the vehicle.

6. 2. The lane boundary line recognition method according to claim 1, wherein the lane boundary lines are road markings that are white solid lines, white broken lines, yellow solid lines, or light blue solid lines.

7. 2. The method of claim 1, wherein the approximation function includes a third-order polynomial.

8. a search range whose width in a direction perpendicular to the traveling direction of the vehicle increases as the distance from the vehicle in the traveling direction of the vehicle increases; calculating the approximation curve using the feature points included in the search range; 2. The lane boundary line recognition method according to claim 1.

9. A lane boundary line recognition device that detects characteristic points of road markings around a vehicle using a sensor mounted on the vehicle and recognizes lane boundary lines based on the detected characteristic points, A process of calculating spatial distribution characteristics and discrete characteristics of the feature points; a process of setting a plurality of coefficients that define an approximation function for calculating an approximation curve of the point group of the feature points based on the spatial distribution characteristic and the discrete characteristic; a process of estimating an approximate curve of the point group of the feature points calculated by the approximate function as the lane boundary line; A lane boundary line recognition device comprising a controller that executes the above.

Images