Calculation device, calculation method, and calculation program

Abstract

To provide a calculation device, a calculation method, and a calculation program capable of accurately calculating a lane width of a lane by suppressing an error.SOLUTION: A calculation device 100 includes an image acquisition unit 110 that acquires a road image in a traveling direction of a vehicle, a lane boundary detection unit 112 that extracts a lane boundary line on a road surface from the road image as a point sequence and outputs an image coordinate value, and an estimation unit 114 that calculates a lane width using at least the image coordinate value and using an assumption that a lane width is constant and an assumption that a longitudinal curvature is constant through a predetermined calculation process.SELECTED DRAWING: Figure 7

Description

[Technical Field]

[0001] The present disclosure relates to a computing device, a computing method, and a computing program. [Background technology]

[0002] There is a conventional technology for lane recognition. In lane recognition using monocular camera images, lane width is calculated from the nearest white line under the assumption of a local plane, and the lane width is used as a scale factor to calculate and reconstruct the three-dimensional structure of the lane.

[0003] For example, there is a technology relating to a bank angle calculation device and vehicle that can calculate the bank angle of the front lane on which the vehicle is traveling (see Patent Document 1). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2022-096427 Summary of the Invention [Problem to be solved by the invention]

[0005] In conventional technology, the local plane assumption breaks down on curves with banking angles, which can lead to errors in the lane width calculated from the nearest white line under the local plane assumption, which can result in estimation errors in the reconstructed three-dimensional structure of the road.

[0006] Specifically, on a curved road with a bank angle, the vehicle body tilts perpendicular to the banked road surface, causing the planar curvature component of the curve observed from the camera coordinates to decrease in proportion to the cosine of the bank angle, and the longitudinal curvature component to increase in proportion to the sine of the bank angle. This causes the local planar assumption to break down, resulting in errors in the lane width calculated based on the local planar assumption, which was the cause of the problem.

[0007] The problem with conventional technology is that lane width is estimated using a local planar assumption. To solve this problem, it is necessary to calculate lane width by taking into account the longitudinal linear component of the road caused by the bank angle.

[0008] The technology disclosed herein has been developed in consideration of the above circumstances, and aims to provide a calculation device, a calculation method, and a calculation program that suppress errors and enable accurate calculation of lane widths. [Means for solving the problem]

[0009] In order to achieve the above object, the calculation device according to the present disclosure includes an image acquisition unit that acquires a road image in the direction of travel of the vehicle, a lane boundary detection unit that extracts lane boundary lines on the road surface from the road image as a sequence of points and outputs image coordinate values, and an estimation unit that uses at least the image coordinate values ​​to calculate the lane width through a predetermined calculation process, assuming a constant lane width and a constant longitudinal curvature. [Effects of the Invention]

[0010] According to the calculation device of the present disclosure, it is possible to suppress errors and calculate the lane width with high accuracy. [Brief explanation of the drawings]

[0011] [Figure 1] Figure 1 shows an example of a model representation and coordinate system for explaining the method. [Figure 2] Figure 2 shows an example of lane width estimation under the local plane assumption. [Figure 3] Figure 3 shows the bank angle and the general shape of the driving lane. [Figure 4] Figure 4 shows the lane width estimates based on the local plane assumption. [Figure 5] Figure 5 shows an example of the radius of curvature as viewed from the 3D camera coordinate system. [Figure 6] FIG. 6 is a block diagram showing the hardware configuration of the computing device. [Figure 7]FIG. 7 is a diagram illustrating the functional configuration of the computing device of the first embodiment. [Figure 8] FIG. 8 shows an example of coordinate values ​​of points on three white lines on the left and right. [Figure 9] FIG. 9 is a flowchart showing the flow of processing by the computing device of the first embodiment. [Figure 10] FIG. 10 is an example of lane width estimation in the method of the present disclosure. [Figure 11] FIG. 11 is a diagram illustrating the functional configuration of a computing device according to the second embodiment. [Figure 12] FIG. 12 is a diagram corresponding to the bank angle and the gradient angle. [Figure 13] FIG. 13 shows an example of coordinate values ​​of points on two white lines, one on the left and one on the right. [Figure 14] FIG. 14 is a flowchart showing the flow of processing by the computing device of the second embodiment. [Figure 15] FIG. 15 is a diagram illustrating the functional configuration of a computing device according to the third embodiment. [Figure 16] FIG. 16 is a flowchart showing the flow of processing by the computing device of the third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0012] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] The present disclosure will be described in detail below with reference to the accompanying drawings. First, the premise of the method of the present disclosure will be described.

[0013] In conventional lane recognition, bank angles are often ignored as small and white lines are assumed to be painted on a local plane. However, on test courses and on sharp curves such as those on the Metropolitan Expressway, bank angles of 10 to 30 degrees can be added. White lines captured by a vehicle's camera on such a banked road appear twisted and raised, making the local plane assumption difficult to apply. When the local plane assumption is broken, errors in the lane width estimated from the image increase. In lane recognition using an on-board monocular camera, lane width is a scale factor, so errors in lane width cause estimation errors in lane recognition. Conversely, improving the accuracy of lane width estimation leads to improved estimation accuracy in lane recognition.

[0014] Figure 1 shows an example of a model representation and coordinate system for explaining the method. The world coordinate system is Y w Cartesian coordinates X with the gravity axis pointing downward w -Y w -Z w The white line of the curve with a bank angle and the 3D position and orientation of the camera (vehicle body) are expressed as follows. However, for convenience, the Z coordinate of the vehicle is used here. v axis and world coordinate Z w The direction of the lines is the same. The white lines on a curve with a bank angle are modeled as a cone with the gravity axis pointing downward, as shown in Figure 1. The white lines are arcs on the cross section of the cone cut horizontally, and the center of the driving lane is an arc with a curvature radius of r. The angle between the generatrix of the cone and the horizontal plane is the bank angle ψ w Represents.

[0015] The camera (vehicle) coordinate 3D is the Cartesian coordinate X fixed to the camera (vehicle) riding vertically on the driving lane inside the cone. v -Y v -Z v Here, it is assumed that the posture of the vehicle (camera) changes along the road surface. The axis extending perpendicular to the road surface from the camera (vehicle) is Y v axis, and the direction of travel of the vehicle is Z v axis, and the vehicle's right hand direction is X v The Y axis vCamera coordinates 3D are those with the origin on the axis set at the camera position, and vehicle coordinates 3D are those with the origin set at the road surface position. Camera coordinates 2D are coordinates ix-iy on the camera's imaging surface, and the results of perspective transformation of the white lines expressed in camera coordinates 3D onto the camera's imaging surface are expressed in camera coordinates 2D. A segment is a line that represents the width of a lane, and its endpoints represent pairs of left and right white line points.

[0016] Next, we will explain the lane width error when assuming a local plane. The calculation formula for lane width when the local plane assumption is adopted is shown in Equation (1) (the same applies to Equation (11) described later). The derivation will be described later.

number

[0017] In this case, r h : horizontal pixel resolution [m / pixel], r v : vertical pixel resolution [m / pixel], h: camera height [m], α r(l) : The inclination of the nearest point (bottom edge of the white line captured on the image) to the right (left) side of the white line on the 2D camera image (image) (see Figure 2). Figure 2 is an example of lane width estimation based on the local plane assumption. The numerical value indicates the bank angle (deg). The lateral position, yaw angle, and pitch angle are all zero.

[0018] Figure 3 shows the approximate shape of the bank angle and the driving lane. Figure 4 shows the results of calculating the lane width using equation (1) for the curve with the bank angle in Figure 3. Figure 4 shows the estimated lane width based on the local plane assumption. (a) in Figure 4 is the true lane width, and (b) is the estimated lane width based on the local plane assumption. In the world coordinate system, Z w -X w The lane width projected onto a plane is set to a constant value (3.5 m). Therefore, the true lane width (a) on a road surface with a bank angle increases in proportion to 1 / cos (bank angle). On the other hand, it can be seen that the lane width (b) calculated using equation (1) based on the local plane assumption has an error from the true value. For example, for a bank angle of 30 degrees, it can be seen that an error of 0.5 m occurs when using the local plane assumption.

[0019] Figure 5 shows an example of the radius of curvature as seen from the camera coordinate 3D. Here, based on Figure 5, we consider the reason why the local plane assumption breaks down on a curve with a bank angle. If we view the camera (vehicle) as a mass point along a curved road with a bank in the world coordinate 3D, its Z w -X w The projection onto the plane is a trajectory g(Z w ) is expressed as the locus g(Z w ) can be approximately expressed by the quadratic equation shown in equation (2). For simplicity, equation (2) assumes that the vehicle is in the center of the lane and (X w =Z w =0), along the track (Z z This assumes a simple situation where the object is facing in a certain direction.

number

[0020] Next, this locus g(Z w ) is assumed to be observed from the camera (vehicle) coordinate system 3D. In this case, Z v -X v The trajectory projected onto the plane is Xv=g1(Z v ) and Y v -Z v The trajectory projected onto the plane is Yv = g2(Z v ) In this case, g1 and g2 can be expressed by equations (3) and (4).

number

number

[0021] In other words, by following the road surface on a banked curve, X v Around the axis, a circular arc with a radius of curvature r / sin(ψ) is generated, and Yv It can be seen that a trajectory around the axis draws an arc with a curvature radius of r / cos(ψ). Furthermore, because curvature is the reciprocal of the radius of curvature, when viewed from the 3D camera (vehicle) coordinate system, a banked curve can be considered to be traveling on a road with a planar curvature of cos(ψ) / r and a longitudinal curvature of sin(ψ) / r. In other words, the local planar assumption does not apply to curves with a banking angle, and it is clear that errors in the lane width will occur if the longitudinal curvature sin(ψ) / r is not estimated and taken into account when calculating the lane width.

[0022] The above is an explanation of the lane width error that is the premise of the technology of the present disclosure. Each embodiment will be described below. The first embodiment is a method for calculating the lane width from the coordinate values ​​of three lines. The second embodiment is a method for calculating the lane width from the coordinate values ​​of two lines. The third embodiment is a method for calculating the lane width from the coordinate values ​​of two lines using the road surface height.

[0023] [First embodiment] Fig. 6 is a block diagram showing the hardware configuration of a computing device. The hardware configuration of the computing device is common to all embodiments. As shown in Fig. 6, a computing device 100 (200, 300) has a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a storage 14, an input unit 15, a display unit 16, and a communication interface (I / F) 17. Each component is connected to each other via a bus 19 so as to be able to communicate with each other.

[0024] The CPU 11 is a central processing unit that executes various programs and controls each component. That is, the CPU 11 reads programs from the ROM 12 or the storage 14 and executes the programs using the RAM 13 as a work area. The CPU 11 controls the above components and performs various arithmetic processing in accordance with the programs stored in the ROM 12 or the storage 14. In this embodiment, the ROM 12 or the storage 14 stores various programs. The various programs include a calculation program.

[0025] The ROM 12 stores various programs and various data. The RAM 13 temporarily stores programs or data as a working area. The storage 14 is configured with a storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive) and stores various programs including the operating system and various data.

[0026] The input unit 15 includes a pointing device such as a mouse and a keyboard, and is used to input various types of information. The display unit 16 is, for example, a liquid crystal display, and displays various types of information. The display unit 16 may also function as the input unit 15 by employing a touch panel system.

[0027] The communication interface 17 is an interface for communicating with other devices such as terminals, and uses standards such as Ethernet (registered trademark), FDDI, and Wi-Fi (registered trademark).

[0028] Next, a description will be given of each functional configuration of the computing device 100. Each functional configuration is realized by the CPU 11 reading out a computing program stored in the ROM 12 or the storage 14, expanding it in the RAM 13, and executing it.

[0029] Fig. 7 is a diagram showing the functional configuration of the computing device of the first embodiment. As shown in Fig. 7, the computing device 100 includes an image acquisition unit 110, a lane boundary detection unit 112, and an estimation unit 114. The estimation unit 114 includes a selection unit 120 and a lane width estimation unit 122. Note that for ease of explanation, some names may be omitted.

[0030] The image acquisition unit 110 is, for example, an in-vehicle camera provided in a vehicle (not shown), and acquires road images in the traveling direction of the vehicle captured by the in-vehicle camera. The image acquisition unit 110 as an in-vehicle camera is set facing forward at the rearview mirror position inside the vehicle, and captures road images in front of the vehicle.

[0031] The lane boundary detection unit 112 extracts lane boundary lines on the road surface from the road image as a sequence of points and outputs image coordinate values. For example, the lane boundary detection unit 112 extracts lane boundary lines such as white lane marks from the input road image and outputs the coordinate values ​​(ix l ,iy l ) and the coordinate value of the point on the right lane boundary line (ix r ,iy r ) is obtained. (ix l ,iy l ), (ix r ,iy r ) may be stored in memory (ROM 12) as a point group, or may be fitted with a B-spline function or the like and treated as a function.

[0032] Furthermore, the lane boundary detection unit 112 extracts lane boundaries, such as white lines and curbs laid on the left and right sides of the lane in which the host vehicle is traveling, from the road image. Generally, lane boundaries are extracted using an edge extraction filter such as a Sobel filter, assuming that there is a difference in brightness on the image at the boundary between the road surface and the lane boundary. Furthermore, within small regions in the image, lane boundaries are assumed to be straight lines, and lane boundaries are extracted using a Hough transform or the like. Alternatively, lane boundary features may be learned using deep learning to extract lane boundaries.

[0033] The estimation unit 114 calculates the lane width using the image coordinate values ​​through a predetermined calculation process, assuming a constant lane width and a constant longitudinal curvature. The following describes the processing of each unit of the selection unit 120 and the lane width estimation unit 122 that perform the predetermined calculation process of the estimation unit 114 of this embodiment.

[0034] The selection unit 120 selects points for three lines by horizontally searching the point sequence data for the lane boundary lines from the lane boundary lines. Fig. 8 shows an example of point coordinate values ​​for the three left and right white lines. As shown in Fig. 8, lines iw0, iw1, and iw2 are selected. This allows the points for the left and right white lines for each line to be obtained.

[0035] The lane width estimation unit 122 calculates the lane width from the coordinate values ​​of each point on three lines selected from the lane boundary line using the following equation (5) based on the coordinate values ​​of each point, camera height, and pixel resolution under the assumption that the lane width and longitudinal curvature are constant.

number

[0036] Also, iw k : lane width in 2D camera coordinates [pixels], iy k : Lane width [pixels] on the camera coordinate system 2D. The assumption of constant lane width means that the lane width on the same lane in the road image is constant. The assumption of constant longitudinal curvature means that the curvature of the road surface on the longitudinal cross section of the same lane in the road image is constant. The derivation of lane width estimation under the assumption of constant longitudinal curvature will be described later. As will be shown in the derivation below, lane width estimation can be adapted to any horizontal linear road shape.

[0037] Next, the operation of the computing device 100 according to the embodiment of the present disclosure will be described. Fig. 9 is a flowchart showing the flow of processing by the computing device 100 according to the first embodiment. In this processing, a series of steps are performed by the CPU 11 reading out a calculation program from the ROM 12 or the storage 14, expanding it in the RAM 13, and executing it (the same applies to the following embodiments).

[0038] In step S100, the CPU 11 acquires an image of the road in the traveling direction of the vehicle, taken by an on-board camera.

[0039] In step S102, the CPU 11 extracts lane boundary lines on the road surface from the road image as a sequence of points and outputs image coordinate values.

[0040] In step S104, the CPU 11 horizontally searches the point sequence data of the lane boundary line from the lane boundary line to select points for three lines.

[0041] In step S106, the CPU 11 calculates the lane width from the coordinate values ​​of each point on three lines selected from the lane boundary line, under the assumption that the lane width is constant and the longitudinal curvature is constant, using equation (5) based on the coordinate values ​​of each point, the camera height, and the pixel resolution.

[0042] FIG. 10 shows an example of lane width estimation using the method disclosed herein. FIG. 10 shows the results of calculating the lane width using Equation (1) for a curve with a bank angle as shown in FIG. 3, with the results of recalculating using Equation (5) under the same conditions as the calculation results in FIG. 4, based on the assumption of a constant longitudinal curvature. (c) shows the lane width estimated using the method disclosed herein, assuming a constant longitudinal curvature. While the error between the true value and the lane width increases with increasing bank angle in the conventional method, the method disclosed herein can be seen to estimate with relatively high accuracy. Note that the lane width is set to a constant value of 3.5 (m) when viewed from above, and is projected onto the curved road surface inclined by the bank angle. Therefore, the true lane width increases with bank angle. On the other hand, in the conventional method based on the local planar assumption, the local planar assumption breaks down on curved roads with a bank angle, causing the lane width to appear smaller. Therefore, the lane width decreases up to a bank angle of 15 degrees and increases at bank angles of 15 degrees or more. In this case, the observed white line is Z v The camera 2D coordinates were selected from those corresponding to 8.5 (m), 13.5 (m), and 18.5 (m).,From Fig. 10, for example, when the bank angle is 30 (deg), a lane width error of 0.5 (m) occurs when assuming a flat surface, but when assuming a constant longitudinal curvature, it can be seen that the lane width error is reduced to 0.03 (m).

[0043] (Derivation of lane width estimation based on local plane assumption) The derivation of the lane width estimation formula under the assumption of a local plane will be explained. The calculation formula for the lane width when the white line dot image shown in Figure 2 is obtained will be derived. In this case, (ix r(l) ,iy): Coordinate value of the right (left) white line point on the camera coordinate system 2D [pixel], α r(l) : The tangent slope of the right (left) white line point on the camera coordinate system 2D.

[0044] The coordinate value of the right (left) white line point on the camera coordinate system 2D (ix r(l), iy) is expressed by the following equation. Because of the local plane assumption, iy is expressed by two variables, h and φ.

number

number

[0045] At this time, φ is the pitch angle, Z v : Distance of the white line point [m], g(Z v ):Z v The lateral position of the center of the driving lane (X v coordinate value) expressing any function [m], ix c ,iy c : The image coordinate value [pixel] of the image center.

[0046] Next, α r and α l is derived.

number

[0047] Equation (6) and equation (7) are Z v By partially differentiating with and substituting into equation (8), we obtain the following equation.

number

number

number

[0048] (Derivation of lane width estimation assuming constant longitudinal curvature) Next, we will explain how to derive the lane width estimation formula assuming a constant longitudinal curvature. r(l) , y) is expressed by the following equation (see Figure 6). Since the longitudinal curvature is assumed to be constant, iy is a constant, and h, φ, and the longitudinal curvature component c v0 It is expressed by three variables:

number

number

[0049] At this time, Z v,k :The depth distance (Z v Coordinate value (m)). ix r,k -ix l,k Calculate g(Z v ) and we obtain the following equation:

number

[0050] At this time, iw k : The lane width (pixels) on the k-th line in camera coordinates 2D. v,k When rearranged into an equation for, the following equation is obtained:

number

number

[0051] Subtracting iy0 from equation (16) and eliminating φ and iy0 gives us the following equation:

number

number

number

number

[0052] Solving equation (20) for lane width w gives us the following equation:

number

number

[0053] As described above, the calculation device according to this embodiment makes it possible to suppress errors and calculate the lane width with high accuracy.

[0054] [Second embodiment] Fig. 11 is a diagram showing the functional configuration of a computing device according to the second embodiment. As shown in Fig. 11, the computing device 200 includes an image acquisition unit 110, a lane boundary detection unit 112, an estimation unit 214, and a sensor data acquisition unit 230. The estimation unit 214 includes a selection unit 120, a longitudinal curvature estimation unit 220, and a lane width estimation unit 222. Note that components similar to those in the above-described embodiments are designated by the same reference numerals, and their description will be omitted.

[0055] The sensor data acquisition unit 230 acquires vehicle motion sensor data including measurement values ​​of the vehicle speed, yaw rate, and acceleration of the host vehicle, and outputs the data to the longitudinal curvature estimation unit 220. The acceleration includes longitudinal G and lateral G measured by a G sensor.

[0056] Next, the processing of each of the selection unit 120, longitudinal curvature estimation unit 220, and lane width estimation unit 222, which perform the predetermined calculation process of the estimation unit 214 of this embodiment, will be described.

[0057] 13 shows an example of coordinate values ​​of points on two white lines, one on the left and one on the right. The selection unit 120 of the second embodiment only needs to search for two lines.

[0058] The longitudinal curvature estimation unit 220 estimates longitudinal curvature components on the camera coordinate system from the vehicle motion sensor data acquired by the sensor data acquisition unit 230. Specifically, the longitudinal curvature estimation unit 220 estimates the bank angle ψ and gradient angle φ (pitch angle) from the measured values ​​of vehicle speed, yaw rate, and acceleration (longitudinal G, lateral G) in the vehicle motion sensor data. FIG. 12 is a diagram corresponding to the bank angle and gradient angle. The bank angle ψ is calculated using the following equation (23), and the gradient angle φ is calculated using the following equation (25).

number

number

number

[0059] where g is the weight acceleration (rad), v is the vehicle speed, ω is the yaw rate (rad) measured by the on-board sensor, and y s : Measured value of lateral acceleration (m / s2), y f :Measured longitudinal acceleration (m / s2).

[0060] The longitudinal curvature estimation unit 220 calculates the longitudinal curvature component c from the vehicle speed, yaw rate, and calculated bank angle ψ using the following equation: v0 Calculate.

number

[0061] The lane width estimation unit 222 estimates the lane width from the coordinate values ​​of the lane boundary for two lines of the lane boundary line and the estimated longitudinal curvature component, under the assumption of a constant lane width and a constant longitudinal curvature.

[0062] Specifically, the lane width estimation unit 222 calculates the coordinate values ​​of the points on two lines selected from the left and right lane boundary lines shown in FIG. 13 and the longitudinal curvature component c v0From this, under the assumptions of constant lane width and constant longitudinal curvature, the lane width w is calculated using the following equation. The two lines are iw0 and iw1. The lane width w is obtained as the solution to the following quadratic equation. A maximum of two solutions can be obtained for the quadratic equation, but an appropriate value can be selected using the known knowledge that the lane width is 2.5 to 3.5 m.

number

[0063] Next, the operation of the computing device 200 according to the embodiment of the present disclosure will be described. Fig. 14 is a flowchart showing the flow of processing by the computing device 200 according to the second embodiment. In the second embodiment, step S200 is executed after step S102.

[0064] In step S200, the CPU 11 horizontally searches the point sequence data of the lane boundary line from the lane boundary line to select points for two lines.

[0065] In step S202, the CPU 11 acquires vehicle motion sensor data including the measured values ​​of the speed, yaw rate, and acceleration of the host vehicle.

[0066] In step S204, the CPU 11 estimates the bank angle and the gradient angle from the vehicle motion sensor data.

[0067] In step S206, the CPU 11 calculates the longitudinal curvature component from the vehicle speed, yaw rate, and the calculated bank angle.

[0068] In step S208, the CPU 11 estimates the lane width from the coordinate values ​​of the lane boundary for two lines of the lane boundary line and the estimated longitudinal curvature component, under the assumptions of constant lane width and constant longitudinal curvature.

[0069] As described above, the calculation device according to this embodiment makes it possible to suppress errors and calculate the lane width with high accuracy.

[0070] [Third embodiment] Fig. 15 is a diagram showing the functional configuration of a computing device according to the third embodiment. As shown in Fig. 15, the computing device 300 includes an image acquisition unit 110, a lane boundary detection unit 112, an estimation unit 314, and a road surface height database 330. The estimation unit 314 includes a selection unit 120, a longitudinal curvature estimation unit 320, and a lane width estimation unit 222. Note that components similar to those in the above-described embodiments are designated by the same reference numerals, and their description will be omitted.

[0071] The road surface height database 330 stores road surface height data for the distance ahead of the vehicle, which is obtained by learning the road surface height using deep learning with map information, a distance measurement sensor, or an image as input.

[0072] The selection unit 120 selects points on two lines in the same manner as in the second embodiment.

[0073] The longitudinal curvature estimating unit 320 uses road surface height data acquired from the road surface height database 330 to estimate the longitudinal curvature component by fitting the points of the road surface height data with a quadratic function.

[0074] Next, the operation of the computing device 300 according to the embodiment of the present disclosure will be described. Fig. 16 is a flowchart showing the flow of processing by the computing device 300 according to the third embodiment. In the second embodiment, step S300 is executed after step S200.

[0075] In step S300, the CPU 11 acquires road surface height data from the road surface height database 330.

[0076] In step S302, the CPU 11 estimates the longitudinal curvature component by fitting the points of the road surface height data with a quadratic function using the road surface height data acquired from the road surface height database 330. After step S302, step S208 is executed.

[0077] As described above, the calculation device according to this embodiment uses road surface height data to suppress errors and enable accurate calculation of lane widths.

[0078] In the above embodiments, the processing performed by the CPU after reading the software (computation program) may be performed by various processors other than the CPU. Examples of processors in this case include programmable logic devices (PLDs) whose circuit configuration can be changed after fabrication, such as field-programmable gate arrays (FPGAs), and dedicated electrical circuits, such as application-specific integrated circuits (ASICs), which are processors with circuit configurations specifically designed to perform specific processing. Furthermore, the processing may be performed by one of these various processors, or by a combination of two or more processors of the same or different types (e.g., multiple FPGAs, or a combination of a CPU and an FPGA). Furthermore, the hardware structure of these various processors is, more specifically, an electrical circuit combining circuit elements such as semiconductor devices.

[0079] In the above embodiment, the calculation program for executing the process is pre-stored (installed) in a ROM or storage, but the present invention is not limited to this. The calculation program may be provided in a form recorded on a non-transitory recording medium such as a CD-ROM (Compact Disk Read Only Memory), a DVD-ROM (Digital Versatile Disk Read Only Memory), or a USB (Universal Serial Bus) memory. The calculation program may also be downloaded from an external device via a network. [Explanation of symbols]

[0080] 100, 200, 300 computing devices 110 Image acquisition unit 112 Lane boundary detection unit 114, 214, 314 Estimation part 120 Selection Section 122, 222 Lane width estimation section 220, 320 Longitudinal curvature estimation section 230 Sensor data acquisition unit 330 Road Surface Height Database

Claims

1. an image acquisition unit that acquires a road image in the traveling direction of the vehicle; a lane boundary detection unit that extracts lane boundary lines on the road surface from the road image as a sequence of points and outputs image coordinate values; an estimation unit that calculates a lane width by a predetermined calculation process using at least the image coordinate values ​​and assuming a constant lane width and a constant longitudinal curvature; A computing device comprising:

2. the estimation unit includes a selection unit and a lane width estimation unit; the selection unit horizontally searches for point sequence data of lane boundary lines from the lane boundary lines to select points for three lines; 2. The calculation device according to claim 1, wherein the lane width estimation unit calculates the lane width from the coordinate values ​​of each point on the three lines selected from the lane boundary line, under the assumption that the lane width is constant and the longitudinal curvature is constant, using a predetermined formula based on the coordinate values ​​of each point, the camera height, and the pixel resolution.

3. Further including a sensor data acquisition unit, the estimation unit includes a selection unit, a longitudinal curvature estimation unit, and a lane width estimation unit; the selection unit horizontally searches for point sequence data of the lane boundary line from the lane boundary line to select points for two lines; the sensor data acquisition unit acquires vehicle motion sensor data including measurement values ​​of a vehicle speed, a yaw rate, and an acceleration of the host vehicle; the longitudinal curvature estimation unit estimates a longitudinal curvature component on a camera coordinate system from the acquired vehicle motion sensor data; the lane width estimation unit estimates the lane width from the coordinate values ​​of the lane boundary lines for the two lines of the lane boundary line and the estimated longitudinal curvature component under the assumption of a constant lane width and a constant longitudinal curvature. The computing device of claim 1 .

4. Further including a road surface height database; the estimation unit includes a selection unit, a longitudinal curvature estimation unit, and a lane width estimation unit; The road surface height database stores road surface height data relative to the distance ahead of the vehicle, the road surface height being learned using deep learning with map information, a distance measurement sensor, or an image as input, the selection unit horizontally searches for point sequence data of the lane boundary line from the lane boundary line to select points for two lines; the longitudinal curvature estimation unit estimates a longitudinal curvature component by fitting points of the road surface height data with a quadratic function using the road surface height data acquired from the road surface height database; the lane width estimation unit estimates the lane width from the coordinate values ​​of the lane boundary lines for the two lines of the lane boundary line and the estimated longitudinal curvature component under the assumption of a constant lane width and a constant longitudinal curvature. The computing device of claim 1 .

5. The computer Obtaining road images in the direction of travel of the vehicle, extracting lane boundary lines on the road surface from the road image as a sequence of points and outputting image coordinate values; Calculating the lane width using at least the image coordinate values ​​through a predetermined calculation process, using the assumption of a constant lane width and a constant longitudinal curvature. Calculation method.

6. On the computer, Obtaining road images in the direction of travel of the vehicle, extracting lane boundary lines on the road surface from the road image as a sequence of points and outputting image coordinate values; Calculating the lane width using at least the image coordinate values ​​through a predetermined calculation process, using the assumption of a constant lane width and a constant longitudinal curvature. A calculation program that executes a process.

Images