Method for creating a three-dimensional road map
By combining an inertial measurement system, a GNSS antenna, and a camera device, vehicle orientation and image data are recorded to generate a high-precision 3D road map. This solves the problem of low accuracy in existing 3D road maps and improves the control and safety of autonomous vehicles.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- AVL LIST GMBH
- Filing Date
- 2025-01-16
- Publication Date
- 2026-07-31
AI Technical Summary
Existing technologies suffer from low accuracy, complex measurement, and inability to fully record the parameters required by autonomous vehicles when creating 3D road maps, which affects the control accuracy and safety of autonomous vehicles.
By using vehicles equipped with inertial measurement systems, GNSS antennas, and cameras, the distance between the lane edge and the tire support point is calculated by recording the vehicle's direction, position, and images. Combined with data synchronization and calibration technologies, a high-precision three-dimensional road map is generated.
It enables the creation of high-resolution 3D road maps in a simple and economical way, improving the control accuracy and safety of autonomous vehicles, and accurately identifying driving conditions and evaluating vehicle behavior.
Smart Images
Figure CN122497850A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for creating a three-dimensional road map, the three-dimensional road map created by this method, and a method for controlling an autonomous vehicle using this three-dimensional road map. Background Technology
[0002] This invention is based on known techniques for measuring roads. These techniques include manually measuring the road, driving over the road with a camera, or scanning the road using a laser.
[0003] The known solutions have drawbacks, namely that they are either not accurate enough, the measurement technology is particularly complex, or they cannot record all relevant parameters for applications such as autonomous driving.
[0004] The demand for high-resolution 3D road maps is growing, especially in the field of autonomous driving. High-resolution maps can take into account important safety factors during autonomous driving and help to accurately identify previously ambiguous driving situations. For vehicle control, high-resolution 3D road maps are also beneficial for better planning driving routes. Summary of the Invention
[0005] The technical problem to be solved by this invention is to overcome, at least partially, the aforementioned disadvantages in a cheap and simple manner. In particular, this invention aims to create three-dimensional road maps in a simple way.
[0006] Another technical problem to be solved by the present invention is to provide a three-dimensional road map that improves the control accuracy of autonomous vehicles.
[0007] Another technical problem to be solved by the present invention is to provide a three-dimensional road map for evaluating the driving behavior of vehicles.
[0008] Another technical problem to be solved by the present invention is to provide a method by means of which the accuracy of three-dimensional road maps can be ensured through verification.
[0009] Another technical problem to be solved by the present invention is to provide a three-dimensional road map that enables vehicles to be located.
[0010] The aforementioned technical problem is solved by a method having the features of claim 1, a road map having the features of claim 13, and a method having the features of claims 14 and 15. Other features and details of the invention will be apparent from the dependent claims, the specification, and the drawings. Herein, the features and details described relating to the method according to the invention naturally also apply to the three-dimensional road map according to the invention, and vice versa; therefore, the disclosures regarding various aspects of the invention are always mutually referenced or can always be mutually referenced.
[0011] According to the present invention, a method for creating a three-dimensional road map will be provided. This method is characterized by the following steps: a) Provide a vehicle, the vehicle including a measuring device, the measuring device including an inertial measurement system, a GNSS antenna and a camera device time-synchronized with the inertial measurement system and the GNSS antenna; b) While driving through the lane, the vehicle's orientation is recorded using an inertial measurement system, the vehicle's position is recorded using a GNSS antenna, and several images are recorded using a camera device, such that at least some of the images show the outer support point of the tires and the edge of the lane. c) Determine the distance between the outer support point of the tire and the lane edge point of the lane edge based on the vehicle direction value recorded simultaneously in step b) and one of the images; d) At this moment, calculate the absolute position of the lane edge point defined in step c) based on the vehicle position and the distance determined in step c); e) Repeat steps c) and d) for multiple consecutive absolute positions in time. f) Determine the first lane edge line of the lane by connecting or interpolating multiple successive absolute positions in time; and g) Repeat steps b) to e) for the second lane edge opposite to the first lane edge, and then... h) Determine the second lane edge line, which is opposite to the first lane edge, by connecting or interpolating the successive absolute positions of the second lane edge over multiple time intervals; and i) Output a 3D road map containing the edge lines of the first lane and the edge lines of the second lane.
[0012] The core concept of this invention is that lanes can be easily measured from vehicles that require camera devices, inertial measurement systems, and GNSS antennas for measurement, and in particular, lane edge lines can be measured with high accuracy and high reliability.
[0013] Vehicles equipped with inertial measurement systems (IMUs) and GNSS antennas are used to record map data. Therefore, these vehicles are reconnaissance vehicles. An inertial measurement system (IMU) comprises a combination of multiple inertial sensors, such as accelerometers and rotational speed sensors. To record the six possible kinematic degrees of freedom, the IMU includes three mutually orthogonal accelerometers (translation sensors) for detecting translational motion along the x, y, and z axes, and three mutually orthogonal rotational speed sensors (gyroscopes) for detecting rotational motion about the x, y, and z axes. Each IMU provides three linear acceleration values for translational motion and three angular velocity values for rotational speed as measurements. Within the inertial navigation system (INS), the linear velocity is determined by integration after compensating for gravitational acceleration, based on the linear acceleration measurements from the inertial measurement system, and the spatial position relative to a reference point is determined by further integration. The integration of the three angular velocities provides the spatial orientation relative to the reference point. An additional magnetometer can be integrated to determine the integration constant, improve accuracy, and correct for zero-point drift and long-term drift of the aforementioned sensors.
[0014] GNSS antennas are used to determine the three-dimensional position of vehicles on Earth. GNSS is a collective term for the use of existing and future global satellite systems, such as NAVSTAR GPS, GLONASS, Galileo, and BeiDou.
[0015] This method can be implemented specifically using a computer.
[0016] The measuring device may have one or more cameras, particularly three cameras, each recording video and / or sequentially recording single frames of images at short time intervals for performing the method. Preferably, the sampling rate is at least 20 frames per second, particularly preferably at least 30 frames per second, and most preferably at least 30 frames per second. However, the method can certainly be performed at significantly lower sampling rates. Preferably, the images recorded in step b) show the road at least every 2 meters. The cameras are time-synchronized with each other, with the inertial measurement system, and with the GNSS antenna, respectively. However, different systems typically have different sampling rates. For example, the GNSS antenna may have a sampling rate of 100 Hz. The camera may record several images at, for example, 20 frames per second (20 Hz) during the execution of the method. The inertial measurement system may record data at, for example, 30 Hz. Therefore, the times of the recorded images and measurements are not usually perfectly synchronized. To improve accuracy, a data synchronization step can be performed. For this purpose, for each individual image, the data points of the GNSS antenna and the data points of the inertial measurement system that match the recording time of the individual image can be calculated. The calculation can be achieved through interpolation of the measurement data, specifically through linear interpolation of the measurements taken by the GNSS antenna and inertial measurement system near the measurement time. The data synchronization step performed in this manner synchronizes the data from the GNSS antenna and inertial measurement system with the image data, i.e., generates data whose timestamps are consistent with the timestamps of each camera. For this purpose, each individual camera preferably has a synchronization channel with the following parameters: image recording time, latitude of the image location, longitude of the image location, elevation angle of the image location, and orientation of the image location.
[0017] Preferably, cameras are positioned on the left and / or right sides of the vehicle, and in the case of two to three cameras, cameras may also be positioned at the center of the vehicle if necessary. The image areas of the cameras or multiple cameras are selected such that the left front wheel and / or right front wheel, and if necessary, the front of the vehicle, are within the entire image area. Using the camera arrangement in conjunction with prior camera calibration, and by employing developed distance calculation logic, the relative, time-synchronized distances, especially those between the wheel bearing points (left and right sides) and / or the outer bearing points of the tires, as well as the relative, time-synchronized distances from the vehicle's center plane along the longitudinal direction to the lane edge via a known or determined intersection point of the plane passing through the vehicle's front axle and the road plane, can also be determined. The wheel bearing points are offset relative to the outer bearing points of the tires by half a tire width. This distance can also be zero, particularly when crossing lane edge lines.
[0018] Preferably, in step c), the calibrated zero point can also be used to determine the distance between the tire's outer support point and the lane edge point. The calibrated zero point is a fixed point with a known location, specifically located on the calibration strip. The calibrated zero point can be used to improve calibration accuracy.
[0019] This method enables the precise geographical location of a vehicle at each image recording moment. To this end, in addition to the recording time and image number, the longitude, latitude, altitude, and accuracy of the GNSS measurements can also be stored. Consistent and synchronized measurement data is obtained through the methodological steps used to record the measurement data, which can be processed automatically.
[0020] Steps c) and d) can be performed as follows: at each image moment of the measured video, the precise geographical location of the vehicle is determined, and the outer support point of the tire is determined by measuring the known three-dimensional offset between the GNSS antenna mounting point and the outer support point of the tire in a Cartesian coordinate system. For this purpose, in addition to storing the recording time and image number, the longitude, latitude, altitude, and accuracy of the GNSS measurement can also be recorded. This yields consistent and synchronized measurement data, which can be processed automatically. If the method involves multiple tires and / or multiple outer support points of the tires, their respective offsets should be considered separately.
[0021] Preferably, a synchronized measurement file of the image position can be stored for each video file.
[0022] Determining the first and / or second lane edges by connecting or interpolating their sequential absolute positions in time can be specifically performed as follows: First, starting from the image recording time, a nearest neighbor search is performed from one lane edge point to the next lane edge point. Then, the time difference to the next lane edge point is determined from the image recording time, and the minimum value is selected. Finally, a second nearest neighbor search is performed for the second nearest lane edge point starting from the image recording time. Each lane edge line is determined by connecting or interpolating the lane edge points. Images of the first and second lane edges can be recorded sequentially or simultaneously if a suitable camera setup is available. The lane edge points thus correspond to their recording times and are then connected or interpolated in this order.
[0023] Advantageously, the recording in step b) is performed at time-synchronized intervals.
[0024] In this way, the validity of the recorded data can be verified by easily identifying missing or extra data. Furthermore, this ensures the uniform accuracy of lane edge lines in the 3D road map, especially in step b) when driving through lanes at a substantially constant speed.
[0025] A further advantage is that the camera device has a first camera on the left side of the vehicle and a second camera on the right side of the vehicle, wherein the first camera is oriented so that its field of view includes the outer support point of the left tire, and the second camera is oriented so that its field of view includes the outer support point of the right tire. In particular, the camera device may also have a third camera whose field of view shows the area in front of the vehicle and the lane.
[0026] Using two cameras oriented towards different tires allows for the parallel (i.e., simultaneous or nearly simultaneous) recording of images of the first and second edges. A third, central camera facilitates data evaluation, as it generates a complete road image. Calibration methods can also be applied here to synchronize all relevant data.
[0027] A further advantage is that the method includes step a1) after step a). a1) Use the calibration device placed on the ground of the lane to calibrate the camera equipment.
[0028] For example, calibration can be performed on the area to be measured on the ground of the lane using a calibration tape with centimeter graduations laid out on the lane. To do this, the following calibration steps are performed on an image of the calibration tape recorded by a camera device and displayed: • The calibration strip is vertically oriented using a laser or laser measuring device. • Set calibration points on the calibration tape • Store calibration files • Verify calibration via distance measurement First, set calibration points on the centimeter scale. Then, calibrate along the lines of the image. Here, the calibration points are set on the calibration lines in the image, within the crosshairs on the calibration band. The first calibration point is the zero point of calibration, and the subsequent calibration points are set manually and / or automatically in ascending order. This determines the ratio of image pixels to the actual distance scale.
[0029] For example, if there are 69 pixels within a 1 cm distance segment in the outer edge region, the measurement accuracy can reach 0.15 cm.
[0030] The calibration accuracy can be particularly preferably performed before and after the measurements (i.e., steps b) and c), to ensure that the distance scale is correct. This step improves the accuracy of the road map.
[0031] Preferably, the calibration device may include a scale device with calibration marks. Instead of a simple centimeter scale as the calibration device, a measurement field may be provided, in which the scale device is arranged in two dimensions on the lane surface.
[0032] A further advantage is that in step c), the lane edge is defined using an edge recognition algorithm.
[0033] Edge detection algorithms are used to separate areas of an image that differ sufficiently in color or grayscale value, brightness, or texture along a straight line or curve. These algorithms identify transitions between areas and mark them as edges. However, individual uniform regions should be identified as separate regions, not divided into two faces by edges. To do this, the color gradient of each individual pixel in the image can be calculated by examining the region surrounding that point. This process is achieved through discrete convolution of the image with a convolution matrix (i.e., an edge operator). The edge operator defines the size of the region to be examined and the weights of each pixel within that region in the calculation. The edge operator determines the average gradient value for the center pixel in the environment. If this is done for all pixels in the image, an edge image can be constructed from the resulting gradient matrix. In this image, edges between uniform regions stand out because the color value gradients at these points are relatively large. Other methods for edge detection can also be employed.
[0034] A further advantage of the method is that it also includes the following steps: Calculate the geographic center point of the absolute position of the first lane edge line, and the geographic center point of the absolute position of the second lane edge line opposite to the absolute position of the first lane edge line; and The lane centerline is determined by connecting or interpolating the successive center positions over time.
[0035] Lane centerlines are another important aspect of 3D road maps, improving their application in autonomous driving. With lane centerlines present, autonomous vehicles can more easily identify the start of deviations from their ideal driving path or overtaking maneuvers. The centerline is calculated from the edges of the left and right lanes. The centerline of each lane serves as a reference and is preferably interpolated at a constant distance scale. The resolution of the centerline can vary arbitrarily, typically depending on lane characteristics.
[0036] Further advantageously, the above method also includes: calculating the lane segment as a polygon, the polygon consisting of two consecutive center positions and two absolute positions of a first lane edge and a second lane edge used to calculate the two center positions, wherein a lane is defined as the sum of lane segments.
[0037] Therefore, a lane segment consists of a total of six support points. These support points include two successive GNSS points on the left and right lane edges and the lane centerline. This geographic information can be integrated into a 3D road map as a geometric model. With this information, the 3D road map can specifically include the following elements: • Polygonal lines along the left lane edge • Polygonal lines along the edge of the right lane • Polygonal lines of lane centerline • Lane edge point on the left side of the lane • Lane edge point on the right side of the lane • Multi-point lane centerline • The polygon of the lane segment • Polygons along the lane boundaries Lane segments also allow for more precise control and / or more accurate verification of the vehicle's behavior. For example, if the vehicle accesses a 3D road map containing polygonal segments, deviations from the expected driving behavior of other road users can be more easily identified. In particular, abnormal braking and acceleration behaviors can be detected more easily and accurately.
[0038] Therefore, each lane has three lines: the left lane edge line, the right lane edge line, and the lane center line. For these three lines, parameters can be calculated for each segment, specifically as listed below: • Lane length • Lane width • Lane markings • Lane line colors • Lane direction • Lane curvature • Lane gradient • Lane lateral slope • Lane line quality Lane line type, lane line color, and lane line quality can be specifically derived from the image or edge recognition algorithm. The remaining parameters are calculated as shown in the table below.
[0039] Lane length m Point-to-point distance Length along the lane Lane width m Point-to-point distance width along the lane Lane direction deg Angle between two points direction of the lane lane curvature 1 / m The angle change between the three points in 1-meter segments lane curvature lane slope % derivative of height lane slope Lane lateral slope % Height difference divided by lane width Lateral slope of the lane
[0040] Furthermore, this method also includes the following steps: The area between the first and second lane edge lines is calculated as the lane center region, which is spaced a certain distance from both lane edge lines. The lane center region also allows for more precise control of autonomous vehicles. For example, if the vehicle accesses a 3D road map that includes the lane center region, deviations from the expected driving behavior of other road users can be identified more easily. Abnormal steering behavior can be detected more easily and accurately. Therefore, lane keeping assist responds faster and more safely in abnormal driving situations.
[0041] Furthermore, this method also includes the following steps: The difference between the surface stretched by the first lane edge line and the second lane edge line and the lane center area is calculated as the lane edge area.
[0042] This generates a 3D map that defines edge and center regions. This further enhances the advantage gained by navigating the center region of the lane.
[0043] Furthermore, this method also includes the step of associating static or dynamic lane parameters with the lane.
[0044] Static lane parameters can specifically be defined as: lane direction, lane width, lane slope, lane lateral inclination, lane length, curvature of the left lane edge line, curvature of the right lane edge line and / or curvature of the lane center line.
[0045] Dynamic lane parameters can specifically be defined as the relative distances of a vehicle to the lane edges. This includes the minimum distances from the left lane edge, right lane edge, and lane centerline, the corresponding current distances, the corresponding average distances, and the corresponding perpendicular distances. The angle between the driving direction and the lane centerline, or between the driving direction and one of the lane edge lines, can also be defined as dynamic lane parameters. Similarly, the current angle, average angle, or maximum angle can also be defined as dynamic lane parameters.
[0046] Furthermore, this method also includes the following steps: Provide the three-dimensional road map obtained above. The vehicle repeatedly crossed the edge of the lane, and in this... The vehicle's orientation is recorded using an inertial measurement system, the vehicle's position is recorded using a GNSS antenna, and several images are recorded using a camera, such that at least some of the images show the tire's outer support point on the lane edge. The lane edge point is determined with optimized accuracy based on the vehicle direction value recorded simultaneously and the image showing the outer support point of the tire on the lane edge; and Verify and / or calibrate 3D road maps using lane edge points with optimized accuracy.
[0047] High-resolution map verification provides the accuracy of the measured route and, therefore, the accuracy of the previously created 3D road map. For this purpose, a vehicle is used as the verification vehicle. Verification is performed with high-precision vehicle position detection, for example, with an accuracy of less than ±5 cm, preferably less than ±3 cm, and more preferably less than ±1.5 cm. By making the known outer support point of the tire relative to the GNSS antenna also the measurement point, the inherent measurement error of the camera when recording distance is reduced. Here, the time measurement grid should be chosen to be as small as possible. In particular, the image recording interval should be 50 ms or less, preferably 20 ms or less, and more preferably 10 ms or less. For each method, image recording can also refer to video recording.
[0048] The vehicle is deliberately driven across the lane edge. The point of crossing is determined from an image, or preferably from video, and compared with the location determination result. From this, the relative distance between the lane edge line of the previously output 3D road map and the now calculated lane edge line value can be calculated. If a lane centerline marker exists, the lane centerline can also be verified.
[0049] According to a second aspect, the present invention provides a three-dimensional road map obtained by the aforementioned method.
[0050] The three-dimensional road map here includes at least the first and second lane edge lines in a three-dimensional coordinate system and a high precision preferably ±3 cm or lower, more preferably ±1.5 cm or lower, and may further include lane center lines and / or lane segments and / or lane center areas and / or lane edge areas.
[0051] According to a third aspect, the present invention provides a method for controlling autonomous vehicles using a three-dimensional road map.
[0052] Therefore, a three-dimensional road map can be integrated into the vehicle's control system, or the vehicle can access the three-dimensional road map via a communication network for control purposes. Lane edge lines and / or lane center lines and / or lane segments and / or lane center areas and / or lane edge areas can be used to perform driving operations to assess the driving behavior of other road users and match the control of the vehicle accordingly. According to a third aspect of the invention, in addition to accessing sensor data relating to the current driving situation, the autonomous vehicle can also use data integrated in the three-dimensional road map to assess road conditions and driving situations, and thus improve its own driving behavior.
[0053] According to a fourth aspect, the present invention provides a method for identifying lane edge lines using the aforementioned three-dimensional road map, particularly during motorsports, the method comprising the following steps: Provide 3D road maps, especially the 3D road maps described above; Enables detection vehicles equipped with GNSS devices to move within road areas covered by three-dimensional road maps; Vehicle position was recorded using GNSS devices; Compare the vehicle's position with the coordinates of the edge lines of the first and / or second lanes; and Crossing of lane edge lines is identified through comparison.
[0054] For this purpose, the detection vehicle preferably also has a GNSS measuring device, and its dimensions (i.e., its maximum extent) are preferably known to the GNSS measuring device. A vehicle trajectory is formed by recording the vehicle's position over a period of time. The maximum vehicle size can be added as an offset. In this way, even if the detection vehicle (e.g., a vehicle participating in a race) crosses the lane edge line with only one or part of its tires, it can be detected. Attached Figure Description
[0055] Other advantages, features, and details of the invention will become apparent from the following description, wherein embodiments of the invention are described in detail with reference to the accompanying drawings. The drawings are as follows: Figure 1 A vehicle with a measuring device for performing the method according to the invention is shown schematically. Figure 2 A calibration device is schematically shown, which carries recorded measurement values for calibrating a measuring device. Figure 3 This schematically illustrates a road measurement method used to measure the distance between the tire's outer support point and the lane edge point. image, Figure 4 A schematic diagram of a 3D road map showing two lane edge lines and a lane center line. Figure 5 A schematic diagram of a 3D road map showing two lane edge lines, lane center lines, and lane segments. Figure 6 A schematic diagram of a 3D road map showing two lane edge lines, a lane center area, and lane edge areas. Figure 7 A flowchart illustrating the method according to the present invention is shown schematically. Figure 8 An alternative flowchart to the method according to the present invention is shown schematically. Detailed Implementation
[0056] Figure 1A vehicle 10 with a measuring device 12 is schematically shown, which is adapted to perform a method for creating a three-dimensional road map according to the invention. The measuring device 12 includes a camera device 13 with three time-synchronized cameras 14. A first camera 14 is arranged on the left side of the vehicle 10, and a second camera 14 is arranged on the right side of the vehicle 10. The first and second cameras 14 can be arranged, in particular, in the area between the B-pillar and the front wheel plane 19 of the vehicle 10. The front wheel plane 19 is perpendicular to the center plane 15 of the vehicle and parallel to the wheel axis. Particularly preferably, the cameras 14 can be arranged perpendicularly above the tires 16, 18 of the vehicle 10. In this arrangement, the field of view of the cameras 14 is parallel to the axial direction, so that perspective distortion of the image occurs only in one direction and is therefore minimized, thereby improving the accuracy of the method. The first camera 14 is oriented so that its field of view includes the outer support point of the left tire 16, and the second camera is oriented so that its field of view includes the outer support point of the right tire 18. The cameras 14 are configured as video cameras to achieve a sufficiently high image recording speed.
[0057] The measuring device 12 also includes a GNSS antenna 20 that operates in time synchronization with the camera device 13. The GNSS antenna includes a transmitting unit, a receiving unit, and an evaluation unit (not shown) for evaluating position data. Time synchronization means that the position data recorded by the GNSS antenna 20 is timestamped, and the images recorded by the camera device 13 are also timestamped, with the timestamps matching when the recording times are the same. This allows for correlation between position data and images, and synchronization of images and data. The time-synchronized inertial measurement system is not explicitly shown in the figure. The third camera 14 is located at the center of the vehicle, with its field of view oriented forward, so that during operation, its field of view shows the area in front of the vehicle and the lane. The third camera can thus simplify evaluation.
[0058] The position of the GNSS antenna relative to the tire's external support point is known.
[0059] The vehicle is positioned on calibration device 22. Calibration device 22 is configured as calibration belt 24 and has scales arranged transversely to the vehicle direction along the axial direction below the front tires. In this position, camera device 13 can be calibrated. Laser measuring device 21 is used to precisely align calibration belt 24 vertically, such that laser measuring device 21, positioned at one end of calibration belt 24, can emit a laser beam 17, thereby precisely aligning calibration belt 24 parallel to the front wheel plane 19 and perpendicular to the vehicle center plane 15. Since the precise alignment of calibration belt 24 affects the calibration of measuring device 12, the accuracy of the calibration process is improved by laser measuring device 21.
[0060] Figure 2 A calibration band 24 is schematically shown. The calibration band 24 has calibration points 26 marked with circles along a line. To calibrate the camera 14, the calibration points are associated with the pixels of the camera image on the photograph taken by the camera 14. This association can be done manually or automatically. For this purpose, the calibration band 24 is preferably arranged so that its distance scale runs along the pixel line of the camera 14. Calibration can be performed by software that interprets mouse clicks as setting calibration points. The calibration points can then be used to determine the ratio of pixel distance to the actual distance scale.
[0061] In one example measurement, there are 69 pixels within a 1 cm distance segment inside the outer edge of the image. In this example, an accuracy of 0.015 cm can therefore be measured.
[0062] exist Figure 3 The image shows road measurements recorded while driving through the lane. The outer support point of the right front tire 14 is shown in the figure.
[0063] The distance between the outer support point 32 of tire 14 and the lane edge point 34 of lane edge 36 is measured. In this example, the lane edge is defined by the lower edge of the shoulder 38 facing lane 30. In other cases, lane edge 36 may be defined by lane markings or otherwise. Lane edge 36 may be determined and / or defined using an edge recognition algorithm. Errors in edge detection performed by the machine may be optionally compensated for to achieve higher accuracy, and errors may be eliminated in particular using human logic. This may compensate for, for example, effects such as difficulty in recognition due to sunlight, shadows, leaves, sand or other objects obscuring lane markings, or the absence of lane markings.
[0064] To determine the lane edge point 34, the distance 42 between the outer support point 32 of the tire 14 and the lane edge point 34 of the lane edge 36 is determined based on the measurement direction values and images recorded at the same time in step b). The known offset between the GNSS antenna 20 and the outer support point 32 of the tire 14 is considered here.
[0065] Distance 42 is determined based on a line 40 connecting the outer support point 32 of tire 14 and the lane edge point 34, the line corresponding to the pixel line orientation of the camera used during calibration. Then, the absolute position of the lane edge point in the GNSS coordinate system at this moment is calculated based on the vehicle direction and distance 42.
[0066] This step is repeated for both lane edges. For this purpose, vehicle 10 may be equipped with only one camera 14, which, for example, records only the right tire 14. In this case, the vehicle travels through lane 30 twice in different directions, so that the absolute positions of the two lane edges 36 can be recorded.
[0067] If the camera device has two cameras 14 arranged on different sides of the vehicle 10, then the images of the edges 36 of the two lanes can be recorded simultaneously only once the vehicle passes through the lane 30.
[0068] The lane edge lines of the first and second lane edges were determined by separately connecting or interpolating the absolute positions of multiple temporally successive first and second lane edges 36.
[0069] exist Figure 4 The diagram illustrates a three-dimensional road map 44, which includes a first lane edge line 46, a second lane edge line 48, and a lane center line 50. The first lane edge line 46 and the second lane edge line 48 are determined by interpolation of their respective absolute positions 52 and 54, and are therefore represented in a three-dimensional GNSS coordinate system.
[0070] The lane centerline 50 is calculated from the two lane edge lines 46, 48 and / or their absolute positions 52, 54. The lane centerline 50 serves as a reference line and is interpolated at a constant distance. In principle, the resolution of the lane centerline 50 can vary arbitrarily, particularly depending on the lane characteristics. In the example shown, the lane centerline 50 is determined as follows: first, the geographic midpoint between the absolute position 52 of the first lane edge line 44 and the absolute position 54 of the second lane edge line 48, which is opposite to the absolute position 52 of the first lane edge line 44, is calculated as the center position 56. The lane centerline 50 can then be determined by connecting or interpolating the successive center positions 56 over time.
[0071] Figure 5 A three-dimensional road map 44 with lane segments 58 is shown. According to the invention, lane segments 58 are defined as polygons consisting of two consecutive center positions 56 and two absolute positions 52 for calculating the two center positions 56, namely a first lane edge line 46 and a second lane edge line 48. A lane is defined as the sum of all lane segments 58. Therefore, the sum of all lane segments 58 forms a surface in three-dimensional space. This surface, and therefore the lane 60, can be further smoothed by three-dimensional interpolation between the lane segments 58 and / or the absolute positions 52, 54 and the center position 56. Thus, lane segments 58 consist of a total of six support points. These support points include two consecutive GNSS points at the left and right lane edges, and the center of lane 60.
[0072] Figure 6 A three-dimensional road map 44 is shown, which includes two lane edge lines 46 and 48, a lane center area 62, and a lane edge area 64. The lane edge area 64 consists of two spatially separated regions.
[0073] The lane center region 62 is the area located between the first lane edge line 46 and the second lane edge line 48, and is separated from these two edge lines by an interval distance. The lane center region 62 can be calculated using simple geometric considerations. The lane edge region 64 can be calculated as the difference between the surface spanned by the first lane edge line 46 and the second lane edge line 48 and the lane center region 62.
[0074] Figure 7 A flowchart of the method according to the present invention is shown.
[0075] In step a), a vehicle is provided, the vehicle including a measuring device comprising an inertial measurement system, a GNSS antenna, and a camera device time-synchronized with the inertial measurement system and the GNSS antenna.
[0076] In step b), while the vehicle is traveling through the lane, the vehicle's orientation is recorded using an inertial measurement system, its position is recorded using a GNSS antenna, and several images are recorded using a camera device. At least some of these images show the outer support points of the tires and the lane edges.
[0077] In subsequent step c), the distance between the outer support point of the tire and the lane edge point is determined. This determination is made using one of the vehicle orientation values and images recorded simultaneously in step b).
[0078] In step d), the absolute position of the lane edge point defined in step c) is calculated. This calculation is performed using the vehicle position at that moment and the distance determined in step c).
[0079] Step e) is a repetitive step. Here, steps c) and d) are repeated for multiple consecutive absolute positions in time. The number of repetitions depends on the length of the lane to be traversed or the total measurement time of the measuring unit. In principle, the number of repetitions n can be freely chosen.
[0080] In step f), the edge of the first lane is determined. This determination is performed by connecting or interpolating multiple successive absolute positions over time.
[0081] Step g) is performed in parallel with, after, or before step f), wherein steps b) through e) are repeated for the second lane edge opposite the first lane edge. Here this is a single repetition; if the camera device only records images on one side of the vehicle, this step is usually performed separately.
[0082] Then, in step h), the edge line of the second lane, opposite to the edge of the first lane, is determined. This determination is performed by connecting or interpolating the successive absolute positions of the second lane edge over multiple time intervals.
[0083] Finally, in step i), the output is a three-dimensional road map that includes at least the edge lines of the first lane and the edge lines of the second lane.
[0084] exist Figure 8 An alternative method according to the invention is shown, the method comprising: Figure 7 All the steps described in [the document]. (and / with) Figure 7 The method described in the text differs from the method described elsewhere. Steps b), c), d), and e) are performed in parallel for the first and second lane edges to obtain data for determining the first and second lane edge lines. These steps can be performed in parallel if the vehicle's camera setup is configured to record images on both sides of the vehicle simultaneously, for example, by using two or more cameras.
[0085] The above explanation of the embodiments describes the present invention by way of example only.
[0086] List of reference numerals 10 vehicles 12 Measuring devices 13 Camera device 14 cameras 15 Vehicle center plane 16. Left tire 17 Laser beams 18 Right tire 19 Front wheel plane 20 GNSS antenna 21 Laser measuring device 22 Calibration Device 24 Calibration Tape 26 calibration points 30 lanes 32 external support points 34 Lane edge points 36 lane edge 38 curb stone 40 connecting cable 42 Distance 44. 3D Road Map 46 First lane edge line 48 Second lane edge line 50 lane center line 52. Absolute position of the edge line of the first lane 54. Absolute position of the edge line of the second lane 56. Central position 58-lane section 60 lanes 62 lane center area Lane 64 edge area
Claims
1. A method (44) for creating a three-dimensional road map, the method comprising the following steps: a) Provide a vehicle (10) including a measuring device (12) including an inertial measurement system, a GNSS antenna (20) and a camera device (13) time-synchronized with the inertial measurement system and the GNSS antenna (20). b) During the passage through the lane (60), the vehicle orientation is recorded using the inertial measurement system, the vehicle position is recorded using the GNSS antenna (20), and several images are recorded using the camera device (13) such that at least some of the images show the outer support point (32) of the tires (16, 18) and the lane edge (36). c) Determine the distance (42) between the outer support point (32) of the tire (16, 18) and the lane edge point (34) of the lane edge (36) based on the vehicle direction value recorded simultaneously in step b) and one of the images. d) At this moment, calculate the absolute position (52) of the lane edge point (34) defined in step c) based on the vehicle position and the distance (42) determined in step c). e) Repeat steps c) and d) for multiple temporally successive absolute positions (52). f) Determine the first lane edge line (46) of the lane by connecting or interpolating multiple temporally successive absolute positions (52); and g) Repeat steps b) to e) for the second lane edge (36) opposite to the first lane edge (36), and then h) Determine the second lane edge line (48) of lane (60) opposite to the first lane edge (36) by connecting or interpolating multiple temporally successive absolute positions (54) of the second lane edge (36); and i) Output a three-dimensional road map (44) containing the first lane edge line (46) and the second lane edge line (48).
2. The method of claim 1, wherein, The recording in step b) is performed at time-synchronized intervals.
3. The method according to any of the preceding claims, wherein, The camera device (13) has a first camera (14) located on the left side of the vehicle (10) and a second camera (14) located on the right side of the vehicle (10), wherein the first camera (14) is oriented such that the field of view of the first camera includes the outer support point (32) of the left tire (16), and the second camera (14) is oriented such that the field of view of the second camera includes the outer support point (32) of the right tire (18).
4. The method according to any one of the preceding claims, the method further comprising step a1) after step a): a1) calibrating the camera device (13) with a calibration device (22) arranged on the road surface.
5. The method of claim 4, wherein, The calibration device (22) includes a scale device with calibration marks on it.
6. The method according to any one of the preceding claims, wherein, In step c), the lane edge is defined using an edge recognition algorithm (36).
7. The method according to any one of the preceding claims, the method further comprising the following steps: Calculate the center position (56) by taking the geographic midpoint between the absolute position (52) of the first lane edge line (46) and the absolute position (54) of the second lane edge line (48) opposite the first lane edge (36) as the center position (56); and The lane centerline (50) is determined by connecting or interpolating the successive center positions (56) in time.
8. The method of claim 7, further comprising the step of: The lane segment (58) is calculated as a polygon, which consists of two consecutive center positions (56) and two absolute positions (52, 54) of the first lane edge line (46) and the second lane edge line (48) used to calculate the center position (56), wherein the lane (60) is defined as the sum of the lane segments (58).
9. The method according to any one of the preceding claims, the method further comprising the following steps: calculating a lane center region (62) as a region located between the first lane edge line (46) and the second lane edge line (48), the center region (62) being spaced apart from the first lane edge line (46) and the second lane edge line (48) by an interval distance value.
10. The method of claim 9, further comprising the step of: The lane edge region (64) is calculated as the difference between the surface spanned by the first lane edge line (46) and the second lane edge line (48) and the lane center region (62).
11. The method according to any one of the preceding claims, further comprising the step of: Associate static or dynamic lane parameters with lane (60).
12. The method according to any one of the preceding claims, the method further comprising the following steps: Provide a three-dimensional road map (44), particularly a three-dimensional road map obtained according to any one of the preceding claims; The vehicle (10) is caused to cross the lane edge (36) multiple times, and in this The vehicle orientation is recorded by the inertial measurement system, the vehicle position is recorded by the GNSS antenna (20), and several images are recorded by the camera device (13) such that at least some of the images show the outer support point (32) of the tires (16, 18) on the edge of the lane (36); Based on the vehicle direction values recorded simultaneously and the image showing the outer support point (32) of the tires (16, 18) on the lane edge (36), a lane edge point (34) with optimized accuracy is determined; and The lane edge points (34) are used to verify and / or calibrate the three-dimensional road map (44) with optimized accuracy.
13. A three-dimensional road map (44) obtained by the method according to any one of the preceding claims.
14. An application of the three-dimensional road map (44) according to claim 13 for controlling an autonomous vehicle.
15. An application of the three-dimensional road map (44) according to claim 13 for identifying lane lines (46, 48) being crossed, particularly in motorsports, comprising the following steps: Provide a three-dimensional road map (44) according to claim 13; The detection vehicle equipped with a GNSS device moves over the road area covered by the three-dimensional road map (44); The GNSS device is used to record the vehicle's location; The vehicle position is compared with the coordinates of the first lane edge line (46) and / or the second lane edge line (48); and The comparison identifies the crossing of the lane edge lines (46, 48).