Determining the course of a lane marking

Abstract

Method (200) for determining the course of a lane boundary (122) along a roadway (115), wherein the lane boundary (122) is characterized by a lane marking (125), the method (200) comprising the following steps: - recording series of observations (130) of the lane marking (125), wherein the series are made by different observers (110) moving along the roadway (115); - determining planes (210) extending perpendicularly to a course (120) of the roadway (115); - determining points of intersection (215) of the planes (210) with each series of observations (130); - summarizing corresponding points of intersection (215) of a plane (210); and - determining the course of the lane boundary (122) based on corresponding summarized points of intersection (220) of different planes (210);characterized in that observations (130) of a lane marking (125) by an observer (110) at different locations along the roadway (115) are each assigned a constant identifier; and that aggregated intersection points (220) of different planes (210) are assigned to each other on the basis of observations (130) with the same identifiers.

Description

[0001] The invention relates to the creation of a highly accurate digital map. In particular, the invention relates to the determination of the course of a lane boundary on a road for vehicles.

[0002] To partially or fully automate the vehicle's longitudinal and / or lateral steering, it is advantageous to have a high-precision geographic map of the surroundings. A standard-definition (SD) map, which can be used, for example, to guide the vehicle to a predetermined destination, typically has an accuracy of approximately one to ten meters. A high-definition (HD) map should typically deviate from reality by less than approximately one meter. HD and SD maps can differ further, for example, in the information they contain.

[0003] If the vehicle is traveling on a multi-lane road, for example, the high-precision map should include lane boundaries to allow differentiation of which lane the vehicle is in. Visual indicators, such as lane markings, are typically used to determine lane boundaries. A lane marking usually consists of a line painted directly onto the road surface.

[0004] Creating a highly accurate map using a survey vehicle is complex and time-consuming. It was therefore proposed to create a highly accurate map based on observations made by a fleet of vehicles already using the road. However, correctly matching observations of lane markings made by different vehicles has proven difficult. In particular, spatially dense lane markings, such as those found in construction zones where new and old lane markings intersect, can be problematic.

[0005] One of the problems underlying the present invention is to provide an improved technique for determining the course of a lane marking. The invention solves this problem by means of the subject matter of the independent claims. Dependent claims describe preferred embodiments.

[0006] A road can have one or more lanes. A lane can be delimited by a lane boundary, which is indicated or marked by one or more lane markings.A method for determining the course of the lane marking comprises the steps of recording series of observations of the lane marking, particularly during traverses of the roadway, wherein the series are made by different observers, each moving along the roadway; determining planes that each extend perpendicular to a course of the roadway; determining the points of intersection of the planes with each series of observations; grouping corresponding intersections, particularly those from different traverses, into a single plane; and determining the course of the lane marking based on corresponding grouped intersections of different, particularly adjacent, planes. Similarly, the points of intersection of each series of observations with the corresponding planes can also be determined.

[0007] An observation generally refers to a lane marking that designates a lane boundary. A series of observations can refer to a single lane marking or to several different lane markings that designate the same lane boundary. For example, a broken line visible on the roadway separating two lanes can be considered a series of observations of separate lane markings indicating the same lane boundary. The series of observations can be made by observers moving in different lanes of the road. Observations within a series refer to the same lane boundary.

[0008] A summarized intersection point can also be called a hypothesis. By determining hypotheses in the planes in a first phase and determining corresponding hypotheses from different planes in a second phase, the problem existing in three-dimensional space can be split into two easier-to-solve subproblems. This method is particularly suitable for determining one or more lane markings on a road based on a large number of observations. Vehicles, preferably motor vehicles such as cars, motorcycles, or trucks, can be used as observers. Even a manageable number of observations or observers may be sufficient for reliably determining the lane marking. For practical purposes, the lane marking can, for example, be determined based on observations from as few as 20 observers who have traveled the road.Furthermore, it has been shown that the specific lane marking can no longer significantly improve in quality when observations from more than approximately 150 observers are processed.

[0009] The levels can be arranged at predetermined intervals from one another. It is preferred that adjacent levels have predetermined, and in particular equal, distances between them. In one embodiment, adjacent levels are approximately two meters apart. The distances between the levels can be related to the alignment of the roadway.

[0010] It is still preferred that intersection points based on different series of observations and lying in the same plane be assigned to each other using the Hungarian method so that they can subsequently be combined. Typically, a series of observations originates from a single vehicle traversing a section of the road. The intersection points of different series of observations can each be assigned to the same traversal. This combining can be achieved, for example, by averaging the positions of the intersection points in the planes. The Hungarian method, also known as the Hungarian algorithm, can ensure an efficient and discriminatory assignment of observations to intersection points in the planes, as it is a "greedy" method.

[0011] The method preferably assumes a predetermined route for the road, which may, for example, be in the form of a geographic map with simple accuracy. Generally, the road can extend along a predetermined curve, with the planes extending perpendicular to the curve. Any structure, particularly one-dimensional, along which the road can extend can be used as the curve. The curve can be specified, for example, as a polyline, route, or contour. The position of an intersection point in a plane can be referenced to an intersection point of the plane with the curve. The curve can, in particular, comprise a polyline. Support points of the polyline can be specified as geographic coordinates, for example, in WGS 84.

[0012] In another embodiment, a composite intersection point supported by fewer than a predetermined number of intersection points is rejected. The predetermined number can, in particular, be specified relative to a number of observers based on whose observations an intersection point with the plane under consideration was determined.

[0013] In other words, a number of observers can be determined who provide observations based on which intersections with the plane are calculated. A predetermined proportion, for example 50%, can be formed from this number. If more intersections than the predetermined proportion are correctly assigned, the intersections can be grouped and processed further. Otherwise, they can be discarded. Misattributions, such as misattributing an observation to an intersection or an intersection to a lane boundary, can be significantly reduced.

[0014] It is still preferred that observations of a lane marking by an observer at different points along the roadway are each assigned a constant identifier; whereby aggregated intersections of different planes are assigned to each other based on observations encompassed by the same identifiers. An identifier initially only needs to be unique per observer. Provided identifiers can, for example, be extended to include a unique identifier for each observer to ensure the uniqueness of the identifiers between observers as well.

[0015] In a preferred refinement of the method, aggregated intersection points of different levels can be assigned to one another based on a Jaccard coefficient derived from the identifiers of the respective observations. Using the Jaccard coefficient, also called the Jaccard index, a measure of the similarity of sets can be determined. By basing the coefficient on the identifiers, it is possible to take into account which continuities between intersection points of the different levels were observed by the observers themselves.

[0016] Associated, aggregated intersection points can be determined using the Hungarian method. This allows for an efficient and low-error determination.

[0017] It is further preferred that a branching or merging of lane markings be determined. Both circumstances can be read from the paths of corresponding aggregated intersection points of different planes along the roadway. It is particularly preferred that a geographical position along the roadway be determined at which a branching or merging occurs. Especially when several lane markings are provided side by side on the roadway, it is preferred that it is additionally specified which of the lane markings branches or merges with which other. A branching or merging can be described in even greater detail.

[0018] Lane markings associated with the same lane boundary can be determined based on observations whose longitudinal positions along the roadway form spatially dense groups. Specifically, the positions of the observations can be projected onto the roadway, and groups of projected observations can be formed. This grouping can be accomplished using a clustering algorithm. For example, this allows for the improved determination of a regularly broken line, such as those used to delineate or separate equal lanes traveling in the same direction on a roadway.

[0019] The invention further relates to a device for determining the course of a lane marking along a road, wherein the device comprises a receiving device for acquiring series of observations of a lane marking of the lane boundary from observers moving along the road; and a processing device. The processing device is configured to carry out a method described herein, in whole or in part.

[0020] The processing device may include a programmable microcomputer or microcontroller, and the method may be in the form of a computer program product with program code. The computer program product may be stored on a computer-readable data carrier. Features or advantages of the method may be transferred to the device or vice versa.

[0021] Furthermore, the invention relates to a system that may include a device as described herein and at least one vehicle which is equipped to observe a lane marking of a roadway traveled by the vehicle and to provide the observation to the device.

[0022] The system can encompass a large number of the vehicles described. Observations of the vehicles can be stored in a database and processed at a later time.

[0023] The invention will now be described in more detail with reference to the attached drawings, in which: Fig. 1 a system; Fig. 2 - Fig. 7 steps of a process illustrate.

[0024] Fig. 1 shows a system 100 , which is an example of a device 105 and a vehicle 110 includes the vehicle 110preferably comprises a motor vehicle, in particular a passenger car, and is equipped to make observations of its surroundings while traveling on a road 115 drives.

[0025] The road 115 follows a predetermined course, defined by a curve 120 It may be specified along which the road runs 115 extends. The curve 120 can in particular be expressed as a polyline with a number of support points, between which straight sections are provided. In one embodiment, the curve is 120 as an approximation of the route of the road 115 specified.

[0026] On the road 115 There is at least one lane marking 122 , which forms a lateral boundary of a lane for a vehicle 110 defined. The course of the lane markings 122It usually follows the curve essentially with a certain lateral offset. 120 The lane markings 122 is indicated by a lane marking 125 , which are linear and painted on the road in, for example, white, yellow or red colors 115 This may be appropriate. The lane marking can be used in this context. 125 They can be formed by different types of lines, for example, a solid or a broken line. Multiple lane markings 125 can use the same lane marking 122 be assigned, for example in the case of a broken line as in the left area of Fig. 1 or in the case of a double line as in the lower right area of Fig. 1.

[0027] Reflective markings, such as reflective beads or other retroreflectors, may be added to the colored line. The lane marking 125This could also include, for example, a roadway 115 These may include a raised threshold or a series of nails (“Botts’ dots”). Other designs are also possible and may be defined, in particular, in a locally applicable road traffic regulation.

[0028] When driving on the road 115 can a lane marking 125 optically scanned, leading to an observation 130 can lead to an observation 130 A geographical position can be assigned. Observations 130 , which share the same lane marking 122 They are usually related to each other, for example by relating the observations 130 A common identifier is assigned. The identifier can be assigned across different, but identical, lane markings. 122 assigned lane markings 125 be the same.

[0029] To determine the course of the lane markings122 To determine with sufficient accuracy, for example to create an HD map, observations of various vehicles can be used. 110 or various crossings of the road 115 They are processed together. It can be difficult to establish an initial lane marking. 122 , which are based on observations 130 was determined, which were collected during a first pass, a second lane restriction 122 to be assigned, which are based on observations 130 was determined, which were collected during a second pass. For example, the roadway 115 comprise multiple lanes, one of which includes a vehicle traveling on the far left or far right 110 It is not possible to visually perceive all of them. A technique described herein particularly facilitates the correct assignment of specific lane markings. 122 to each other.

[0030] On board the vehicle 110 is preferably a device 135 provided, which is set up to make the observations 130 to provide. The observations presented 130 Examples include point observations that can be connected to form a line observation. 130 to result. The device 135 preferably includes a processing facility 140 , equipped with at least one scanning device 145 , and preferably a communication device 150 is connected.

[0031] The scanning device 145 is for contactless, preferably optical, scanning of the vehicle's surroundings 110 It is set up and includes, for example, a photo or video camera. Alternatively, a radar or LiDAR sensor can be used. The communication device 150is used for preferably wireless communication with a device outside the vehicle 110 The device is installed at a designated location and can operate, for example, via Wi-Fi or mobile network. The scanning device is located in an execution unit. 145 set up to observe 130 to assign identifiers, whereby observations 130 , which are on the same lane marking 122 preferably the same identifier is assigned. A typical sampling frequency of the scanning device 145 The frequency is approximately 9 Hz.

[0032] To determine the geographical position of the vehicle 110 It can also include a positioning device 155The position can be determined, in particular, by means of a receiver for signals from, for example, a satellite-based positioning system. Other usable methods of position determination include measurements using an inertial system or using speed sensors on the vehicle's wheels. 110 Based on the vehicle's position 110 can a position of an observation 130 be determined.

[0033] The device 135 is preferably set up for observations 130 one or more lane markings 125 to determine and preferably outside the vehicle 110 arranged device 105 to transmit. A location for the device to be installed. 105 is generally not decisive; in one embodiment it can also be on board the vehicle 110 appropriate.

[0034] The device 105preferably includes a processing facility 160 as well as an optional data storage device 165 Observations 130 can be done using a communication device 170 to be received. The device is usually 105 implemented as a server or service, for example in a cloud, and accepted observations 130 a large number of vehicles 110 The device 105 is designed to define the course of a lane boundary 122 based on observations 130 of lane markings 125 to determine which are driven by several vehicles 110 from each time when driving on the road 115 The procedures were carried out. The accuracy of the determination is preferably sufficiently high to accurately define the lane markings. 122 to be recorded in an HD map.

[0035] Fig. 2 to Fig. Figure 7 illustrates the steps of an exemplary procedure 200, which is based in particular on the system 100 to determine the course of a lane marking 125 can be used.

[0036] A geographical map 205 describes the routes of a multitude of roads 115 and their relationships to each other. The map 205 can be used in particular for guiding a vehicle to a destination 110 can be used to travel from a current position to a predetermined target position. The accuracy of the map 205 included routes of roads 115 is usually limited and can be in the range of one or more meters (SD).

[0037] Based on the map 205 can a road 115 , on which the vehicle 110 drives, identifies and its at least approximate course based on the curve 120 be determined. The curve 120It can take any shape and is usually defined in three dimensions. A common representation of the curve. 120 This is done using a polyline.

[0038] Along the curve 120 can have a multitude of levels 210 are formed, whereby the levels 210 each perpendicular to the curve 120 stand and preferably have predetermined distances from each other. The planes 210 can regarding the curve 120 be equidistant, with adjacent planes 210 can maintain a distance of, for example, approximately two meters.

[0039] Fig. Figure 3 shows a compilation of observations 130 different lane markings 125 the same road 115 by different observing vehicles 110 . In the left area of Fig. The three images, from top to bottom, show observed lane markings. 125different vehicles 110 shown, with all vehicles 110 the same section of a predetermined route 115 scan. Three vehicles are shown here as examples. 110 planned, which are designated 110.1, 110.2, etc. Observations can be made. 130 of n vehicles 110 are processed, where generally 2 < n and more preferably 20 ≤ n ≤ 120. The identity of a vehicle 110 This is irrelevant. Should a vehicle 110 the road 115 If the area is travelled multiple times, observations can be made. 130 as from two different vehicles 110 are considered to originate from.

[0040] Every vehicle 110 represents one or more series of observations 130 ready. An observation 130 is a lane marking 125 assigned, whereby observations 130 a series of the same lane marking122 is assigned to the lane marking 122 can be achieved by one or more lane markings 125 must be marked. For the purposes of this discussion, it is assumed by way of example that every vehicle 110 each a series of observations 130 for four adjacent lane boundaries 122 provides. Based on a series of observations. 130 a vehicle 110 can a polyline 212 to be determined. For each level 210 can be a point of penetration or intersection 215 with the polyline 212 to be determined. The intersection points are shown here. 215 Each is indicated by a plus sign.

[0041] In the right area of Fig. 3 is a superposition of the observed polylines of the vehicles 110 depicted in one plane 210 Intersection points 215 , which share the same lane marking 122refer to, with a certain degree of dispersion. Intersection points assigned to each other 215 are now to be found and considered as a hypothesis 220 to be summarized.

[0042] Fig. 4 shows the determination of consolidated intersection points 215 in a predetermined plane 210 based on information from various vehicles 110 The determination preferably works iteratively according to the Hungarian method, whereby in each iteration the information of a vehicle is used. 110 are processed. Steps shown vertically correspond to each other. The representation of Fig. 4 again provide exemplary information from three vehicles. 110 to the ground.

[0043] The determination preferably begins with the information from that vehicle. 110 , whose observations 130 to the fewest number of intersections 215 in the plane under consideration 210have led to the information from the other vehicles 110 can later be done one after the other in ascending order of the number of intersection points determined in each case. 215 be added.

[0044] In the example shown, the observations provide 130 of the first vehicle 110 three intersection points 215 Since there are no other intersection points yet 215 of another vehicle 110 considered, for each of the intersection points 215 directly a hypothesis 220 in the plain 210 be determined. In Fig. 4 are hypotheses 220 exemplified as circles; intersections 215 are represented as plus signs.

[0045] The hypothesis 220 The following will be based on information from other vehicles 110 will be improved successively. In ascending order of the number of specific intersection points. 215The second vehicle will follow. 110 with four intersection points 215 In the second line of Fig. The four intersection points on the left are... 215 of the second vehicle 110 together with the currently valid hypotheses 220 depicted.

[0046] In the middle representation, each hypothesis 220 an intersection 215 of the added vehicle 110 assigned. The intersection points can be used for this purpose. 215 and hypotheses 220 according to the nature of the observations on which they are based 130 to be assigned to each other. A hypothesis 220 can that intersection point 215 of the added vehicle 110 to be assigned the one that has the smallest geometric distance to it.

[0047] After the assignment, the hypotheses are 220 based on the added intersection points 215updated. This involves adjusting the geographical position of the updated hypothesis. 220 preferably a weighted average over the positions of the included intersection points 215 in the plain 210 formed so that each intersection 215 , which is used to formulate the hypothesis 220 was used, had an equally large influence on the hypothesis 220 has.

[0048] In Fig. In the right-hand representation of the second row, there are four updated hypotheses. 220 shown, of which the three on the right each represent an average value from previously determined intersection points 215 the vehicles 110 and 110.2, while the left one, which is powered by information from the first vehicle 110 is not supported, but based solely on an intersection point 215 from observations 130 of the second vehicle 110 is educated.

[0049] Similarly, in the third line of Fig. 4 shows how the determined intersection points 215 of the third vehicle 110 for updating or forming further improved hypotheses 220 can be used. Based on information from the third vehicle. 110 are five intersection points 215 determined, so that the existing four hypotheses 220 a fifth one is added.

[0050] The hypothesis shown furthest to the left in the right-hand representation of the third row 220 is of only one in three vehicles 110 supported (1 / 3). The hypothesis to the right 220 is of two out of three vehicles 110 supported (2 / 3) and the remaining hypotheses 220 are of all three of the three vehicles 110 supported (3 / 3).

[0051] Hypotheses 220 , which are observed by fewer observers 110Perceptions that are considered predetermined can be rejected. In one embodiment, an absolute threshold is specified, for example, 20 or 50 observers. 110 Preferably, a relative threshold is used, which is based in particular on the number of observers. 110 can be related, based on whose information at least one intersection 215 in the plane under consideration 210 could be determined. For the purposes of this discussion, a threshold of 50% is assumed as an example, so that hypotheses can be formulated. 220 , which are observed by less than 1.5 of the observing vehicles 110 The hypotheses that were recorded can be discarded. In the present example, this only applies to the hypothesis shown on the left. 220 , which is marked with an X for this reason.

[0052] Fig. Figure 5 shows the merging of results relating to different levels 210were determined in the manner described. There are several levels in the upper area. 210 along the curve 120 depicted, with each level 210 the previously determined hypotheses 220 are marked, which henceforth serve as intersection points 215 can be treated.

[0053] Starting from a predetermined level 210 can an adjacent plane 210 to be considered, and those from the levels 210 included intersections 215 can be assigned to each other in pairs. Connections between the intersection points 215 This can include a representation of a lane marking. 125 The best connections between two intersection points result. 215 different levels 210can be determined using the Hungarian method. This method employs a distance metric based on the Jaccard distance between identifiers that correspond to the intersection points. 215 , each based on a hypothesis 220 Identifiers were formed and assigned. Identifiers can be found in the system. 100 already on board a vehicle 110 , alternatively by a scanning device 145 or the processing facility 140 , can be determined. This allows for the selection of two intersection points. 215 different levels 210 The connection or assignment with the strongest evidence is determined in each case. In the lower range of Fig. 5 are lane marking routes 125 shown, which in this way along the curve 120 were determined. It should be noted that this involves merging two different lane markings. 125 was determined.

[0054] Fig. Figure 6 shows the determination of a lane marking interrupted at regular intervals. 125 The top section contains original observations. 130 along the curve 120 The positions of the observations are shown. 130 can now turn towards the curve 120 are projected and longitudinal positions of the observations 130 , that is, their positions along the curve 120 , can be determined.

[0055] Groups or clusters can be identified between the specific longitudinal positions. A single observation 130 An entry that cannot be assigned to any group can be discarded. A specific group can be an observable section of a lane marking. 125 include, with the lane marking between two such groups 125 It can continue invisibly.

[0056] Fig. Figure 7 shows the result of the above with reference to Fig. 5 described procedure in combination with the determination of broken lines according Fig. 6. It can be seen how different types of lane markings are used. 125 could be detected and how the detected lane markings 125 along the curve 120 extend. Reference symbol list 100 System 105 Device 110 vehicles 115 Road 120 Curve 122 Lane marking 125 Lane marking 130 observations 135 Device on board a vehicle 140 processing equipment 145 Scanning device, in particular optical 150 communication equipment 155 Positioning device 160 processing unit 165 data storage 170 Communication device 205 Geographic map, especially in simple accuracy (SD) Level 210 212 polyline 215 Intersection point, point of penetration 220 Hypothesis

Claims

[1] Method (200) for determining the course of a lane boundary (122) along a roadway (115), wherein the lane boundary (122) is characterized by a lane marking (125), the method (200) comprising the following steps: - Recording series of observations (130) of the lane marking (125), wherein the series are made by different observers (110) moving along the roadway (115); - Determining planes (210) that each extend perpendicular to a course (120) of the roadway (115); - Determining the intersection points (215) of the planes (210) with each a series of observations (130); - Combining related intersection points (215) of a plane (210); and - Determining the course of the lane boundary (122) on the basis of mutually assigned aggregated intersection points (220) of different planes (210). [2] Method (200) according to claim 1, wherein adjacent planes (210) each have predetermined distances. [3] Method (200) according to claim 1 or 2, wherein intersection points (215) based on different series of observations (130) are assigned to each other using the Hungarian method. [4] Method (200) according to one of the preceding claims, wherein the roadway (115) extends along a predetermined curve (120), wherein the planes (210) extend perpendicular to the curve (120). [5] Method (200) according to any of the preceding claims, wherein a combined intersection point supported by less than a predetermined number of intersection points (215) is rejected. [6] Method (200) according to one of the preceding claims, wherein observations (130) of a lane marking (125) by an observer (110) at different locations along the roadway (115) are each assigned a constant identifier; and aggregated intersection points (220) of different planes (210) are assigned to each other on the basis of aggregated observations (130) with identical identifiers. [7] Method (200) according to claim 6, wherein aggregated intersection points (220) of different planes (210) are assigned to each other on the basis of a Jaccard coefficient of identifiers of observations (130) encompassed. [8] Method (200) according to one of claims 6 or 7, wherein mutually associated aggregated intersection points (220) are determined using the Hungarian method. [9] Method (200) according to one of the preceding claims, wherein a branching or merging of lane markings (125) is determined. [10] Method (200) according to one of the preceding claims, wherein lane markings (125) which are assigned to the same lane boundary (122) are determined on the basis of observations (130) whose longitudinal positions along the roadway (115) form spatially dense groups. [11] Device (105) for determining the course of a lane marking (122) along a roadway (115), the device comprising a receiving device for recording series of observations (130) of a lane marking (125) of the lane marking (122) by observers (110) moving along the roadway (115); and a processing device (160) configured to perform a method (200) according to any of the preceding claims. [12] System (100) comprising a device (105) according to claim 11 and at least one vehicle (110) which is configured to observe a lane marking (125) of a roadway on which the vehicle (110) is driving and to provide the observation (130) to the device (105).

Images