Object localization from vehicle sensor data

Abstract

The invention relates to a method for determining a position of a stationary object (1), comprising the steps of: detecting (S1) by a plurality of vehicles (2) a distance of the object (1) from the respective vehicle (2); determining (S2) a distance between the detected stationary object (1) and a building (3) running along the road; determining (S3) an absolute longitudinal position component of the stationary object (1) along the direction of extension of the road and a relative lateral position component of the stationary object (1), wherein the absolute longitudinal position component is determined from a respective geodetic position of a vehicle (2), and wherein the determined distance of the stationary object (1) from the building (3) is used as the lateral position component.

Description

[0001] The invention relates to a method for determining a position of a stationary object in the area of a road used by a plurality of vehicles, as well as a data processing system for carrying out the method.

[0002] For partially and highly automated driving of passenger cars and trucks, it is useful to geodetically record the positions of objects such as traffic signs, for example, to incorporate them into a frequently updated digital map. This also makes it possible to track changes in infrastructure, such as the installation of traffic signs in a construction zone for short periods of time compared to permanently installed traffic signs.

[0003] Vehicles equipped with appropriate sensors can automatically detect objects such as traffic signs. If the current position of the recording vehicle is also recorded via satellite-based positioning, the position of the detected object can be determined, particularly by measuring the distance between the detected object and the vehicle itself. Random errors in such a calculation can be at least partially averaged out by collecting corresponding data from a large number of vehicles and calculating an average of this data or the result of the position determination of the respective detected object.

[0004] However, due to the low accuracy of conventional satellite-based positioning in relation to the requirements of this procedure, there remains too much uncertainty in the determination of the position of the object, especially when two objects are close to each other and there are incorrect assignments between these two objects in their position or a merging of the two objects into a supposed single one.

[0005] Against this background, the state of the art attempts to address the problem of false detections when determining the position of traffic signs on the basis of a large number of measurement data from one or more vehicles by evaluating an existence probability that is incrementally increased each time the traffic sign is seen and decreased each time it is not seen.

[0006] In this context, DE 10 2020 201 280 A1 relates to a method for determining positions and validating traffic signs along a road section, comprising the steps of: providing a plurality of measurement data along the road section to an evaluation unit of one or more vehicles, wherein the measurement data were recorded by means of environmental sensors of the one or more vehicles, wherein the measurement data of a vehicle each comprise a traffic data set which comprises the type of detected traffic signs and the GPS information along the road section as well as the position of the respective detected traffic sign in relation to the respective vehicle coordinate system, clustering the one or more detected traffic signs along the road section into clusters which each comprise the same traffic sign, determining an existence probability for each traffic sign,where the existence probability indicates the probability of the actual presence of a detected traffic sign, the existence probability being determined incrementally for each traffic sign by recalculating the existence probability after each positive detection of the traffic sign in a newly added measurement data set and after each negative detection of the traffic sign in a newly added measurement data set, taking into account the previous existence probability.

[0007] However, the cluster separation proposed in DE 10 2020 201 280 A1 could frequently fail when traffic signs are positioned close together, or oscillate between correct and incorrect states. The incremental method proposed there also has the disadvantage of requiring relatively high computational resources, since a world state must be permanently stored in memory.

[0008] US 2021 / 0190536 A1 also relates to a system for precisely locating the position of at least one road object connected to a part of a road network. The general principle is based on determining the position of a road object using an unsupervised classification based on distribution density. In the disclosure, the classification is applied in two phases. In the first phase, it is applied for the first time to the plurality of geographical coordinates associated with a road object to automatically group it into homogeneous classes. In the second phase, it is applied a second time to each class generated in the first phase to automatically group it into homogeneous subclasses based on the azimuth angles of the road object linked to the geographical coordinates of the class.

[0009] From DE 10 2013 009 856 A1 a method for determining the position of traffic objects, for example traffic signs, is known, wherein it is provided that a plurality of motor vehicles, when driving along the route, transmit position estimates of the traffic objects to a central database, which determines the position of the traffic objects from the position estimates by statistical processing, for example clustering.

[0010] DE 10 2018 008 904 A1 discloses a method for locating a vehicle using static landmarks, such as traffic signs. The method involves discarding the lateral position of the landmarks and using only the longitudinal position of the landmark relative to the roadway to determine the position by means of map matching. For this purpose, this position of the landmark is projected onto a reference line, such as the road boundary.

[0011] From US 11 113 545 B2 a method for detecting a roadway, in particular the road edge and road course, is known, wherein it is provided to detect traffic signs arranged next to the roadway and then, by using attributes of the traffic signs recorded in a database, to infer the relative position of the road edge from the position of the traffic signs and known information regarding lateral offset, etc. and, based thereon, to infer the offset of the road edge to the vehicle.

[0012] The object of the invention is to further improve the accuracy and reliability in determining the position of objects, in particular traffic signs.

[0013] The invention is based on the features of the independent claims. Advantageous developments and refinements are the subject of the dependent claims.

[0014] A first aspect of the invention relates to a method for determining a position of a stationary object in the area of a road traveled by a plurality of vehicles, comprising the steps: - detecting, by the plurality of vehicles, a respective longitudinal and a respective lateral distance of the stationary object from the respective vehicle, assigning a geodetic position of the respective vehicle and a lateral distance of the vehicle from a building running along the road, in each case at the time of detecting the stationary object; - for each detection by a respective vehicle: determining a distance between the detected stationary object and the buildings along the road, - Determining an absolute longitudinal position component of the stationary object related to the direction of extension of the road and a relative lateral position component of the stationary object that is transverse to the longitudinal position component, wherein the absolute longitudinal position component is determined from the respective geodetic position of the respective vehicle and the longitudinal distance between the respective vehicle and the stationary object, and wherein the determined distance of the stationary object to the building is used as the lateral position component.

[0015] The longitudinal position component is related to the course of the road, the lateral position component is aligned transversely to the longitudinal position component and thus always describes the shortest distance between the stationary object and the building.

[0016] The plurality of vehicles has respective sensors to detect and ideally recognize the object as it travels along a road. The object is preferably a traffic sign such as a speed limit or other traffic sign. Furthermore, the vehicles are all designed to continuously record their geodetic position. This is preferably done by a satellite-based positioning unit, which allows the position of the respective vehicle to be determined within certain accuracy limits. The geodetic position of the vehicle indicates an absolute position, preferably specified in WGS-84 coordinates, and therefore comprising at least a longitude component and a latitude component.

[0017] In addition, the vehicles determine a longitudinal and a lateral distance from the object when they detect and ideally recognize it, for example as a traffic sign. The longitudinal distance indicates a distance along the direction of extension of the road, whereas the lateral distance indicates the distance between the object and the vehicle perpendicular to the longitudinal distance, i.e. how far the object is offset to the side. Therefore, if the geodetic position of the vehicle is determined by the tracking unit, the absolute position of the object can be determined by adding the longitudinal and lateral distances of the object to the vehicle. However, the absolute position of the object determined in this way can only be determined with the same accuracy as the geodetic position of the vehicle.

[0018] To avoid this error, the specified method determines at least the absolute position in a lateral component relative to a structure running along the road. Such a structure could be, for example, a guardrail or a noise barrier. The directions of extension of the road and the structure are therefore essentially parallel, at least in some sections. Since the lateral distance of the object from the structure can be determined much more accurately than in absolute coordinates based on the geodetic position of the vehicle, this makes it much easier to recognize closely spaced traffic signs as separate and to determine their position much more accurately.

[0019] The development thus makes it possible to express at least the lateral component with respect to the course of the road to describe the position of the stationary object by a relative position component, and to eliminate the inaccuracy, for example, in the position of the vehicle determined by GPS, at least in the lateral component.

[0020] By further advantageously considering the detection of a large number of vehicles, a statistical evaluation can be carried out and random errors can thus be essentially averaged out, for example by using a median or a mean value of the naturally scattering sensor values of the vehicles or the position components of the object calculated therefrom.

[0021] According to an advantageous embodiment, the structure is a guardrail, a noise barrier, or a concrete retention system.

[0022] According to a further advantageous embodiment, the development extends along a left side of the road and a right side of the road, wherein the respective lateral distance between the vehicle and the development on the left side of the road and the development on the right side of the road is detected, and wherein a distance between the detected stationary object and the development on the left side of the road and the development on the right side of the road is determined.

[0023] According to a further advantageous embodiment, the absolute longitudinal position of the stationary object and / or the relative lateral position of the stationary object is determined by determining a mean value or a median.

[0024] According to a further advantageous embodiment, the detections from the individual vehicles are clustered. Clustering can be applied both to the determined geoposition of the stationary object and to the relative lateral distance to the buildings (left + right). The relative lateral distance thus becomes a further dimension in the clustering.

[0025] According to a further advantageous embodiment, a "nearest-neighbor" algorithm is used to assign detections with insufficient quality of the lateral distance to the built-up area to a cluster. For clustering, normally only the detections of individual vehicles with sufficient quality are considered. The lower-quality detections / measurement points, for which the lateral distance to the built-up area is greater than the mean or median, are assigned to the nearest cluster using a "nearest-neighbor" algorithm.

[0026] According to a further advantageous embodiment, for an object detected by the vehicles from two streets, at least the respective absolute longitudinal position of the object is standardized from the detections from the two streets.

[0027] According to a further advantageous embodiment, each time the stationary object is detected, it is checked whether the building can be detected from the vehicle in order to determine a relative lateral position of the object to the building only if the building is successfully detected, and otherwise to determine an absolute lateral position of the stationary object.

[0028] According to a further advantageous embodiment, the stationary object is a traffic sign.

[0029] A further aspect of the invention relates to a data processing system comprising means for carrying out the method as described above and below.

[0030] Advantages and preferred developments of the proposed system result from an analogous and analogous transfer of the statements made above in connection with the proposed method.

[0031] Further advantages, features, and details will become apparent from the following description, in which at least one embodiment is described in detail—possibly with reference to the drawings. Identical, similar, and / or functionally equivalent parts are provided with the same reference numerals.

[0032] They show: Fig. 1: A method for determining a position of a stationary object in the road area according to an embodiment of the invention. Fig. 2: An exemplary situation in which the procedure according to Fig. 1 can be applied.

[0033] The representations in the figures are schematic and not to scale.

[0034] Fig. 1 shows a method for determining the position of a stationary object 1 in the area of a road traveled by a plurality of vehicles 2. Such a road with an exemplary vehicle 2 and an exemplary object 1, designed as a traffic sign, is shown in Fig. 2. For this reason, the following explanations of the process steps can also be used Fig. 2 for a closer understanding. In a first step of the method, the detection S1, by a respective one of the plurality of vehicles 2, of a respective longitudinal and a respective lateral distance of the stationary object 1 to the respective vehicle 2 takes place. The data provided by each of the vehicles 2 in this process comprise a geodetic position of the respective vehicle 2, which is estimated, for example, by satellite-based positioning, a relative distance between the detected object 1 with a lateral and a longitudinal component perpendicular thereto, as well as a relative distance of the vehicle 2 to a structure 3 that extends essentially parallel to the main direction of extension of the road, such as a guardrail. The distance of the respective vehicle 2 to the structure 3 is determined, for example, by ultrasonic sensors, optionally supported by a camera or radar unit.The data are related to one another accordingly, so that for each vehicle 2 a detected object 1 can be assigned to a current geodetic position of the respective detecting vehicle 2 as well as the associated distance between the vehicle 2 with lateral and longitudinal components to the object 1. In this case, it can already be determined whether there is any suitable development 3 along the road in order to be used as a one-dimensional reference for determining the lateral position of the stationary object 1 relative to the development 3. From the respective geometric position of the detecting vehicle 2 as well as the lateral and longitudinal distance between vehicle 2 and stationary object 1, outlined in . Fig.2 by two dashed double arrows leading away from the vehicle 2, as well as by the detected distance between the vehicle 2 and the building 3, a distance between the detected stationary object 1 and the building 3 can also be determined S2. This takes place in a computing unit, preferably locally in the vehicle 2, but can also be done in a central computing unit. The stationary object 1 can then be localized after it has been detected, i.e. a position can be assigned to the individual stationary object 1. The position has two different components, whereby the position component along the course of the road, i.e. the longitudinal position component, is given in absolute, geodetic coordinates such as longitude and latitude, whereas the lateral position component running transversely to it is only given in relative coordinates, namely as the distance to the building 3.The uncertainty in the lateral component of object 1 introduced by the geodetic coordinates of vehicle 2, which would be transferred to the position determination of stationary object 1 using the aforementioned calculation method, is thus eliminated. This method for determining the position of stationary object 1 is not carried out solely on the basis of the sensor data of a vehicle 2 detecting stationary object 1, but rather using a plurality of vehicles 2, which, in particular, detect stationary object 1 at different times as described. This naturally results in a point cloud of scattered data points, since the position determination of each individual vehicle 2 is not carried out solely for an individual detection location, but also introduces its own uncertainties and deviations.Through clustering, the data points can be summarized and statistically evaluated, in particular by calculating a median or a mean value of the data. This particularly applies to the position determination of the stationary object 1 already carried out for the respective detection of the stationary object 1. A special method can also be used here to determine whether, in a scattering point cloud of estimated positions with longitudinal and lateral position components as described above, the estimated positions are standardized to a resulting estimated position of the stationary object 1, or whether it can be assumed that two different stationary objects 1 are present, such as two traffic signs standing close to one another. In this case, a threshold value is advantageously used, in particular two meters.If the estimated individual positions of stationary object 1 fall below this threshold, it can be assumed that a single stationary object 1, such as a single traffic sign, is present. If this threshold is exceeded, two separate stationary objects 1 can be identified, which might not be possible without the significantly more accurate lateral relative position component instead of an absolute position component in geodetic coordinates.

[0035] Although the invention has been illustrated and explained in detail by preferred embodiments, the invention is not limited by the disclosed examples, and other variations may be derived therefrom by those skilled in the art without departing from the scope of the invention. It is therefore clear that numerous variations exist. It is also clear that exemplary embodiments are truly only examples and should not be construed as limiting the scope, possible applications, or configuration of the invention in any way.Rather, the preceding description and the description of the figures enable the person skilled in the art to implement the exemplary embodiments in concrete terms, whereby the person skilled in the art, with knowledge of the disclosed inventive concept, can make various changes, for example with regard to the function or the arrangement of individual elements mentioned in an exemplary embodiment, without departing from the scope of protection defined by the claims and their legal equivalents, such as further explanations in the description. List of reference symbols 1 object 2 vehicles 3 Development S1 Capture S2 Determine S3 Determine

Claims

[1] Method for determining a position of a stationary object (1) in the area of a road traveled by a plurality of vehicles (2), comprising the steps: - detecting (S1), by the plurality of vehicles (2), a respective longitudinal and a respective lateral distance of the stationary object (1) to the respective vehicle (2) while assigning a geodetic position of the respective vehicle (2) and a lateral distance of the vehicle (2) to a building (3) running along the road, in each case at the time of detecting the stationary object (1); - for each detection by a respective vehicle (2): determining (S2) a distance between the detected stationary object (1) and the buildings (3) running along the road, - Determining (S3) an absolute longitudinal position component of the stationary object (1) along the direction of extension of the road and a relative lateral position component of the stationary object (1), wherein the absolute longitudinal position component is determined from the respective geodetic position of the respective vehicle (2) and the longitudinal distance between the respective vehicle (2) and the stationary object (1), and wherein the determined distance of the stationary object (1) to the building (3) is used as the lateral position component. [2] Method according to claim 1, wherein the structure (3) is a guard rail, a noise barrier, or a concrete retention system. [3] Method according to one of the preceding claims, wherein the building (3) extends along a left side of the road and a right side of the road, wherein the respective lateral distance between the vehicle (2) and the building (3) on the left side of the road and the building (3) on the right side of the road is detected, and wherein a distance between the detected stationary object (1) and the building (3) on the left side of the road and the building (3) on the right side of the road is determined. [4] Method according to one of the preceding claims, wherein the absolute longitudinal position of the stationary object (1) and / or the relative lateral position of the stationary object (1) is determined by determining a mean value or a median from detections of the individual vehicles (2). [5] Method according to one of the preceding claims, wherein the detections by the individual vehicles (2) are clustered. [6] Method according to claim 5, wherein a "nearest neighbor" algorithm is used to assign detections with insufficient quality of the lateral distance to the building (3) to a cluster. [7] Method according to one of the preceding claims, wherein for an object (1) detected from two roads by the vehicles (2), at least the respective absolute longitudinal position of the object (1) is standardized from the detections from the two roads. [8] Method according to one of the preceding claims, wherein it is checked each time the stationary object (1) is detected whether the building (3) can be detected from the vehicle (2), in order to determine a relative lateral position of the object (1) to the building (3) only if the building (3) is successfully detected, and otherwise to determine an absolute lateral position of the stationary object (1). [9] Method according to one of the preceding claims, wherein the stationary object (1) is a traffic sign. [10] A data processing system comprising means for carrying out the method according to any one of the preceding claims.

Images