Generating non-semantic reference data for determining the position of a motor vehicle

Abstract

According to a computer-implemented method for generating non-semantic reference data for determining the position of a motor vehicle (6), a set of raw data points (7) is provided, which describes a predetermined environmental region (11). A predetermined descriptor is determined for each raw data point (7), which characterizes the characteristics of the environmental region (11). At least one point cluster (9a, 9b, 9c, 9d, 9e, 9f) is generated in such a way that the raw data points (7) are grouped according to their descriptors. Based on the descriptors of the raw data points (7), a characteristic parameter is assigned to a first point cluster, which relates to an information gain for determining the position of the motor vehicle (6). Based on the characteristic parameter, characteristic information of the first point cluster is stored in a storage unit (10) as non-semantic reference data for determining the position.

Description

Technical Field

[0001] The present invention relates to a computer-implemented method for generating non-semantic reference data for determining the position of a motor vehicle, wherein a set of raw data points is provided, which describes a predetermined surrounding area. The present invention also relates to a method for determining the position of a motor vehicle, a mapping system for determining the position of a motor vehicle, a motor vehicle having a mapping system, a computer system for generating non-semantic reference data for determining the position of a motor vehicle, a computer program, and a computer-readable storage medium. Background Art

[0002] For example, highly automated or autonomous vehicles require precise knowledge of their own location for navigation or trajectory planning. In known methods for determining location, semantic structures and patterns in the vehicle's surroundings, so-called landmarks, are detected by vehicle sensors and compared with corresponding entries in the vehicle's digital map. These semantic structures are always assigned to predefined categories, for example, with information about the type of object, such as a traffic sign or a house edge.

[0003] However, landmark-based localization has the disadvantage that only those environmental features that can be assigned to more or less universal categories are used for localization. This limits the number of environmental features that can be used for localization, so that in areas where no such landmarks are available, localization cannot be performed or can only be performed with low accuracy.

[0004] The position of a vehicle can also be determined using satellite signals from the Global Navigation Satellite System (GNSS). However, the accuracy of the satellite receivers typically installed in vehicles is too low to enable highly automated or autonomous driving. Furthermore, high-precision satellite receivers are associated with considerable costs.

[0005] EP 3 290 864 A1 describes a driver assistance system for determining a vehicle's position. Approximate position data for the vehicle is acquired based on GPS signals. Furthermore, an image of the vehicle's surroundings is recorded and compared with stored image data. By combining this information, the vehicle's position can be determined even in conditions of poor satellite reception. Summary of the Invention

[0006] Against this background, the object of the present invention is to provide an improved approach for determining the position of a motor vehicle, in particular for determining the position of a motor vehicle based on a map, which is independent of the availability of semantic surrounding structures.

[0007] The improved solution is based on the idea of ​​generating non-semantic reference data in the following way, that is, the original data points describing the environment are clustered according to preset descriptors for describing the environment characteristics, and the corresponding feature information is stored according to information gain, and the corresponding point clusters (punktcluster) can make this information gain helpful for position determination.

[0008] According to an independent aspect of the improved concept, a computer-implemented method for generating non-semantic reference data for determining the position of a motor vehicle is described. A set of raw data points, in particular generated by an environmental sensor, is provided, which depict a predetermined environmental area. A computing unit determines a predetermined descriptor for each raw data point, which characterizes the environmental area at the location of the corresponding raw data point or at a location corresponding to the corresponding raw data point in the environment. The computing unit generates at least one point cluster by, in particular, grouping the raw data points into point clusters according to their respective descriptors. The computing unit assigns a characteristic parameter (Kennzahl or characteristic factor) to a first point cluster of at least one point cluster based on the descriptors of the raw data points of the first point cluster, which relates to the information gain (information gain) for determining the position of the motor vehicle. Based on the characteristic parameter, the computing unit stores characteristic information of the first point cluster in a storage unit as non-semantic reference data for determining the position.

[0009] The surrounding area is, in particular, the surrounding area of ​​an environmental sensor or a data acquisition system or a data acquisition vehicle equipped with an environmental sensor. Raw data points are generated before the vehicle uses the reference data for determining its position. The surrounding area is thus the surrounding area of ​​the possible position of the vehicle.

[0010] Environmental sensors can be designed, for example, as radar sensors or lidar sensors, also known as laser scanners. To characterize the surrounding area, the environmental sensor system generates a point cloud from scanning points, which are typically presented as three-dimensional coordinate tuples. The raw data points representing the surrounding area are typically such a point cloud or a portion thereof.

[0011] In particular, the set of raw data points is provided in a computer-readable form so that the raw data points can be read out by means of a computing unit. In particular, the generation of the raw data points is not necessarily part of the method for generating non-semantic reference data according to the improved concept.

[0012] The fact that the raw data points are generated beforehand and then made available has the advantage, in particular, that a highly precise measuring device can be used when generating the raw data points.

[0013] Several raw data points of the set of raw data points can exist, for example, as corresponding coordinate tuples in a predefined reference coordinate system, for example a global coordinate system or a world coordinate system, for example a geodetic coordinate system such as WGS84.

[0014] Descriptors are, in particular, properties of the surrounding area that can be measured based on the raw data points, such as geometric properties. Geometric properties that can be used as descriptors include, in particular, the curvature or average curvature of the scanned surface or area in the surrounding area. Statistical properties of the raw data points, or their distribution, or distribution properties of the raw data points can also be used as descriptors. Optical properties of the environment can also be reflected by the raw data points. In particular, in addition to spatial coordinates, the raw data points can also contain intensity information, or intensity information can be assigned to the raw data points and provided. In the case of a lidar sensor, this is, for example, the intensity of the reflected laser beam. Since the emission characteristics, in particular the wavelength distribution of the emitted laser beam, are known, the spectral reflectance or color of the corresponding point in the environment corresponding to the raw data point can be determined based on the intensity. These properties can also be used as descriptors.

[0015] A preset descriptor may also contain several values ​​suitable as descriptors, or one or more variables derived from these values.

[0016] The use of descriptors that are not associated with the level of meaning of objects in the environment enables a non-semantic description of the environment.

[0017] In order to group several raw data points of the group of raw data points into at least one point cluster, in particular each raw data point of the group of raw data points is either assigned to exactly one point cluster of the at least one point cluster, or the corresponding raw data point is cleaned or discarded, i.e. not further considered for generating reference data.

[0018] The original data points of the first point cluster are in particular those original data points of the group of original data points which form the first point cluster due to grouping.

[0019] The fact that the raw data points are grouped according to their descriptors can be understood to mean, in particular, that the individual descriptors of the respective raw data points are used for grouping, or that the descriptors of the raw data points are statistically evaluated, for example, by forming a mean value of the distribution of the descriptors, forming a local mean value, or other analysis, and that the grouping is performed based on the result of the statistical evaluation. These two aspects can also be combined or performed sequentially to generate at least one point cluster.

[0020] In particular, the raw data points can first be grouped according to their individual descriptors, for example, into descriptor clusters, and then, for example, based on the descriptors of their corresponding raw data points, the descriptor clusters can be assigned cluster descriptors, which correspond, for example, to statistical characteristic values ​​of the descriptors of the raw data points of the corresponding descriptor cluster, such as the mean, median, etc. In a second step, the descriptor clusters can be grouped into point clusters according to further criteria, in particular according to their cluster descriptors.

[0021] The fact that the characteristic parameter of the first point cluster relates to the information gain for determining the position of the motor vehicle can be understood, for example, as indicating the influence of using the first point cluster to determine the position of the motor vehicle on the accuracy of the position determination. The characteristic value can indicate, for example, how clearly the first point cluster is identified, how many other clusters are present in the immediate vicinity of the first point cluster, how different the cluster descriptors or descriptors of the raw data points of the first point cluster are from the other point clusters of at least one point cluster, etc.

[0022] In other words, the feature parameter can express how obvious the feature of the environment area represented by the first point cluster is. The higher the uniqueness (Einzigartigkeit) or singularity of the feature The higher the value, and the lower the density of other features near the corresponding feature, the more obvious the feature can be considered. These characteristics can be quantified by predefined rules or regulations, so that the description of how obvious a feature is can also be quantified according to predefined rules and can therefore be measured.

[0023] In particular, each point cluster can be assigned to a corresponding feature in the surrounding area or can be understood as a corresponding feature.

[0024] The features represented by point clusters are non-semantic properties, so there is no need to assign values ​​to these features.

[0025] The feature information may include, for example, the position of the first point cluster, such as the center or representative position of the raw data points of the first point cluster, the spatial extent of the raw data points of the first point cluster, or other geometric characteristics of the first point cluster or the raw data of the first point cluster. The feature information may also include descriptors of the raw data points and / or cluster descriptors of the first point cluster.

[0026] The characteristic information can be stored based on the characteristic parameters, for example, by storing the characteristic information together with the associated characteristic parameters, or by storing the characteristic information including the characteristic parameters themselves. Alternatively or additionally, the characteristic information of the first point cluster can be stored only if the characteristic parameter is above a predefined limit value. Alternatively or additionally, if the characteristic parameter is below a predefined limit value, the characteristic information can be stored in the form of a flag to indicate that the corresponding point cluster is only of limited suitability for position determination.

[0027] By storing the characteristic information according to the label, a high quality of the reference data is ensured and, in particular, a high accuracy of the position determination according to the reference data can be achieved.

[0028] Since the surrounding area can be described based on non-semantic features, the method according to the improved concept can generate reference data for determining the position largely independent of the content of the surrounding area, i.e., which objects or structures are located within the surrounding area. Therefore, the improved concept can be used universally and flexibly and is particularly independent of the presence of semantic landmarks.

[0029] In particular, the features represented by the point clusters may also involve features in the environment that cannot be intuitively recognized by humans.

[0030] According to a further development, the descriptors of the raw data points serve as a means of identifying features in the environment and taking into account particularly distinctive features for determining their location or storing them as reference data. For example, this utilizes the distribution, especially spatial distribution, of the descriptors of the raw data points for distinctive features, i.e., particularly clearly identifiable features, which differ from non-distinctive features, such as objects with very complex surfaces. Features in the form of feature information are displayed and further used in a form without any meaningful content.

[0031] According to at least one embodiment of the method for providing reference data for determining a position, the method includes acquiring sensor measurement data by means of an environmental sensor system and generating raw data points based on the sensor measurement data.

[0032] According to at least one embodiment, a computing unit assigns a corresponding characteristic parameter to each of the at least one point cluster based on corresponding descriptors of the raw data points of the corresponding point cluster, the characteristic parameter relating to a corresponding information gain for determining the position of the motor vehicle using the corresponding point cluster. Corresponding feature information of the at least one point cluster is stored in a storage unit based on the corresponding characteristic parameters of the point cluster.

[0033] As a result, a large number of features in the surrounding area are identified in a non-semantic manner and provided as reference data for determining the position.

[0034] According to at least one embodiment, a computing unit analyzes the spatial distribution of all point clusters of at least one point cluster. Based on the results of the spatial distribution analysis, a positioning characteristic value is determined for a first point cluster, and, if necessary, positioning characteristic values ​​are determined for all other point clusters of the at least one point cluster. The computing unit then determines characteristic parameters of the first point cluster based on the positioning characteristic values ​​of the first point cluster. The same applies to the characteristic parameters of the other point clusters, if necessary.

[0035] The localization characteristic value quantifies in particular how high the density of the point clusters of the at least one point cluster is at the location of the first point cluster, ie in particular in the vicinity of the first point cluster.

[0036] The more point clusters there are near the first point cluster, the less suitable the first point cluster is for determining the position, or the lower the information gain is, and the first point cluster can make the information gain helpful for position determination.

[0037] Correspondingly, the smaller the density of the point cluster at the position of the first point cluster, the larger the positioning feature value, for example.

[0038] In particular, the larger the positioning feature value for the first point cluster is, the larger the feature parameter of the first point cluster is.

[0039] By prioritizing more strongly localized features in the surrounding area, the reliability or accuracy with which the position can be determined using the reference data is increased.

[0040] According to at least one embodiment, the number of point clusters of the at least one point cluster located in a predetermined subregion of the surrounding area where the first point cluster is located is determined by means of a calculation unit, and the positioning characteristic value is determined based on the number of point clusters in the predetermined subregion.

[0041] For example, the positioning feature value may be additionally determined based on the total number of point clusters of the at least one point cluster, for example, determined as a ratio of the number of point clusters in a preset sub-region to the total number.

[0042] For example, the environment area may be completely divided into a plurality of preset sub-areas, including the sub-area where the first point cluster is located.

[0043] Then, the positioning feature value can be determined based on the average number of point clusters in different sub-regions. For example, the positioning feature value can be determined as the ratio of the number of point clusters in the sub-region where the first point cluster is located to the average number of point clusters in all sub-regions.

[0044] In these embodiments, in particular the average density of the point clusters is taken into account in order to determine the positioning feature value.

[0045] According to at least one embodiment, the greater the number of point clusters located in the preset sub-region of the first point cluster, the smaller the positioning feature value.

[0046] According to at least one embodiment, a calculation unit determines singular feature values ​​of the first point cluster based on descriptors of the raw data points of the first point cluster and based on descriptors of the raw data points of the at least one second point cluster. Feature parameters are determined based on the singular feature values ​​of the first point cluster and, in particular, based on the positioning feature values.

[0047] The singularity or uniqueness of a point cluster or a feature corresponding to a point cluster can be understood as a characteristic value of a deviation of the descriptors of the original data points of the corresponding point cluster or the cluster descriptor of the corresponding point cluster from other point clusters of the at least one point cluster. In particular, the descriptors or cluster descriptors of all point clusters of the at least one point cluster can be used to determine the singularity code of the first point cluster.

[0048] By prioritizing unique or singular features or point clusters, those features for determining position that are as little identical as possible to other features in the surrounding area are particularly preferred. Accordingly, the quality of the reference data, ie, in particular, the accuracy of position determination that can be achieved using the reference data, can be further improved.

[0049] According to at least one embodiment, the greater the difference between the descriptor of the original data point of the first point cluster and the descriptor of the second point cluster, or the greater the difference between the cluster descriptor of the first point cluster and the cluster descriptor of the second point cluster, the greater the singularity eigenvalue of the first point cluster.

[0050] According to at least one embodiment, the singularity characteristic values ​​and the positioning characteristic values ​​are weighted by means of a calculation unit, and the characteristic parameters are determined based on the weighted singularity characteristic values ​​and the weighted positioning characteristic values.

[0051] This makes it possible to assign greater weight to singularity characteristic values ​​or positioning characteristic values, depending on the application or the type of descriptor used.

[0052] According to at least one embodiment, at least one descriptor cluster is generated by means of a calculation unit by grouping the raw data points according to their respective descriptors and independently of their respective spatial positions. The at least one point cluster is generated by spatial grouping of the raw data points, wherein each descriptor cluster of the at least one descriptor cluster is identical to one of the at least one point cluster or is separated to form at least two of the at least one point cluster.

[0053] To generate at least one descriptor cluster, raw data points whose descriptors have similar values ​​are grouped together. Since the spatial position of the raw data points is not taken into account for this, the descriptor clusters can also be spatially discontinuous according to predefined criteria.

[0054] Therefore, a descriptor cluster may be understood to be, in particular, a subset of the raw data points of the group of raw data points which have similar descriptors according to a predefined definition and are formed, in particular, according to known methods for cluster analysis.

[0055] Accordingly, a descriptor cluster can represent one or more features in an environment area. By using descriptors and ignoring spatial location, two spatially separated but otherwise identical or similar objects can, for example, result in raw data points from the same descriptor cluster. For example, two spatially separated walls in an environment area can be assigned to the same descriptor cluster.

[0056] According to at least one embodiment, a cluster descriptor is determined for each descriptor cluster by means of a calculation unit based on the descriptors of the raw data points of the descriptor cluster, for example, by a statistical evaluation of the descriptors of the raw data points. For example, the cluster descriptor may correspond to an average value or another statistical variable of the descriptors of the raw data points of the descriptor cluster.

[0057] Thereby, the raw data points of the descriptor cluster can be further processed jointly, thereby reducing memory and / or computation requirements.

[0058] According to at least one embodiment, all raw data points of the set of raw data points that cannot be assigned to one of the descriptor clusters according to a predefined rule are discarded and are no longer considered for generating reference data.

[0059] According to at least one embodiment, a computing unit determines distinctive feature values ​​of the first point cluster based on descriptors of the original data points of the first point cluster. Feature parameters are determined based on the distinctive feature values, and in particular based on the positioning feature values, and for example based on the singular feature values ​​of the first point cluster.

[0060] Depending on the type of descriptor used or the original data points, the distinctive feature values ​​can be determined differently in each case.

[0061] In particular, in order to determine the distinctive feature value, the distribution of the values ​​of the descriptors of the original data points of the first point cluster can be determined and compared with a predefined standard.

[0062] For example, it can be determined from the distribution whether the descriptor of the first point cluster is unimodal or multimodal, how many local maxima the distribution has, how large the maximum of the local maxima is, how wide the distribution or individual subdistributions of the distribution are, etc. For example, jumps in the distribution can also be taken into account for determining the distinctive feature value.

[0063] According to at least one embodiment, a histogram of the descriptors of the raw data points of the first cluster of points is generated by means of a calculation unit and the histogram is analyzed in order to determine the distinctive feature value.

[0064] The distinctive feature value describes, in particular, how well the first point cluster can be described, for example how clearly the features described by the first point cluster can be recognized in the environment.

[0065] By giving priority to features having a higher distinctive characteristic value, a higher reliability of the position determination can be achieved.

[0066] According to at least one embodiment, a corresponding distinctive feature value is determined for each descriptor cluster of at least one descriptor cluster.The distinctive feature value of a first point cluster then corresponds to the distinctive feature value of the descriptor cluster from which the first point cluster was generated.

[0067] Therefore, the first point cluster inherits the unique feature values ​​of the related descriptor cluster to a certain extent.

[0068] According to at least one embodiment, a distribution of the descriptors of the raw data points of the first point cluster is determined by means of a calculation unit, and the distinctive feature value is determined as a function of the distribution.

[0069] According to at least one embodiment, the characteristic information of the first cluster of points is stored in a storage unit using a computing unit in accordance with distinctive characteristic values.

[0070] In particular, the characteristic information can only be stored on the memory unit, for example, if the distinctive characteristic value is greater than a predetermined further limit value.

[0071] According to at least one embodiment, the distinctive feature values, the singular feature values ​​and the localization feature values ​​are weighted by means of a calculation unit, and feature parameters are determined based on the weighted distinctive feature values, the weighted singular feature values ​​and the weighted localization feature values.

[0072] According to another independent aspect of the improved concept, a method for determining the position of a motor vehicle is described. Here, image data of the motor vehicle's surroundings are generated by an environmental sensor of the motor vehicle. Using a further computing unit of the motor vehicle, the image data is compared with predefined reference data for determining the position, which are stored, in particular, on a digital map or mapping system of the motor vehicle. The position of the motor vehicle is determined based on the comparison result using the further computing unit. The reference data for determining the position are generated using a method for determining reference data for determining the position according to the improved concept.

[0073] According to another independent aspect of the improved concept, a map system, in particular a digital map system, in particular a digital map or a high-definition map, is described for determining the position of a motor vehicle. The map system has a further storage unit, wherein reference data for determining the position is stored on the further storage unit, the reference data being generated by the method according to the improved concept for determining reference data for determining the position of the motor vehicle.

[0074] According to the method according to the embodiment, the memory unit in which the characteristic information is stored is, in particular, a further memory unit of a mapping system.

[0075] According to a further independent aspect of the improved concept, a motor vehicle is specified which has a mapping system for determining a position according to the improved concept.

[0076] According to another independent aspect of the improved solution, a computer system for generating reference data for determining the position of a motor vehicle is described. The computer system has a computing unit and a storage unit. The computing unit is designed to obtain a set of raw data points that depict a predetermined environmental area. The computing unit is designed to determine a descriptor for each raw data point, which characterizes the characteristics of the environmental area at the position of the corresponding raw data point. The computing unit is designed to generate at least one point cluster by grouping the raw data points according to their descriptors. The computing unit is designed to assign characteristic parameters related to information gain for determining the position of the motor vehicle to a first point cluster of at least one point cluster based on the descriptors of the raw data points of the first point cluster. The computing unit is designed to store characteristic information of the first point cluster as reference data for determining the position on the storage unit based on the characteristic parameters.

[0077] The fact that the computing unit is designed to obtain the set of raw data points is to be understood in particular to mean that the raw data points can be read out by means of the computing unit.

[0078] Other embodiments of the computer system according to the improved concept flow directly from different embodiments of the method according to the improved concept for generating non-semantic reference data for determining a position, and vice versa. In particular, the computer system according to the improved concept is designed or programmed to perform the method according to the improved concept for generating non-semantic reference data, or the computer system performs such a method.

[0079] According to another independent aspect of the improved concept, a computer program is described, which includes instructions that, when executed by a computer system, in particular a computer system according to the improved concept, in particular a processing unit of the computer system, cause the computer system to perform the method according to the improved concept for determining non-semantic reference data.

[0080] According to a further independent aspect of the improved concept, a computer-readable storage medium is specified, on which the computer program according to the improved concept is stored.

[0081] The invention also includes combinations of features of the described embodiments.

[0082] Next, embodiments of the present invention will be described. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] In the attached figure:

[0084] Figure 1An exemplary embodiment of a computer system according to an improved solution is shown;

[0085] Figure 2 A flow chart showing an exemplary embodiment of a method for determining non-semantic reference data according to an improved solution; and

[0086] Figure 3 A motor vehicle is shown with an exemplary embodiment of a mapping system according to an improved concept. DETAILED DESCRIPTION

[0087] The embodiments described below are preferred embodiments of the present invention. In the embodiments, the components described in the embodiments each represent features of the present invention that are considered independent of one another, each independently furthering the present invention, and therefore also considered as components of the present invention, either individually or in combinations other than those shown. Furthermore, the described embodiments may be supplemented by other features of the present invention that have already been described.

[0088] In the figures, functionally identical elements are each provided with the same reference numerals.

[0089] exist Figure 1 , a computer system 16 according to an improved embodiment is schematically shown. The computer system 16 comprises a calculation unit 8 and a memory unit 10. A set of raw data points is stored on the memory unit 10, for example, which describes a predetermined surrounding area, ie, a potential surrounding area of ​​the motor vehicle 6.

[0090] The functionality of the computer system 16 is explained in more detail below with reference to an exemplary embodiment of a method for generating reference data according to an improved concept, as for example in Figure 2 As shown in .

[0091] Figure 2 A flow chart of an exemplary embodiment of a method for generating non-semantic reference data for determining the position of motor vehicle 6 is shown.

[0092] In step 1 of the method, the set of raw data points 7 is provided by being stored in a computer-readable form, i.e., in a manner readable in particular by means of a computing unit 8, on a memory unit 10. The set of raw data points 7 corresponds, for example, to or is equivalent to a point cloud previously generated by a lidar system.

[0093] A predetermined descriptor is assigned to each raw data point 7 by means of a calculation unit 8, or a corresponding descriptor is calculated for each raw data point 7 by means of the calculation unit 8, wherein the descriptor particularly characterizes geometric properties of the surrounding area at the position represented by the corresponding raw data point, such as the curvature or average curvature of the environment or an object in the environment at the position of the corresponding raw data point. Additionally or alternatively, depth information may also be used as a descriptor.

[0094] In step 2 of the method, the original data points 7 are grouped into descriptor clusters 12a, 12b, 12c by combining the original data points 7 with similar descriptors. That is, descriptors of the same type are grouped into descriptor clusters 12a, 12b, 12c by clustering.

[0095] In this step, for example, features of the surrounding area 11 that relate to similar objects can be grouped together in a common descriptor cluster 12a, 12b, 12c. For example, all descriptions similar to a wall can be grouped together in a descriptor cluster 12a, 12b, 12c.

[0096] In step 2 , raw data 7 that cannot be assigned to one of the descriptor clusters 12 a , 12 b , 12 d according to predefined criteria can also be cleared and then no longer used.

[0097] In step 2, a distinctive feature value (or also called unique feature value) can be determined for each descriptor cluster 12a, 12b, 12c, in particular based on the descriptors of the corresponding raw data points 7 of the corresponding descriptor cluster 12a, 12b, 12c. The distinctive feature value quantifies the writeability and recognizableness of the descriptors of the corresponding descriptor cluster 12a, 12b, 12c.

[0098] For example, in step 2, the descriptor clusters 12a, 12b, 12c whose distinctive feature values ​​are below a preset limit value may be cleared and not considered further.

[0099] For illustrative purposes, Figure 2 The schematic diagram shows distributions 17a and 17b for two descriptor clusters 12a, 12b, and 12c. For example, distribution 17a may represent the distribution of curvature values ​​as descriptors for an advertising column, while distribution 17b may represent the corresponding distribution for a tree. While distribution 17a indicates high uniqueness because it has three distinct maxima, distribution 17b may not be suitable for position determination because it indicates a relatively uniform distribution of descriptors and, therefore, a low distinctive feature value.

[0100] In step 3 of the method, the descriptor clusters 12a, 12b, 12c are spatially separated from one another, so that corresponding point clusters 9a, 9b, 9c, 9d, 9e, 9f are generated by means of the calculation unit 8. For example, two house walls located at different points in the environment can fall into the same descriptor cluster 12a, 12b, 12c, but into different point clusters 9a, 9b, 9c, 9d, 9e, 9f.

[0101] The point clusters 9a, 9b, 9c, 9d, 9e, 9f take over, in particular, the distinctive feature values ​​of the corresponding descriptor clusters 12a, 12b, 12c, from which the point clusters are respectively generated.

[0102] In step 3, a uniqueness characteristic value or a singularity characteristic value can also be determined for each point cluster 9a, 9b, 9c, 9d, 9e, 9f by means of the calculation unit 8. The uniqueness or singularity of a point cluster 9a, 9b, 9c, 9d, 9e, 9f in particular quantifies how different the corresponding point cluster 9a, 9b, 9c, 9d, 9e, 9f is from the other point clusters 9a, 9b, 9c, 9d, 9e, 9f.

[0103] In particular, the singularity eigenvalues ​​can be determined by using the calculation unit 8 to describe the differences between the descriptors of the individual point clusters 9a, 9b, 9c, 9d, 9e, 9f according to a predefined mathematical rule.

[0104] For example, if there are many house walls in the surrounding area 11 , the singularity eigenvalues ​​of the respective associated point clusters 9a , 9b , 9c , 9d , 9e , 9f are smaller than if there is only one house wall.

[0105] In step 3, in addition to point clusters 9a, 9b, 9c, 9d, 9e, and 9f, localization characteristic values ​​can also be assigned. For this purpose, the spatial distribution 18 of point clusters 9a and 9b can be analyzed. In particular, the computing unit 8 can be used to divide the surrounding area 11 into a predetermined number of subareas, and the number of point clusters within a subarea can be determined. The more point clusters a subarea contains, the smaller the localization characteristic values ​​of the point clusters 9a, 9b, 9c, 9d, 9e, and 9f within that subarea.

[0106] In step 4 of the method, a calculation unit 8 determines, based on the positioning characteristic values, singularity characteristic values, and distinctive characteristic values, a characteristic parameter for each point cluster 9a, 9b, 9c, 9d, 9e, 9f, which relates to the information gain for determining the position of the vehicle. In particular, the higher the positioning characteristic value, singularity characteristic value, and / or distinctive characteristic value of the corresponding point cluster 9a, 9b, 9c, 9d, 9e, 9f, the higher the characteristic parameter.

[0107] The positioning feature value, the singularity feature value, and the distinctive feature value together describe how significant the feature described by the corresponding point cluster 9a, 9b, 9c, 9d, 9e, 9f in the surrounding area 11 is. The more significant the feature, the higher the information gain, and the more valuable the corresponding point cluster 9a, 9b, 9c, 9d, 9e, 9f is for determining the position of the vehicle.

[0108] In step 5 of the method, feature information for each point cluster 9a, 9b, 9c, 9d, 9e, 9f, in particular the corresponding spatial position, range or other geometric information of the point clusters 9a, 9b, 9c, 9d, 9e, 9f, and, for example, the corresponding feature parameters or corresponding feature values ​​of the point clusters 9a, 9b, 9c, 9d, 9e, 9f are stored on the storage unit 10 if the feature parameters related to the information gain of the corresponding point clusters 9a, 9b, 9c, 9d, 9e, 9f are greater than the relevant preset limit value.

[0109] In particular, only those features or point clusters 9a, 9b, 9c, 9d, 9e, 9f which are sufficiently distinct are thus stored on the memory unit as non-semantic reference data for determining the position, which is defined by the limit value.

[0110] Figure 3 A motor vehicle 6 is shown, which has a further processing unit 14 and an environmental sensor system 13 , for example a camera, a lidar system, or a radar system.

[0111] The motor vehicle 6 also has a digital map 15, in particular a further memory unit on which the digital map 15 is stored. The digital map 15 contains reference data for determining a position, which are generated according to the method according to the improved concept.

[0112] Image data of the surroundings of motor vehicle 6 can be generated by environmental sensors 13 of motor vehicle 6 . The image data can be compared with reference data on a digital map 15 by means of a computer unit 14 .

[0113] Based on the comparison in the image data, features represented by the reference data can be identified, so that the position of motor vehicle 6 can be determined.

[0114] As described, the improved concept provides the possibility of providing and using non-semantic reference data for determining the position of a motor vehicle.

[0115] Due to the improved concept, highly precise raw data points can be used to generate distinct reference data for a digital map, in order to enable precise positioning of the motor vehicle.

[0116] In various embodiments, the storage space of the digital map is significantly reduced by combining similar descriptors of the raw data points compared to direct storage of the raw data points.

[0117] Therefore, the improved scheme can build a highly accurate global reference map, in which non-semantic features are stored together with their descriptions.

[0118] Reference Signs List

[0119] 1 Method steps

[0120] 2 Method steps

[0121] 3 Method Steps

[0122] 4 Method Steps

[0123] 5 Method Steps

[0124] 6 Motor vehicles

[0125] 7 original data points

[0126] 8 calculation units

[0127] 9a, 9b, 9c, 9d, 9e, 9f point cluster

[0128] 10 storage units

[0129] 11 Environmental Areas

[0130] 12a, 12b, 12c descriptor clusters

[0131] 13 Environmental Sensors

[0132] 14 computing units

[0133] 15Digital Map

[0134] 16 Computer Systems

[0135] 17a distribution

[0136] 17b distribution

[0137] 18 Spatial distribution

Claims

1. A computer-implemented method for generating non-semantic reference data for determining the position of a motor vehicle (6), wherein: providing a set of raw data points (7), wherein the set of raw data points depicts a predetermined environmental region (11); It is characterized by: - determining, by means of a calculation unit (8), for each raw data point (7) a predetermined descriptor, which characterizes the properties of the surrounding area (11) at the location of the respective raw data point (7); - generating at least one point cluster (9a, 9b, 9c, 9d, 9e, 9f) by means of a calculation unit (8) in such a way that the original data points (7) are grouped according to their descriptors; - assigning, by means of a calculation unit (8), characteristic parameters to a first point cluster of at least one point cluster (9a, 9b, 9c, 9d, 9e, 9f) based on the descriptors of the raw data points (7) of the first point cluster, the characteristic parameters relating to the information gain for determining the position of the motor vehicle (6); and - Based on the feature parameters, the feature information of the first point cluster is stored in a storage unit (10) as non-semantic reference data for determining the position, and with the aid of the calculation unit (8), Analyzing the spatial distribution of all point clusters (9a, 9b, 9c, 9d, 9e, 9f) of at least one point cluster (9a, 9b, 9c, 9d, 9e, 9f), determining a positioning feature value for the first point cluster based on the analysis result of the spatial distribution, wherein the positioning feature value quantifies how high the density of the point clusters of the at least one point cluster is at the position of the first point cluster, and determining the characteristic parameter based on the positioning feature value; and / or Determining a singularity feature value of the first point cluster based on the descriptors of the original data points (7) of the first point cluster and based on the descriptors of the original data points (7) of the second point cluster of at least one point cluster, and determining the feature parameter based on the singularity feature value; and / or Determine the distinctive feature value of the first point cluster based on the descriptor of the original data points (7) of the first point cluster and determine the feature parameter based on the distinctive feature value.

2. The method according to claim 1, characterized in that By means of the calculation unit (8), - determining the number of point clusters located in a predetermined sub-area of ​​the environment area (11) where the first point cluster is located; and - determining a positioning feature value based on said quantity.

3. The method according to claim 1, characterized in that By means of the calculation unit (8), - generating at least one descriptor cluster (12a, 12b, 12c) by grouping the raw data points (7) according to their descriptors and independently of their respective spatial positions; and - at least one point cluster (9a, 9b, 9c, 9d, 9e, 9f) is generated by spatial grouping of raw data points (7), wherein each descriptor cluster (12a, 12b, 12c) is identical to one of the point clusters or is separated to form at least two of the point clusters.

4. The method according to claim 1, wherein By means of the calculation unit (8), - determining the distribution of descriptors of the original data points (7) of said first cluster; and - determining a distinctive feature value based on said distribution.

5. The method according to claim 1, wherein With the aid of the calculation unit (8), feature information of the first point cluster is stored in a storage unit (10) based on the distinctive feature value.

6. The method according to claim 1, wherein Only when the distinctive feature value is greater than or equal to a predetermined limit value is the feature information of the first point cluster stored in the storage unit (10) by means of the calculation unit (8).

7. The method according to claim 1, characterized in that The characteristic parameter indicates how the use of the first cluster of points to determine the position of the motor vehicle (6) affects the accuracy of the position determination.

8. A method for determining the position of a motor vehicle (6), wherein: - generating image data of the surroundings of the motor vehicle (6) by means of an environmental sensor (13) of the motor vehicle (6); - comparing the image data with predefined reference data for determining the position by means of a further computing unit (14) of the motor vehicle (6); and - determining the position of the motor vehicle (6) based on the comparison result by means of a further computing unit (14); It is characterized in that reference data for determining the position are generated by means of a method according to any one of claims 1 to 7 .

9. A mapping system for determining the position of a motor vehicle (6), said mapping system having a further storage unit (15), characterized in that Reference data for determining a position are stored in the further memory unit (15), which reference data are generated by means of a method according to any one of claims 1 to 7.

10. A motor vehicle having a mapping system (15) for determining a position according to claim 9.

11. A computer system for generating non-semantic reference data for determining the position of a motor vehicle (6), the computer system (16) having a computing unit (8) and a memory unit (10), wherein: The calculation unit (8) is designed to obtain a set of raw data points, wherein the set of raw data points describes a preset environmental area (11); Characterized in that the calculation unit (8) is designed to: - determining a descriptor for each raw data point (7), said descriptor characterizing the characteristics of the surrounding area (11) at the location of the respective raw data point; - generating at least one point cluster (9a, 9b, 9c, 9d, 9e, 9f) by grouping the original data points (7) according to their descriptors; - assigning characteristic parameters relating to information gain for determining the position of the motor vehicle (6) to a first point cluster of at least one point cluster (9a, 9b, 9c, 9d, 9e, 9f) based on the descriptors of the original data points (7) of the first point cluster; and - based on the characteristic parameters, storing the characteristic information of the first point cluster as reference data for determining the position in a storage unit (10), Analyze the spatial distribution of all point clusters (9a, 9b, 9c, 9d, 9e, 9f) of at least one point cluster (9a, 9b, 9c, 9d, 9e, 9f), and determine a positioning feature value for the first point cluster based on the analysis result of the spatial distribution, wherein the positioning feature value quantifies how high the density of the point clusters of at least one point cluster is at the position of the first point cluster, and the feature parameter is determined based on the positioning feature value; and / or Determining a singularity feature value of the first point cluster based on the descriptors of the original data points (7) of the first point cluster and based on the descriptors of the original data points (7) of the second point cluster of at least one point cluster, and determining the feature parameter based on the singularity feature value; and / or Determine the distinctive feature value of the first point cluster based on the descriptor of the original data points (7) of the first point cluster and determine the feature parameter based on the distinctive feature value. 12 . A computer program product comprising instructions which, when the computer program is executed by a computer system, cause the computer system to carry out the method according to claim 1 .

13. A computer-readable storage medium having stored thereon the computer program product according to claim 12.

Images