Method for extending an object detected in a parking space of motor vehicles

Abstract

The invention relates to a method for expanding an object detected in a parking space of motor vehicles (2), wherein at least a partial area of ​​the parking space is detected by at least one environmental sensor (7, 9, 18) of at least one motor vehicle (2), and first sensor data (19) are generated in the process, wherein on the basis of the first sensor data (19) a plurality of first data points (pk) and a number of second data points (pk) are generated in a digital map (30), each of which corresponds to a detection of an object at a specific location in the parking space, wherein a first object cluster (CL1) is generated from the first data points (pk) by a density-based clustering method, wherein at least one first scatter parameter (s 1,xx / yy , s 1,jk) of the first data points (pk) of the first object cluster (CL1) is calculated, and wherein the number of second data points (pk) is added to the first object cluster (CL1) to form an extended object cluster (CLE), and based on the first scatter parameter (s 1,xx / yy , s 1,jk ) and based on the number of second data points (pk) without directly using the first data points, at least one extended scatter parameter (s 12,jk ) of the extended object cluster (CLE) is calculated.

Description

[0001] The invention relates to a method for expanding an object detected in a parking space of motor vehicles, wherein data points generated by an environment sensor are arranged with each other using a density-based clustering method.

[0002] For the increasing automation of motor vehicle driving functions, it is desirable to have the most detailed digital maps of parking spaces possible. On the one hand, the process of parking in and out of parking spaces is particularly suitable for automation due to the low speeds involved and the usually very limited interaction with other road users. On the other hand, precise digital maps of parking spaces, such as (large) parking lots with a large number of individual parking spaces or parking garages, enable efficient parking management.

[0003] To improve automation, precise detection of the structural boundaries of parking spaces, especially walls and pillars (the latter primarily in parking garages), is particularly necessary. Precise detection of vehicles in parking spaces can, in turn, be used to improve the mapping of individual parking spaces.

[0004] The aforementioned objects - walls, pillars, and vehicles - can be recognized using data generated by vehicle environmental sensors. Each environmental sensor scans the vehicle's surroundings and uses this data to generate data points in the space surrounding the vehicle, which can then be recorded in a digital two- or three-dimensional map of the parking space. For the actual recognition of the aforementioned objects, a density-based clustering of the data points is usually performed. This means that spatially adjacent data points are first grouped into common objects, and the grouped objects are then recognized according to their object type (wall, pillar, vehicle).The data points belonging together and representing a specific object are examined as clusters for the specific spatial dimensions of the respective detected object, and the associated information about the boundaries of the object is finally used to create the digital map.

[0005] In the clustering described above, grouping and assignment to a specific object is essentially based on the distances between neighboring data points and the resulting density of the data points. However, for complete recognition of a specific object, its spatial orientation must also be captured. For this purpose, a straight line is often placed within a cluster according to a mathematical formula (usually using statistical methods), for example, to detect the orientation of a wall in an xy plane (parallel to the road plane) or the boundary of a vehicle.

[0006] However, this presents the problem that some objects, especially walls, are rarely fully detectable simultaneously by a vehicle's sensors due to their dimensions. This can result in delayed detections of seemingly different objects, which in reality all belong to the same object (e.g., the same wall). This raises the problem of how such delayed, different detections of the same object can be combined into a digital map.

[0007] US 2023 / 0 103 178 A1 describes the merging of so-called "object tracks" in a digital map depending on the distance of a prediction for an extension of an object already detected and recorded in the digital map from a newly detected object. If the newly detected object is sufficiently close to the said prediction for the "continuation" of the existing object, the two objects are merged, or the existing object is extended by the new object. US 2019 / 0 293 782 A1 also discloses a method for merging objects detected by radar or lidar sensors of a motor vehicle, depending on the distance between the said detected objects.

[0008] US 2022 / 0 178 718 A1 mentions a fusion of corresponding detections of a large number of vehicles by means of a global nearest neighbor method executed on a central system across the different detections, for which the individual detections of the different vehicles are previously transformed from the vehicle coordinates into a stationary coordinate system depending on the vehicle movement.

[0009] However, one problem that remains when merging objects is that for each new object, especially for each newly added data point, the statistical methods for calculating the spatial orientation must be completely re-executed. Due to the high computing power and time required, the aforementioned methods are only partially suitable for real-time applications (e.g. for parking assistance during operation).

[0010] The invention is therefore based on the object of specifying a method for expanding objects detected by vehicle sensors, which has the lowest possible computational complexity.

[0011] The stated object is achieved according to the invention by a method for expanding an object detected in a parking space of motor vehicles, wherein at least a partial area of ​​the parking space is detected by at least one environmental sensor of at least one motor vehicle, and first sensor data is generated in the process, wherein a plurality of first data points and a number of second data points are generated in a digital map on the basis of the first sensor data, each of which corresponds to a detection of an object at a specific location in the parking space, and wherein a first object cluster is generated from the first data points by a density-based clustering method.

[0012] According to the method, at least one first scatter parameter of the first data points of the first object cluster is calculated, and the number of second data points is added to the first object cluster to form an extended object cluster. Based on the first scatter parameter and the number of second data points, at least one second scatter parameter of the extended object cluster is calculated without directly using the first data points. Advantageous and, in part, inventive embodiments are the subject of the dependent claims and the following description.

[0013] A parking space for motor vehicles (motor vehicles) is understood in particular to mean a spatially definable area which is intended and equipped for the parking of a plurality of motor vehicles, and preferably has a plurality of parking spaces, each of which is intended and equipped for the parking of a single motor vehicle, i.e. in particular is accessible and passable for a motor vehicle, and is preferably provided with at least one marking in order to demarcate the parking space at least from general traffic routes in the parking space. The parking of a motor vehicle is understood in particular to mean the temporary and / or longer-term parking of a motor vehicle, whereby the engine of the motor vehicle is switched off and, in particular, the driver or all passengers leave the vehicle.

[0014] A parking space is therefore particularly provided by a (large) parking lot with a large number of parking spaces arranged in parking bays or similar, or by a parking garage.

[0015] An environment sensor here includes any device that is configured to detect the environment of the motor vehicle using physical means, i.e. in particular with the aid of electromagnetic waves, and from this to detect at least a distance from the environment sensor for individual objects in the environment. The environment sensor is preferably configured to generate a two- or three-dimensional point cloud from the determined distances of objects in the environment of the vehicle. In particular, an environment sensor here is an active sensor in that the electromagnetic radiation used to detect objects and / or distances to objects is emitted by the sensor itself or by an auxiliary device associated with the environment sensor, for example a radar or lidar sensor.However, a sensor that determines distances to objects based on ultrasound, structure-from-motion, VisualSLAM, and / or direct sparse odometry can also be considered as an environmental sensor.

[0016] A data point in a digital map which corresponds to a detection of at least one partial object at a specific location in the parking space is to be understood in particular as meaning that the environment sensor initially detects at least part of an object as being “present” at a definable point in the parking space, without the object being completely detected in terms of its dimensions or type, and that the detected part of the object is assigned the location or point in the parking space and is represented by the data point in the digital representation of the parking space which the digital map represents.

[0017] In other words, when any part of any object is detected by the environment sensor, the point in the parking space detected by the environment sensor is assigned to it, and this point is recorded as a data point in the digital map. To generate a data point, the first sensor data can be transformed from polar coordinates in the vehicle's reference system into stationary Cartesian coordinates (of the parking space) so that, for a distance from the said partial object detected by the environment sensor in a specific angular direction, location and / or movement information of the vehicle can be taken into account. In particular, a "pre-clustering" of the "raw" first sensor data can also be carried out in such a way that a data point is formed based on a plurality of raw detections recorded by the environment sensor at the same location.

[0018] The ordering of data points to one another using a density-based clustering method specifically means that related objects in the digital map are delineated based on the spatial density of the data points, and areas of lower density separate an object of greater density (i.e., with a higher density of data points) from its surroundings. This specifically includes a DBSCAN method (Density-Based Spatial Clustering of Applications with Noise).

[0019] A first dispersion parameter is now calculated for the first object cluster from the first data points associated with the first object cluster. A dispersion parameter is to be understood in particular as a measure of the dispersion of the respective data points in at least one spatial direction and can in particular be vector- or matrix-valued (and in this case, reflect the dispersion of the respective data points in two Cartesian spatial directions, in particular of the digital map). The dispersion parameter can in particular be used to determine a main direction of the respective object cluster based on a degree of dispersion. For example, if there is a direction in two-dimensional space (parallel to the road plane) in which the dispersion of the data points of an object cluster (i.e. in particular of the first or second object cluster or the extended object cluster) is minimal (see below, Fig. 1), and in particular is smaller than a predefined threshold (e.g., to filter out the "common clustering" of essentially different objects), then, due to the nature of the sensor data, which reflects reflections from surfaces in the parking space, it can be assumed that this is a main direction of a reflective surface (where the main direction is preferably defined perpendicular to the z-direction). A dispersion parameter of an object cluster can, in particular, and in each case in different spatial directions, comprise: a variance, a covariance, a sample covariance, or a corrected sample covariance.

[0020] If the second data points of the second object cluster are now to be added to the first object cluster to combine or create an extended object cluster, e.g. if the corresponding conditions for such a combination are present, an extended scatter parameter, which describes the scatter of the data points of the extended object cluster, is determined according to the invention based on the first scatter parameter of the first object cluster and on the second data points, but without a renewed, direct use of the first data points themselves. In other words, the extended scatter parameter is not calculated using exactly the same method as the calculation of the first scatter parameter, but using the same method and the newly added second data points (possibly also an associated second scatter parameter of the second object cluster). Thus, there is no simple "recalculation" of the scatter parameter.

[0021] By calculating the extended scatter parameter and ultimately the main direction of the extended object cluster based on the first scatter parameter of the first object cluster and the second data points, a complete recalculation can be avoided, which saves computational complexity and thus time.

[0022] The method can, on the one hand, be carried out entirely in a vehicle, so that the digital map only contains the detections of the vehicle in question. The vehicle can also transmit a fully detected object (including all calculated information on its position in the digital map, its dimensions and, if applicable, the object type) to a server, e.g. a backend of the vehicle manufacturer or similar, in which all objects fully detected by all participating vehicles are entered into a digital master map, which, after post-processing of the detected objects using the proposed method (and, if applicable, further post-processing steps), is then made available to all participating vehicles. Individual vehicles can, however, carry out partial steps of the proposed method, for example up to and including the detection of a particular object cluster, and then transmit the result as an intermediate result to the said server ortransferred to the backend, where further calculations for the complete recognition of objects (including the aforementioned post-processing) are carried out, and ultimately the digital master map is created.

[0023] For the density-based clustering method to generate the first object cluster, a DBSCAN algorithm (“Density-Based Spatial Clustering of Applications with Noise”) and / or an OPTICS algorithm (“Ordering Points To Identify the Clustering Structure”) is preferably used. With a DBSCAN algorithm, it is determined for each data point whether a specified minimum number of data points (minPts) is contained within an ε-neighborhood (with a specified distance parameter ε). The data points for which this is the case are “dense” in the sense of the algorithm. Clusters thus extend within the coverage of dense data points.The OPTICS algorithm determines a core distance for each data point, which corresponds to the distance parameter ε of the smallest possible ε-neighborhood containing the specified minimum number of data points (minPts). The data points are then sorted from a starting point according to their next smallest core distance. Both methods are particularly suitable for density-based clustering of the available data points due to the consideration of the ε-neighborhoods and the nature of the initial sensor data.

[0024] Advantageously, a RANSAC (Random Sample Consensus) algorithm is applied to the data points of at least one object cluster, and in particular, a number of data points are eliminated from the object cluster as invalid. In the RANSAC algorithm, a subset, preferably two data points, is randomly selected from the data points of a cluster, and the model parameters—in this case, a line representing the orientation of a wall of the associated object cluster—are calculated from this subset. The data points of the original object cluster that lie within an error bound with respect to the said line (the so-called "consensus set") and those that lie outside the dimensions are then determined. The latter are defined as outliers and eliminated from the object cluster. The described procedure can be iterated until the consensus set no longer changes.By applying the RANSAC algorithm, on the one hand, the boundary of an object can be better recognized, and in particular, in an object cluster which includes two essentially different objects (e.g. two wall sections adjacent to each other at an angle), the RANSAC algorithm can be used to distinguish between the two objects.

[0025] On the other hand, it is advantageous to first apply the RANSAC algorithm to an object cluster and not to all (unclustered) data points, since, for example, the proportion of data points actually belonging to a wall in a DBSCAN cluster covering the wall is significantly higher than the proportion of data points belonging to such a wall in the total set of data points. This also significantly reduces the runtime of the RANSAC algorithm. Even if the RANSAC algorithm is applied multiple times to each detected object cluster, this application is carried out on a much smaller set of data points than if it were applied to the (unclustered / unordered) set of all generated data points. Likewise, the number of possible iterations can be reduced, since, for example, walls can be found more quickly due to the aforementioned significantly higher proportion of data points belonging to the wall in the object cluster.

[0026] Advantageously, a distance of an edge point of the first object cluster to a second data point or to an edge point of a second object cluster formed from the second data points is determined, and depending on the said distance, the second data point or the second object cluster is added to the first object cluster to form the extended object cluster. This comprises in particular: A check is carried out to determine whether the distance to an edge point of the first object cluster for a second data point which is not yet part of the first object cluster does not exceed a predetermined limit. In this case, it can be decided that the second data point still belongs to the first object cluster, and thus the extended object cluster is formed from the first object cluster and the said second data point by merging them. If the predetermined limit for the distance is exceeded, no merging takes place.

[0027] If a DBSCAN and / or an OPTICS method is used to determine the object cluster, the specified limit value is preferably chosen to be larger than the distance parameter ε (DBSCAN) or a maximum reachable distance (OPTICS).

[0028] It is further advantageous if the second object cluster is only added to the first object cluster to form the extended object cluster if the distance is smaller than a specified threshold. If there are multiple object clusters that satisfy the aforementioned distance condition, the second object cluster can first be merged with the first object cluster. The merger of the newly created object cluster with other object clusters can then be tested.

[0029] In an advantageous embodiment, a principal component analysis is calculated for the first object cluster to detect a main direction, and in the process a first covariance matrix is ​​calculated which comprises the at least one first dispersion parameter, wherein a principal component analysis is calculated for the extended object cluster to detect a main direction, and in the process an extended covariance matrix is ​​calculated only on the basis of the mean values ​​of the first object cluster, the entries of the first covariance matrix of the first object cluster, the cardinality of the first and second data points, and the second data points (or the second data point), which comprises the at least one extended dispersion parameter.

[0030] This means in particular: The first dispersion parameter of the first object cluster is given by a first (2x2) covariance matrix, in which the variances or covariances of the xy coordinates of the first data points are present as entries, whereby the sample covariance (normalization over n for n elements) or the corrected sample covariance (normalization over n-1 for n elements) can be used as the covariance. Using this covariance matrix, a main direction of the first object cluster can then be determined as the direction of smallest dispersion of the first data points in the xy plane, whereby this direction can be determined using an eigenvector to the larger eigenvalue of the first covariance matrix.

[0031] If the first object type cluster is now combined with the number of second data points, the extended covariance matrix, i.e. the covariance matrix of the extended object cluster formed by the combination, is formed as the corresponding dispersion parameter based on the mean values ​​of the first object cluster and the entries of its covariance matrix (the first covariance matrix), as well as on the basis of the second data points (or the second data point if there is only one second data point), and on the basis of the cardinality, i.e. the respective numerical size of the sets of first and second data points.

[0032] In the case of a plurality of second data points, the use of the second data points can preferably be carried out by using the mean values ​​of the second data points or an associated second object cluster as well as by using the corresponding entries of the second covariance matrix.

[0033] In an advantageous embodiment, the entries s 12,jk the extended covariance matrix of the extended object cluster calculated as s12,jk=n1−1n1+n2−1s1,jk+n2−1n1+n2−1s2,jk+n1n2(n1+n2−1)(n1+n2)(m1,j−m2,j)(m1,k−m2,k) with the spatial direction indices j,k ∈{x,y}, the entries s 1,jk and s 2,jk the respective covariance of the first or second object cluster (or the second data points), the mean values ​​m 1,j / k and m 2,j / k of the first and second object clusters (or the second data points), and the cardinality n1, n2 of the first and second data points.

[0034] For only a single second data point p n+1 with n1 = n, n2 = 1, equation (i) becomes s12,jk=n−1ns1,jk+1(n+1)(m1,j−pn+1,j)(m1,k−pn+1,k).

[0035] The geometric meaning of the parameters from equation (i) is given by Fig. 1, in which two object clusters CL1 and CL2 are schematically represented from respective data points pk. The first object cluster CL1 has a mean value m1 with the coordinates m 1,x , m 1,y the second object cluster CL2 has a mean value m2 with the coordinates m 2,x , m 2,y (see corresponding dashed lines from the respective center to the coordinate axes x, y). In free units, the variances s 1,xx and s 1,yy of the first object cluster CL1 and the variances s 2,xx and s 2,yy of the second object cluster CL2 are plotted as the respective scatter parameters using corresponding scatter bars. Additionally, the boundary points pr1 of the first object cluster CL1 and pr2 of the second object cluster, along with their distance d, are plotted.

[0036] A proof for equation (i) can be motivated in essential features as follows: The mean value m12 of the extended object cluster can be calculated from the mean values ​​m1, m2 of the first and second object clusters and their thicknesses n1 and n2 respectively to m12=1n1+n2(n1 m1+n2 m2)

[0037] When using the corrected sample variance, the entries s 12,jk the extended covariance matrix given by s12,jk=1n1+n2−1∑i=1n1+n2(pij−m12,j)(pik−m12,k)

[0038] By splitting the sum into a first sum over the first object cluster (i ≤ n1) and a second sum over the second object cluster (n1 +1 ≤ i ≤ n1 + n2) and transforming the mean values ​​m 12,j / k in equation (iii) according to equation (ii) the sums can be put into the following form: 1n1+n2−1∑i=1n1(pij−m1,j+n2n1+n2(m1,j−m2,j))(pik−m1,k+n2n1+n2(m1,k−m2,k))+1n 1+n2−1∑i=n1+1n1+n2(pij−m2,j+n1n1+n2(m2,j−m1,j))(pik−m2,k+n1n1+n2(m2,k−m1,k))

[0039] Taking advantage of the circumstance Σ n1 (p i - m1) = Σ n2 (p i - m2) = 0, only the covariances s remain in the products of the respective sums in equation (iv). 1,jk or s 2,jk corresponding terms (p ij - m 1,j ) (p ik - m 1,k ) or (p ij - m 2,j ) (p ik - m 2,k ) as well as the terms of the differences of the means, which, after ordering the prefactors, result in the form of equation (i).

[0040] Advantageously, the first sensor data is generated using a radar and / or a lidar as the at least one environmental sensor. These environmental sensors exhibit high accuracy (especially with regard to distance and angular direction) in detecting the data points and can also be manufactured cost-effectively.

[0041] Preferably, the partial area of ​​the parking space is detected by at least one environmental sensor of another motor vehicle, generating second sensor data, with the majority of data points also being generated based on the second sensor data. This includes the previously mentioned case where the digital master map can be created on a backend based on the detections from multiple vehicles.

[0042] The invention further relates to a motor vehicle comprising at least one first environmental sensor and a control unit, wherein the motor vehicle is configured to carry out the method described above. In particular, the control unit is configured to carry out the calculations performed in the method by means of a processor and a working memory addressable by the processor, as well as by means of corresponding program instructions in the working memory or in a non-volatile memory.

[0043] The motor vehicle according to the invention shares the advantages of the method according to the invention. The advantages stated for the method and its further developments can be applied analogously to the motor vehicle.

[0044] An embodiment of the invention is explained in more detail below with reference to the accompanying drawings, each of which shows schematically: Fig. 1 data points recorded by a vehicle sensor system, which are grouped into two object clusters, Fig. 2 a motor vehicle with a vehicle sensor system, which has several radar sensors for recording the data points according to Fig. 1 includes, and Fig. 3 shows in a block diagram the sequence of a method for the extension of objects which are detected by means of the vehicle sensor system according to Fig. 2 were recorded.

[0045] Corresponding parts and sizes are provided with the same reference numerals in all figures.

[0046] Fig. Figure 2 schematically shows a motor vehicle 2 with a vehicle sensor system 4. The vehicle sensor system 4 is designed as a vehicle sensor system by means of which the surroundings of the motor vehicle 2 are monitored. The vehicle sensor system 4 has an environment sensor 7 designed as a front radar 6 and two further environment sensors 9 each designed as rear radars 8. In the Fig. 1, the radar areas 10, 12 of the front radar 6 and the rear radar 8 are also shown schematically. In addition, the vehicle sensor system 4 has further environmental sensors 18, with two such environmental sensors 18 being arranged at the front and two at the rear. The environmental sensors 18 are also designed as radar sensors, but can also be designed as lidar sensors. The vehicle sensor system 4 can also have further Fig. 2 sensors not shown, which can be provided, for example, by one or more vehicle cameras, ultrasonic sensors, etc., which, however, do not play a direct role in the method described above.

[0047] The individual environmental sensors 7, 9, 18 (and possibly the additional sensors not shown in detail) of the vehicle sensor system 4 are controlled by a controller 14, which in particular controls the transmission of radar signals by the front radar 6 and the rear radars 8 (and of additional radar signals and / or laser pulses from the environmental sensors 18) and receives the associated sensor signals as first sensor data 19. The sensor signals or the first sensor data 19 are forwarded, possibly after low-threshold preprocessing such as analog preamplification and / or digitization in the controller 14, to an evaluation unit 20, which is connected to the controller 14. However, the evaluation unit 20 can also be connected directly to the radar and the additional environmental sensors 7, 9, 18, and receive the respective sensor signals from them for evaluation.The evaluation unit 20 thus records the sensor signals of the environmental sensors 7, 9, 18 and, when driving into a parking space, which can be determined, for example, using the data of a navigation system (not shown), examines these for the presence of relevant objects in the parking space, in particular walls.

[0048] In Fig. 3 is a schematic block diagram of the sequence of a method by means of which objects which are detected by the vehicle sensor system 4 according to Fig. 2 can be expanded.

[0049] In a first step S1, the radar sensors 6, 8, 18 of the motor vehicle 2 detect the surroundings of the motor vehicle 2 in a parking space (not shown), and from this, data points pk are generated in a digital map 30. In a step S2, two object clusters CL1, CL2 are detected in the data points pk using a density-based clustering method such as DBSCAN. In particular, in an intermediate step S2a, the detection can be further improved by post-processing the clustered data points, e.g., by applying a RANSAC algorithm to better detect the boundaries or edges of the object clusters CL1, CL2.

[0050] In a step S3, a covariance matrix s 1,jk of the first object cluster CL1, which thus comprises a first dispersion parameter of the first object cluster CL1. Based on the covariance matrix s 1,jk(or the dispersion parameter(s), a first principal direction HR1 of the real object in the parking space corresponding to the object cluster CL1 can now be determined. Furthermore, in step S3, a covariance matrix s 2,jk of the second object cluster CL2, which thus comprises a second scatter parameter of the second object cluster CL2.

[0051] For the second object cluster CL2, the distance d of an edge point pr2 to an edge point pr1 of the first object cluster CL1 is determined in a step S4. If the determined distance d is below a predetermined limit value G, i.e., d < G, the two object clusters CL1, CL2 are sufficiently close to one another, which is why they can be combined into an extended object cluster CLE in a step S5, possibly after checking additional criteria not described in detail here, such as an angular deviation of the two main directions of the object clusters and / or an orthogonal distance of a center point of an object cluster from the vector of the main direction HR of the other object cluster.

[0052] In a step S6, the Fig. 1 described type from only the first and second covariance matrix s 1,jk , s 2,jk and the corresponding mean values ​​m 1 / 2of the two object clusters CL1, CL2 and the numbers n1, n2 of the data points pk contained in them, the extended covariance matrix s 12,jk of the extended object cluster CLE is calculated, from which the main direction HRE can be determined in a step S7.

[0053] Although the invention has been illustrated and described in detail by the preferred embodiment, the invention is not limited to the disclosed examples and other variations may be derived therefrom by those skilled in the art without departing from the scope of the invention. List of reference symbols 2 Motor vehicle (motor vehicle) 4 Vehicle sensor system 6 front radar 7 Environment sensor 8 rear radar 9 Environment sensor 10 radar range 12 radar range 14 controllers 18 Environment sensor 19 first sensor data 20 Evaluation unit 30 digital cards CL1 / 2 first / second object cluster CLE extended object cluster d distance G limit HR1 first main direction HRE main direction (of the extended object cluster) m 1 / 2 mean m 1 / 2,x / y x / y coordinate of the mean pk data point pr 12 edge point S1-7 Process steps S2a Process step s 1,xx / yy Variance, first dispersion parameter s 1,jk (first) covariance matrix, first dispersion parameter s 2,xx / yy Variance, second dispersion parameter s 2,jk (second) covariance matrix, second dispersion parameter s 12,jk extended covariance matrix, extended dispersion parameter QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] US 2023 / 0 103 178 A1

[0007] US 2019 / 0 293 782 A1

[0007] US 2022 / 0 178 718 A1

[0008]

Claims

[1] Method for extending an object detected in a parking space of motor vehicles (2), - wherein at least one environmental sensor (7, 9, 18) of at least one motor vehicle (2) detects at least a partial area of ​​the parking space, and first sensor data (19) are generated, - wherein, based on the first sensor data (19), a plurality of first data points (pk) and a number of second data points (pk) are generated in a digital map (30), each of which corresponds to a detection of an object at a specific location in the parking space, - wherein a first object cluster (CL1) is generated from the first data points (pk) by a density-based clustering method, - wherein at least one first scattering parameter (s 1,xx / yy , s 1,jk ) of the first data points (pk) of the first object cluster (CL1) is calculated, and - wherein the number of second data points (pk) is added to the first object cluster (CL1) to form an extended object cluster (CLE), and based on the first scatter parameter (s 1,xx / yy , s 1,jk ) and based on the number of second data points (pk) without directly using the first data points, at least one extended scatter parameter (s 12,jk ) of the extended object cluster (CLE) is calculated. [2] Method according to claim 1, wherein a DBSCAN algorithm and / or an OPTICS algorithm is used for the density-based clustering method for generating the first object cluster (CL1). [3] Method according to claim 1 or claim 2, wherein a RANSAC algorithm is applied to the data points (pk) of the first object cluster (CL1). [4] Method according to one of the preceding claims, wherein a distance (d) of an edge point (pr1) of the first object cluster (CL1) to a second data point (pk) or to an edge point (pr2) of a second object cluster (CL2) formed on the basis of the second data points (pk) is determined, and wherein, depending on said distance (d), the second data point (pk) or the second object cluster (CL2) is added to the first object cluster (CL1) to form the extended object cluster (CLE). [5] Method according to claim 4, wherein the second object cluster (CL2) is added to the first object cluster (CL1) to form the extended object cluster (CLE) only in the event that the distance (d) is smaller than a predetermined limit value (G). [6] Method according to one of the preceding claims, wherein a principal component analysis is calculated for the first object cluster (CL1) to detect a main direction (HR1), and a first covariance matrix is ​​calculated which contains the at least one first scatter parameter (s 1,xx / yy , s 1,jk ), and where a principal component analysis is calculated for the extended object cluster (CLE) to detect a principal direction (HRE), and an extended covariance matrix is ​​calculated only based on - the mean values ​​of the first object cluster, - the entries of the first covariance matrix (s 1,xx / yy , s 1,jk ) of the first object cluster, - the thicknesses of the first and second data points (pk), and - the second data points (pk) are calculated which represent the at least one extended scatter parameter (s 12,jk ) includes. [7] Method according to claim 6, where the entries s 12,jkthe extended covariance matrix of the extended object cluster (CLE) can be calculated as s12,jk=n1−1n1+n2−1s1,jk+n2−1n1+n2−1s2,jk +n1n2(n1+n2−1)(n1+n2)(m1,j−m2,j)(m1,k−m2,k) with - the spatial direction indices j,k ∈{x,y}, - the entries s 1,jk and s 2,jk the respective covariance of the first or second object cluster (CL1, CL2), - the mean values ​​m 1,j / k and m 2,j / k of the first or second object cluster (CL1, CL2), and - the thicknesses n1, n2 of the first and second data points (pk). [8] Method according to one of the preceding claims, wherein the first sensor data (19) are generated using a radar (6, 8) and / or a lidar as the at least one environment sensor (7, 9, 18). [9] Method according to one of the preceding claims, - wherein the partial area of ​​the parking space is detected by at least one environmental sensor (7, 9, 18) of another motor vehicle, and second sensor data are generated in the process, - wherein the plurality of first and / or second data points (pk) are also generated based on the second sensor data. [10] Motor vehicle (2), comprising at least one first environment sensor (7, 9, 18) and a control unit (14, 20), wherein the motor vehicle (2) is configured to carry out the method according to one of claims 1 to 8.

Images