DE102021132024A1 - METHOD OF MEASURING A VICINITY OF A MOTOR VEHICLE - Google Patents

Abstract

A method is proposed for measuring the surroundings of a motor vehicle (1) which has a number of ultrasonic transceivers (2). The method comprises: a) activation (S1) of the number of ultrasound transceivers (2) for emitting ultrasound transmission signals into the area surrounding the motor vehicle (1) and receiving ultrasound reception signals (7) from the area surrounding the motor vehicle (1); b) identifying (S2) a plurality of ultrasonic echoes (8) in the received ultrasonic signals (7) and detecting a respective echo signal representation (9) for each of the identified ultrasonic echoes (8); c) determining (S3) a classification (32, 33) of each of the plurality of echo signal representations (9) as a reflection of the ultrasonic transmission signal on a punctiform object (10) or as a reflection of the ultrasonic transmission signal on a linear object (11); and d) generating (S4) a number of object features (18, 19), which indicate a respective object (10, 11) in the area surrounding the motor vehicle (1), on the basis of the plurality of ultrasonic echoes (8) taking into account the classifications (32, 33) of the multiple echo signal representations (9).

Description

The present invention relates to ultrasonic measuring technology in motor vehicles and, in particular, to a method for measuring an area surrounding a motor vehicle with one or more ultrasonic transceivers.

Motor vehicles are equipped with ultrasonic transceivers, which emit ultrasonic transmission signals into the surroundings of the motor vehicle and receive ultrasonic reception signals from the surroundings of the motor vehicle. A distance to the object in the vicinity of the motor vehicle is determined on the basis of a signal propagation time between the transmission of an ultrasonic transmission signal and the occurrence of an ultrasonic echo in the ultrasonic reception signal, which is due to a reflection of the ultrasonic transmission signal on an object in the vicinity of a motor vehicle. Possible positions of the object, or more precisely of the reflection point, form an ellipse whose focal points form the positions of the transmitting and the receiving ultrasonic transceiver and whose semimajor axis corresponds to the determined distance. The actual position of the reflection point is then determined by trilateration or the like using a number of ellipses obtained at different measurement positions. With trilateration and the like, the problem arises that an object that is essentially point-shaped when viewed in two dimensions, such as a post, is essentially located at the intersection of the ellipses determined from different measurement positions, since in this case the reflections in the multiple measurements are all in the take place essentially at the same point. On the other hand, an extended object such as a wall is usually arranged along a tangent to the ellipses obtained at the various measurement positions, since in this case the multiple reflections usually occur at different points of the extended object. It is often not possible to tell from the resulting image of the multiple ellipses whether reflections are at the same point of a point-like object or reflections at different points of an extended object. Thus, the positions of some reflection points cannot be determined with sufficient accuracy.

In particular when the motor vehicle is parked, a very precise determination of the position of the objects in the vicinity of the motor vehicle is necessary for controlling a door opening assistant, for example, in order to avoid damaging a vehicle door when it is opened.

Against this background, the object of the invention is to further improve the measurement of an environment of a motor vehicle by means of ultrasound.

According to a first aspect, a method for measuring the surroundings of a motor vehicle is proposed, which has a number of ultrasonic transceivers. The method comprises: a) controlling the number of ultrasonic transceivers for emitting ultrasonic transmit signals into the area surrounding the motor vehicle and receiving ultrasonic received signals from the area surrounding the motor vehicle; b) identifying a plurality of ultrasound echoes in the received ultrasound signals received and acquiring a respective echo signal representation for each of the identified ultrasound echoes; c) determining a classification of each of the plurality of echo signal representations as a reflection of the ultrasonic transmission signal on a punctiform object or as a reflection of the ultrasonic transmission signal on a linear object; and d) generating a number of object features, which indicate a respective object in the area surrounding the motor vehicle, based on the plurality of ultrasonic echoes, taking into account the classifications of the plurality of echo signal representations.

The proposed method is based on the idea that, based on a signal form of an individual echo signal (a section of the ultrasonic received signal arranged around an ultrasonic echo), conclusions can be drawn as to whether the ultrasonic echo is due to a reflection of the ultrasonic transmitted signal on an essentially punctiform object or on a reflection of the ultrasonic transmission signal is due to an extended object. This can be caused, for example, by differences in the geometry, in the surface texture and the like of typical two-dimensional objects that are approximately point-shaped, such as posts, corners of buildings and the like, on the one hand, and linear objects, such as extensive walls and walls and the like, on the other.

Due to the proposed pre-classification of the respective echo signal curves - and thus the associated ultrasonic echoes - as reflections either on point-like or on line-like objects, it is advantageously possible in the further course of the method to identify the object features, which at least indicate a position of the respective object in the area surrounding the motor vehicle include, to be determined correctly either on the basis of tangents to or on the basis of intersections of the particular ellipses. Thus, an accuracy of the surroundings measurement can be advantageously improved.

A "number" is here synonymous with "one or more".

An ultrasonic transmission signal can in particular be an impulse applied to a carrier frequency. A received ultrasound signal can be received and preferably recorded over a predefined period of time after the transmission of the ultrasound signal.

With a respective measurement, the ultrasonic transmission signal is transmitted with one of the ultrasonic transceivers. The associated ultrasound reception signal can be received with the same or with a different one of the ultrasound transceivers. That is, a transmitting and a receiving ultrasonic transceiver can be identical or different in a respective measurement. Several measurements can be carried out one after the other or, for example using different carrier frequencies, also simultaneously with different ultrasonic transceivers.

In order to be able to determine a position of an object in the area surrounding the motor vehicle, it is sufficient if at least two measurements are carried out, with at least one of the transmitting ultrasonic transceiver and the receiving ultrasonic transceiver being arranged at different positions for different measurements. It is conceivable that the motor vehicle is equipped with only a single ultrasonic transceiver. In this case, at least two measurements are carried out while the motor vehicle is moving, so that the ultrasonic transceiver is arranged at different positions relative to a stationary object in the vicinity of the motor vehicle for the different measurements. However, the motor vehicle preferably has several, for example four, eight or twelve, ultrasonic transceivers on each side of the vehicle. In this case, at least two measurements can also be carried out when the motor vehicle is stationary using different ultrasonic transceivers.

An ultrasonic echo is to be understood in particular as a reflection of the ultrasonic transmission signal on an object in the vicinity of the motor vehicle. The ultrasonic echo can be identified in the measured ultrasonic received signal or in a recording thereof based on the occurrence of an amplitude maximum, for example using a threshold filter. However, further and complex signal analysis techniques can also be used for identification, which, in addition to an amplitude, also take into account a signal form of a peak at the amplitude maximum and the like.

The result of the identification can be a point in time at which the ultrasonic echo occurs or a transit time of the ultrasonic transmission signal between the transmission of the ultrasonic transmission signal by the transmitting ultrasonic transceiver and the receipt of the ultrasonic echo by the receiving ultrasonic transceiver. Insofar as further steps of the method take place "based on an ultrasonic echo", this also means in particular "based on the temporal occurrence of the ultrasonic echo in the associated ultrasonic received signal" and/or "based on a signal propagation time of the ultrasonic transmitted signal until the ultrasonic echo occurs" and/or "by a measured distance".

An “echo signal representation” is in particular a digital representation of a section of the ultrasound reception signal in a predetermined area in which the ultrasound echo occurs in the ultrasound reception signal. For example, the echo signal representation can include a number of sampled values of the received ultrasound signal and/or an amplitude envelope of the received ultrasound signal. "Capturing" the echo signal representation is in particular the generation of the digital representation based on the ultrasonic received signal. In some embodiments, the identification of the ultrasound echo can also already take place on a digital representation of the ultrasound reception signal; in this case, “capturing” the echo signal representation can be understood to mean the provision of a corresponding section of the digital representation of the received ultrasound signal.

The terms "point" and "line" refer to an idealized shape of the actual object when viewed two-dimensionally in a horizontal plane. This means that the object actually located in the environment does not actually have to be point-shaped or line-shaped. Rather, classification as a “point object” or as a “line object” means that the actual object is described by a two-dimensional point object feature or by a two-dimensional line object feature for the purposes of the method and is therefore described with sufficient accuracy.

Examples of objects classified as point-like are posts, lamp posts, street signs, and the like. Another example of an object to be classified as a point is a corner of a building. In other words, objects that are essentially describable by a single two-dimensional coordinate are classified as "point-like". To put it another way, “point objects” are objects that are at under different measurements performed with ultrasonic transceivers at different positions, the ultrasonic transmit signal is reflected at substantially the same point on the object. “Substantially the same” is understood to mean all points that are identifiable as the same point within an acceptable measurement uncertainty.

Examples of objects classified as line-shaped are walls, walls, and the like. In other words, objects are classified as “lines” that can be described by two two-dimensional coordinates, or by one one-dimensional coordinate, a length specification and an angle specification, or the like. To put it another way, “linear objects” are objects for which the ultrasonic transmission signal is reflected at different points of the object in different measurements that are carried out at different positions of the ultrasonic transceivers involved.

It should also be noted that the term “object in the area surrounding the motor vehicle” is also based on the representations of the area formed by the object features and not on the object that is actually present. For example, a single object that is actually present, such as a building positioned diagonally to the motor vehicle, could be used in the proposed method as a first line-shaped object (first wall), a point-shaped object (corner of the building) and a second line-shaped object (second wall on the other side of the corner) are described. A non-rectilinear object could be approximated by several linear objects.

The classification can be done using heuristic databases, by comparison with reference signal representations, by identifying characteristic signal shapes and/or using a neural network, with artificial intelligence and the like.

An object feature (also "feature") can be understood as a data set that includes a number of data elements with information about at least one position and preferably further information about the position and orientation of a respective object in the area surrounding the motor vehicle. In particular, an object feature is a data record that enables the reconstruction of a two-dimensional representation of a punctiform or linear object on a two-dimensional map of the area surrounding the motor vehicle.

A map of the measured surroundings of the motor vehicle can thus be interpreted as a two-dimensional plot of all specific object features. A map determined in this way can be used, for example, to determine whether and to what extent a door of the motor vehicle can be opened. Thanks to the improved accuracy of the proposed method, door damage can be avoided more reliably.

Within the scope of generating the object features in step d), the position and the like of a respective object in the area surrounding the motor vehicle can be determined by trilateration or the like of a plurality of measurements. A respective “measurement” includes in particular an identified ultrasonic echo, the associated echo signal representation, and preferably topological parameters of an associated measurement arrangement consisting of a transmitting and a receiving ultrasonic transceiver, with which the ultrasonic echo was generated and measured. The classification of the echo signal representation can be taken into account here, for example, when selecting the measurements used for the trilateration or the like and/or when selecting a suitable trilateration technique, for example intersection point formation or tangent fitting. In this way, the object features can advantageously be determined particularly precisely.

According to one embodiment, the respective echo signal representation includes a number of samplings of an envelope curve of the corresponding ultrasonic received signal before and after a temporal occurrence of the corresponding ultrasonic echo.

Such a data-reduced echo signal representation of the respective ultrasonic echo advantageously transports enough information to enable the echo signal representation to be classified as an ultrasonic echo of a punctiform or line-shaped object, while on the other hand it saves enough data to efficiently enable classification using neural networks and the like.

The envelope is in particular an amplitude envelope. The number of scans of the envelope curve before the ultrasonic reception signal is preferably 50, particularly preferably 25 and particularly preferably 10. The number of samplings of the envelope curve after the ultrasonic reception signal is preferably 100, particularly preferably 50 and particularly preferably 20. A sampling rate can in this case be chosen appropriately, so that the echo signal display essentially the entire ultrasonic echo from Moment of increase in the envelope includes until the envelope decays below a predefined threshold.

According to a further embodiment, step c) takes place using a classification device trained by means of machine learning, which is provided with at least the respective echo signal representation as input data and whose output data represent the classification.

The inventors have recognized that by using a classification device trained using machine learning, it is possible to classify echo signal representations as ultrasonic echoes from punctiform or linear objects, which is superior to the human eye and thus to a knowledge-based approach based on heuristic databases and the like.

According to a further embodiment, one or more topological parameters of a measurement arrangement consisting of a transmitting one of the number of ultrasonic transceivers and a receiving one of the ultrasonic transceivers, using which the respective echo signal representation was recorded, are also provided as input data for the classification device.

The inventors have recognized that a classification device trained using machine learning provides more reliable results if it takes into account the topology of the measurement arrangement used, since this can have an influence on the signal shapes contained in the echo signal representations.

According to a further embodiment, the respective topological parameter comprises an installation height of the transmitting and/or the receiving ultrasonic transceiver, an elevation angle of a measuring axis of the transmitting and/or the receiving ultrasonic transceiver and an azimuth angle enclosed by the measuring axes of the transmitting and the receiving ultrasonic transceiver.

If the input data for the classification device include the parameters mentioned, particularly precise classifications can result.

A measuring axis of a respective ultrasonic transceiver is in particular the axis along which a transmitted signal lobe has a maximum. An elevation angle is to be understood in particular as the angle that is enclosed by the measuring axis projected onto a vertical plane to a roadway on which the motor vehicle is standing or driving, with the roadway. An azimuth angle is to be understood in particular as meaning that angle which the two measurement axes projected onto the roadway enclose.

According to a further embodiment, the classification device comprises an artificial neural network trained by supervised machine learning.

The monitored machine learning can take place using reference measurements, in which a plurality of predefined reference objects of known geometry are arranged in an otherwise object-free or otherwise predefined environment of the motor vehicle and associated reference echo signal representations are recorded. Alternatively, the monitored machine learning can also be based on simulation results.

According to a further embodiment, the artificial neural network comprises an input neuron layer, a number of hidden convolutional neuron layers, a number of hidden densely connected multilayer perceptron layers and an output neuron layer.

Such a structure of the neural network delivers particularly good results if the input data of the neural network includes a respective echo signal representation on the one hand and the topological parameters of the associated measurement arrangement on the other. The convolutional neuron layers are advantageous in identifying similarities based on the echo signal representations comprising a plurality of continuously adjacent samples, while the multilayer perceptron layers can particularly advantageously take into account the disjoint topological parameters.

The neural network may conform to an hourglass model where the input neuron layer and the output neuron layer have more neurons than the intermediate hidden layers. A total number of neurons of the neural network can be from 1000 to 10000, particularly preferably from 2000 to 5000, very particularly preferably from 2000 to 3000.

According to a further embodiment, step d) comprises: selecting a first ultrasound echo and one or more further ultrasound echoes from the plurality of identified ultrasound echoes; Determine a locus of possible reflection points for each of the selected Ultra sound echoes; Determining an intersection of the determined loci and/or determining a tangent to the determined loci; validating the determined intersection point and/or the determined tangent; and generating a point feature based on the validated intersection and/or a line feature based on the validated tangent. The classification of at least the echo signal representation for the first ultrasonic echo is used in a decision as to whether an intersection point or a tangent is determined and/or in a decision as to whether a point object feature or a line object feature is generated and/or in a decision which one or which several further ultrasonic echoes are selected is taken into account.

In operation, there can be a large number of current and older measurements that were recorded with a large number of different measurement arrangements. The selection of the measurements (ultrasonic echoes and associated echo signal representations, topological parameters and the like) on the basis of which an attempt is to be made to construct an object feature can be based on spatial criteria - selection of measurements with neighboring ultrasonic transceivers - or temporal criteria - selection of only the most recent measurements - or a combination of these.

If a measurement arrangement was used for a given ultrasonic echo, in which the transmitting ultrasonic transceiver and the receiving ultrasonic transceiver are identical, the locus of possible reflection points for a given ultrasonic echo can be a circle whose center is the position of the ultrasonic transceiver, whose radius is given by the measured distance is, or more precisely a semicircle section of this circle outside the motor vehicle.

If a measurement arrangement was used for a given ultrasonic echo, in which the transmitting ultrasonic transceiver and the receiving ultrasonic transceiver are not identical, the locus of possible reflection points can be an ellipse whose focal points are the positions of the two ultrasonic transceivers and whose semimajor axis is given by the measured distance , or more precisely a semi-ellipse section of this ellipse outside the motor vehicle.

The point of intersection and/or the tangent can be determined by fitting, for example using the least squares method or the like. Because, in a general case where more than two measurements are selected, there may not exist an exact intersection of more than two ellipses and there may not exist a tangent drawn exactly to more than two ellipses. Thus, when determining the locus curve, in addition to the coordinates of the point of intersection or the tangent, a measurement uncertainty can result from the adjustment, which can serve as a measure of the accuracy of the adjustment or the measurement.

The intersection point and/or the tangent can thus be validated, for example, based on the measurement uncertainty determined when determining the intersection point and/or the tangent. The respective measurement uncertainties can, for example, be compared with a threshold value that must not be exceeded. However, other types of validation are conceivable, for example plausibility checks. If, for example, multiple executions of step d) result in object features that contradict one another, the contradictory object features cannot be validated, for example. Heuristic databases can also be used for validation in order to compare the determined object features with combinations of object features that are known to be plausible or implausible.

The validation does not necessarily take place before a respective object feature is generated. A number of object feature hypotheses can also initially be generated by repeatedly selecting and determining loci and points of intersection and/or tangents without validating them. Validation can only take place after several object feature hypotheses have been generated. The individual object feature hypotheses can then either be discarded or confirmed or accepted as generated object features.

A point-object feature is in particular a data set that includes an x-coordinate, a y-coordinate and an indication of a measurement uncertainty. The measurement uncertainty can also be understood as an indication of an "extension" of the "point", or in other words the radius of a circle around the position determined by the x and y coordinates. The point-object feature can thus enable the reconstruction of a two-dimensional circle that indicates or delimits a position of the point-like object.

A line object feature is in particular a data record that includes an x-coordinate, a y-coordinate (start of the line), an angle specification (direction of the line) and a length specification (length of the line). As an alternative to this, a line object feature can, for example, also have two x-coordinates and two y-coordinates (beginning and end of the line). In addition to this, the line object feature include a measurement uncertainty, which can be understood as a "thickness" of the "line", or in other words, as a distance between two lines parallel to the line. The line object feature can thus allow the reconstruction of a two-dimensional bar (length: distance between the reflection points 16, 17; thickness: measurement uncertainty) or a two-dimensional ellipse (major semi-axis: half the distance between the reflection points 16, 17; minor semi-axis: half the measurement uncertainty) that indicates or delimits a position of the linear object.

If only identically classified echo signal representations are selected in the selection step, the generation of erroneous object features can advantageously be further contained. Further possibilities as to how the classification of the echo signal representation of the first ultrasonic echo and preferably also of the further echo signal representations can advantageously be taken into account are discussed in more detail on the basis of the following embodiments.

According to a further embodiment, in step d) both the intersection and the tangent are determined and, if only the intersection is successfully validated, the point feature is created; if only the tangent is successfully validated, the line feature is created; and if both the intersection and the tangent are successfully validated, generating either the point object feature or the line object feature depending on the classification of at least the echo signal representation for the first ultrasonic echo.

Accordingly, the decision of whether to generate a line feature based on the tangent or an intersection based on the tangent can be made using computational validation techniques. The classification therefore advantageously only needs to be carried out and taken into account if a computational validation technique does not provide a clear result, and in this case can advantageously bring about a decision for the correct type of object feature.

According to a further embodiment, in step d) depending on the classification of at least the echo signal representation for the first ultrasonic echo, either the intersection point is determined and, if the intersection point is successfully validated, the point object feature is generated, or the tangent is determined and, if the tangent is successfully validated, the line feature is created.

Particularly preferably, only those ultrasound echoes are selected as the one or more further ultrasound echoes whose echo signal representations have the same classification as the echo signal representation of the first ultrasound echo.

Intersections and tangents are thus advantageously only constructed and validated if this makes sense based on the classification of the echo signals. Calculation effort can be reduced and the quality of the measurement can be improved.

According to a further embodiment, the method also includes e) controlling an operating process of the motor vehicle as a function of the generated object features.

The operating process is not restricted and can denote any partially or fully automatic operating process that depends directly or indirectly on the presence and position of objects in the area surrounding the motor vehicle.

According to a further embodiment, the proposed method is carried out when the motor vehicle is stationary.

When the motor vehicle is stationary, validation techniques based, for example, on the movement of the motor vehicle and the resulting shifts in the positions of the objects in the area surrounding the motor vehicle are not available. It is therefore particularly advantageous here, as proposed, to generate point or line-shaped object features, taking into account the classification of individual echo signal representations.

According to a second aspect, a method for training a classification device for use in the method of the first aspect when using the classification device is proposed. The method for training comprises: arranging a punctiform or linear object in a real or simulated environment of a motor vehicle; Determination of ultrasonic received signals, each of which includes a number of ultrasonic echoes that have arisen through reflection of a respective transmitted ultrasonic signal on the object in the vicinity of the motor vehicle; identifying a plurality of ultrasound echoes in the received ultrasound signals received and acquiring a respective echo signal representation for each of the identified ultrasound echoes; Perform supervised machine learning with the classifier using at least each echo signal representation as training input data, and using classification as reflection on a point-like object or reflection on a line-like object, in accordance with the kind of the placed object, as training output data.

The step of determining the ultrasonic received signals can be carried out, for example, by controlling a plurality of ultrasonic transceivers in the motor vehicle to emit ultrasonic transmitted signals into the real environment of the motor vehicle and receiving ultrasonic received signals from the real environment. In this case, the real environment can be specially prepared and, apart from the arranged object, cannot have any further objects within a measuring range of the ultrasonic transceiver. The real environment can also be an acoustic absorber chamber.

As an alternative to this, the step of determining the ultrasonic received signals can take place by simulating the propagation of ultrasonic transmitted signals in the simulated environment of the motor vehicle and the reception of the ultrasonic received signals.

According to a third aspect, there is proposed a classification device for use in the method of the first aspect using the classification device trained according to the method of the second aspect.

The classification device can be embodied in hardware or in software and can comprise a neural network, for example.

According to a fourth aspect, there is proposed a motor vehicle having a plurality of ultrasonic transceivers and the classifying device of the third aspect.

The motor vehicle can also include a control unit that is set up to carry out the method of the first aspect. The motor vehicle may be an automobile, truck, motorcycle, electric bicycle, and the like.

According to a fifth aspect, a computer program product is proposed which comprises instructions which, when executed by a control unit of a motor vehicle of the fourth aspect, cause the control unit to carry out the method of the first aspect.

The features, advantages and embodiments described for the proposed method according to the first aspect also apply correspondingly to the proposed training method of the second aspect, the proposed classification device of the third aspect, the proposed motor vehicle of the fourth aspect and the proposed computer program product of the sixth aspect.

Further possible implementations of the invention also include combinations of features or embodiments described above or below with regard to the exemplary embodiments that are not explicitly mentioned. The person skilled in the art will also add individual aspects as improvements or additions to the respective basic form of the invention.

• 1 zeigt ein beispielhaftes Kraftfahrzeug mit mehreren Ultraschallsendeempfängern;
• 2 veranschaulicht Details einer Steuereinheit des beispielhaften Kraftfahrzeugs;
• 3 veranschaulicht Schritte eines Verfahrens gemäß einem ersten Ausführungsbeispiel;
• 4 zeigt eine Auftragung einer Hüllkurve eines Ultraschall-Empfangssignals;
• 5 veranschaulicht Messungen an einem punktförmigen Objekt in der Umgebung eines Kraftfahrzeugs;
• 6 veranschaulicht Messungen an einem linienförmigen Objekt in der Umgebung eines Kraftfahrzeugs;
• 7 veranschaulicht ein Punkt-Objektmerkmal;
• 8 veranschaulicht ein Linien-Objektmerkmal;
• 9 veranschaulicht den Schritt des Klassifizierens einer Echosignaldarstellung mit einer beispielhaften Klassifizierungsvorrichtung;
• 10 veranschaulicht Schritte eines Verfahrens gemäß einem zweiten Ausführungsbeispiel;
• 11 veranschaulicht Schritte eines Verfahrens gemäß einem dritten Ausführungsbeispiel;
• 12 veranschaulicht ein beispielhaftes Verfahren zum Trainieren einer Klassifizierungsvorrichtung;
• 13 veranschaulicht ein beispielhaftes Trainingsverfahrens; und
• 14 veranschaulicht Schritte des beispielhaften Trainingsverfahrens.

Further advantageous refinements and aspects of the invention are the subject matter of the dependent claims and of the exemplary embodiments of the invention described below. The invention is explained in more detail below on the basis of preferred exemplary embodiments with reference to the attached figures.

• 1 Fig. 12 shows an exemplary motor vehicle with multiple ultrasonic transceivers;
• 2 illustrates details of a control unit of the example motor vehicle;
• 3 illustrates steps of a method according to a first embodiment;
• 4 shows a plot of an envelope of an ultrasonic received signal;
• 5 illustrates measurements on a point-like object in the vicinity of a motor vehicle;
• 6 illustrates measurements on a linear object in the vicinity of a motor vehicle;
• 7 illustrates a point feature;
• 8th illustrates a line object feature;
• 9 Figure 12 illustrates the step of classifying an echo signal representation with an exemplary classifier;
• 10 illustrates steps of a method according to a second embodiment;
• 11 illustrates steps of a method according to a third embodiment;
• 12 illustrates an example method for training a classifier;
• 13 illustrates an example training procedure; and
• 14 illustrates steps of the example training method.

Elements that are the same or have the same function have been provided with the same reference symbols in the figures, unless otherwise stated.

1 1 shows an exemplary motor vehicle 1 with a plurality of ultrasonic transceivers 2 and a control unit 3. The plurality of ultrasonic transceivers 2 are arranged along one side of the motor vehicle 1 and are connected to the control unit 3 via a communication line 4. FIG.

2 illustrates details of the control unit 3. The control unit 3 comprises a control device 5 and a classification device 6. The control device 5 is set up to carry out the method described below. For this purpose, the control device 5 carries out processing steps, sends control signals to the ultrasonic transceivers 2 and receives measurement signals from them, as described in detail below. The classification device 6 is used for the classification of echo signal representations, which will be described later. The control unit 3 (control device 5, classification device 6) is implemented by one or more control units or ECUs of the motor vehicle 1, for example.

3 1 illustrates steps of a method for measuring surroundings of motor vehicle 1 according to a first exemplary embodiment. It will be on 1 until 3 referenced.

In step S1, the control device 5 controls the ultrasonic transceivers 2 and thereby causes them to emit an ultrasonic transmission signal into the lateral surroundings of the motor vehicle 1 in each case. For example, each of the ultrasonic transceivers 2 is caused to emit a control signal at a different carrier frequency. As an alternative to this, the ultrasonic transceivers 2 can be controlled at different times. In addition, the control device 5 receives an ultrasonic reception signal from each of the ultrasonic transceivers 2 . Each of the ultrasonic transceivers 2 delivers to the control device 5 an ultrasonic received signal that contains essentially only ultrasound with that carrier frequency with which the same ultrasonic transceiver 2 previously transmitted the ultrasonic transmitted signal.

It will be on 1 until 4 referenced. 4 shows a plot of an ultrasonic received signal 7 supplied by one of the ultrasonic transceivers 2. More precisely, FIG 4 the amplitude envelope of the ultrasound received signal 7 is plotted. The time t since the transmission of the ultrasonic transmission signal is plotted on the horizontal axis, and the amplitude A is plotted on the vertical axis. At a time t0=0 a high amplitude is measured; this is a reverberation of the transmitted ultrasonic signal, which is caused by the after-oscillation of an ultrasonic membrane of the ultrasonic transceiver 2 and the like. At time t1 the reverberation has subsided. At time t2 the amplitude rises again, reaches a maximum at time t3 and then falls again. At time t4 the peak has subsided again.

In step S2, the control device 5 identifies the maximum in the ultrasound reception signal 7 at t3 as an ultrasound echo 8, which was created by reflection of the transmitted ultrasound transmission signal on an object in the vicinity of the motor vehicle 1. The control device 5 detects a number of scans of the envelope of the ultrasound received signal 7 in the area of the ultrasound echo 8, more precisely from time t2 to time t4 as an echo signal representation 9 of the ultrasound echo 8. For example, 10 scans before and 20 scans after the maximum are detected.

Based on the temporal occurrence of the maximum of the ultrasonic echo 8 in the ultrasonic received signal 7, a distance to a point on the object in the vicinity of the motor vehicle 1, from which the ultrasonic echo was reflected, is determined. Specifically, in the present example, in which the ultrasonic echo is received by the same ultrasonic receiver 2 that sent out the ultrasonic transmission signal, the distance is a=0.5*(t3−t0)*c, with the speed of sound c in air .

The position of objects in the area surrounding motor vehicle 1 is to be determined by trilateration using a number of such measurements, which are carried out from different positions, as described below.

5 and 6 illustrate measurements on a punctiform object 10 or on a linear object 11 in an area surrounding motor vehicle 1. 5 and 6 show a schematic projection of the motor vehicle 1 and the objects 10 and 11 onto a two-dimensional road plane when viewed in the vertical direction z from above.

Suppose a first ultrasonic transceiver 2a has identified an ultrasonic echo 8 ( 4 ) a distance a to the object 10, 11 is determined. A point 16, 17 of the object 10, 11, at which the ultrasonic echo 8 ( 4 ) has been reflected is on a circle 12, the center of which is the position of the ultrasonic transceiver 2a and the radius of which is the measured distance a. The circle 12 is an example of a locus of possible reflection points 16, 17. Assuming that a second ultrasonic transceiver 2b has determined a distance b from the object 10, 11. A circle 13 with the second ultrasound transceiver 2b as the center and the distance b as the radius forms another locus of possible reflection points 16, 17 of the ultrasound transmission signal emitted by the ultrasound transceiver 2b.

If the object is the in 5 If the punctiform object 10 shown is - like a post or the like - then the reflection points 16, 17 where the circles 12, 13 touch the punctiform object 10 are close to each other and close to the intersection 14 of the circles 12 and 13. The intersection 14 therefore represents a good approximation for the position of the object 10, which is in the form of a point in a first approximation.

If, on the other hand, the object is the in 6 shown extended, linear object 11 - like a wall or the like - acts, but otherwise, as in 6 shown, the same distances a and b were determined, then the reflection points 16, 17, at which the circles 12, 13 touch the line-shaped object 11, do not coincide or even approximately, but are far apart. In this case, the point of intersection 14 of the circles 12 and 13 does not represent a good approximation for the position of the linear object 11. Rather, the position of the linear object 11 is better described in this case by a line on a tangent 15 to the loci 12, 13 .

It will be on 2 until 6 referenced. It should be noted that the measured distances a and b in the examples in 5 and 6 are the same in each case and therefore do not give any indication as to whether the punctiform object 10 or the linear object 11 is in the vicinity of the motor vehicle 1 .

Typical point objects 10, i. H. Objects whose position is well approximated by the intersection 14 of several loci 12, 13, since all conceivable reflection points 16, 17 are close together, are posts, lanterns, street signs, trees, corners of walls and walls, and the like. These each have a characteristic surface structure. Typical linear objects 10, whose position is better approximated by the tangent 15 to several loci 12, 13, are walls, walls, hedges and the like, and also each have a characteristic surface structure that differs from the surface structure of the punctiform objects. The surface structure of a particular type of object affects the amplitude and the specific signal form of the ultrasonic echo 8 in the received ultrasonic signal 7 . The amplitude and signal form can be read from the echo signal representation 9 recorded in step S3.

In step S4 the control device 5 therefore uses the classification device 6 in order to classify the recorded echo signal representations 9 . The classification device 6 receives at least one respective echo signal representation 9 as an input and provides a classification as an output that indicates whether the echo signal representation 9 contains an ultrasonic echo 8 that was reflected from a point-shaped object 10, or contains an ultrasonic echo 8 that was reflected from a line-shaped object 11 was reflected. The classification device 6 can determine the classification by comparison with previously stored reference echo signal representations and/or by means of artificial intelligence or the like.

In step S5, the control device 5 creates a number of object features 18, 19 ( 7 , 8th ), which indicate a respective one of the objects 10, 11 in the area surrounding the motor vehicle 1. It calculates how based on 5 and 6 was described, points of intersection 14 and/or tangents 15 and takes into account the classification determined in step S4 when deciding whether, based on a respective point of intersection 14, a point object feature 18 ( 7 ) or based on a respective tangent 15 a line object feature 19 ( 8th ) is created.

7 FIG. 12 illustrates a point object feature 18 that indicates the point-like object 10. FIG. It will be on 5 and 7 referenced. The point-object feature 19 includes the x-coordinate 20 and the y-coordinate 21 of the point of intersection 14 of the loci 12, 13 as well as a measurement uncertainty 22 of the coordinates 20, 21. The measurement uncertainty 22 is determined, for example, as the least square of an adjustment of the point of intersection 14 in a case where more than two loci 12, 13 do not intersect exactly at a single intersection 14. The point-object feature 18 thus allows the reconstruction of a circle with the coordinates 20, 21 of the point of intersection 14 and a radius determined by the measurement uncertainty 22 in a two-dimensional map of the area surrounding the motor vehicle 1. As can be seen in 5 sees, a circle around the intersection point 14 would be slightly closer to the motor vehicle 1 than the actual punctiform object 10. However, it is never further away from the motor vehicle 1 than the actual punctiform object 10. It is therefore the circle that can be reconstructed from the point object feature 18 around a conservative estimate of the position of the point-shaped object 10.

8th FIG. 12 illustrates a line object feature 19 which indicates the line-shaped object 11. FIG. It will continue on 6 and 8th referenced. The line object feature 20 includes the x-coordinate 23 and the y-coordinate 24 of the first reflection point 16 at which the tangent 15 touches the locus 12, and also includes an angle specification 25 and a length specification 26. The angle specification 25 and the length specification 26 are set in such a way that a line from the first reflection point 16 to the second reflection point 17 can be reconstructed from the line object feature 19 . Furthermore, the line object feature includes a measurement uncertainty 27. The measurement uncertainty 27 is determined, for example, as the square error of an adjustment of the tangent 15 in a case with more than two loci 12, 13, in which the tangent 15 with the smallest possible error square to the plurality of loci 12, 13 is adjusted. A bar or an ellipse can thus be constructed from the line object feature 19 which extends from the first to the second reflection point 16 to the second reflection point 17 and has a thickness determined by the measurement uncertainty 27 . Due to the consideration of the measurement uncertainty 27, the line object feature 19 also represents a conservative estimate of the position of the extended or linear object 10.

It should be noted that the corners 28 , 29 of the linear object 11 are not covered by the two-dimensional feature (bar or ellipse) that can be reconstructed from the line object feature 19 . These corners would, however, be detected as separate point-like objects if a measurement were actually carried out with a sufficient number of ultrasonic transceivers 2, so that at the location of the corners 28, 29 further circles would be reconstructed from further point-object features and overall a sufficiently exact representation of the object 11 caused by several object features.

With reference to 1 , 4 , 5 , 6 it should be noted that, for the sake of simplicity, it was described above that a separate ultrasound transceiver 2a, 2b emits an ultrasound transmission signal and receives the ultrasound reception signal with the associated ultrasound echo therein. In fact, however, it is also conceivable for a measurement arrangement to include a first of the ultrasonic transceivers 2 and a second ultrasonic transceiver 2 that differs therefrom. The first ultrasound transceiver 2 transmits an ultrasound transmission signal, and the second ultrasound transceiver 4, which is different therefrom, receives an ultrasound reception signal which contains the associated ultrasound echo 8 of the transmitted ultrasound transmission signal. In this case, a locus curve 12, 13 of possible reflection points results as an ellipse with the positions of the two ultrasonic transceivers 2 involved as the focal point, the large semi-axis of which is equal to the specific "distance" (here more correctly as "half the distance of the ultrasonic signal from transmission to reception". denote) a, b is. In this way, by suitably permuting the ultrasonic transceivers 2, which can be done sequentially or simultaneously in frequency multiplex, with a number n ultrasonic transceivers 2, a total of n*(n-1) different measurement arrangements can be formed, each consisting of a transmitting and an associated receiving ultrasonic transceiver 2. In this way, more locus curves can be determined and, correspondingly, more object features can be generated in the area surrounding motor vehicle 1 for more precise measurement.

9 illustrates the step of classifying an echo signal representation 9 with an exemplary classification device 6 according to an advantageous development of the first exemplary embodiment.

The classification device 6 in 9 includes an artificial neural network that is trained using monitored machine learning and is therefore set up to classify the echo signal representation 9 based on an input data vector 30, which includes at least the echo signal representation 9, and as an output data vector 31, either a classification 32 as a reflection on a punctiform object or a classification 33 output as a reflection of a linear object.

In addition to the echo signal representation 9, the input data vector 30 preferably also includes at least three topological parameters 34-36 that specify the measurement arrangement with which the echo signal representation 9 was recorded.

The first topological parameter 34 preferably includes an installation height of the transmitting and/or the receiving ultrasonic transceiver 2 ( 1 ) of the measurement arrangement. The second topological parameter 36 preferably includes an elevation angle of a measurement axis of the transmitting and/or the receiving ultrasonic transceiver 2 ( 1 ). The third topological parameter 37 preferably comprises one of the measurement axes of the transmitting and receiving ultrasonic transceiver 2 ( 1 ) included azimuth angle. In this way, when classifying The topology of the measuring arrangement can also be advantageously taken into account.

The artificial neural network of the classification device 6 preferably comprises, in addition to an input neuron layer to which the input data vector 31 is input and an output neuron layer which outputs the classification 32 or 33, at least three hidden, convolutional neuron layers and at least three hidden, densely connected multilayer perceptron layers. The layers may correspond to an hourglass model, in which intermediate hidden layers have fewer neurons than outer layers. Good results have been obtained with a total number of neurons in the low four-digit range, e.g. B. from 2000 to 3000 achieved.

10 illustrates steps of a method according to a second embodiment. The second exemplary embodiment is based on the method of the first exemplary embodiment, with the proviso that not just two but a large number of measurements are carried out with a large number of different measuring arrangements. Sub-steps S11-S16 of a specific embodiment of step S4 ( 3 ) determining the object features 18, 19 ( 7 , 8th ) described. It will be in particular 10 and 4 until 6 referenced.

In (partial) step S11, a first measurement and one or more further measurements are selected from the plurality of measurements. A measurement in this case includes an ultrasonic echo 8 and the associated echo signal representation 9. In the present example, a set of measurements is selected in this way that was recorded or updated in a most recent measurement run.

In step S12, a locus curve 12, 13 of possible reflection points 16, 17 is determined for each selected measurement in the manner described for the first exemplary embodiment.

In step S13, on the basis of 5 described manner, a best possible adapted intersection 14 of the loci 12, 13 of all selected measurements and an associated measurement uncertainty of the intersection 14 is determined.

In step S14, on the basis of 6 manner described, a tangent 15 is adapted to the loci 12, 13 of all selected measurements and an associated measurement uncertainty of the tangent 15 is determined.

In step S15, the tangents 15 and points of intersection 14 determined are validated at least on the basis of the measurement uncertainties determined. Further heuristic considerations can be included in the validation.

If in step S15 only the intersection 14 was successfully validated and the tangent 15 was not successfully validated, step S15 branches to step S161. In step S161, a point object feature 18 ( 7 ) is generated based on the intersection point 14 and the measurement uncertainty of the intersection point 14.

If in step S15 only the tangent 15 was successfully validated and the intersection was not successfully validated, step S15 branches to step S162. In step S162, a line object feature 19 ( 8th ) based on the tangent 15, the reflection points 16, 17 and the measurement uncertainty of the tangent 15.

If neither the tangent 15 nor the intersection 14 has been successfully validated, the in 10 Processing flow shown without creating an object feature.

In addition to 10 and 4 until 6 will also be on 9 referenced. If both the tangent 15 and the point of intersection 14 were successfully validated in step S15, the process branches to step S163. In step S163, recourse is made to the classifications 32, 33 of the echo signal display 9 of the selected measurements. In this case, only the classification 32, 33 of the echo signal representation 9 of the first selected ultrasonic echo 8 can be taken into account, or that classification 32, 33 which occurs most frequently among the selected measurements can be taken into account. If the classification 32, 33, which is taken into account in this way, is a classification 32 as a reflection on a point-like object, a point-object feature 18 ( 7 ) he follows. If the classification 32, 33 taken into account is a classification 33 as a reflection on a linear object, a line object feature 19 ( 8th ) generated.

According to the second exemplary embodiment, the classifications 32, 33 determined by the classification device 6 are used to decide ambiguous cases if the validation in step S15 alone does not indicate whether a point object feature 18 or a line object feature 19 is to be generated.

11 and 12 illustrates steps of a method according to a third embodiment. The third embodiment is based on the method of the first embodiment in the same manner as the second embodiment. Based on the third embodiment Sub-steps S11-S16 of a further possible embodiment of step S4 ( 3 ) of determining the object features.

It will be on 11 , 9 and 4 until 6 Reference will be made and a first portion of the method of the third embodiment described. In step S11, a first measurement and one or more further measurements are selected from the plurality of measurements. In the third exemplary embodiment, further measurements are selected from the same or earlier measurement runs, which were acquired by means of ultrasonic transceivers 2 which are in close proximity to the ultrasonic transceiver 2 with which the first selected measurement was acquired. As a further condition, in the first section of the method in step S11 only measurements are selected whose respective echo signal representation 9 has a classification 32 as a reflection on a point-like object.

In step S12, a locus curve 12, 13 of possible reflection points 16, 17 is determined for each selected measurement in the manner described for the first exemplary embodiment.

In step S13, on the basis of 5 described manner, a best possible adapted intersection 14 of the loci 12, 13 of all selected measurements and a measurement uncertainty of the intersection 14 is determined.

In step S15, the point of intersection 14 determined is validated at least on the basis of the measurement uncertainties determined. If the intersection point 14 was successfully validated in step S15, the process continues with step S16. Otherwise no feature is generated.

In step S16, a point object feature 18 ( 7 ) is generated based on the intersection point 14 and the measurement uncertainty of the intersection point 14.

It will be on 12 , 9 and 4 until 6 referenced. The second portion of the method of the third embodiment 12 corresponds to the first section 11 with the proviso that in step S11 only measurements are selected whose echo signal representation 9 has a classification 33 as a reflection on a line-shaped object, that step S14 is carried out instead of step S13, in which a tangent 15 is adapted to the loci 12, 13 and a measurement uncertainty of the tangent 15 is determined, that the tangent 15 is validated in step S15, and that a line object feature 19 ( 8th ) is generated based on the tangent 15, the reflection points 16, 17 and the measurement uncertainty of the tangent 14.

According to the third exemplary embodiment, the classification 32, 33 of each echo signal representation 9 is taken into account from the start. Point object features 18 are generated only from measurements classified as reflection on point objects, and line object features 19 are generated only from measurements classified as reflection on line objects. The accuracy can be further improved.

13 illustrates an exemplary method for training a classifier 6, and 14 illustrates steps of the example method for training. It will be on 13 , 14 , and 4 until 6 referenced.

In step S21, a reference object, which can be a punctiform object 10 or a linear object 11, is arranged in an area surrounding motor vehicle 1. The arrangement can take place physically, or a simulated reference object 10, 11 can be arranged in a simulated environment of the motor vehicle 1.

In step S22, received ultrasound signals 7 are determined, each of which includes one or more ultrasound echoes 8 that were created by reflection of a transmitted ultrasound signal on reference object 10, 11 in the vicinity of motor vehicle 1. The ultrasonic reception signals 7 can be measured in the same way as described for the first exemplary embodiment of the measurement method. As an alternative to this, the propagation of the ultrasonic transmission signals, their reflections on the reference object 10, 11 and the respective resulting ultrasonic reception signal 7 can be simulated.

In step S23, ultrasonic echoes 8 are identified in each of the ultrasonic received signals 7 determined in step S22 and the associated echo signal representations 9 are recorded in the same way as described with reference to the first exemplary embodiment of the measurement method.

In step S24, the artificial neural network of the classification device 6 is trained using monitored machine learning. This means that the artificial neural network of the classification device 6 is provided with a respective echo signal representation 9 and preferably also the associated topological parameters 34-36 as training input data 37, and as training output data 38 the classification device 6 is provided with whether the reference object is a punctiform object 10 or a linear object 11, either the classification 32 as a reflection on a punctiform object or the classification 33 as a reflection on a linear object is provided. A supervised machine learning step is then performed on the training input data 37 and the training output data 38 .

Steps S21 to S24 are repeated many times with different reference objects, different measurement arrangements and so on until the artificial neural network of the classification device 6 is sufficiently trained.

Although the invention has been described on the basis of exemplary embodiments, it can be modified in many ways within the scope of the appended claims.

Various exemplary embodiments of the method for measuring the surroundings of motor vehicle 1 have been described. The respective method can be carried out when the motor vehicle 1 is moving or stationary.

Using the object features generated according to the method, a map of the surroundings of the motor vehicle 1 can be created, which consists of circles (points with radii that correspond to measurement inaccuracy) and bars (lines with a thickness that corresponds to measurement inaccuracy or ellipses with a minor axis, which corresponds to half the measurement uncertainty). This map of the surroundings of motor vehicle 1 can then be used to determine whether and to what extent a door (not shown) of the motor vehicle can be opened, whether it is safe for moving motor vehicle 1 to change lanes, whether parking maneuvers for motor vehicle 1 are possible, The parking process of the motor vehicle 1 can be controlled partially or fully automatically using the map created in this way, and the like. The above-mentioned operating processes of the motor vehicle can thus be controlled as a function of the object features 18, 19 generated.

Reference List

1

motor vehicle

2, 2a, 2b

ultrasonic transceiver

3

control unit

4

communication line

5

control device

6

classification device

7

Ultrasonic receiving signal

8th

ultrasonic echo

9

echo signal representation

10

point object

11

line object

12, 13

Locus of possible reflection points

14

Intersection of the locus curves

15

Tangent to the locus curves

16, 17

reflection points

18

point feature

19

line feature

20, 21

x, y coordinate of the point feature

22

Measurement uncertainty of the point feature

23, 24

x, y coordinate of the line feature

25

Angle specification of the line feature

26

Length specification of the line feature

27

Measurement uncertainty of the line feature

28, 29

corners

30

input data

31

output data

32

Classification as a reflection from a point object

33

Classification as a reflection on a line-shaped object

34-36

topological parameters

37

training input data

38

training output data

S1-S163

process steps

A

amplitude

away

distances

t

Time

t0-t4

times

x, y

Directions parallel to a road plane

e.g

vertical direction

Claims (15)

Method for measuring an area surrounding a motor vehicle (1) which has a number of ultrasonic transceivers (2), the Ver driving comprises: a) activating (S1) the number of ultrasound transceivers (2) for emitting ultrasound transmission signals into the area surrounding the motor vehicle (1) and receiving ultrasound reception signals (7) from the area surrounding the motor vehicle (1); b) identifying (S2) a plurality of ultrasonic echoes (8) in the received ultrasonic signals (7) and detecting a respective echo signal representation (9) for each of the identified ultrasonic echoes (8); c) determining (S3) a classification (32, 33) of each of the plurality of echo signal representations (9) as a reflection of the ultrasonic transmission signal on a punctiform object (10) or as a reflection of the ultrasonic transmission signal on a linear object (11); and d) generating (S4) a number of object features (18, 19), which indicate a respective object (10, 11) in the area surrounding the motor vehicle (1), on the basis of the plurality of ultrasonic echoes (8) taking into account the classifications (32, 33) of the multiple echo signal representations (9). procedure after claim 1 , wherein the respective echo signal representation (9) comprises a number of samples of an envelope curve of the corresponding ultrasonic received signal (7) before and after a temporal occurrence of the corresponding ultrasonic echo (8). Method according to one of the preceding claims, wherein step c) takes place using a classification device (6) trained by means of machine learning, which is provided with at least the respective echo signal representation (9) as input data (30) and whose output data (31) contains the classification (32, 33). procedure after claim 3 , wherein as input data (30) for the classification device (6) also one or more topological parameters (34-36) of a measurement arrangement from a transmitting of the number of ultrasonic transceivers (2) and a receiving of the number of ultrasonic transceivers (2), using which the respective Echo signal representation (9) was detected, are provided. procedure after claim 4 , wherein the respective topological parameter (34-36) is an installation height of the transmitting and/or the receiving ultrasonic transceiver (2), an elevation angle of a measuring axis of the transmitting and/or the receiving ultrasonic transceiver (2) and one of the measuring axes of the transmitting and the receiving Includes ultrasonic transceiver (2) included azimuth angle. Procedure according to one of claims 3 until 5 , wherein the classification device (6) comprises an artificial neural network trained by supervised learning. procedure after claim 6 wherein the artificial neural network comprises an input neuron layer, a number of hidden convolutional neuron layers, a number of hidden densely connected multilayer perceptron layers, and an output neuron layer. A method according to any one of the preceding claims, wherein step d) comprises: - selecting (S11) a first ultrasonic echo (8) and one or more further ultrasonic echoes (8) from the plurality of identified ultrasonic echoes (8); - Determining (S12) a locus (12, 13) of possible reflection points for each of the selected ultrasonic echoes (8); - Determining (S13) an intersection (14) of the determined loci (12, 13) and/or determining (S14) a tangent (15) to the determined loci (12, 13); - validating (S15) the determined intersection (14) and/or the determined tangent (15); and - Generating (S16) a point object feature (18) based on the validated intersection (14) and/or a line object feature (19) based on the validated tangent (15), the classification (32, 33) at least the echo signal representation (9) for the first ultrasonic echo (8) in a decision as to whether an intersection (14) or a tangent (15) is determined, and/or in a decision as to whether a point object feature (18) or a line object feature ( 19) is generated and/or is taken into account in a decision as to which one or which several further ultrasonic echoes (8) are selected. procedure after claim 8 , wherein in step d): both the point of intersection (14) and the tangent (15) are determined; if only the intersection (14) is successfully validated, creating the point feature (18); if only the tangent (15) is successfully validated, the line object feature (19) is created; and if both the intersection (14) and the tangent (15) are successfully validated, depending on the classification (32, 33) of at least the echo signal representation (9) for the first ultrasonic echo (8) either the point object feature (18) or the line object feature (19) is generated. procedure after claim 8 , wherein in step d): depending on the classification of at least the echo signal representation (9) for the first ultrasound cho (8) either the intersection (14) is determined and, if the intersection (14) validates successfully, the point feature (18) is created, or the tangent (15) is determined and if the tangent (15) succeeds is validated, the line object feature (19) is generated. A method according to any one of the preceding claims, further comprising: e) controlling an operating process of the motor vehicle (1) as a function of the generated object features (18, 19). Method for training a classification device (6) for use in the method according to any one of claims 3 until 11 , comprising: - arranging (S21) a punctiform or linear object (10, 11) in a real or simulated environment of a motor vehicle (1); - Determination (S22) of ultrasonic received signals (7), each comprising a number of ultrasonic echoes (8) which have arisen by reflection of a respective transmitted ultrasonic signal on the object (10, 11) in the vicinity of the motor vehicle (1). ; - Identifying (S23) a plurality of ultrasonic echoes (8) in the ultrasonic received signals (7) and detecting a respective echo signal representation (9) for each of the identified ultrasonic echoes (8); - Performing (S24) supervised machine learning with the classification device (6) using at least one respective echo signal representation (9) as training input data (37) and using a classification (32) as a reflection on a point-like object or (33) as a reflection a line-shaped object (11), in accordance with the kind of the arranged object (10, 11), as training output data (38). Classification device (6) for use in the method according to one of claims 3 until 11 , which after the procedure after claim 12 is trained. Motor vehicle (1), comprising a number of ultrasonic transceivers (2) and the classification device (6). Claim 13 . Computer program product comprising instructions which, when executed by a control unit (3) of a motor vehicle (1). Claim 14 cause the control unit (3) to carry out the method according to one of Claims 1 until 11 performs.

Images