METHOD AND DEVICE FOR DETECTING ENVIRONMENTAL OBJECTS WITH AN ULTRASOUND SENSOR SYSTEM AND ULTRASOUND SENSOR SYSTEM

Description

Technical field

[0001] The invention relates to ultrasonic sensor systems for detecting objects in the environment. Technical background

[0002] Ultrasonic sensors are used in a variety of applications for detecting objects in the environment and determining their distance. For example, in motor vehicles, ultrasonic sensor systems with multiple sensors are used as parking aids on the front and rear bumpers. These systems detect the distance to an approaching object and thus assist the driver when maneuvering.

[0003] However, current ultrasonic sensor systems only evaluate the distance to a detected object. When multiple objects are present, the shortest distance to each object is used to signal a warning to the driver. However, objects that only slightly impede maneuvering, such as curbs, can also trigger detection, potentially warning the driver even if no object hindering maneuvering is present or the distance to such an object is significantly greater. Providing a distance profile of objects in the vicinity of the vehicle can also be useful for other applications, particularly visualization applications.In particular, such distance profile information can also be used for applications of autonomous or semi-autonomous driving.

[0004] When evaluating ultrasonic sensor signals, various types of information can be used to locate environmental objects. Signal time-of-flight and attenuation measurements can be used to determine the distance to the detected object. Furthermore, the transmit directional characteristic of an ultrasonic transmitter and the receive directional characteristic of an ultrasonic receiver can also affect the amplitude of the received ultrasonic signal and provide an indication of the position of an environmental object within the detection range.

[0005] When locating multiple environment objects using such determined distances, a disadvantage is that the assignment of multiple distances to a position is not unambiguous, and therefore incorrect positions of the environment objects can be determined.

[0006] Furthermore, particularly when using a suitable ultrasonic sensor assembly consisting of multiple ultrasonic sensors, the direction in which the object to be detected is located can also be determined based on the phase shift between the received signals of several ultrasonic sensors within the assembly. A disadvantage of this method is that unambiguous direction determination requires distances between the receiving ultrasonic sensors within the assembly that are less than half a wavelength. These small ultrasonic sensors would be very small and inefficient.

[0007] Furthermore, when using an ultrasonic sensor assembly consisting of several ultrasonic sensors with different reception directional characteristics, the position of the detected environmental object can also be deduced from amplitude differences between the received signals of several ultrasonic sensors.

[0008] From publication DE 10 2017 011 329 A1, a signal processing arrangement for classifying objects based on the signals of a plurality of different sensors is known, wherein at least some of the sensors are assigned more than one neural network, wherein the neural networks are designed and trainable for processing different characteristics of the object to be classified, and the outputs of the neural networks form a feature vector, wherein the feature vector is connected to another neural network that is used for

[0009] Determination of associations from the feature vector and designed and trainable for generating association information, wherein the association information is connected to identification and classification means by which a final identification information can be generated.

[0010] German patent application DE 10 2017 121581 A1 discloses the use of a neural network for processing received signals from an ultrasonic sensor, in particular by converting the ultrasonic data for each ultrasonic sensor into a one-dimensional vector of values ​​for the discrete ultrasonic data from successive reception times, and combining the vectors thus created into a two-dimensional matrix. This matrix serves as input information for the neural network.

[0011] The publication by Popperli M. et al., "Capsule Neural Network based Height Classification using Low-cost Automotive Ultrasonic Sensors", 2019, IEEE Intelligent Vehicle Symposium, pages 661-666, discloses a method for determining the height classes of objects within the detection range of ultrasonic sensors. The raw sensor data is preprocessed by sampling, low-pass filtering, and reorganization before being fed into the neural network.

[0012] The object of the present invention is to provide an improved method for detecting a distance profile with respect to one or more environmental objects, which avoids the disadvantages of the individual measurement methods and provides improved information about the position of environmental objects in the detection range of an ultrasonic sensor system. Disclosure of the invention

[0013] This problem is solved by the method for operating an ultrasonic sensor system with multiple ultrasonic transmitters and multiple ultrasonic receivers for detecting environmental objects in a detection range according to claim 1, as well as by the device and the ultrasonic sensor system according to the dependent claims.

[0014] Further details are specified in the dependent claims.

[0015] According to a first aspect, a method implemented in a control unit for operating an ultrasonic sensor system with multiple ultrasonic transmitters and multiple ultrasonic receivers for detecting environmental objects in a detection area is provided, wherein in particular the ultrasonic transmitters and ultrasonic receivers are designed as ultrasonic transducers, with the following steps: Emitting an ultrasound transmission signal by one of the ultrasound transmitters and receiving a respective ultrasound reception signal by the ultrasound receivers, whereby a signal profile is provided in each case, indicating the temporal course of the signal strength with reference to the transmission time of the ultrasound transmission signal; determining a distance profile depending on the signal profile information using a neural network that is trained to output a distance profile over environmental objects located in the detection range, depending on the signal profile information.

[0016] Determining the distance profile includes the following further steps: Creating volume maps representing the detection area for each combination of a transmitting ultrasonic transducer or transmitter and a receiving ultrasonic transducer or receiver, depending on the respective signal waveform specifications, wherein a volume map indicates possible positions of environmental objects depending on the signal waveform specification, wherein each element of the volume map is assigned a signal strength specification as a value of the signal waveform specification at a specific time after the time of transmission of the ultrasonic transmit signal, wherein the volume map corresponds to a 3D matrix and in the volume map each element is assigned to a spatial point or spatial area of ​​the detection area to be represented, wherein the signal strength specification assigned to the element corresponds to that value of the signal waveform specification at a time corresponding to the signal propagation time along a path from the transmitting ultrasonic transducer or transmitter.The ultrasound transmitter corresponds to the spatial point representing the element within the detection range and back to the receiving ultrasound transducer or ultrasound receiver, wherein the volume map is created depending on the positions of the respective ultrasound transmitter and the respective ultrasound receiver, wherein the elements of the volume map are each assigned the signal strength of the respective signal waveform specification that corresponds to a time point with respect to the transmission of the ultrasound signal, which corresponds to the signal propagation time of the ultrasound signal between transmission and reception; determining a distance profile depending on the volume maps and using the neural network that is trained to assign a distance profile over environmental objects located in the detection range to the volume maps.

[0017] The distance profile can, on the one hand, specify the distance to an environmental object or a reflection point on an environmental object for any spatial direction, starting from a reference point of the ultrasonic sensor system. This can be specified, for example, by an environmental map representing a vertical surface of the detection range perpendicular to the detection direction, the entries of which indicate the respective distance to a reflection point.

[0018] Alternatively, the distance profile can be a surface profile that assigns an elevation value or elevation classes (classes of elevation ranges) to the surrounding objects above, for example, a horizontal surface. One idea behind the above method is to use the diverse information from the ultrasonic sensors of an ultrasonic sensor system in a uniform way to create a distance profile of surrounding objects within the detection range, or more precisely, of reflection points on the surrounding objects.

[0019] An ultrasonic sensor system typically comprises several ultrasonic transmitters and receivers, with the transmitter and receiver usually integrated within a single ultrasonic transducer. Such a transducer uses a diaphragm to both transmit an ultrasonic signal and receive an ultrasonic signal reflected from one or more reflection points on one or more surrounding objects.

[0020] In an ultrasonic sensor system, the ultrasonic transducers are activated sequentially to output a corresponding ultrasonic transmit signal. This ultrasonic transmit signal can be an ultrasonic pulse or a frequency-modulated or phase-modulated ultrasonic signal. When an object is present in the detection range, the ultrasonic transducers receive a reflected ultrasonic signal as a receiving ultrasonic signal. The received ultrasonic signal detected by the transducer is converted into an electrical transducer signal. Typically, the transducer processes this received signal and generates a waveform display that shows the time course of the signal strength (amplitude) of the received ultrasonic signal. In the case of frequency-modulated or phase-modulated signals, suitable filters can be used to generate the waveform display.This signal waveform description corresponds to an envelope of the received raw signal (ultrasound reception signal). The envelope indicates the temporal amplitude profile of the ultrasound reception signal over a reception period.

[0021] Typically, the ultrasonic transducers are activated one after the other to transmit the ultrasonic signal during a transmission phase, so that each transducer receives a corresponding ultrasonic signal and provides a corresponding transducer-receiver signal during each transmission phase. When operating the ultrasonic sensor system, the signal paths of the ultrasonic signal between a transducer and the same and other receiving transducers can therefore be evaluated. During the evaluation, for example, the signal propagation times from the transmitting transducer (ultrasound transmitter) to the reflection point and from the reflection point to each receiving transducer (ultrasound receiver), as well as any attenuation information, can be taken into account.The attenuation information depends on the attenuation of the ultrasound signal along the signal path from the ultrasound transducer to the reflection point and from the reflection point to the ultrasound transducer, as well as on the transmit and receive directional characteristics of the ultrasound transducer for transmitting and receiving ultrasound signals. Thus, the respective signal waveform is determined by the signal propagation time information and the attenuation information, whereby the signal waveform can also be determined by the reception of multiple ultrasound signals from one or more reflection points of one or more surrounding objects.

[0022] To evaluate the received ultrasound reception signals, it is proposed, according to the above procedure, to use an artificial neural network that is trained, depending on input data which may correspond to, include, or be derived from the signal history information, to output a distance profile indicating the position of one or more environmental objects.

[0023] It can be advantageous to preprocess the signal waveform data to provide a better representation of possible positional information. For this purpose, signal waveform data provided by the ultrasound transducers is used to create a corresponding volume map. Specifically, a volume map is created for each combination of a transmitting ultrasound transducer (ultrasound transmitter) and a receiving ultrasound transducer (ultrasound receiver) by assigning a value from the signal waveform data at a specific time point after the transmission of the received ultrasound signal to each voxel in the volume map. The volume map corresponds to a 3D matrix in which each element is assigned to a spatial point or region within the detection area.

[0024] The volume map represents the detection range of the ultrasonic sensor system and is appropriately rasterized. The volume map thus contains volume points / voxels, each assigned a signal strength value as a signal waveform at a specific time. The signal strength value assigned to the voxel corresponds to the signal waveform value at a given time, which represents the signal propagation time along the path from the transmitting ultrasonic transducer (ultrasound transmitter) to the spatial point within the detection range representing the voxel, and back to the receiving ultrasonic transducer (ultrasound receiver). This path is determined by the spatial coordinates of the respective ultrasonic transducers and the spatial point within the detection range representing the voxel.The timing is determined by the transmission time of the ultrasound signal and the signal propagation time, which depends on the speed of sound in the surrounding air. In this way, volume maps can be created for each signal profile, i.e., for each ultrasound transmitter-receiver combination. Such volume maps significantly facilitate subsequent analysis using an artificial neural network.

[0025] In other words, each voxel represents a point in space within the detection range. The signal propagation time of the ultrasound signal from the transmitting ultrasound transducer to the point in space and from the point in space to the receiving ultrasound transducer is determined and assigned a signal amplitude corresponding to the signal propagation time. The corresponding signal amplitude is then assigned to the respective voxel.

[0026] Thus, the neural network can be trained to output a height profile map as a distance profile. This can be created in the form of a two-dimensional matrix that specifies the horizontal extent of the detection area. The matrix can indicate height classes or height values ​​at the positions within the detection area where environmental objects have been detected, as assigned by the matrix elements. This is sufficient for a motor vehicle application, since floating environmental objects are generally not found in the vehicle's vicinity. Of course, a distance profile in the form of a three-dimensional environmental map is also possible, provided a sufficient number of ultrasonic transducers (or ultrasonic transmitters and receivers) are available.

[0027] The neural network can be designed as a fully convolutional network.

[0028] According to one embodiment, one or more ultrasonic transmitters and one or more ultrasonic receivers, in particular with identical or different directional characteristics, can be arranged in a sensor assembly, wherein a phase shift profile is provided in each case, which indicates the temporal course of the phase shift between each pair of ultrasonic reception signals from at least two ultrasonic receivers of the sensor assembly; wherein at least one further volume map representing the detection area is created depending on at least one of the phase shift profiles, wherein the further volume map indicates possible positions of surrounding objects depending on the phase shift profile.

[0029] In particular, the further volume map can be created depending on the positions of the ultrasound transmitter and the ultrasound receiver in question, whereby the voxels of the volume map are each assigned the phase offset of the relevant phase offset profile that is assigned to a time point with reference to the emission of the ultrasound transmit signal, which corresponds to the signal propagation time of the ultrasound transmit signal and the ultrasound receive signal.

[0030] According to another aspect, an ultrasonic sensor system with multiple ultrasonic transmitters and multiple ultrasonic receivers for detecting environmental objects in a detection range and a control unit is provided, wherein in particular the ultrasonic transmitters and ultrasonic receivers are designed as ultrasonic transducers, and wherein the control unit is designed to: Controlling the transmission of an ultrasound transmit signal by one of the ultrasound transmitters and the reception of a respective ultrasound receive signal by the ultrasound receivers, wherein a signal waveform is provided in each case, indicating the temporal course of the signal strength with respect to the transmission time of the ultrasound transmit signal; determining a distance profile based on the signal waveforms using a neural network trained to output a distance profile over environmental objects located in the detection range based on the signal waveforms; creating volume maps representing the detection range for each combination of a transmitting ultrasound transducer or ultrasound transmitter and a receiving ultrasound transducer or an ultrasound receiver based on the respective signal waveforms, wherein a volume map indicates possible positions of environmental objects based on the signal waveforms.wherein each element of the volume map is assigned a signal strength value as a value of the signal waveform at a specific time after the time of transmission of the ultrasound signal, wherein the volume map corresponds to a 3D matrix and in the volume map each element is assigned to a spatial point or area of ​​the detection range to be represented, wherein the signal strength value assigned to the element corresponds to the value of the signal waveform at a time that corresponds to the signal travel time along a path from the transmitting ultrasound transducer or transmitter to the spatial point representing the element in the detection range and back to the receiving ultrasound transducer or receiver, wherein the volume map is created depending on the positions of the respective ultrasound transmitter and receiver.wherein each element of the volume map is assigned the signal strength of the respective signal waveform that corresponds to a time point relative to the transmission of the ultrasound signal, which corresponds to the signal propagation time of the ultrasound signal and the ultrasound reception signal; determining a distance profile based on the volume maps and using the neural network trained to assign a distance profile to the volume maps over environmental objects located in the detection range.

[0031] It may be provided that one or more of the ultrasonic transmitters of a sensor assembly and one or more of the ultrasonic receivers of a sensor assembly are spaced apart from each other by no more than five times the wavelength of the emitted ultrasonic transmission signal, in particular that they are spaced apart from each other by more than half a wavelength of the emitted ultrasonic transmission signal. Brief description of the drawings

[0032] The embodiments are described in more detail below with reference to the accompanying drawings. These show: Figure 1 shows an exemplary ultrasonic sensor system with three ultrasonic transducers and signal paths for an environmental object with two reflection points; Figures 2a and 2b show exemplary representations of a directional characteristic for an ultrasonic transducer and directional characteristics for a sensor assembly consisting of several ultrasonic transducers; Figure 3 shows an ultrasonic sensor system with three sensor assemblies, each with three ultrasonic transducers, for the additional detection of phase differences between ultrasonic receiving signals in a sensor assembly; Figure 4 shows a flowchart illustrating a method for determining a distance profile with respect to environmental objects in a detection area using the ultrasonic sensor system; Figure 5 shows a block diagram illustrating a possible topology of a convolutional neural network for creating the environmental map; and Figures 6a and 6b show schematic representations of exemplary distance profiles as height profiles and height profiles, respectively.Elevation class profile. Description of embodiments

[0033] Figure 1 Figure 1 shows an exemplary ultrasonic sensor system 1 with a number of ultrasonic transducers 2, in the present embodiment three ultrasonic transducers 2, for detecting environmental objects O within a detection range E. The ultrasonic transducers 2 combine the function of an ultrasonic transmitter and ultrasonic receiver in a single component. Alternatively, the ultrasonic sensor system can be designed with separate ultrasonic transmitters and ultrasonic receivers.

[0034] The ultrasonic transducers 2 have a known vibrating diaphragm 3 which can be excited by means of a piezoelectric actuator 4 to generate an ultrasonic transmission signal. The frequency of the ultrasonic transmission signal can be, for example, between 30 kHz and 100 kHz, preferably between 40 kHz and 60 kHz. The same vibrating diaphragm 3 can be used to receive an ultrasonic reception signal, which excites the vibrating diaphragm 3 to vibrate, with the piezoelectric actuator 4 providing a corresponding electrical signal.

[0035] The ultrasonic transducers 2 are each equipped with a control electronics unit 5, which is configured to generate a pulse of an ultrasonic transmission signal by controlling the piezo actuator 4. Furthermore, the control electronics unit 5 is configured to preprocess an electrical signal from the piezo actuator 4, which is generated by excitation from a received ultrasonic reception signal.

[0036] The operation of the ultrasonic transducers 2 is controlled by a control unit 6. Specifically, the control unit 6 triggers the transmission of an ultrasonic transmit signal. This ultrasonic transmit signal is based on a square wave drive for a transmission time known to the control unit 6. From the time of transmission of the ultrasonic transmit signal, e.g., at the end of the signal pulse, the control electronics 5 of the respective ultrasonic transducer 2 detects the excitations of the vibrating diaphragm 3 and provides the control unit with a time-dependent amplitude profile of the vibration signal. In other words, the signal profile corresponds to a time series of amplitude values ​​of the electrical received signal from the ultrasonic transducer.Since the acquisition of the ultrasound reception signal is started at a known time (transmission time), a temporal relationship between the ultrasound transmission signal and the ultrasound reception signal is known in the control unit 6.

[0037] When operating the ultrasonic sensor system 1, the control unit 6 controls the ultrasonic transducers 2 or the ultrasonic transmitters (in the case of separate designs of ultrasonic transmitters and ultrasonic receivers) separately one after the other and receives the signal profile information for the corresponding received ultrasonic reception signals from the ultrasonic transducers 2 or the ultrasonic receivers.

[0038] The presentation of Figure 1The diagram shows the signal paths for the transmitted ultrasound signal (solid arrow) during transmission and for the received ultrasound signals (dashed arrows) reflected at reflection points R on the surrounding object. The received ultrasound signals reflected at the various reflection points superimpose at the receiving ultrasound transducers 2, so that the signal profile of the received ultrasound signal typically exhibits several signal strength maxima.

[0039] Given known signal strengths of the transmitted ultrasound signal, each received ultrasound signal thus contains several pieces of information about the surrounding objects O located within the detection range E. Firstly, the signal propagation time of the ultrasound signal between the transmission and reception times can provide information about the distance of the received ultrasound signal reflected by the surrounding object O. Furthermore, the signal strength of the received ultrasound signal in the ultrasound transducer 2 can, due to attenuation over the signal path, also provide information about the distance of the reflection point R on the surrounding object O. In addition, each ultrasound transducer 2 has a directional characteristic, such as that described, for example, in Figure 2aThe directional characteristic relates to both a transmit directional characteristic, where the signal strength of the emitted ultrasound signal is direction-dependent with respect to the orientation of the ultrasound transducer, and a receive directional characteristic, where the signal strength of the received ultrasound signal is direction-dependent. These effects overlap and are reflected in the time course of the signal strength of the received ultrasound signal, so that the resulting signal profile can be evaluated accordingly.

[0040] In Figure 2b The directional characteristics of ultrasound transducers 2 are shown.

[0041] Figure 3Figure 1 shows an alternative embodiment of an ultrasonic sensor system 1', in which ultrasonic transducers 2 are grouped into sensor assemblies 7. In addition to the signal waveform measurement described above, these assemblies also enable the determination of a phase difference between the ultrasonic received signals from ultrasonic transducers of the same sensor assembly 7. The phase difference can also indicate the position of the reflection point within the detection range. Thus, in addition to the signal waveform measurement, a sensor assembly 7 can also provide the control unit with phase shifts between the ultrasonic received signals of the ultrasonic receivers of a sensor assembly 7 according to a time-dependent profile.

[0042] The ultrasonic transducers of the sensor assembly 7 can be arranged in close proximity, and the arrangement is essentially arbitrary. In particular, the ultrasonic transducers 2 do not need to be arranged symmetrically to each other. A structural limitation of conventional ultrasonic sensor assemblies, namely that the ultrasonic transducers 2 must be arranged with a distance of less than half a wavelength relative to the ultrasonic transmission frequency with respect to their centers, does not apply here, since, due to the additional information sources from the signal waveform data, unambiguous position determination using the phase shift between the ultrasonic transmission signal and the ultrasonic reception signal by a sensor assembly 7 is not required.

[0043] In Figure 4 A flowchart illustrating a procedure for determining an environmental map using ultrasonic sensor signals is described in more detail.

[0044] First, for the ultrasonic sensor system 1 of the Figure 1The corresponding signal waveform data, which describe the temporal evolution of a signal strength relative to a reference point, such as the transmission time of the ultrasound signal, are provided. Each of the ultrasound transducers 2 acts sequentially as an ultrasound transmitter to emit the ultrasound signal, and each transducer also acts as an ultrasound receiver for one reflected transmitted ultrasound signal, i.e., one received ultrasound signal. Thus, when using three ultrasound transducers 2, nine different signal waveform data sets can be recorded. These signal waveform data sets include information about the signal propagation time, the attenuation over the signal path, and the transmit and receive directional characteristics with respect to the transmitting and receiving ultrasound transducers 2. Furthermore, each signal waveform data set can include information about one or more reflection points R on one or more surrounding objects O.

[0045] In step S2, a volume map is created from each signal waveform dataset. The volume map corresponds to a three-dimensional grid whose grid points represent voxels (space points) and which maps the detection area. Each grid node in the three-dimensional grid corresponds to a space point within the detection area. The volume maps are then populated with the measured values ​​from the signal waveform datasets. For this purpose, a volume map is generated from each signal waveform dataset.

[0046] In the volume map V i The voxels are assigned x signal strength values as,r from the signal waveform data as,r ( t ) assigned. V i x = a s , r t = τ s , r x , τ s , r x = c − 1 x s − x + x r − x , where x s = ( xs , yszs ) T< a spatial position of the transmitting ultrasound transducer and x r = ( xryrzr ) T< corresponds to a spatial position of the receiving ultrasound transducer 2.

[0047] For each voxel of the volume map, the signal path between the receiving and transmitting ultrasonic transducers considered for the volume map in question is determined, and the resulting signal propagation time is calculated using the speed of sound c.

[0048] The signal propagation time τ s,r ( x ) now corresponds to the time period for which the signal strength is extracted from the signal waveform and assigned to the corresponding voxel. The speed of sound c can additionally be taken into account as a function of temperature and humidity.

[0049] In this way, a volume map can be created for each ultrasound transmitter-ultrasound receiver combination.

[0050] Are sensor assemblies 7 installed according to the Figure 3If this is used, further volume maps can be created based on additional ultrasound transmitter-receiver combinations and based on a phase shift profile, each specifying the time course of the phase differences between the ultrasound receiver signals of a sensor assembly 7. In other words, instead of the signal strengths, the resulting phase shifts are assigned to each voxel of the volume map to be created.

[0051] In step S3, the volume maps thus created are fed into an artificial convolutional neural network (CNN). The neural network is trained to assign a distance profile of surrounding objects O within the detection area E to the volume maps.

[0052] Alternatively, the created volume maps can also be processed in an artificial deep neural network (DNN) that receives the volume maps as input data and provides a corresponding distance profile.

[0053] The distance profile can specify the distance to an environmental object or a reflection point on an environmental object for any spatial direction, starting from a reference point of the ultrasonic sensor system. This can be specified, for example, by an environmental map representing a vertical surface of the detection range perpendicular to the detection direction, the entries of which indicate the respective distance to a reflection point.

[0054] In principle, a wide variety of neural network topologies are conceivable for processing the volume maps.

[0055] In Figure 5The diagram schematically illustrates the structure of a convolutional neural network for evaluating volume maps. The neural network is used for 2D image reconstruction, with the volume maps having a size of, for example, 128 x 128 x 8 pixels. Each volume map can be provided as a separate input channel to the convolutional neural network.

[0056] In a series of down blocks, the number of voxels per volume map is halved in each spatial dimension, and the number of channels is doubled. Specifically, the down block operation continues until the height dimension equals 1.

[0057] The number of pixels in the 2D map is then expanded by a factor of 2 in each of the two spatial dimensions in successive up-blocks. Both the down-blocks and the up-blocks use residual units, which consist of a batch normalization layer, followed by a convolutional layer with n channels and a ReLU layer as the activation function.

[0058] Using a sigmoid activation function for a height map as an output, or a softmax activation function in the case of a height classification, a corresponding distance profile can be output as the last layer.

[0059] The distance profile can alternatively represent a reduced environmental map, depicting a primarily horizontal area of ​​the detection range and assigning elevation information or an elevation class of a local object to each grid point on the map. The depicted area of ​​the detection range can be determined by a horizontal reception direction and the dimension of the distance relative to a reference point of the sensor system. Such distance profiles can be used as elevation profiles, for example, as in the Figures 6a and 6bThese distance profiles are presented in a representative manner. Such distance profiles reduce the spatial positions of surrounding objects to a position in the plane and only indicate the height of the surrounding object. A height profile of this type has a reduced information content compared to a complete distance profile. This reduces the complexity of the neural network used, enabling real-time application with reduced computational requirements.

[0060] The height profiles may be sufficient for applications such as a maneuvering aid in a motor vehicle, since the surrounding area of ​​a motor vehicle usually only contains objects that are on the ground and not free-floating.

Claims

1. A method implemented in a control unit (6) for operating an ultrasonic sensor system (1) having a plurality of ultrasonic transmitters and a plurality of ultrasonic receivers for detecting environmental objects (O) in a detection zone (E), wherein in particular the ultrasonic transmitters and ultrasonic receivers are in the form of ultrasonic transducers (2), having the following steps: - emitting an ultrasonic transmission signal by one of the ultrasonic transmitters and receiving a respective ultrasonic reception signal by the ultrasonic receivers, wherein in each case a signal profile information is provided which indicates the time profile of the signal strength with respect to the instant of transmission of the ultrasonic transmission signal; - determining a distance profile as a function of the signal profile information with the aid of a neural network which is trained to output a distance profile relating to environmental objects (O) situated in the detection zone (E) as a function of the signal profile information, characterized in that the determination of the distance profile comprises the following further steps: - creating volume maps representing the detection range (E) for each combination of a transmitting ultrasonic transducer or ultrasonic transmitter and a receiving ultrasonic transducer or an ultrasonic receiver depending on the respective signal profile information, wherein a volume map indicates possible positions of environmental objects (O) depending on the signal profile information, wherein elements of the volume map are each assigned a signal strength indication as a value of the signal profile information at a specific time instance after the time instance of transmission of the ultrasonic transmission signal, wherein the volume map corresponds to a 3D matrix and in the volume map each element is assigned to a spatial point or spatial region of the detection range to be represented, wherein the signal strength indication associated with the element corresponds to the value of the signal profile information at a time instance that corresponds to the signal propagation time along a path from the transmitting ultrasonic transducer or ultrasonic transmitter to the spatial point in the detection area that represents the element and back to the receiving ultrasonic transducer or ultrasonic receiver, wherein the volume map is generated in dependence on the positions of the relevant ultrasound transmitter and the relevant ultrasound receiver, wherein the elements of the volume map are each assigned that signal strength of the relevant signal profile information assigned to a time instance with respect to the transmission of the ultrasonic transmission signal that corresponds to the signal transit time of the ultrasonic transmission signal between transmission and reception; - determining a distance profile as a function of the volume maps and with the aid of the neural network, which is trained to assign a distance profile to the volume maps relating to environmental objects (O) situated in the detection zone (E), - wherein the distance profile represents an area profile which assigns a height indication above a base surface to the environmental objects (O), or wherein the distance profile indicates the distance of an environmental object or of a reflection point on an environmental object for a plurality of spatial directions starting from a reference point of the ultrasonic sensor system.

2. The method according to claim 1, wherein the neural network is configured as a deep neural network or as a convolutional neural network.

3. An ultrasonic sensor system (1) having a plurality of ultrasonic transmitters and a plurality of ultrasonic receivers for detecting environmental objects (O) in a detection zone (E) and a control unit (6), in particular the ultrasonic transmitters and ultrasonic receivers being in the form of ultrasonic transducers (2), the control unit (6) being designed for: - controlling the transmission of an ultrasonic transmission signal by one of the ultrasonic transmitters and the reception of a respective ultrasonic reception signal by the ultrasonic receivers, in each case a signal profile information item being provided which indicates the time profile of the signal strength indication with respect to the transmission time of the ultrasonic transmission signal; - creating volume maps representing the detection zone (E) as a function of at least one of the signal profile information, a volume map indicating possible positions of environmental objects (O) as a function of the signal profile information; - determining a distance profile as a function of the volume maps and a neural network which is trained to assign a distance profile to the volume maps relating to environmental objects (O) situated in the detection zone (E), characterized in that the control unit (6) is furthermore designed to determine the distance profile in order to: - create volume maps representing the detection range (E) for each combination of a transmitting ultrasonic transducer or ultrasonic transmitter and a receiving ultrasonic transducer or an ultrasonic receiver, depending on the respective signal profile information, a volume map indicating possible positions of environmental objects (O) depending on the signal profile information, wherein elements of the volume map are each assigned a signal strength indication as a value of the signal profile information at a specific time instance after the time instance of transmission of the ultrasonic transmission signal, wherein the volume map corresponds to a 3D matrix and in the volume map each element is assigned to a spatial point or spatial region of the detection range to be represented; wherein the signal strength indication associated with the element corresponds to the value of the signal profile information at a time instance that corresponds to the signal transit time along a path from the transmitting ultrasonic transducer or ultrasonic transmitter to the spatial point in the detection area that represents the element and back to the receiving ultrasonic transducer or ultrasonic receiver, wherein the volume map is generated in dependence on the positions of the relevant ultrasound transmitter and the relevant ultrasound receiver, wherein the elements of the volume map are each assigned that signal strength indication of the relevant signal profile information assigned to a time instance with reference to the transmission of the ultrasonic transmission signal, which corresponds to the signal transit time of the ultrasonic transmission signal between transmission and reception; - determine a distance profile dependent on the volume maps and with the aid of the neural network, which is trained to assign a distance profile to the volume maps relating to environmental objects (O) situated in the detection zone (E), wherein the distance profile represents an area profile, which assigns the environmental objects (O) a height indication above a base area, or wherein the distance profile indicates the distance of an environmental object or a reflection point on an environmental object for a plurality of spatial directions starting from a reference point of the ultrasonic sensor system.