Method and device for classifying objects, especially in a motor vehicle environment

Abstract

A method for classifying objects, especially objects in the environment of a motor vehicle, by means of an ultrasonic sensing device is proposed, wherein the ultrasonic sensing device has a plurality of spatially distributed ultrasonic sensors. A plurality of measurements are carried out, especially continuously. In one measurement, an ultrasonic signal is emitted by one of the ultrasonic sensors of the ultrasonic sensors, and a signal having a plurality of reflected echo signals, i.e., so-called multiple echoes, is received by at least one of the ultrasonic sensors, and the received echo signals are assigned to the objects. A plurality of features can be determined from the received echo signals. According to the invention, the objects are classified according to a combination of at least two of the features, especially as pedestrians.

Description

Technical Field

[0001] The present invention relates to a method and a device for classifying objects, in particular objects in the environment of a motor vehicle, by means of an ultrasonic sensing device, wherein the ultrasonic sensing device has a plurality of spatially distributed ultrasonic sensors. Background Art

[0002] From EP2908153 A2, a device for identifying dynamic objects in the environment of a motor vehicle by means of ultrasonic waves is known. The device includes a sensor array having at least two sensor elements arranged at a pre-given distance from each other, and the sensor elements operate both as transmitters and as receivers. The detected moving object is classified by comparing its trajectory and speed with a pre-given model, for example, classified as a pedestrian.

[0003] From DE 112016003462 T5, a vehicle control device is known, which includes an ultrasonic sensor and a monocular camera. The ultrasonic sensor is designed to detect obstacles in front of the vehicle, and the monocular camera is designed to capture an image of the area in front of the vehicle. The controller determines whether there is an obstacle in front of the vehicle based on the image captured during the vehicle's travel using the monocular camera, and the controller adjusts the driving force of the vehicle according to the combination of whether an object is detected or not detected by the ultrasonic sensor and whether an obstacle is present or not based on the image captured using the monocular camera.

[0004] DE 102016124157 A1 shows a method for determining the braking necessity for a motor vehicle according to possible collision objects, such as pedestrians, in the surrounding area of the motor vehicle.

[0005] The task of the present invention can be regarded as achieving as reliable as possible pedestrian recognition in the environment of a motor vehicle based on an ultrasonic sensing device. Summary of the Invention

[0006] A method for classifying objects, in particular objects in the environment of a motor vehicle, by means of an ultrasonic sensing device is proposed, wherein the ultrasonic sensing device has a plurality of spatially distributed ultrasonic sensors.

[0007] The ultrasonic sensing device can, for example, be part of a driving assistance system or a parking assistance system provided on a motor vehicle. Here, the ultrasonic sensors can, for example, be distributed on the front bumper and / or the rear bumper of the motor vehicle.

[0008] Each ultrasonic sensor in the ultrasonic sensors can be configured to transmit an ultrasonic signal and receive the ultrasonic signal reflected on an object. The distance to the object can be determined in a known manner by means of the propagation time of the received ultrasonic signal reflected on the object. Extended and structured objects may generate so-called multiple echoes. In this case, the transmitted ultrasonic signal is reflected at different points of the object, and for one transmitted ultrasonic signal, multiple echo signals with different propagation times may be received by the ultrasonic sensor respectively.

[0009] The method includes the following steps:

[0010] Performing, in particular continuously performing, multiple measurements, wherein in one measurement, an ultrasonic signal is respectively transmitted by one of the ultrasonic sensors, and a signal is received by at least one of the ultrasonic sensors, the signal having multiple reflected echo signals, i.e., so-called multiple echoes, and the received echo signals are assigned to the object.

[0011] Multiple features can be determined from the received echo signals:

[0012] - A first feature that represents the frequency of the distance d exceeding a pre-given distance threshold with respect to the number of performed measurements, wherein the distance d corresponds to the distance between the first received echo signal in time and the last received echo signal in time in one measurement;

[0013] - A second feature that represents the variance of the distance d;

[0014] - A third feature that represents the distribution of the number of received echo signals on the ultrasonic sensors in each measurement;

[0015] - A fourth feature that represents the orientation of the ultrasonic sensor relative to the object;

[0016] - A fifth feature that represents the variance of the first object distance in multiple measurements, wherein a first object distance belonging to each determined first received echo signal in time is calculated, and the approach of the object is particularly considered when determining the variance;

[0017] - A sixth feature that represents the correlation between the received echo signal and the transmitted ultrasonic signal;

[0018] - A seventh feature that represents the amplitude of the echo signal;

[0019] - An eighth feature that represents the distribution of the reflection points, wherein each reflection point indicates a measured object position.

[0020] All of the features mentioned enable classification of an object with a certain probability, and in particular enable separation of a pedestrian from a curbstone or other objects which are not relevant, for example, for an automated or assisted emergency braking function of a motor vehicle. None of the features alone is sufficient to reliably infer whether the object is a pedestrian without adding additional features. Therefore, the features mentioned are combined logically or statistically according to the present invention.

[0021] According to the present invention, the object is classified, in particular as a pedestrian, according to a combination of at least two of the features. Here, combination is understood as analyzing and processing the features as follows: for example, whether it is higher or lower than a determined threshold, or whether a determined distribution or variance of the distribution of the features is observed in, for example, multiple successive measurements. Each feature observed can then make a specific contribution to the determination of the classification, for example, by indicating the probability that the object is a pedestrian.

[0022] Alternatively or additionally, the features can be used as input signals for a so-called classifier (such as a neural network, decision tree, etc.) trained by a database for machine learning.

[0023] The first feature and the third feature can in particular be combined in such a way that the object is reliably classified as a pedestrian if the following frequency according to the first feature is particularly high: the distance d between the first detected echo signal and the last detected echo signal of multiple echoes exceeds a pre-given distance threshold, i.e., a frequency threshold, and the distribution according to the third feature indicates that only certain ultrasonic sensors, in particular only one or only a few ultrasonic sensors, especially adjacent ultrasonic sensors, among the ultrasonic sensors receive multiple echo signals in each measurement.

[0024] The present invention is based on the following knowledge: Pedestrians, as structured objects, reflect ultrasonic waves at different body parts, such as at the feet, at the torso, head, and / or arms. Due to this property, multiple echo signals (reflections, multiple echoes) are usually obtained with a single ultrasonic measurement, and these echo signals are formed successively in time or in distance. Characteristically, the distance d between the first received signal in time and the last received signal in time is usually relatively large, especially larger than in the case of low objects, such as curbstones. This may be determined, for example, by the fact that different body parts reflect the transmitted ultrasonic signals, and these ultrasonic signals are at a relatively large distance from each other as seen from the ultrasonic sensor. In addition, due to the movement of the pedestrian, this relative distance d may fluctuate strongly, especially when observing a sequence consisting of successive measurements - this effect is furthermore determined by the fact that the transmitted ultrasonic signal is not reflected by the same point of the pedestrian in every measurement.

[0025] Therefore, the following features can be considered as the first feature for classifying the reflecting object: how frequently multiple echoes are measured compared to the total number of measurements assigned to the object, and how frequently the distance d is large, i.e., greater than a determined threshold (first feature).

[0026] Furthermore, the following feature can be determined as the second feature for classification: how strongly the distance d fluctuates, i.e., how large the variance of d is.

[0027] A determined curbstone structure, such as a curbstone with grass pavers behind it, may produce echo signals similar to those of pedestrians, especially multiple echoes. However, pedestrians only have an aggregation of the received multiple echoes in the case of the following ultrasonic sensors The main measurement axis of the ultrasonic sensor is basically aligned with the direction of the pedestrian, or the main measurement axis of the ultrasonic sensor has the smallest possible angle with the pedestrian. For an ultrasonic sensor with a main measurement axis at a large angle to the pedestrian, the multiply reflected echo signals are usually weak due to the signal propagation path, so that multiple reflections are not received or only received in small amounts. In this regard, in the case of a pedestrian, typical multiple echoes can be determined particularly at only one ultrasonic sensor or at two adjacent ultrasonic sensors in the ultrasonic sensor, whereas such a distribution is not observed in the case of, for example, a curbstone as the reflecting object. Therefore, the distribution of the number of echo signals received in each measurement on the ultrasonic sensors is detected as a third feature. That is, it is detected which ultrasonic sensors in the ultrasonic sensor receive multiple echo signals in each measurement and which do not. Preferably, if the distribution according to the third feature indicates that only the determined ultrasonic sensors in the ultrasonic sensor, in particular only one ultrasonic sensor or only a small number of ultrasonic sensors, in particular adjacent ultrasonic sensors, receive multiple echo signals contributing to the object to be classified in each measurement, the object is classified as a pedestrian. In particular, it is considered here that in a complex scenario, although other ultrasonic sensors can also detect echo signals, they cannot detect the following echo signals: the echo signals are assigned to the object to be classified due to their spatial position, which can be determined, for example, by trilateration.

[0028] According to the fourth feature, the orientation of the ultrasonic sensor relative to the object can be determined. Therefore, it can also be considered that the multiply reflected echo signals are usually very weak and can only be recognized by a sensor system set to be very sensitive. Therefore, the reception of these ultrasonic signals (multiple echoes) can vary strongly between different ultrasonic sensors. Therefore, it is optionally proposed that, in order to obtain statistical values, that is, in particular, to determine the distance d, to determine the frequency at which the distance d according to the first feature exceeds a determined threshold, and / or to determine the variance of d according to the second feature, in particular, only the measurement data of those ultrasonic sensors aligned with the object, that is, having a good viewing angle, are analyzed, or preferably only those ultrasonic sensors having the best viewing angle to the object are used. Here, in particular, the angle of the main axis of each ultrasonic sensor relative to the recognized object can be considered.

[0029] Also determined by the complex structure of the pedestrian, especially compared to a reflector with a simple geometric shape and stability, such as a curbstone, in a measurement sequence, the object distance derived from the first received reflected echo signal (first reflection) in terms of time also varies strongly. Thus, the variance of the object distance derived from the first received reflected echo signal (first reflection) in terms of time can be used as an additional fifth feature. In particular, if this variance exceeds a second variance threshold, the object can be classified as a pedestrian.

[0030] Since there are always multiple points reflecting in the case of a pedestrian, the echo signals are acoustically superimposed. If the ultrasonic sensor emits ultrasonic signals in a coded manner, for example by using a characteristic frequency change trend, and the receiving ultrasonic sensor analyzes the change trend of the received echo signals to obtain the correlation, the superimposition of the echo signals results in an interfering effect on the correlation, that is, reduces the correlation. In contrast, in the case of, for example, a curbstone, a very high correlation that is never possible in the case of a pedestrian can be observed. Therefore, the correlation between the received echo signal and the transmitted ultrasonic signal is determined as the sixth feature. If a very high correlation value is received, this can be used as an exclusion criterion for classifying as a pedestrian. Preferably, if the correlation value is greater than a determined correlation threshold for a determined number of measurements, the classification of the object as a pedestrian is excluded.

[0031] According to the seventh feature, the amplitude of at least one received echo signal can be analyzed. Since the clothes of a pedestrian usually strongly absorb acoustic signals, the reflected acoustic energy of the ultrasonic signal reflected by the pedestrian is rather low to moderate. In this regard, the amplitude can be analyzed in particular such that a particularly high amplitude reduces the classification probability of classifying as a pedestrian. In particular, if the amplitude of at least one received ultrasonic signal is greater than a determined amplitude threshold, the classification of the object as a pedestrian can be excluded.

[0032] In addition, in each measurement, a so-called reflection point can be determined, which indicates the following position in space: the ultrasonic signal received has been reflected from this position. The reflection point can be determined, for example, in a known manner by trilateration. Additionally, it is typical for pedestrians that the reflection points received from one measurement to the next change strongly in terms of location, i.e., the coordinates assigned to the pedestrian as an object in space change in a determined manner during the measurement. This is caused by the complex geometric shape structure of the pedestrian and the fact that the pedestrian can move. Therefore, in a measurement sequence, the spatial distribution of the reflection points can be used as an additional eighth feature, and the reflection points represent the location-based assignment of multiple measurements by trilateration. In particular, if the spatial distribution of the reflection points according to the eighth feature has a characteristic shape, especially the clustering of the reflection points at possible object positions and the dispersion in the lateral direction, then the object can be classified as a pedestrian.

[0033] All the features mentioned can separate pedestrians from curbs or other objects not significantly related to braking with a certain probability. None of the features alone enables a reliable inference to be made clearly enough without adding additional features. Therefore, according to the present invention, the features mentioned are combined with each other especially logically or statistically. Similarly, these features can be used as input signals for classifiers (neural networks, decision trees, etc.) trained through a database for machine learning. What is primarily decisive is not the way of logical combination, but the features that are decisive for pedestrian recognition.

[0034] In addition, it should be noted that not all features are necessary to achieve classification, but the classification quality can be improved through the combination of features.

[0035] The more of the described features are combined when classifying an object, the greater the reliability of correctly identifying an actual pedestrian as a pedestrian and achieving a low false recognition rate, where the false recognition rate refers to misidentifying an object that is not a pedestrian (such as a curb) as a pedestrian (false positive).

[0036] According to another aspect of the present invention, a device is proposed, which is configured to classify an object, especially an object in the environment of a motor vehicle. The device includes:

[0037] An ultrasonic sensing device, wherein the ultrasonic sensing device has a plurality of ultrasonically sensors arranged in a spatially distributed manner, and the ultrasonic sensors are especially arranged on the vehicle body of the motor vehicle.

[0038] An analysis and processing device, which is configured to perform the steps of the method configured as described above.

[0039] The analysis and processing device is in particular configured to determine two or more of the above-mentioned features from the measurement data detected by the ultrasonic sensor and to combine at least two of the features in order to determine the classification of an object in the environment of the device.

[0040] The device can be part of a driver assistance system of a motor vehicle. The ultrasonic sensor is preferably arranged on the bumper of the motor vehicle and is oriented in such a way that the ultrasonic sensor can detect the area in front of or behind the motor vehicle in the driving direction.

[0041] The device can in particular be part of a braking assistance system, wherein, if the object is classified as a pedestrian and the object is detected, for example, at a distance closer than a determined minimum distance in the driving direction of the motor vehicle, emergency braking can be triggered, for example.

[0042] According to another aspect of the invention, a motor vehicle having a device according to the invention is proposed. Description of the Drawings

[0043] Embodiments of the invention are described in detail with reference to the accompanying drawings.

[0044] Figure 1 The device according to a possible embodiment of the invention in the case of detecting a pedestrian is shown.

[0045] Figure 2 Distance data detected during a plurality of measurements successively in time are shown exemplarily according to the invention.

[0046] Figure 3 Distance data detected during a plurality of measurements successively in time taking into account multiple echoes are shown exemplarily according to the invention, wherein a distance d is determined for each measurement corresponding to the distance between the first echo signal received in time and the last echo signal received in time in one measurement. i .

[0047] Figure 4 A motor vehicle and a pedestrian constructed according to the invention are shown schematically, and the distribution of reflection points from multiple measurements is shown in a superimposed manner. Detailed Description of the Embodiments

[0048] In the following description of the embodiments of the invention, identical or similar elements are denoted by the same reference signs, wherein, where necessary, the repeated description of these elements in individual cases is omitted. The drawings only schematically show the subject matter of this aspect.

[0049] In Figure 1The front part of a motor vehicle 1 is schematically shown. The motor vehicle 1 includes a device 10 for classifying an object 70 in the environment of the motor vehicle 1. The device 10 includes an ultrasonic sensing device, wherein the ultrasonic sensing device has four ultrasonic sensors 12.1, 12.2, 12.3 and 12.4, and the ultrasonic sensors are arranged along the front of the motor vehicle 1. The device 10 further includes an analysis and processing device 11, which is configured to analyze and process the measurement data of the ultrasonic sensors 12.1 to 12.4 and classify the object 70 based thereon. The analysis and processing device is configured to control each of the ultrasonic sensors 12.1 to 12.4 such that the ultrasonic sensors 12.1 to 12.4 transmit ultrasonic signals, receive the ultrasonic signals reflected on the object 70, and assign the received signals to the object 70.

[0050] In the example shown, the object 70 is a pedestrian 80.

[0051] To classify the object 70 as a pedestrian 80, different characteristics of the received echo signals are determined.

[0052] Due to the characteristic structure and shape of the pedestrian 80, the probability that at least one of the ultrasonic sensors 12.1 to 12.4 receives multiple reflected echo signals is high. Thus, for example, the hand 82 and the foot 84 of the pedestrian 80 reflect the ultrasonic signal transmitted by the ultrasonic sensor 12.3. The foot 84 and the ultrasonic sensor 12.3 may have a distance 24, for example, which is less than the distance 22 between the hand 82 and the ultrasonic sensor 12.3. The ultrasonic sensor 12.3 thus receives at least two echo signals in the measurement. Assuming that the foot 84 has the minimum distance to the ultrasonic sensor 12.3 among all the reflection points of the pedestrian 80, and the hand 82 has the maximum distance to the ultrasonic sensor 12.3 among all the reflection points of the pedestrian 80, the echo signal reflected by the foot 84 is received as the first echo signal in time, and the echo signal reflected by the hand 82 is received as the last echo signal in time.

[0053] From the echo signals, it is possible to determine the distance d between the first received echo signal in time and the last received echo signal in time. For this purpose, for example, first the propagation time difference of these echo signals is determined, and from this propagation time difference, the spatial distance d can be calculated in a known manner given the known speed of sound of the ultrasonic signal. Now, by means of multiple measurements, it is possible to observe how the separately determined distances d behave. If, for example, a distance d is determined in a determined minimum share of the number of measurements that exceeds a determined threshold, this fact can be used as an indicator that the object 70 is a pedestrian. In addition, due to, for example, the movement of the arms and legs and / or the different orientations of the pedestrian 80 relative to the ultrasonic sensor, strong fluctuations in the distance d occur. That is, if the variance of the distance d is observed, for example, when this variance exceeds a determined threshold, this variance can be used as a further indicator that "the object 70 is a pedestrian".

[0054] In Figure 3 it is exemplarily shown in the graph 200 for multiple measurements i the first received echo signal 210 in time and the last received echo signal 220 in time and the distance d determined thereby i . Here, the measurement time points t are plotted on the x-axis and the measured distance s from the ultrasonic sensor is plotted on the y-axis. It is possible to compare for each measurement whether the distance d exceeds a pre-given distance threshold, and thus it is possible to determine the frequency with which the determined distance d exceeds the pre-given distance threshold with respect to the number of measurements performed.

[0055] Another feature is that the number of echo signals received by the ultrasonic sensor in each measurement is determined. In Figure 1 the situation shown, compared to the two outer ultrasonic sensors 12.1 and 12.4, the two middle ultrasonic sensors 12.2 and 12.3 receive stronger echo signals due to their proximity to the pedestrian 80 and their orientation relative to the pedestrian 80. Correspondingly, the two middle ultrasonic sensors 12.2 and 12.3 are more likely to receive multiple echoes. That is, in the case where the ultrasonic sensors 12.2 and 12.3 are arranged adjacent to each other, the distribution of the number of echo signals received in each measurement has an aggregation.

[0056] Likewise, due to the complex structure of the pedestrian 80, especially compared to geometrically simple and stable reflectors - such as curbstones - the distance between the pedestrian 80 and the ultrasonic sensor determined by the respective first received echo signal in time can also vary strongly in a sequence of measurements. This is shown in Figure 2is exemplarily shown in the measurement chart 100. The measurement time points t are plotted on the x-axis, and the measured distance s to the ultrasonic sensor is plotted on the y-axis. At each measurement time point, the distance 90 between the object 70 that reflects and the ultrasonic sensor that performs the measurement is determined by means of the first echo signal received in time. Overall, a linear relative approach can be seen between the ultrasonic sensor that performs the measurement and the object 70 that reflects, as shown by the straight line 95. If the vehicle speed is known and the object is assumed to be stationary, the straight line 95 can be calculated. If the motion state of the object 70 is not known, the straight line 95 can be determined from the measured values 90 (e.g., by common fitting methods). To determine this fifth feature, the variance of the measured values 90 relative to the straight line 95 is determined by multiple measurements, and the measured values represent the corresponding object distances, thereby taking into account the approach of the object 70 to the vehicle 1 or the ultrasonic sensor that performs the measurement. If the variance thus determined according to the fifth feature exceeds the second variance threshold, this can be judged as an additional indication that "the object 70 that reflects is a pedestrian 80".

[0057] Figure 4 Schematically shows a possible example of how the eighth feature according to the invention that represents the distribution of the reflection points 60 can be determined. In Figure 4 the front part of the vehicle 1 is schematically shown, which has an ultrasonic sensing device 12, and the ultrasonic sensing device has four ultrasonic sensors 12.1, 12.2, 12.3, and 12.4. The object 70 is located in front of the vehicle 1. By means of multiple measurements performed by the ultrasonic sensors, for example, the coordinates of the reflection points 60 are determined by means of trilateration, and each of the reflection points represents a measured position of the object 70. A spatial distribution of the reflection points is formed, as marked by the coordinate system in Figure 4 As shown. It has been found that in the case of a pedestrian 80, a characteristic distribution of the reflection points 60 is generated, which on the one hand has an aggregation at possible object positions, and on the other hand has a certain dispersion of the reflection points 60 in the lateral direction - i.e., in the direction perpendicular to the main measurement direction.

Claims

1. A method for classifying an object (70) by means of an ultrasonic sensing device (12), wherein, The ultrasonic sensing device (12) has a plurality of ultrasonically sensors (12.1, 12.2, 12.3, 12.4) arranged in a spatially distributed manner, and the method includes the following steps: Performing a plurality of measurements, wherein, in one measurement, - transmitting an ultrasonic signal through one of the ultrasonic sensors (12.1, 12.2, 12.3, 12.4), - receiving a signal through at least one of the ultrasonic sensors (12.1, 12.2, 12.3, 12.4), the signal having a plurality of reflected echo signals (210, 220), - assigning the received echo signals to an object (70), wherein a plurality of features are determined from the received echo signals, the features including at least two of the following features: - A first feature representing the frequency at which the distance (d) exceeds a pre-given distance threshold with respect to the number of measurements performed, wherein the distance (d) corresponds to the distance between the first received echo signal (210) and the last received echo signal (220) in time in one measurement, - A second feature representing the variance of the distance (d), - A third feature representing the distribution of the number of received echo signals on the ultrasonic sensors (12.1, 12.2, 12.3, 12.4) in each measurement, - A fourth feature representing the orientation of the ultrasonic sensors (12.1, 12.2, 12.3, 12.4) relative to the object (70); - A fifth feature representing the variance of the first object distance in a plurality of measurements, wherein a respective first object distance is calculated for each determined first received echo signal (210) in time, - A sixth feature representing the correlation between the received echo signals and the transmitted ultrasonic signals, - A seventh feature representing the amplitude of the echo signals, - An eighth feature representing the distribution of the reflection points (60), wherein each reflection point (60) indicates a measured object position; wherein the object (70) is classified according to a combination of at least two of the eight features, wherein if at least the frequency according to the first feature exceeds a frequency threshold and the distribution according to the third feature indicates that only one or only a few of the ultrasonic sensors (12.1, 12.2, 12.3, 12.4) receive a plurality of echo signals (210, 220) in each measurement, then the object (70) is classified as a pedestrian (80).

2. The method according to claim 1, wherein, If in addition the variance according to the second feature exceeds a first variance threshold, then the object (70) is classified as a pedestrian (80).

3. The method according to claim 1 or 2, wherein According to the fourth feature, only those ultrasonic sensors (12.1, 12.2, 12.3, 12.4) aligned with the object (70) are considered when determining the distance (d).

4. The method according to claim 1 or 2, wherein If, in addition, the variance according to the fifth feature exceeds a second variance threshold, the object (70) is classified as a pedestrian (80).

5. The method according to claim 1 or 2, wherein The emitted ultrasonic signal has a defined frequency variation trend, and a frequency variation trend is determined for at least one received echo signal, and a correlation value of the echo signal and the transmitted ultrasonic signal is calculated according to the sixth feature.

6. The method according to claim 5, wherein If the correlation value is greater than a defined correlation threshold for a defined number of measurements, the classification of the object (70) as a pedestrian (80) is excluded.

7. The method according to claim 1 or 2, wherein According to the seventh feature, an amplitude is determined for at least one received echo signal, and if the amplitude of at least one received ultrasonic signal is greater than a defined amplitude threshold, the classification of the object (70) as a pedestrian (80) is excluded.

8. The method according to claim 1 or 2, wherein If, in addition, the spatial distribution of the reflection points (60) has a characteristic shape according to the eighth feature, the object (70) is classified as a pedestrian (80).

9. The method according to claim 1 or 2, wherein An optimized combination of at least one of the features and / or the thresholds is pre-determined by means of a machine learning method for classifying the object (70) as a defined object type.

10. The method according to claim 1, wherein The object (70) is an object in the environment of a motor vehicle (1).

11. The method according to claim 1, wherein, The proximity of the object (70) is taken into account when determining the fifth feature.

12. The method according to claim 1, wherein The object (70) is classified as a pedestrian (80) according to a combination of at least two of the eight features.

13. The method according to claim 1, wherein, The only a small number of ultrasonic sensors are ultrasonic sensors arranged adjacent to each other.

14. The method according to claim 3, wherein, According to the fourth feature, when determining the distance (d), only these ultrasonic sensors (12.1, 12.2, 12.3, 12.4) are considered: the ultrasonic sensors have a main measurement direction that is optimally aligned with the object (70).

15. The method according to claim 8, wherein The characteristic shape includes the clustering of the reflection points (60) at possible object positions and the discreteness in the lateral direction.

16. The method according to claim 9, wherein The defined object type is a pedestrian (80).

17. A device (10) for classifying an object (70), the device comprising: an ultrasonic sensing device (12), wherein the ultrasonic sensing device (12) has a plurality of spatially distributed ultrasonic sensors (12.1, 12.2, 12.3, 12.4), an analysis and processing device (11), the analysis and processing device being configured to perform the steps of the method according to any one of claims 1 to 16.

18. The apparatus (10) according to claim 17, wherein The object (70) is an object in the environment of a motor vehicle (1).

19. The apparatus (10) according to claim 17, wherein, The ultrasonic sensors are arranged on the vehicle body of the motor vehicle.

20. A motor vehicle (1) having a device (10) according to any one of claims 17 to 19.

Images