Method for classifying objects

Abstract

The invention relates to a method for classifying objects (20) by means of a plurality of ultrasonic sensors (11, 12, 13, 14), wherein an overall image is generated by carrying out the following steps for each of the ultrasonic sensors (11, 12, 13, 14): a) the respective ultrasonic sensor (11, 12, 13, 14) emits a signal; b) the emitted signal is reflected by at least one object (20); c) the reflected signal is picked up by all ultrasonic sensors (11, 12, 13, 14); d) a first image channel is generated from the picked-up signals and inserted into the overall image; e) a second image channel is generated from the picked-up signals and inserted into the overall image; f) a third image channel is generated from the picked-up signals and inserted into the overall image; steps a) to f) are carried out at successive times for each of the ultrasonic sensors (11, 12, 13, 14); the generated overall image is supplied to a classification unit; and at least one object (20) is classified by the classification unit, to which previously generated training images have been supplied.

Description

[0001] Methods for classifying objects

[0002] Description:

[0003] The invention relates to a method for classifying objects using a plurality of ultrasonic sensors.

[0004] In a technical facility, such as a factory, safety technology is used to reduce the risk of physical injury from an autonomous mobile robot to an acceptable level. This technology can typically only distinguish, in binary terms, whether something is in front of the autonomous mobile robot, but not what it is. In outdoor applications, such as on a street, current safety technology cannot differentiate between curbs and people lying down. Furthermore, dust, leaves, rain, snow, and grass are recognized as objects and cause the autonomous mobile robot to stop. Therefore, existing technology cannot be used outdoors.

[0005] Document DE 10 2021 001 282 A1 discloses a mobile system and a method for operating the mobile system in a technical plant. The mobile system comprises a first sensor, designed as a laser scanner, and a second sensor, designed as a monocular camera. By combining the data acquired by the two sensors, objects in the technical plant are detected and classified.

[0006] German patent application DE 102017 121 581 A1 discloses a method for processing data obtained by means of ultrasound. In this process, ultrasound data is converted into a one-dimensional vector, and the resulting vector is used to generate input information for a neural network.

[0007] From DE 102012 020413 A1, a method and a device for sensing the environment of a vehicle are known. This involves fusing measurement results from several measuring devices for object location in the vehicle's vicinity.

[0008] From DE 102016210 534 A1, a method for classifying a vehicle's environment is known. A representation of the environment is provided, and the environment is classified into predefined categories using a machine learning method.

[0009] ISI \ EIDOPAT 10.07.2025 The invention is based on the objective of further developing a method for classifying objects.

[0010] The problem is solved by a method for classifying objects with the features specified in claim 1. Advantageous embodiments and further developments are the subject of the dependent claims.

[0011] A method for classifying objects using multiple ultrasonic sensors is proposed. An overall image is generated by performing the following steps for each ultrasonic sensor: a) the respective ultrasonic sensor emits a signal; b) the emitted signal is reflected by at least one object; c) the reflected signal is received by all ultrasonic sensors; d) a first image channel is generated from the received signals and inserted into the overall image; e) a second image channel is generated from the received signals and inserted into the overall image; f) a third image channel is generated from the received signals and inserted into the overall image.

[0012] Steps a) to f) are performed sequentially for each of the ultrasonic sensors. The resulting image is fed to a classification unit, which then classifies at least one object. The classification unit was previously fed with training images.

[0013] Unlike other sensors such as laser scanners or cameras, ultrasonic sensors are relatively insensitive to interference, for example, from rain, snow, leaves, and other environmental conditions. The method according to the invention is therefore also applicable outside of enclosed buildings, for example, in a pedestrian zone or a parking lot. The method according to the invention is also applicable in backlighting and in darkness. By classifying a detected object, it is possible, for example, to distinguish between a curb and a person lying down, thus avoiding unnecessary stops of a mobile robot. The method is also applicable when objects are moving relative to the ultrasonic sensors. The method according to the invention is suitable for classifying several different types of objects simultaneously. According to an advantageous embodiment of the invention, the classification unit comprises a neural network.The classification unit is trained to classify a wide variety of real-world objects, particularly those found in residential areas. These objects include, for example, adults, children, bicycles, walkers, motor vehicles, curbs, shrubs, bushes, trees, dolls, ride-on cars, and other toys. The previously generated training images were fed into the neural network of the classification unit.

[0014] According to an advantageous embodiment of the invention, the classification unit additionally localizes at least one object. Localization comprises determining the distance to a detected object and determining the direction to the detected object relative to the ultrasonic sensors. Localization can be performed by evaluating the transit times of the transmitted and received signals. The method is suitable for simultaneously localizing several different types of objects.

[0015] According to an advantageous embodiment of the invention, the ultrasonic sensors are each arranged at a defined position on a mobile robot. This enables the mobile robot to classify objects in its environment. This allows mobile robots to operate, for example, for transporting packages, food, or other cargo, outside of enclosed buildings, such as in residential areas.

[0016] According to an advantageous embodiment of the invention, the ultrasonic sensors are arranged on a front side of the mobile robot in a transverse offset from one another, preferably at equidistant intervals. This allows the mobile robot to detect objects, particularly those traveling in the direction of travel. The transverse offset, i.e., perpendicular to the direction of travel, increases the resolution of the acquired signals and improves the quality and accuracy of the classification.

[0017] When the mobile robot travels straight ahead, its direction of travel corresponds to a longitudinal direction. The transverse direction is perpendicular to the longitudinal direction. A vertical direction is perpendicular to both the longitudinal and transverse directions. The vertical direction is perpendicular to the surface on which the mobile robot is located. The transverse and longitudinal directions are parallel to the surface. According to an advantageous embodiment of the invention, before one of the ultrasonic sensors transmits a signal in step a), the image channels previously generated for that ultrasonic sensor are removed from the overall image. Thus, a generated overall image always comprises the image channels of exactly four ultrasonic sensors and is therefore current. Image channels that were generated previously and are therefore potentially outdated are not considered for classification purposes.For example, a person moving perpendicular to the direction of travel of the mobile robot will no longer be detected once they have left the detection range of the ultrasonic sensors.

[0018] According to an advantageous embodiment of the invention, in step d) the recorded signals are inserted into the overall image as elliptical arcs with the respective distance, wherein the width of an elliptical arc corresponds to a distance from when the dynamic threshold of reflection is first exceeded until the dynamic threshold of reflection is again undershot, and wherein the brightness of a pixel is proportional to the number of elliptical arcs that lie over the respective pixel. A pixel corresponds to a defined area in front of the ultrasonic sensors, wherein said area lies in a plane perpendicular to the vertical direction.

[0019] According to an advantageous embodiment of the invention, in step e), the recorded signals are inserted into the overall image as elliptical arcs with the respective distances, wherein the width of an elliptical arc corresponds to a distance from when the dynamic threshold of the reflection is first exceeded until the dynamic threshold of the reflection is again undershot, and wherein the brightness of a pixel is proportional to an average amplitude of superimposed reflections. A pixel corresponds to a defined area in front of the ultrasonic sensors, wherein said area lies in a plane perpendicular to the vertical direction.

[0020] According to an advantageous embodiment of the invention, in step f) all reflections are inserted into the overall image as intersection points of the elliptical arcs at the respective distances, wherein the brightness of a pixel is proportional to the maximum amplitude of a reflection. A pixel corresponds to a defined area in front of the ultrasonic sensors, said area lying in a plane perpendicular to the vertical direction.

[0021] According to an advantageous embodiment of the invention, if a signal emitted by one of the ultrasonic sensors is received exclusively by the emitting ultrasonic sensor, the corresponding ultrasonic sensor is marked accordingly. If an interfering object, for example, a hand, is located in an area of ​​approximately 0 cm to 25 cm in front of an ultrasonic sensor, signals received by said ultrasonic sensor are blocked. Similarly, reflected signals from the other ultrasonic sensors are blocked for the aforementioned ultrasonic sensor. In this case, the ultrasonic sensor is marked by a near-field detection flag. As long as the near-field detection flag is active, said ultrasonic sensor is not used for classification. Additionally, a cone, which, for example, has a length of approximately 25 cm, is drawn into the overall image.

[0022] According to an advantageous embodiment of the invention, at least one training image is generated in preparation during a training phase by performing the following steps for each of the ultrasonic sensors:

[0023] 1) The respective ultrasonic sensor emits a signal;

[0024] 2) the emitted signal is reflected by the object;

[0025] 3) The reflected signal is received by all ultrasonic sensors;

[0026] 4) A first image channel is generated from the recorded signals and inserted into the training image;

[0027] 5) A second image channel is generated from the recorded signals and inserted into the training image;

[0028] 6) A third image channel is generated from the recorded signals and inserted into the training image.

[0029] Steps 1) to 6) are performed sequentially for each of the ultrasonic sensors. An object class is assigned to the object contained in the training image, and the classification unit is trained using the generated training image.

[0030] Preferably, a very large number, in particular at least 1000, of training images are generated, wherein each object contained in the training images is assigned an object class, and wherein the classification unit is trained with the generated training images.

[0031] Preferably, in each generated training image, an area containing the object is additionally marked, and the classification unit is trained using these generated training images. If multiple objects are present in the same training image, each object is assigned an object class, and the area containing that object is marked. The classification unit is then trained using these generated training images.

[0032] By generating training images and training the classification unit with these images, the classification unit is enabled to classify a wide variety of real-world objects, particularly those found in residential areas. Such objects include, for example, adults, children, bicycles, walkers, motor vehicles, curbs, shrubs, bushes, trees, dolls, ride-on cars, and other toys.

[0033] According to an advantageous embodiment of the invention, in step 4) the recorded signals are inserted into the training image as elliptical arcs with the respective distance, wherein the width of an elliptical arc corresponds to a distance from when the dynamic threshold of reflection is first exceeded until the dynamic threshold of reflection is again undershot, and wherein the brightness of a pixel is proportional to the number of elliptical arcs that lie over the respective pixel. A pixel corresponds to a defined area in front of the ultrasonic sensors, wherein said area lies in a plane perpendicular to the vertical direction.

[0034] According to an advantageous embodiment of the invention, in step 5) the recorded signals are inserted into the training image as elliptical arcs with the respective distance, wherein the width of an elliptical arc corresponds to a distance from when the dynamic threshold of the reflection is first exceeded until the dynamic threshold of the reflection is again undershot, and wherein the brightness of a pixel is proportional to an average amplitude of superimposed reflections. A pixel corresponds to a defined area in front of the ultrasonic sensors, wherein said area lies in a plane perpendicular to the vertical direction.

[0035] According to an advantageous embodiment of the invention, in step 6) all reflections are inserted into the training image as intersection points of the elliptical arcs at the respective distances, wherein the brightness of a pixel is proportional to the maximum amplitude of a reflection. A pixel corresponds to a defined area in front of the ultrasonic sensors, the said area lying in a plane perpendicular to the vertical direction.

[0036] The invention is not limited to the combination of features stated in the claims. For a person skilled in the art, further meaningful combinations of claims and / or individual claim features and / or features of the description and / or the figures will become apparent, in particular from the problem statement and / or the problem arising from a comparison with the prior art.

[0037] The invention will now be explained in more detail with reference to the illustrations. The invention is not limited to the embodiments shown in the illustrations. The illustrations only depict the subject matter of the invention schematically. They show:

[0038] Figure 1: a schematic side view of a mobile robot with an object to be classified and

[0039] Figure 2: a schematic top view of the mobile robot and the object to be classified.

[0040] Figure 1 shows a schematic side view of a mobile robot 10 with an object 20 to be classified. In this case, the object 20 is a cylindrical object, such as a pole, a post, or a lamppost. In this illustration, the mobile robot 10 and the object 20 are located on an approximately level surface outside of enclosed buildings, for example, in a pedestrian zone or a parking lot.

[0041] In the diagram shown, the mobile robot 10 moves in a longitudinal direction X towards object 20 and is at a distance from object 20 in the longitudinal direction X. The longitudinal direction X corresponds to the direction of travel of the mobile robot 10. A vertical direction Z runs perpendicular to the ground and perpendicular to the longitudinal direction X. A transverse direction Y runs perpendicular to the vertical direction Z and perpendicular to the longitudinal direction X. The longitudinal direction X and the transverse direction Y extend parallel to the ground and represent horizontal directions. Any direction that extends perpendicular to the vertical direction Z and parallel to the ground represents a horizontal direction.

[0042] The mobile robot 10 is designed as an autonomously driving vehicle and has a drive unit, an electrical energy storage device to supply the drive unit, and a control unit to control the drive unit. The drive unit includes, among other things, an electric motor, a gearbox, and several wheels 17. The wheels 17 are in contact with the ground.

[0043] The mobile robot 10 is used, for example, to transport packages, food, or other cargo. The mobile robot 10 also features a position sensor for determining its position. Furthermore, the mobile robot 10 has a communication device for wireless communication via a network with a server and with other mobile robots 10.

[0044] The mobile robot 10 has a first ultrasonic sensor 11, a second ultrasonic sensor 12, a third ultrasonic sensor 13, and a fourth ultrasonic sensor 14. The ultrasonic sensors 11, 12, 13, and 14 are arranged on a front side of the mobile robot 10, offset from each other in the transverse Y direction. In the present illustration, the first ultrasonic sensor 11, the second ultrasonic sensor 12, and the third ultrasonic sensor 13 are obscured by the fourth ultrasonic sensor 14.

[0045] The ultrasonic sensors 11, 12, 13, 14 emit signals that are reflected by objects 20. The reflected signals are detected by the ultrasonic sensors 11, 12, 13, 14. The distance to the object 20 that reflects the signal can be determined from the signal's travel time.

[0046] The mobile robot 10 also features a classification unit. This unit is used to classify objects 20 that reflect signals emitted by the ultrasonic sensors 11, 12, 13, and 14. The classification unit also serves to locate objects 20 that reflect signals emitted by the ultrasonic sensors 11, 12, 13, and 14. The classification unit incorporates a neural network. Previously generated training images were fed to the classification unit, and in particular to the neural network.

[0047] Figure 2 shows a schematic top view of the mobile robot 10 and the object 20 to be classified. The mobile robot 10 has an approximately rectangular cross-section. The ultrasonic sensors 11, 12, 13, 14 are arranged offset from one another in the transverse direction Y on a longitudinally X front side, which is referred to as the front of the mobile robot 10. The ultrasonic sensors 11, 12, 13, 14 are arranged at equidistant intervals from each other.

[0048] To classify object 20 using the ultrasonic sensors 11, 12, 13, 14, a composite image is generated. This composite image is then fed to the classification unit.

[0049] The classification unit classifies object 20. To do this, in step a), the first ultrasonic sensor 11 first emits a signal. In step b), the signal emitted by the first ultrasonic sensor 11 is reflected by object 20. In step c), the reflected signal is received by all ultrasonic sensors 11, 12, 13, and 14.

[0050] The emitted by the first ultrasonic sensor 11 and by the first ultrasonic sensor

[0051] The recorded signal is represented in the form of a first elliptical arc 111. The radii of the first elliptical arc 111 both correspond to a distance between the object 20 and the first ultrasonic sensor 11. The first elliptical arc 111 thus corresponds to a circular arc.

[0052] The signal emitted by the first ultrasonic sensor 11 and by the second ultrasonic sensor

[0053] The signal recorded is represented in the form of a second elliptical arc 112. The radii of the second elliptical arc 112 correspond to the distance of the object 20 to the first ultrasonic sensor 11 and the distance of the object 20 to the second ultrasonic sensor 12.

[0054] The emission from the first ultrasonic sensor 11 and from the third ultrasonic sensor

[0055] The recorded signal is represented in the form of a third elliptical arc 113. The radii of the third elliptical arc 113 correspond to the distance of the object 20 to the first ultrasonic sensor 11 and the distance of the object 20 to the third ultrasonic sensor 13.

[0056] The emission from the first ultrasonic sensor 11 and from the fourth ultrasonic sensor

[0057] The recorded signal 14 is represented in the form of a fourth elliptical arc 114. The radii of the fourth elliptical arc 114 correspond to the distance of the object 20 to the first ultrasonic sensor 11 and the distance of the object 20 to the fourth ultrasonic sensor 14.

[0058] The elliptical arcs 111, 112, 113, 114 each have a width in the radial direction. The width of an elliptical arc 111, 112, 113, 114 corresponds to a distance from when the dynamic threshold of reflection is first exceeded until it falls below the dynamic threshold again. In step d), a first image channel is generated from the recorded signals and inserted into the overall image. The recorded signals are inserted into the overall image as elliptical arcs 111, 112, 113, 114 with their respective distances. The brightness of a pixel is proportional to the number of elliptical arcs 111, 112, 113, 114 that lie above the respective pixel. A pixel corresponds to a defined area in front of the ultrasonic sensors, where said area lies in a plane perpendicular to the vertical direction Z.

[0059] In step e), a second image channel is generated from the recorded signals and inserted into the overall image. The recorded signals are inserted into the overall image as elliptical arcs 111, 112, 113, 114 at their respective distances. The brightness of a pixel is proportional to the average amplitude of superimposed reflections. A pixel corresponds to a defined area in front of the ultrasonic sensors, with this area lying in a plane perpendicular to the vertical direction Z.

[0060] In step f), a third image channel is generated from the recorded signals and inserted into the overall image. All reflections are inserted into the overall image as intersection points of the elliptical arcs 111, 112, 113, 114 at their respective distances. The brightness of a pixel is proportional to the maximum amplitude of a reflection. A pixel corresponds to a defined area in front of the ultrasonic sensors, where this area lies in a plane perpendicular to the vertical direction Z.

[0061] Then, in step a), the second ultrasonic sensor 12 emits a signal. The signal emitted by the second ultrasonic sensor 12 is reflected by the object 20 in step b). The reflected signal is received by all ultrasonic sensors 11, 12, 13, 14 in step c).

[0062] The signals emitted by the second ultrasonic sensor 12 and received by all ultrasonic sensors 11, 12, 13, 14 can also be represented as elliptical arcs. However, these elliptical arcs are not shown in the diagram.

[0063] In step d), a first image channel is generated from the recorded signals and inserted into the overall image. The recorded signals are inserted into the overall image as elliptical arcs with their respective distances. The brightness of a pixel is proportional to the number of elliptical arcs that lie above the respective pixel. A pixel corresponds to a defined area in front of the ultrasonic sensors, with this area lying in a plane perpendicular to the vertical direction Z.

[0064] In step e), a second image channel is generated from the recorded signals and inserted into the overall image. The recorded signals are inserted into the overall image as elliptical arcs with their respective distances. The brightness of a pixel is proportional to the average amplitude of superimposed reflections. A pixel corresponds to a defined area in front of the ultrasonic sensors, with this area lying in a plane perpendicular to the vertical direction Z.

[0065] In step f), a third image channel is generated from the recorded signals and inserted into the overall image. All reflections are incorporated into the overall image as intersection points of the elliptical arcs at their respective distances. The brightness of a pixel is proportional to the maximum amplitude of a reflection. A pixel corresponds to a defined area in front of the ultrasonic sensors, where this area lies in a plane perpendicular to the vertical direction Z.

[0066] Then, in step a), the third ultrasonic sensor 13 emits a signal. The signal emitted by the third ultrasonic sensor 13 is reflected by the object 20 in step b). The reflected signal is received by all ultrasonic sensors 11, 12, 13, 14 in step c).

[0067] The signals emitted by the third ultrasonic sensor 13 and received by all ultrasonic sensors 11, 12, 13, 14 can also be represented as elliptical arcs. However, these elliptical arcs are not shown in the illustration.

[0068] In step d), a first image channel is generated from the recorded signals and inserted into the overall image. The recorded signals are inserted into the overall image as elliptical arcs with their respective distances. The brightness of a pixel is proportional to the number of elliptical arcs that lie above the respective pixel. A pixel corresponds to a defined area in front of the ultrasonic sensors, with this area lying in a plane perpendicular to the vertical direction Z.

[0069] In step e), a second image channel is generated from the recorded signals and inserted into the overall image. The recorded signals are inserted into the overall image as elliptical arcs with their respective distances. The brightness of a pixel is proportional to the average amplitude of superimposed reflections. A pixel corresponds to a defined area in front of the ultrasonic sensors, with this area lying in a plane perpendicular to the vertical direction Z.

[0070] In step f), a third image channel is generated from the recorded signals and inserted into the overall image. All reflections are incorporated into the overall image as intersection points of the elliptical arcs at their respective distances. The brightness of a pixel is proportional to the maximum amplitude of a reflection. A pixel corresponds to a defined area in front of the ultrasonic sensors, where this area lies in a plane perpendicular to the vertical direction Z.

[0071] Then, in step a), the fourth ultrasonic sensor 14 emits a signal. The signal emitted by the fourth ultrasonic sensor 14 is reflected by the object 20 in step b). The reflected signal is received by all ultrasonic sensors 11, 12, 13, 14 in step c).

[0072] The signals emitted by the fourth ultrasonic sensor 14 and received by all ultrasonic sensors 11, 12, 13, 14 can also be represented as elliptical arcs. However, these elliptical arcs are not shown in the illustration.

[0073] In step d), a first image channel is generated from the recorded signals and inserted into the overall image. The recorded signals are inserted into the overall image as elliptical arcs with their respective distances. The brightness of a pixel is proportional to the number of elliptical arcs that lie above the respective pixel. A pixel corresponds to a defined area in front of the ultrasonic sensors, with this area lying in a plane perpendicular to the vertical direction Z.

[0074] In step e), a second image channel is generated from the recorded signals and inserted into the overall image. The recorded signals are inserted into the overall image as elliptical arcs with their respective distances. The brightness of a pixel is proportional to the average amplitude of superimposed reflections. A pixel corresponds to a defined area in front of the ultrasonic sensors, with this area lying in a plane perpendicular to the vertical direction Z. PT7293 POT

[0075] - 14 -

[0076] In step f), a third image channel is generated from the recorded signals and inserted into the overall image. All reflections are incorporated into the overall image as intersection points of the elliptical arcs at their respective distances. The brightness of a pixel is proportional to the maximum amplitude of a reflection. A pixel corresponds to a defined area in front of the ultrasonic sensors, where this area lies in a plane perpendicular to the vertical direction Z.

[0077] Steps a) to f) are thus performed at sequentially sequential times for each of the ultrasonic sensors 11, 12, 13, 14. This generates the overall image. As already mentioned, the resulting overall image is then fed to the classification unit.

[0078] The classification unit classifies object 20. The classification unit also locates object 20. The resulting image is fed to a classification unit, which then classifies at least one object. Previously generated training images were used to guide this process.

[0079] Reference symbol list

[0080] 10 Mobile robot 11 First ultrasonic sensor

[0081] 12 second ultrasonic sensor

[0082] 13 third ultrasonic sensor

[0083] 14 fourth ultrasonic sensor

[0084] 17 Wheel 20 Object

[0085] 111 first elliptical arc

[0086] 112 second elliptical arc

[0087] 113 third elliptical arc

[0088] 114 fourth elliptical arc X longitudinal direction

[0089] Y transverse direction

[0090] Z Vertical direction

Claims

Patent claims:

1. A method for classifying objects (20) using a plurality of ultrasonic sensors (11, 12, 13, 14), wherein an overall image is generated by performing the following steps for each of the ultrasonic sensors (11, 12, 13, 14): a) the respective ultrasonic sensor (11, 12, 13, 14) emits a signal; b) the emitted signal is reflected by at least one object (20); c) the reflected signal is received by all ultrasonic sensors (11, 12, 13, 14); d) a first image channel is generated from the received signals and inserted into the overall image; e) a second image channel is generated from the received signals and inserted into the overall image; f) a third image channel is generated from the received signals and inserted into the overall image; Steps a) to f) are performed at sequentially sequential times for each of the ultrasonic sensors (11, 12, 13, 14);the generated overall image is fed to a classification unit; and at least one object (20) is classified by the classification unit, to which previously generated training images were fed.; 2. The method according to claim 1, characterized in that the classification unit comprises a neural network.

3. Method according to one of the preceding claims, characterized in that at least one object (20) is located by the classification unit.

4. Method according to one of the preceding claims, characterized in that the ultrasonic sensors (11, 12, 13, 14) are each arranged at a defined position on a mobile robot.

5. Method according to claim 4, characterized in that the ultrasonic sensors (11, 12, 13, 14) are arranged offset from each other in a transverse direction (Y) on a front side of the mobile robot, preferably at equidistant distances from each other.

6. Method according to one of the preceding claims, characterized in that before one of the ultrasonic sensors (11, 12, 13, 14) sends out a signal in step a), the image channels previously generated for the respective ultrasonic sensor (11 , 12, 13, 14) are removed from the overall image.

7. Method according to one of the preceding claims, characterized in that in step d) the recorded signals are inserted into the overall image as elliptical arcs (111 , 112, 113, 114) at the respective distance, wherein the width of an elliptical arc (111, 112, 113, 114) corresponds to a distance when the dynamic threshold of reflection is exceeded for the first time, until the dynamic threshold of reflection is again undershot, and wherein a brightness of a pixel is proportional to a number of elliptical arcs (111 , 112, 113, 114) which lie above the respective pixel. - 18 - 8. Method according to one of the preceding claims, characterized in that in step e) the recorded signals are inserted into the overall image as elliptical arcs (111 , 112, 113, 114) at the respective distance, wherein the width of an elliptical arc (111, 112, 113, 114) corresponds to a distance when the dynamic threshold of the reflection is exceeded for the first time, until the dynamic threshold of the reflection is again undershot, and wherein a brightness of a pixel is proportional to an average amplitude of superimposed reflections.

9. Method according to one of the preceding claims, characterized in that in step f) all reflections are inserted into the overall image as intersection points of the elliptical arcs (111, 112, 113, 114) at the respective distance, wherein a brightness of a pixel is proportional to a maximum amplitude of a reflection.

10. Method according to one of the preceding claims, characterized in that if a signal emitted by one of the ultrasonic sensors (11, 12, 13, 14) is received exclusively by the emitting ultrasonic sensor (11, 12, 13, 14), the corresponding ultrasonic sensor (11, 12, 13, 14) is marked accordingly. - 19 - 11. Method according to one of the preceding claims, characterized in that at least one training image is generated in preparation during a training phase by performing the following steps for each of the ultrasonic sensors (11 , 12, 13, 14): 1) The respective ultrasonic sensor (11, 12, 13, 14) emits a signal; 2) the emitted signal is reflected by the object (20); 3) the reflected signal is received by all ultrasonic sensors (11 , 12, 13, 14); 4) A first image channel is generated from the recorded signals and inserted into the training image; 5) A second image channel is generated from the recorded signals and inserted into the training image; 6) A third image channel is generated from the recorded signals and inserted into the training image; steps 1) to 6) are carried out at sequentially sequential times for each of the ultrasonic sensors (11, 12, 13, 14); an object class is assigned to the object (20) contained in the training image; the classification unit is trained with the generated training image.

12. Method according to claim 11, characterized in that in step 4) the recorded signals are inserted into the training image as elliptical arcs (111, 112, 113, 114) with the respective distance, wherein the width of an elliptical arc (111, 112, 113, 114) corresponds to a distance when the dynamic threshold of reflection is exceeded for the first time, until the dynamic threshold of reflection is again undershot, and wherein a brightness of a pixel is proportional to a number of elliptical arcs (111, 112, 113, 114) which lie above the respective pixel. - 20 - 13. Method according to one of claims 11 to 12, characterized in that in step 5) the recorded signals are inserted into the training image as elliptical arcs (111 , 112, 113, 114) with the respective distance, wherein the width of an elliptical arc (111, 112, 113, 114) corresponds to a distance when the dynamic threshold of the reflection is exceeded for the first time, until the dynamic threshold of the reflection is again undershot, and wherein a brightness of a pixel is proportional to an average amplitude of superimposed reflections.

14. Method according to one of claims 11 to 13, characterized in that in step 6) all reflections are inserted into the training image as intersection points of the elliptical arcs (111, 112, 113, 114) at the respective distance, wherein a brightness of a pixel is proportional to a maximum amplitude of a reflection.

Images