Method and apparatus for combining classification results of a plurality of classification models for object identification by means of ultrasound-based sensor systems in mobile devices
Patent Information
- Application Number
- EP2023703757
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2022-02-21
- Filing Date
- 2023-02-06
- Publication Date
- 2025-01-01
AI Technical Summary
Ultrasonic sensor systems in mobile devices, such as vehicles, face challenges in accurately classifying collision-relevant surrounding objects due to different transducer arrangements and separately trained classification models, leading to varying and potentially safety-critical classification results when objects are detected by multiple sensors during maneuvers.
A method for merging classification results from multiple ultrasonic sensor devices using a weighted averaging of classification quality and novelty, where the overall quality is determined by combining the classification quality, detection quality, feature quality, and model quality, and selecting the classification result based on the highest overall quality or novelty threshold.
Improves classification accuracy by selecting the most reliable and recent classification result, enhancing safety by ensuring consistent and accurate detection of collision-relevant objects during vehicle maneuvers.
Smart Images
Figure 1.1
Abstract
Description
[0001] Description
[0002] title
[0003] Method and device for fusion of classification results of multiple classification models for object identification using ultrasound-based sensor systems in mobile devices
[0004] Technical area
[0005] The invention relates to ultrasonic sensor systems for mobile devices, such as motor vehicles, and in particular to the fusion of classification results from several ultrasonic sensor devices with regard to collision-relevant environmental objects.
[0006] Technical background
[0007] Vehicles are typically equipped with ultrasonic sensor systems for object detection. These often have multiple ultrasonic sensor devices, each with multiple ultrasonic transducers for different detection ranges in which surrounding objects are to be detected. Such ultrasonic sensor devices are often arranged on the front and rear bumpers of a vehicle. The ultrasonic sensor devices are usually designed differently and, due to the different arrangement of the individual ultrasonic transducers, have a separate classification model for detecting / identifying surrounding objects.
[0008] Thus, the ultrasonic transducer arrangements of the ultrasonic sensor devices for the detection of collision-relevant environmental objects are evaluated with different classification models, and these classification models classify the environmental objects with regard to their height, so that a distinction can be made as to whether the environmental objects can be driven over or not, ie are collision-relevant.
[0009] This distinction between high and low ambient objects has so far been handled using rule-based methods, whereby ambient objects that only cause a low signal strength of the ultrasonic reception signal are masked out very early in the signal chain. Alternatively, other approaches can evaluate a variety of sensor parameters or object properties using heuristic approaches and empirical values.
[0010] Disclosure of the invention
[0011] According to the invention, a method for merging classification results of several classification models for object identification using ultrasonic sensor devices in mobile devices according to claim 1 and a corresponding device according to the independent claim are provided.
[0012] Further embodiments are specified in the dependent claims.
[0013] According to a first aspect, a method for merging classification results of multiple classification models for classifying environmental objects using ultrasonic sensor devices in a mobile device, comprising the following steps:
[0014] Detecting sensor signals from ultrasonic transducers of the ultrasonic sensor devices;
[0015] Tracking a position of detected environmental objects by the ultrasonic sensor devices and classifying an object type of the detected environmental objects by a classification model assigned to the ultrasonic sensor devices in order to obtain a respective classification result;
[0016] Upon determining that a specific environmental object has entered a detection range of one of the ultrasonic sensor devices and has previously been located in a detection range of another of the ultrasonic sensor devices, selecting one of the classification results for the specific environmental object depending on a classification quality based on the classification model and on a degree of novelty of the classification result.
[0017] The use of machine learning to create data-based classification models requires the collection of training data for various environmental objects. This training data is determined from signal characteristics of the received ultrasonic signals and assigned a corresponding object type of the environmental object. The object type specifies, among other things, the distinction between traversable and collision-relevant environmental objects. Signal characteristics can be aggregated variables from the sensor signals, such as maximum amplitude, time of maximum amplitude, phase position between two received signals, as well as other attributes, such asa currently measured amplitude, an extent of the object, a detection rate, a number of average received signal peaks / transmission pulses (multi-reflectivity), a number of ultrasonic transducers that can receive the surrounding object, a signal stability during the movement of the mobile device, a quotient of the distance traveled and the measured object distance, a number of found intersection points of echo circles of neighboring ultrasonic transducers and the like.
[0018] The characteristics of signal features often depend crucially on the installation position of the individual ultrasonic transducers in an ultrasonic sensor device. The installation height and vertical adjustment angle are particularly important here. In a vehicle as a mobile device, where ultrasonic sensor devices are arranged on the front and rear bumpers, the ultrasonic transducer arrangements of the ultrasonic sensor devices typically differ considerably due to the different bumper geometries. Consequently, separately trained classification models are assigned to the ultrasonic sensor devices.
[0019] However, the classification quality for each of the ultrasonic sensor devices depends on a variety of influencing factors. In addition to the distance to the environmental object to be detected, the detection angle at which the environmental object in question is detected by an ultrasonic transducer, the relative speed between the environmental object and the mobile device, and the duration during which an ultrasonic sensor device detects a corresponding environmental object, the configuration of the ultrasonic transducer arrangement is also crucial for the classification quality. For example, if the ultrasonic transducers are installed very low, the difference in the calculated signal characteristics, based on amplitude or multi-reflectivity, for example, between high and low environmental objects is smaller than with a higher installation position.
[0020] An ultrasonic sensor system on a mobile device, such as a vehicle, can have multiple ultrasonic devices oriented in different directions. The ultrasonic sensor devices detect the surroundings of the mobile device in different directions and map detected objects. Additionally, the detected objects are classified according to the characteristics of the received ultrasonic signal to determine whether the object is, for example, a drivable object or a collision-relevant object.
[0021] However, during movement of the mobile device, it often occurs that a specific environmental object is detected successively by different ultrasonic sensor devices (e.g., in a motor vehicle, by the ultrasonic sensor devices on the front and rear bumpers) under different conditions. Such cases occur, for example, during turning maneuvers, passing the environmental object in question, driving over low environmental objects, and the like.
[0022] If a specific environmental object is detected by multiple ultrasonic sensor devices during a corresponding maneuver of the mobile device, separate classifications are usually performed for that environmental object. Due to the different ultrasonic transducer arrangements of the ultrasonic sensor devices and the separately trained classification models, separate, specific classification results are obtained for the relevant environmental object by evaluating the classification models. The classification results can therefore also differ, which can be safety-critical, especially for collision-relevant environmental objects.
[0023] In addition, the classification models used offer the possibility of evaluating the classification quality of the environmental object in question. The classification quality is based on the classification confidence, which can be specified by the classification model. Especially in conjunction with the different classification qualities of the classification models, a specific environmental object is detected and classified with different classification results and different classification qualities depending on the specific movement situation of the vehicle.
[0024] Therefore, to improve the classification accuracy of classification with multiple ultrasonic sensor devices, it may be useful to fuse the classification evaluations.
[0025] Furthermore, the classification models can be data-based, whereby the classification qualities are determined as values of an element of an output vector of the classification models with the highest value.
[0026] The classification quality can be determined directly by the classification model used as classification confidence. This can implicitly take into account the detection quality, which is derived from the reliability of the detections of the relevant environmental object by the respective classification model, the feature quality, which assesses the quality and availability of individual signal features, and the model quality of the classification model, which assesses the overall quality of the classification model. Alternatively, the classification quality can also be determined from these separately determinable individual qualities.
[0027] It can be provided that the degree of novelty is assigned to a novelty quality using an assignment function, which indicates a higher quality the shorter the time since the acquisition of the underlying sensor data, whereby an overall quality is determined from the classification quality and the novelty quality, whereby the classification result is selected depending on the overall qualities of the classification results of the classification models.
[0028] In particular, the overall quality sigma_ges of a classification process by an ultrasonic sensor device can be determined as a weighted average using the classification quality sigma_klass of the relevant classification model and a novelty quality sigma_neu resulting from a novelty degree of the last determined classification result for the relevant environmental object, where: sigma_ges= (sigma_klass*a1 +sigma_neu*a2) / 2 with sigma_klass, sigma_neu =0 ...1 , where a1 and a2 represent additional predeterminable weighting factors.
[0029] Alternatively, depending on the classification model used, if it cannot implicitly provide classification confidence or if it is not trusted, the aforementioned individual qualities—namely, the detection quality sigma_det, the feature quality sigma_merk, and the model quality sigma_mod—which relate to the classification model and the relevant environment object, can be determined separately and calculated to form an overall quality sigma_ges using a weighted average. sigma_ges = (sigma_det*a1+sigma_merk*a2+sigma_mod*a3+sigma_neu*a4) / 4, where a1 to a4 represent additional predefined weighting factors.
[0030] If a specific environmental object is detected and classified by multiple ultrasonic sensor devices, the classification result of the classification model with the highest overall quality sigma_ges can be used. If the classification quality of the classification model is determined using the detection quality and / or the feature quality and / or the model quality, the individual qualities can be determined as described below:
[0031] For example, the detection quality can be determined depending on a quotient of the number of correct classifications of the specific environmental object and the total number of classifications of the specific environmental object.
[0032] For detection quality, a quotient of the number of correct classifications and the total number of detections is calculated using predefined validation data sets. This quotient can be assigned to a corresponding detection quality value using a mapping function, such as a lookup table.
[0033] The number of correct classifications is determined by evaluating a predefined database of labeled data. These form validation data sets that assign sensor signals and / or sensor characteristics generated from them of known environmental objects to a classification (object property: traversable, collision-relevant). These validation data sets are then used to determine the number of object detections in which the classification result of the actual object property is recognized, and a corresponding probability is determined.
[0034] The feature quality can be determined depending on the contributions of signal features resulting from the sensor signals for determining the classification result for the classification of the specific environmental object and a detection feature, in particular a relative position of the specific environmental object to the ultrasonic transducer associated with the signal feature.
[0035] Thus, the feature quality can be determined from the availability of the signal features used. This evaluates how the different signal features are available and how they can contribute to the classification of a specific environmental object. Since the quality of signal features often depends on the detection distance and angle and / or the nature of surfaces (reflectivity), the signal features to be evaluated, which form the basis for the classification by the respective classification model, can be used accordingly to assess the quality of the classification result. For example, the signal feature of amplitude provides very good differentiation at close range, while at long range it has only a low value due to similar vertical detection angles.Furthermore, high and low amplitude values can be a good identification for the object height, while values in the middle range are less meaningful.
[0036] To determine the overall feature quality, each individual signal feature to be evaluated and the detection conditions, such as the distance to the object to be classified and the detection angle, are considered, and a corresponding individual feature quality is determined using a suitable assignment function. Furthermore, the individual feature qualities for the individual signal features are aggregated into an overall feature quality, in particular by averaging, particularly with a weighted consideration of the individual feature qualities.
[0037] Furthermore, the model quality of the classification model used can be considered. This takes into account that different model qualities can be achieved depending on the respective sensor configuration, which results from the installation situation of the ultrasonic transducers of the respective ultrasonic sensor device. The model quality is independent of the environmental situation and only determines the quality of a classification based on the configuration (arrangement and geometric orientation) of the ultrasonic transducers of the respective ultrasonic sensor device.
[0038] The novelty level of a specific environmental object can preferably be determined based on the timestamp of the last detection for classifying the environmental object in question. For environmental objects detected by multiple ultrasonic sensor devices, the respective timestamp of the last classification of the environmental object in question is translated into a corresponding novelty level, in particular using a suitable assignment function that takes into account the probability that changes in the environment occur within a specific time window.
[0039] It may be provided that the classification result of the classification model whose degree of novelty indicates a more recent classification is selected if the corresponding overall quality exceeds a predetermined overall quality threshold.
[0040] In an alternative embodiment, the decision regarding the classification result to be used can always be made dependent on the novelty of the classification result as soon as the corresponding overall quality of the classification model in question exceeds a predetermined overall quality threshold. As a result, when the mobile device moves, the classification result of that classification model is generally trusted until the surrounding object enters the detection range of another of the ultrasonic sensor devices (and thus exhibits a higher novelty) and there produces a classification result whose overall quality for the surrounding object in question exceeds a predetermined threshold.
[0041] Depending on the determined or selected classification result, the operation of the mobile device can be controlled or performed. In particular, maneuvers can be planned or warnings can be signaled based on the classification result.
[0042] According to a further aspect, an apparatus for carrying out one of the above methods is provided.
[0043] Brief description of the drawings
[0044] Embodiments are explained in more detail below with reference to the attached drawings. They show:
[0045] Figure 1 is a schematic representation of a vehicle with a
[0046] Ultrasonic sensor system with two ultrasonic sensor devices and separate classification models; and
[0047] Figure 2 is a flow chart illustrating a method for operating the ultrasonic sensor system of Figure 1; and
[0048] Figure 3 is a flow chart illustrating another
[0049] Method for operating the ultrasonic sensor system of Figure 1.
[0050] Description of embodiments
[0051] Figure 1 shows a schematic representation of a vehicle 1 in a vehicle environment in which one or more environmental objects U are located. The vehicle 1, as an example of a mobile device, comprises an ultrasonic sensor system 2 with two ultrasonic sensor devices 3a, 3b, which are arranged on a front bumper 4a and a rear bumper 4b.
[0052] Each of the ultrasonic sensor devices 3a, 3b comprises a plurality of ultrasonic transducers 5 for emitting an ultrasonic signal with signal pulses and for receiving ultrasonic signals reflected from the surrounding objects U. The arrangement or orientation of the ultrasonic transducers 5 on the various ultrasonic sensor devices is generally different for the front and rear bumpers 4a, 4b of the motor vehicle 1.
[0053] A control unit 6 is provided for evaluating the sensor signals of the ultrasonic transducers 5 of the ultrasonic sensor devices 3a, 3b. A corresponding data-based classification model 61a, 61b is implemented in the control unit 6 for each of the ultrasonic sensor devices 3a, 3b, which is trained separately for the respective ultrasonic sensor device 3a, 3b.
[0054] The sensor signals are further evaluated in a conventional manner using ultrasound-based localization methods to create a virtual map of the environment in the control unit 6 and enter the positions of detected environmental objects U there. Classification results are assigned to the detected environmental objects U using the respective classification model 61a, 61b. The classification results classify the environmental objects U according to properties relevant to the vehicle's operation. The classification results can indicate, for example, that the relevant environmental object U is drivable and relevant to a collision.
[0055] The data-based classification model assigns a classification vector to an input variable vector, which includes signal features from the sensor signals of the ultrasonic transducers. The classification vector contains an element for each possible class of the classification result. Using an argmax function, the specific class can be output as the classification result for model evaluation. The value of the element determined by argmax corresponds to the classification confidence.
[0056] When performing certain driving maneuvers, a specific environmental object U can be detected successively by the two ultrasonic sensor devices 3a, 3b when it leaves the detection range of one of the ultrasonic sensor devices 3a, 3b and enters the detection range of the other ultrasonic sensor devices 3a, 3b. Since both ultrasonic sensor devices 3a, 3b each perform a classification using the associated classification models 61a, 61b, two classification results result for the specific environmental object U. This is not problematic as long as the classification results are identical. However, if the classification results are different, it is necessary to select the most accurate classification result.
[0057] In this regard, a method is provided in the control unit, which is described in more detail below using the flow chart in Figure 2.
[0058] In step S1, continuous detections are performed using both ultrasonic sensor devices 3a, 3b. For this purpose, detected environmental objects U are localized in a conventional manner based on the sensor signals to be evaluated and entered or tracked in an environmental map. In step S2, for each detection of an environmental object U by one of the ultrasonic sensor devices 3a, 3b, a classification is performed using the associated classification model 61a, 61b. The classification leads to corresponding classification results.
[0059] In step S3, a check is carried out to determine whether one or more surrounding objects have been detected in chronological order by both ultrasonic sensor devices 3a, 3b. This can occur particularly during corresponding driving maneuvers, such as turning maneuvers or the like. If it is determined that a specific surrounding object U has been detected by both ultrasonic sensor devices (alternative: yes), the method continues with step S4; otherwise (alternative: no), the method returns to step S1.
[0060] In step S4, it is checked whether the classification results of both classifications are identical with the classification models 61a, 61b. If this is the case (alternative: yes), the method continues with step S1. Otherwise (alternative: no), the method continues with step S5.
[0061] In step S5, the corresponding classification confidences are determined for the classifications. Classification using the data-based classification model is performed by calculating the argmax over the available classes, whereby the value of the element determined by argmax corresponds to or correlates with the classification confidence, i.e., the reliability of the classification result. The classification confidence can be provided as the classification quality.
[0062] In step S6, the novelty of the most recently determined classifications is evaluated as a further aspect. The novelty can be determined using a timestamp of the last detection of the relevant environmental object by the ultrasonic sensor devices 3a, 3b. The difference between the respective timestamp and the current time can thus be translated into a novelty quality according to a predefined assignment function. The novelty quality is higher the shorter the time period between the respective timestamp and the current time. An overall quality can now be determined for each of the classification models 3a, 3b in step S7 as a weighted average of the novelty quality and the classification quality for each of the classification models.
[0063] In particular, the overall quality sigma_ges of a classification process by the respective ultrasonic sensor device 3a, 3b can be determined as a weighted average using the classification quality sigma_klass of the respective classification model and a novelty quality sigma_neu resulting from a novelty degree of the last determined classification result for the respective environmental object U, where: sigma_ges= (sigma_klass*a1 +sigma_neu*a2) / 2 where a1 and a2 represent additional predeterminable weighting factors.
[0064] In step S7, the classification result determined by the classification model assigned the highest overall quality is assumed as the classification result. In this way, the resulting classification result is not always automatically adopted whenever a first detection of the relevant environmental object U is performed by one of the ultrasonic sensor devices 3a, 3b after it has already been detected by another of the ultrasonic sensor devices 3a, 3b. Instead, the classification quality and the novelty quality are also taken into account.
[0065] Depending on the assumed classification result, driving maneuvers can be planned or warnings can be signaled to the driver.
[0066] According to a further embodiment of the method, which is described in more detail in the flow chart of Figure 3, the degree of novelty of the data can be taken more into account.
[0067] In step S11, continuous detections are performed using both ultrasonic sensor devices 3a, 3b. For this purpose, detected environmental objects U are localized in a conventional manner based on the sensor signals to be evaluated and entered or tracked in an environmental map.
[0068] In step S12, for each detection of an environmental object U by one of the ultrasonic sensor devices 3a, 3b, a classification is performed using the associated classification model 61a, 61b. The classification leads to corresponding classification results.
[0069] In step S13, a check is carried out to determine whether one or more surrounding objects have been detected in chronological order by both ultrasonic sensor devices 3a, 3b. This can occur particularly during corresponding driving maneuvers, such as turning maneuvers or the like. If it is determined that a specific surrounding object U has been detected by both ultrasonic sensor devices (alternative: yes), the method continues with step S14; otherwise (alternative: no), the method returns to step S11.
[0070] In step S14, it is checked whether the classification results of both classifications are identical with the classification models 61a, 61b. If this is the case (alternative: yes), the method continues with step S11. Otherwise (alternative: no), the method continues with step S15.
[0071] In step S15, corresponding classification confidences are determined for the classifications. Classification using the data-based classification model is performed by calculating the argmax over the available classes, whereby the value of the element determined by argmax corresponds to or correlates with the classification confidence, i.e., the reliability of the classification result. The classification confidence can be provided as the classification quality.
[0072] Subsequently, in step S16, it is checked whether the classification quality of the classification model whose detection has a higher degree of novelty exceeds a predefined quality threshold. If this is the case (alternative: yes), the method continues with step S17 and the classification result of the respective classification model is signaled. Otherwise (alternative: no), the method continues with step S11. The classification confidence can also be determined based on one or more individual qualities. The individual qualities can include a detection quality, a feature quality, and / or a model quality.
[0073] For detection quality, a quotient of the number of correct classifications and the total number of detections is calculated using predefined validation data sets. This quotient can be assigned to a corresponding detection quality value using a mapping function, such as a lookup table.
[0074] The number of correct classifications is determined by evaluating a predefined database of labeled data. These form validation data sets that assign sensor signals and / or sensor characteristics generated from them of known environmental objects to a classification (object property: traversable, collision-relevant). These validation data sets are then used to determine the number of object detections in which the classification result of the actual object property is recognized, and a corresponding probability is determined.
[0075] Furthermore, feature quality can result from the availability of the signal features used. This assesses the availability of the various signal features that contribute to determining the classification result for the classification of a specific environmental object.
[0076] Some signal characteristics cannot be calculated in certain situations or do not provide any added value. For example, a quotient between the distance traveled and the measured distance to the object cannot be evaluated if the vehicle has not moved far enough. The evaluation of reflections from the detection channel at the point of contact between the object and the ground cannot be evaluated at close range, as reflections from near the ground cannot be detected due to the sensor's aperture angle. Whether a signal characteristic can be evaluated or not depends on geometric or temporal conditions. If a signal characteristic cannot be evaluated, it is marked with an invalid value and ignored by the classification.Since the quality of signal features often depends on the detection distance and angle and / or the nature of surfaces (reflectivity), the signal features to be evaluated, which form the basis for classification by the respective classification model, can be used to assess the quality of the classification result. For example, the signal feature of amplitude provides very good differentiation in the near range, i.e. the surrounding object in question is very close to the ultrasonic transducer arrangement, while in the far range it has only a low value due to similar vertical detection angles. Furthermore, high and low amplitude values can be a good identifier for the object height, whereas values in the middle range are less meaningful. Thus, the feature quality can be specified depending on the relative position of the surrounding object to the respective ultrasonic sensor device.This can be done using a predefined feature assignment function, for example, in the form of a lookup table, so that a specific individual feature quality can be assigned to each signal feature and a detection feature associated with the signal feature. The detection feature associated with the signal feature can, for example, result from a spatial relationship between the relevant surrounding object and the ultrasonic transducer involved in determining the signal feature.
[0077] To determine the overall feature quality, each individual signal feature to be evaluated and the detection conditions, such as the distance to the object to be classified and the detection angle, are considered, and a corresponding individual feature quality is determined using a suitable assignment function. Furthermore, the individual feature qualities for the individual signal features are aggregated into an overall feature quality, in particular by averaging, particularly with a weighted consideration of the individual feature qualities with predefined weightings.
[0078] Furthermore, the model quality of the classification model used can be considered. This takes into account the general suitability of the ultrasonic sensor devices for classifying environmental objects, so that different model qualities can be assumed based on the respective sensor configuration, which results from the installation situation of the ultrasonic transducers of the respective ultrasonic sensor device. The model quality is independent of the environmental situation and only determines the quality of a classification based on the configuration (arrangement and geometric orientation) of the ultrasonic transducers of the respective ultrasonic sensor device.
[0079] An overall quality can now be determined by taking a weighted average of the detection accuracy, the feature quality, and the model quality. Alternatively, the overall quality can be determined by a weighted average of the detection accuracy, the feature quality, the model quality, and the novelty quality.
Claims
Claims 1. Method for fusing classification results of several classification models (61a, 61b) for the classification of environmental objects (U) using ultrasonic sensor devices (3a, 3b) in a mobile device, comprising the following steps: Acquisition (S1 , S11) of sensor signals from ultrasonic transducers (5) of the ultrasonic sensor devices (3a, 3b); Tracking the position of detected environmental objects (U) using the ultrasonic sensor devices (3a, 3b) and classifying the object type of the detected environmental objects (U) using a classification model (61a, 61b) assigned to each ultrasonic sensor device (3a, 3b) in order to to obtain the classification result; Upon detection that a specific environmental object (U) has entered a detection range of one of the ultrasonic sensor devices (3a, 3b) and was previously located in a detection range of another of the ultrasonic sensor devices (3a, 3b) (S3, S4), selections (S7) of one of the classification results for the specific environmental object (U) depending on a classification quality (sigma_klass) based on the classification model (61a, 61b) and on a novelty level of the classification result.
2. Method according to claim 1, wherein the classification result of the classification model (61a, 61b) whose degree of novelty indicates a more recent classification is selected if the corresponding overall quality exceeds a predetermined threshold.
3. Method according to claim 1 or 2, wherein the degree of novelty is assigned by means of an assignment function of a novelty quality (sigma_new), which indicates a higher quality the more recent the acquisition of the underlying sensor data, wherein an overall quality (sigma_total) is derived from the The classification accuracy (sigma_class) and the novelty accuracy (sigma_new) are determined, wherein the classification result is selected depending on the overall accuracy (sigma_total) of the classification results of the classification models (61a, 61b). A method according to claim 3, wherein the overall accuracy (sigma_total) is determined by weighted averaging the classification accuracy (sigma_class) and the novelty accuracy (sigma_new). A method according to any one of claims 1 to 4, wherein the classification models are data-based, and wherein the classification accuracy (sigma_class) is determined as the value of an element of an output vector of the classification models (61a, 61b) with the highest value.A method according to any one of claims 1 to 2, wherein the degree of novelty is assigned using an assignment function to a novelty quality (sigma_new), which indicates a higher quality the more recently the underlying sensor data has been acquired, wherein an overall quality (sigma_total) is determined from individual qualities and the novelty quality (sigma_new), and wherein the classification result is selected depending on the overall qualities (sigma_total) of the classification results of the classification models (61a, 61b). A method according to claim 6, wherein the classification qualities (sigma_class) are determined by weighted averaging of the individual qualities, wherein the individual qualities comprise at least one of the following: a detection quality, a feature quality (sigma_mark), and a model quality (sigma_mod).A method according to claim 7, wherein the detection accuracy (sigma_det) is determined as a function of the quotient of the number of correct classifications of the specified environmental object (U) and the total number of classifications of the specified environmental object (U). A method according to claim 7 or 8, wherein the feature accuracy (sigma_merk) is determined as a function of the contributions of signal features resulting from the sensor signals to the classification result. Classification of a specific environmental object (U) and a detection feature, in particular a relative position of the specific environmental object to the ultrasonic transducer (5) associated with the signal feature, is determined. Method according to any one of claims 7 to 9, wherein the model quality (sigma_mod) is determined by the selection of the classification model (61a, 61b). Method according to any one of claims 6 to 10, wherein the overall quality (sigma_ges) is determined by a weighted average of the individual quality values and, optionally, the novelty quality. Device for carrying out any one of the methods according to any one of claims 1 to 11. Computer program product comprising instructions which, when the program is executed by a computer, cause the computer to execute the steps of the method according to any one of claims 1 to 11.Machine-readable storage medium comprising instructions which, when executed by a computer, cause the computer to perform the steps of the method according to any one of claims 1 to 11.