Method and device for processing ultrasonic signal

Abstract

The present application relates to a method for processing an ultrasonic signal. The method comprises: a vehicle making multiple trips in a first scenario, obtaining a first ultrasonic sample set through an ultrasonic sensor with a first installation arrangement; a vehicle making multiple trips in the first scenario, obtaining a second ultrasonic sample set through an ultrasonic sensor with a second installation arrangement, wherein each ultrasonic sample in the first ultrasonic sample set and the second ultrasonic sample set comprises multiple ultrasonic signal features, and each ultrasonic signal feature has multiple feature representations; calculating the distribution of each feature representation of each ultrasonic signal feature in the first scenario based on the first ultrasonic sample set and the second ultrasonic sample set respectively; and determining an ultrasonic signal feature set for characterization based at least on the similarity of the distribution of each feature representation of each ultrasonic signal feature in the first scenario between the first ultrasonic sample set and the second ultrasonic sample set.

Description

[0001] METHOD AND DEVICE FOR PROCESSING ULTRASONIC SIGNAL

[0002] Technical Field

[0003] The present application relates to ultrasonic signal processing, and more particularly, to a method, a system, a computer readable medium, and a computer program product for processing an ultrasonic signal to determine an object in the surroundings of a vehicle.

[0004] Background

[0005] Accurately identifying objects in a vehicle’s surroundings is crucial for implementing various vehicle auxiliary functions in applications such as autonomous driving systems, automatic parking systems, and driver assistance systems. For instance, sensor signals can be used to detect objects in a vehicle’s surroundings, enabling the vehicle to identify obstacles in real-time and adjust its driving route accordingly. Similarly, in a parking environment, sensor signals can detect obstacles and accurately determine available parking spaces.

[0006] Ultrasonic sensors (USSs) offer a cost-effective solution for object detection. However, the low resolution of data captured by ultrasonic radar prevents conventional detection methods using ultrasonic signals from being effectively applied in complex scenarios. Additionally, the variety of objects present in different vehicle application scenarios necessitates improvements in the accuracy of object detection using ultrasonic signals. Enhancing this accuracy is crucial for optimizing the performance of assisted driving and parking.

[0007] Summary of the Invention

[0008] The following introduction is provided in order to introduce selected concepts in a simple manner, and these concepts will be further described in the detailed description below. The introduction is not intended to highlight the key or necessary features of the claimed subject matter, nor is it intended to limit the scope of the claimed subject matter.

[0009] Generally, to detect objects in the surroundings of a vehicle, features can be extracted from ultrasonic data obtained by ultrasonic sensors, which serves as input to a machine learning model to predict the position and type of the object, among other details. The machine learning model can be trained in a supervised manner based on ultrasonic data collected in various scenarios during the training stage. However, ultrasonic sensors may have different installation arrangements during the training stage and the inference stage, such as being installed in various locations on the same vehicle model or on different vehicle models. Due to the sparse characteristics of ultrasonic signal itself, different installation arrangements may lead to significant variations in the feature representations extracted based on the ultrasonic data. This can cause the performance of the machine learning model, which performs well during the training stage, to drop significantly in the inference stage, ultimately affecting the accuracy of subsequent object detection tasks.

[0010] Therefore, it is desirable to provide a method for processing ultrasonic signals. The method can identify ultrasonic signal features and corresponding feature representations insensitive to changes in installation arrangements, allowing the characterization of ultrasonic data based on these identified ultrasonic signal features as input to the machine learning model for object prediction. Processing ultrasonic signals using the method disclosed in the present invention can extract more universal feature representations, thereby enhancing the generalization ability of the machine learning model. This improvement leads to greater accuracy and robustness across various vehicle models and driving scenarios, ultimately enhancing the performance of vehicle- assisted driving and parking.

[0011] In one aspect, examples of the present disclosure provides a method for processing an ultrasonic signal. The method comprises: a vehicle making multiple trips in a first scenario, obtaining a first ultrasonic sample set through an ultrasonic sensor with a first installation arrangement; a vehicle making multiple trips in the first scenario, obtaining a second ultrasonic sample set through an ultrasonic sensor with a second installation arrangement, wherein each ultrasonic sample in the first ultrasonic sample set and the second ultrasonic sample set comprises multiple ultrasonic signal features, and each ultrasonic signal feature has multiple feature representations; calculating the distribution of each feature representation of each ultrasonic signal feature in the first scenario based on the first ultrasonic sample set and the second ultrasonic sample set respectively; and determining an ultrasonic signal feature set for characterization based at least on the similarity of the distribution of each feature representation of each ultrasonic signal feature in the first scenario between the first ultrasonic sample set and the second ultrasonic sample set.

[0012] In one aspect, examples of the present disclosure provide a method for training a machine learning model to determine an object in the surroundings of a vehicle based on an ultrasonic signal. The method comprises: selecting ultrasonic signal features and corresponding feature representations from the ultrasonic signal feature set for characterization determined by the method described in any example of the present disclosure; inputting a training ultrasonic sample set processed based on the selected ultrasonic signal features and corresponding feature representations into a machine learning model set to output prediction information associated with the object in the surroundings of the vehicle; and training the machine learning model set based on the prediction information and corresponding labels associated with the object in the surroundings of the vehicle.

[0013] In one aspect, examples of the present disclosure provide a method for determining an object in the surroundings of a vehicle based on an ultrasonic signal. The method comprises: obtaining an ultrasonic sample set by an ultrasonic sensor during movement of a first vehicle, wherein each ultrasonic sample in the ultrasonic sample set comprises multiple ultrasonic signal features, and each ultrasonic signal feature has multiple feature representations; performing clustering based on the ultrasonic data set to obtain one or more sample clusters, wherein each sample cluster corresponds to an object in the surroundings of the first vehicle; processing the ultrasonic samples corresponding to each sample cluster based on the ultrasonic signal features and the corresponding feature representations to input a trained machine learning model set, wherein the ultrasonic signal features and the corresponding feature representations are selected from the ultrasonic signal feature set for characterization determined by the method described in any example of the present disclosure; and outputting prediction information associated with the object in the surroundings of the first vehicle by the machine learning model set.

[0014] In another aspect, examples of the present disclosure provide a system, comprising: at least one processor; a memory coupled to the at least one processor, the memory storing executable instructions, wherein the executable instructions, when executed by the at least one processor, allow the at least one processor to implement the method according to any example of the present disclosure.

[0015] In another aspect, examples of the present disclosure provide a computer-readable medium which stores a computer program comprising instructions that, when executed by a processor, cause one or more units to implement the method according to any example of the present disclosure.

[0016] In another aspect, examples of the present disclosure provide a computer program product, comprising a computer program, wherein the computer program, when executed by the processor, implement the method according to any example of the present disclosure.

[0017] In another aspect, examples of the present disclosure provides a vehicle comprising an ultrasonic sensor for transmitting and receiving an ultrasonic signal; and one or more units for implementing the method according to any example of the present disclosure.

[0018] Brief Description of the Drawings

[0019] The nature and advantages of the present disclosure may be further implemented by referring to the following accompanying drawings. In the drawings, similar components or features may have the same reference signs.

[0020] FIG. 1 is a schematic diagram of an exemplary vehicle according to examples of the present disclosure.

[0021] FIG. 2 is a schematic diagram of an exemplary control system in a vehicle according to examples of the present disclosure.

[0022] FIG. 3 is a schematic diagram of a module for determining an object in the surroundings of a vehicle based on an ultrasonic signal according to examples of the present disclosure.

[0023] FIG. 4 is a schematic diagram of a scenario for processing an ultrasonic signal according to examples of the present disclosure.

[0024] FIG. 5 is a schematic diagram of a process for processing an ultrasonic signal according to examples of the present disclosure.

[0025] FIG. 6 is a flow chart of a method for processing an ultrasonic signal according to examples of the present disclosure.

[0026] FIG. 7 is a flow chart of a method for training a machine learning model to determine an object in the surroundings of a vehicle based on an ultrasonic signal according to examples of the present disclosure.

[0027] FIG. 8 is a flow chart of a method for determining an object in the surroundings of a vehicle based on an ultrasonic signal according to examples of the present disclosure.

[0028] FIG. 9 is a block diagram of a system according to examples of the present disclosure.

[0029] Detailed Description of the Embodiments

[0030] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that discussions about these embodiments are provided to aid those skilled in the art in better understanding and thereby implementing the subject matter described herein rather than limiting the scope of protection, applicability, or examples described in the Claims. Changes may be made to the functions and arrangements of the elements discussed without departing from the scope of protection of the content of the present disclosure. Various processes or components may be omitted, substituted, or added in the various examples as needed. For example, the described method may be performed in a different order than that described, and various steps may be added, omitted, or combined. In addition, features described in relation to some examples may also be combined in other examples.

[0031] As used herein, the term “comprising” and its variations are open terms, which mean “including but not limited to”. The term “based on” indicates “at least partially based on”. The terms “one example” and “an example” indicate “at least one example”. The term “another example” indicates “at least one other example”. The terms “first”, “second”, etc. may refer to different or same objects. Unless explicitly stated in the context, the definition of one term is consistent throughout the description.

[0032] Ultrasonic sensors have garnered significant attention and research in the automotive field due to their advantages of low cost, rapid response, and ease of integration. However, effectively utilizing ultrasonic sensors for object detection in the automotive field remains one of the current challenges. To address this issue, the examples of the present disclosure offer a technical solution for processing an ultrasonic signal to detect an object in the surroundings of a vehicle. A detailed description is given below with reference to the specific examples.

[0033] FIG. 1 is a schematic diagram of an exemplary vehicle according to examples of the present disclosure. It should be understood that the following examples are provided solely to enhance understanding of the present disclosure and do not impose any limitations on the scope of the present disclosure.

[0034] In the example of FIG. 1 , at least one ultrasonic sensor 110 (simply represented as a black dot in FIG. 1) may be installed on a vehicle 100. For example, as shown in FIG. 1 , the ultrasonic sensor 110 may be installed on the front side, rear side, left side, and right side of the vehicle 100, that is, corresponding to the front, rear, left side, and right side of the vehicle 100, respectively. FIG. 1 shows 16 ultrasonic sensors. However, the number and installation arrangements of the ultrasonic sensors 110 are not limited to those shown in FIG. 1. In various implementations, the vehicle 100 may be equipped with either more or fewer ultrasonic sensors, and the installation arrangements of the ultrasonic sensors may vary as well. Moreover, the installation arrangements of ultrasonic sensors may also be different on vehicles of different models.

[0035] The ultrasonic sensor 110 may transmit an ultrasonic signal toward the surroundings of the vehicle 100. For example, two ultrasonic sensors 110 located on the right front side of the vehicle 100 may transmit an ultrasonic signal 120. The ultrasonic signal 120 will be described below as an example. The ultrasonic signal 120 may encounter various objects during transmission, such as obstacles (e.g., other vehicles, ground locks, trees, railings, fences, etc.), objects related to the driving operation of the vehicle (e.g., curbs, etc.), pedestrians, etc. When encountering an object, the ultrasonic signal 120 may be reflected. After the ultrasonic signal 120 is reflected, the ultrasonic sensor 110 may receive the reflected signal (also commonly referred to as an echo signal).

[0036] Various ultrasonic data may be obtained based on the transmitted ultrasonic signal and the received echo signal. These operations may be implemented in various ways. For example, in some implementations, the vehicle 100 may comprise an ultrasonic sensor module. The ultrasonic sensor module may comprise an ultrasonic sensor 110 and a processing unit. The ultrasonic sensor 110 may comprise a transmitter that transmits an ultrasonic signal and a receiver that receives an echo signal. The processing unit may obtain various ultrasonic data based on the transmitted ultrasonic signal and the received echo signal. In addition, the processing unit may also control the transmission and reception of signals by each ultrasonic sensor.

[0037] For example, in FIG. 1 , two ultrasonic sensors 110 located on the right front side of the vehicle 100 may constitute an ultrasonic sensor module 105, which comprises two ultrasonic sensors for transmitting and receiving ultrasonic signals and a processing unit (not explicitly shown in FIG. 1 ). The processing unit may obtain various ultrasonic data based on the ultrasonic signals transmitted by the two ultrasonic sensors and the echo signals received. Similarly, the two ultrasonic sensors 100 located on the right rear side, the two ultrasonic sensors 110 located on the left front side, the two ultrasonic sensors 110 located on the left rear side, the four ultrasonic sensors 110 located on the front side, and the four ultrasonic sensors 110 located on the rear side of the vehicle 110 may all constitute an ultrasonic sensor module with the corresponding processing units.

[0038] Or, for another example, a single ultrasonic sensor 110 of the vehicle 100 may constitute an ultrasonic sensor module with a corresponding processing unit, and the processing unit may obtain various ultrasonic data based on the ultrasonic signal transmitted by the single sensor and the echo signal received. For another example, in FIG. 1 , the four ultrasonic sensors 110 located on the right side of the vehicle 100 and the corresponding processing units may constitute an ultrasonic sensor module, and the processing units may obtain various ultrasonic data based on the ultrasonic signals transmitted by the four sensors and the echo signals received. Similarly, the four ultrasonic sensors 110 located on the left side of the vehicle 100 may each constitute an ultrasonic sensor module with the corresponding processing unit. For another example, all the ultrasonic sensors 110 shown in FIG. 1 may constitute an ultrasonic sensor module with corresponding processing units, and the processing units may obtain various ultrasonic data based on the ultrasonic signals transmitted by all these ultrasonic sensors and the echo signals received.

[0039] Although the ultrasonic sensor and the ultrasonic sensor module are described separately above, according to the context, the ultrasonic sensor may narrowly represent a sensing unit for transmitting an ultrasonic signal and receiving an echo signal, or may broadly represent an ultrasonic sensor module. Those skilled in the art are able to distinguish the meaning of the ultrasonic sensor in a specific context.

[0040] As described above, the processing unit may obtain various ultrasonic data based on the ultrasonic signal transmitted by the ultrasonic sensor 110 and the echo signal received, and the ultrasonic data may comprise various related data.

[0041] In some examples, the ultrasonic data may comprise echo data. The echo data may comprise information related to the echo signal. For example, the echo data may comprise an echo timestamp, an echo amplitude, an echo significance, an echo distance, an echo height, echo point coordinates (for example, two plane axis coordinates may be comprised), sensor coordinates, an echo point deflection (for example, the angle with the vehicle’s forward direction), etc. of the echo signal. In some examples, the echo data may be obtained using the center line method. For example, while the vehicle 100 is traveling along the direction 140, the ultrasonic sensor 110 located on the right front side of the vehicle 100 may transmit an ultrasonic signal and receive an echo signal. In the center line method, it is assumed that the reflection point on the detected object is on the center line 150 of the ultrasonic arc. Based on this, the processing unit may detect the reflection point of the corresponding ultrasonic signal and use it as the position of the echo signal. This position may be denoted, for example, by the echo coordinates.

[0042] In some examples, the ultrasonic data may comprise echo intersection data. The echo intersection data may comprise information related to intersections between different echo signals. For example, the echo intersection data may comprise echo intersection coordinates (for example, two plane axis coordinates may be comprised), an echo intersection distance, an echo intersection deflection, an echo intersection height, adjacent echo intersection coordinates, an adjacent echo intersection height, an adjacent echo intersection distance, an adjacent echo intersection deflection (for example, the angle with the vehicle’s forward direction), sensor coordinates, etc. For example, while the vehicle 100 is traveling along the direction 140, the ultrasonic sensor 110 located on the right front side of the vehicle 100 may transmit an ultrasonic signal and receive an echo signal. The two sensors 110 may transmit two ultrasonic signals 120 and receive corresponding echo signals. According to the known positional relationship of the two ultrasonic sensors 110, the transmission time of the two ultrasonic signals 120, the reception time of the corresponding echo signals, and other information, the processing unit may calculate an echo intersection 160 of the two ultrasonic signals 120, and the echo intersection 160 may represent the reflection point on the detected object. Accordingly, the echo intersection data may comprise information about the intersection 160. In some examples, the echo intersection is not limited to being the intersection of the two echoes of ultrasonic signals transmitted by the two different ultrasonic sensors. For example, the echo intersection may also be the intersection of the two echoes of ultrasonic signals transmitted by the same ultrasonic sensor at different times during movement. Moreover, the echo intersection data may be calculated based on the echo data. For example, based on the ultrasonic arcs of any two echo data within a certain range, the corresponding echo intersection data may be obtained.

[0043] In some examples, the ultrasonic data may comprise the above-mentioned echo data and / or echo intersection data. Of course, the ultrasonic data may optionally comprise other data related to the ultrasonic signal, such as data related to the ultrasonic signal obtained based on methods known in the art or possible methods in the future.

[0044] In some examples, the processing unit of the ultrasonic sensor module may provide the obtained ultrasonic data to a control unit 130 of the vehicle 100. For example, the control unit 130 may be an Electronic Control Unit (ECU) of the vehicle. In some examples, part or all of the operations performed by the processing unit of the ultrasonic sensor module may also be performed by the control unit 130. For example, the control unit 130 may obtain echo data and / or echo intersection data based on the ultrasonic signal transmitted by the ultrasonic sensor 110 and the echo signal received. In some examples, the vehicle 100 may also comprise other processing units to perform such operations.

[0045] In some implementations, the sensors, processing units, and control units mentioned above may be comprised in the control system of the vehicle. The control system of the vehicle may perform various controls on the vehicle. For ease of understanding, FIG. 2 is a schematic diagram of an exemplary control system in a vehicle according to examples of the present disclosure.

[0046] In the example of FIG. 2, the same reference signs are used for the same components as those in FIG. 1. Moreover, it should be understood that FIG. 2 only shows some components related to the technical solution of the present disclosure. In actual implementation, the control system of the vehicle may also comprise various other components, which are not limited by the present disclosure.

[0047] In the example of FIG. 2, the control system 200 of the vehicle may comprise ultrasonic sensor modules 105-1 to 105-N. The ultrasonic sensor modules 105-1 to 105-N comprise one or more ultrasonic sensors 110 and corresponding processing units 115-1 to 115-N, respectively. As described above in conjunction with FIG. 1 , in some implementations, the vehicle may comprise only one ultrasonic sensor module. For example, the ultrasonic sensor module comprises multiple ultrasonic sensors and processing units installed on the vehicle 100; in different implementations, the vehicle may comprise one or more ultrasonic sensor modules, each of which may comprise one or more ultrasonic sensors and corresponding processing units.

[0048] The control system 200 may further comprise a control unit 130. The control unit 130 may control the operation of any one of the ultrasonic sensor modules 105-1 to 105-N. For example, the control unit 130 may control the ultrasonic sensor 110 in any one of the ultrasonic sensor modules 105-1 to 105-N to transmit ultrasonic signal and receive echo signal; the control unit 130 may receive ultrasonic data from any one of the ultrasonic sensor modules 105-1 to 105-N, and perform further operations or control the vehicle based on the ultrasonic data, etc. As described above, some or all of the operations performed by the processing units in the ultrasonic sensor modules 105-1 to 105-N may also be performed by the control unit 130 or other processing units.

[0049] The control system 200 may further comprise a human-machine interface 180. The control unit 130 may output information that may be understood by a user (e.g., a driver) via the humanmachine interface 180, and may receive information input by the user from the human-machine interface. In some examples, the user may input a selection regarding entering the assisted parking mode or an autonomous driving mode via the human-machine interface 180. The control unit 130, upon receiving user input to enter the assisted parking mode or the autonomous driving mode, may control the vehicle to operate in the assisted parking mode or the autonomous driving mode. For example, in the assisted parking mode, the vehicle is controlled to automatically find a parking space and automatically park. For another example, in the autonomous driving mode, the vehicle is controlled to plan a route and avoid obstacles. In some examples, the control unit 130 may automatically control the vehicle to enter the assisted / automatic parking mode or the assisted / autonomous driving mode without user input. Regardless of which mode is entered by which method, the control unit 130 may detect an object in the surroundings of the vehicle by processing ultrasonic data, such as determining object information such as the position, type, and height of the object. Based on the object information, the control unit 130 may perform various operations related to vehicle parking or path planning, such as detecting available parking spaces for the vehicle, planning a parking route for the vehicle based on the detected parking spaces, planning a driving route for the vehicle based on the detected obstacles, and the like.

[0050] Typically, to detect an object in the surroundings of a vehicle, features may be extracted based on ultrasonic data to serve as input to a machine learning model to predict relevant information of the object, such as position, type, height, etc. FIG. 3 is a schematic diagram of a module for determining an object in the surroundings of a vehicle based on an ultrasonic signal according to examples of the present disclosure.

[0051] 310 in FIG. 3 is a schematic diagram of an ultrasonic sample set obtained based on ultrasonic data according to examples of the present disclosure. While the vehicle is traveling, an ultrasonic sensor in an ultrasonic sensor module, such as the ultrasonic sensor 110 in the ultrasonic sensor module described with reference to FIG. 1 , may transmit an ultrasonic signal to collect ultrasonic data and obtain an ultrasonic sample based on the ultrasonic data. The ultrasonic sample may comprise an echo sample and an echo intersection sample. An ultrasonic sample in 310 of FIG. 3 may represent an echo sample point or an echo intersection sample point, and a circular sample represents an echo sample point, and a triangular sample represents an echo intersection sample point.

[0052] An ultrasonic sample set may comprise multiple ultrasonic samples. For example, the ultrasonic sample set may correspond to all ultrasonic samples obtained by using the ultrasonic sensor during a single movement of the vehicle for a period of time. For another example, the ultrasonic sample set may correspond to all ultrasonic samples obtained by using the ultrasonic sensor during multiple movements of the vehicle for a period of time. For example, the ultrasonic sample set may correspond to all ultrasonic samples obtained by the ultrasonic sensor during a single movement of the vehicle for a distance. For another example, the ultrasonic sample set may correspond to all ultrasonic samples obtained by the ultrasonic sensor during multiple movements of the vehicle for a distance. What 310 in FIG. 3 shows may be an ultrasonic sample set obtained during a single movement of the vehicle. In one example, the ultrasonic sample set may be obtained by fusing data captured by the ultrasonic sensor during multiple sliding windows based on a sliding window mechanism. The size of the sliding window may be fixed or variable, and the size of the sliding window may correspond to the actual spatial range covered by the ultrasonic signal 120. Those skilled in the art will appreciate that FIG. 3 is merely a schematic diagram of an ultrasonic sample set for illustrative purposes. In various practical applications, the number of samples in the ultrasonic sample set within a specific spatial range may be greater and the distribution may be more complex.

[0053] Each ultrasonic sample may comprise a variety of ultrasonic signal features. For example, an echo sample may comprise echo point coordinates, sensor coordinates, an echo point deflection, an echo amplitude, an echo significance, an echo distance, an echo height, echo time, etc. For example, the echo intersection sample may comprise echo intersection coordinates, sensor coordinates, an echo intersection deflection, an echo intersection amplitude, an echo intersection significance, an echo intersection distance, an echo intersection height, adjacent echo intersection coordinates, an adjacent echo intersection deflection, an adjacent echo intersection amplitude, an adjacent echo intersection significance, an adjacent echo intersection distance, an adjacent echo intersection height, etc., wherein features such as echo intersection height, echo intersection amplitude and echo intersection significance may be described by corresponding features of two or more echoes that generate the echo intersection. Each ultrasonic signal feature may be multi-dimensional. For example, the echo point coordinates may be represented by two coordinate values, and the echo distance may be represented by two values, namely, a lateral distance and a longitudinal distance from a sensor. Therefore, each ultrasonic sample may be mathematically represented as a high-dimensional array, for example, echo sample 1 : [echo point 1st coordinate, echo point 2nd coordinate, echo significance 1 st feature, ..., echo significance Nth feature, echo distance 1st feature, ..., echo distance Nth feature, ...], echo intersection sample 1 : [echo intersection 1 st coordinate, echo intersection 2nd coordinate, echo intersection height 1st feature, ..., echo significance Nth feature, ...]. Those skilled in the art will appreciate that, in addition to the features of the echo sample and echo intersection sample exemplified above, the echo sample and echo intersection sample may also comprise other features related to the ultrasonic signal, and the features of the echo sample and echo intersection sample known in the art and that may be adopted in the future may all be applied to the technical solutions disclosed herein.

[0054] In one example, the ultrasonic data collected by the ultrasonic sensor may be cached in a cache or memory of the vehicle’s processing system, and the vehicle’s control unit 130 or other processing units may subsequently process the cached ultrasonic data to obtain ultrasonic samples. In one example, the ultrasonic samples may be cached in a cache or memory of the vehicle’s processing system, and the vehicle’s control unit 130 or other processing unit may subsequently process the cached ultrasonic samples to determine the position and / or type of the object. In one example, the control unit 130 or other processing unit of the vehicle may plan a vehicle driving route based on the determined position and / or type of the object, such as assisting the vehicle in parking or driving.

[0055] Furthermore, to determine the object in the surroundings of the vehicle based on the ultrasonic signal, multiple exemplary steps such as preprocessing, clustering, feature extraction, and object detection may be performed. As shown in FIG. 3, a preprocessing module 320, a clustering module 330, a feature extraction module 340 and an object detection module 350 may be comprised, wherein the dashed block represents an optional operation.

[0056] In the example shown in FIG. 3, for example, the ultrasonic sample set 310 shown in FIG. 3 may be used as an input of a preprocessing module 320, and the preprocessing module 320 performs preprocessing based on the ultrasonic sample set 310.

[0057] For example, preprocessing may optionally comprise up / down-sampling. In one example, a down-sampling algorithm may be used to process the ultrasonic sample set. To address the problem of disequilibrium in the classification of ultrasonic echo raw data, part of data can be selected from a majority set to be recombined with a minority set into a new data set, and such a manner is known as down-sampling. For example, commonly used down-sampling algorithms include, but are not limited to, random down-sampling, EasyEnsemble, BalinceCascade, or NearMiss, etc. In another example, an up-sampling algorithm may be used to process the ultrasonic sample set. To address the problem of disequilibrium in the classification of ultrasonic echo raw data, data from a minority set can be expanded to the same number as data from a majority set, and such a manner is known as up-sampling. For example, commonly used up- sampling algorithms include, but are not limited to, interpolation, transposation convolution, up- pooling, etc.

[0058] For example, the preprocessing may optionally comprise gridding. In one example, the ultrasonic sample set may be gridded based at least on the position-related information in the features of the ultrasonic samples. The size of the grid may be fixed, or may be flexibly adjusted based on a predetermined threshold condition. The gridding processing of the data helps integrate the high-dimensional feature data for easy input into the machine learning model for processing.

[0059] For example, the preprocessing may optionally comprise smoothing. In one example, the gridded ultrasonic sample set may be smoothed to remove redundant information and further integrate feature data, thereby reducing the number of parameters. For example, commonly used smoothing algorithms include, but are not limited to, maximum pooling, minimum pooling, average pooling, overlapping pooling, pyramid pooling, bilinear pooling, etc.

[0060] For example, the preprocessing may optionally comprise normalization. In one example, the gridded ultrasonic sample set may be normalized to avoid order of magnitude differences between several multi-dimensional features at the sample points. For example, commonly used normalization algorithms include but are not limited to min-max normalization, z-normalization, nonlinear normalization, etc.

[0061] Those skilled in the art will appreciate that the above-mentioned preprocessing operations are only used as examples, and the preprocessing module 320 may also comprise other preprocessing operations known in the art or that may be adopted in the future.

[0062] In the example shown in FIG. 3, the ultrasonic sample set that has been optionally preprocessed may be used as an input to the clustering module 330, and the clustering module 330 performs clustering based on the ultrasonic samples to form one or more sample clusters, wherein each sample cluster corresponds to an object in the surroundings of the vehicle.

[0063] For example, Density-Based Spatial Clustering of Applications with Noise (DBSCAN) may be used to cluster at least part of the ultrasonic samples in the ultrasonic sample set, such as one or more of echo samples or echo intersection samples, to form one or more sample clusters.

[0064] In the example shown in FIG. 3, the clustered ultrasonic sample set may be used as an input to the feature extraction module 340, and the feature extraction module 340 extracts feature representations of ultrasonic signal features based on the clustered ultrasonic samples.

[0065] For example, feature representations may be extracted for ultrasonic signal features based on a single sample cluster. For example, the feature representation may comprise statistics of ultrasonic signal features of multiple ultrasonic samples, including but not limited to minimum value, maximum value, mean value, variance, median, quantile, kurtosis, etc. For another example, the feature representation may comprise transformation or approximation of ultrasonic signal features of multiple ultrasonic samples, including but not limited to Fourier transform, piecewise approximation, autocorrelation, cepstrum analysis, etc.

[0066] In the example shown in FIG. 3, the extracted feature representations may be used as an input to the object detection module 350 which determines the object in the surroundings of the vehicle based on the feature representations to output prediction information 360 associated with the object in the surroundings of the vehicle, such as position, height, type, etc.

[0067] For example, the object detection module 350 may be implemented by a trained machine learning model. In one example, the machine learning model may be implemented based on a classification algorithm to achieve type recognition of an object in the surroundings of the vehicle. For example, the machine learning model may use an Extended Gradient Boost (XG Boost) model to determine the object type corresponding to each sample cluster based on a classification algorithm. For another example, the machine learning model may be implemented based on a regression algorithm to achieve position and / or height recognition of an object in the surroundings of the vehicle. For example, the machine learning model may use an Extended Gradient Boost (XG Boost) model to determine the position and / or height of the object corresponding to each sample cluster based on a regression algorithm. Additionally or alternatively, the object detection module 350 may comprise a machine learning model set, the machine learning model set comprising multiple machine learning models, wherein each machine learning model is for a specific object type. For example, an applicable machine learning model may be selected from the machine learning model set based on the object type.

[0068] Those skilled in the art will appreciate that the above-mentioned machine learning models and algorithms are only used as examples, and the object detection module 350 may also comprise other machine learning models known in the art and that may be adopted in the future, such as Convolutional Neural Networks (CNNs), transformer models, etc.

[0069] Although an object in the surroundings of the vehicle may be determined based on the ultrasonic signal after the processing described above, in actual applications, extracting all feature representations of all ultrasonic signal features as inputs to the object detection module in the feature extraction stage consumes a lot of computing and storage resources. Further, the object detection module is usually implemented based on a machine learning model, which further involves computing and storage resource consumption in both the training and inference stages. Therefore, compared with all feature representations of all ultrasonic signal features, selecting partial feature representations of some ultrasonic signal features for feature extraction is more conducive to improving computing efficiency and saving storage resources.

[0070] Accordingly, it is necessary to select partial feature representations of some ultrasonic signal features for feature extraction. However, since ultrasonic sensors may have different installation arrangements during the training and inference stages, such as being installed at different locations on the same vehicle model or on different vehicle models, different installation arrangements may result in large differences in feature representations based on ultrasonic samples obtained in similar scenarios. This may thereby reduce the accuracy of subsequent object detection by the machine learning model, which will affect the performance of the vehicle’s assisted driving or parking, and also affect the user experience.

[0071] In the technical solution of the present disclosure, partial feature representations of some ultrasonic signal features that are insensitive to changes in installation arrangements may be selected from all feature representations of all ultrasonic signal features for feature extraction, thereby saving computing resources, and even if the ultrasonic sensor has an installation arrangement different from that in the training stage in actual application, the performance of the trained machine learning model will not be affected. Therefore, more accurate and reliable object detection may be achieved across multiple scenarios and multiple vehicle models, thereby improving vehicle assisted driving / parking performance and user experience.

[0072] As described above, in order to select partial feature representations of some ultrasonic signal features that are insensitive to changes in installation arrangements from all feature representations of all ultrasonic signal features, the present invention discloses that ultrasonic data may be first collected separately in the same scenario using ultrasonic sensors with different installation arrangements, and all feature representations of all ultrasonic signal features of different groups of collected ultrasonic data may be compared, and the above process may be repeated in multiple different scenarios to select feature representations of ultrasonic signal features that are stable in different installation arrangements and different scenarios.

[0073] FIG. 4 is a schematic diagram of a scenario for processing an ultrasonic signal according to examples of the present disclosure.

[0074] As shown in FIG. 4, ultrasonic data may be collected from a first vehicle 400-1 equipped with an ultrasonic sensor 410-1 in a first scenario, and an ultrasonic sample set may be obtained based on the ultrasonic data. For example, the first vehicle 400-1 may make a single trip in the first scenario to obtain an ultrasonic sample set, such as the ultrasonic sample set 310 described above in conjunction with FIG. 3. For another example, the first vehicle 400-1 may make multiple trips in the first scenario to obtain an ultrasonic sample set corresponding to ultrasonic data collected multiple times.

[0075] In one example, the scenario deployment may comprise arrangements regarding the following items: the vehicle driving route, the vehicle driving speed, the number of objects in the vehicle’s surroundings, the type of object in the vehicle’s surroundings, the position of object in the vehicle’s surroundings, etc. As shown in FIG. 4, the first vehicle 400-1 will travel along a path 430 at a speed of 10 km / h passing a lateral vehicle 440, a ground lock 450, and a vertical vehicle 460. In other examples, the surroundings may comprise a variety of other types of objects (not shown), such as walls, railings, columns, trees, etc.

[0076] In one example, the ultrasonic sample set may comprise multiple ultrasonic samples, such as echo samples and / or echo intersection samples. Each ultrasonic sample may comprise multiple ultrasonic signal features. For example, each echo sample may comprise one or more ultrasonic signal features of echo point coordinates, echo amplitude, echo significance, echo distance, echo height, echo point deflection, sensor coordinates, and echo time; each echo intersection sample may comprise one or more ultrasonic signal features of echo intersection coordinates, echo intersection distance, echo intersection height, echo intersection deflection, echo intersection amplitude, echo intersection significance, adjacent echo intersection coordinates, adjacent echo intersection height, adjacent echo intersection distance, adjacent echo intersection deflection, adjacent echo intersection significance, adjacent echo intersection amplitude, adjacent echo intersection height, sensor coordinates, and echo time.

[0077] In one example, each ultrasonic signal feature may have multiple feature representations. For example, the feature representation may comprise statistics of ultrasonic signal features of multiple ultrasonic samples, including but not limited to minimum value, maximum value, mean value, variance, median, quantile, kurtosis, etc. For another example, the feature representation may comprise transformation or approximation of ultrasonic signal features of multiple ultrasonic samples, including but not limited to Fourier transform, piecewise approximation, autocorrelation, cepstrum analysis, etc.

[0078] In one example, each feature representation for each ultrasonic signal feature may be obtained based on the ultrasonic sample obtained by the first vehicle 400-1 during a single trip in the first scenario. For example, if M ultrasonic signal features are used and each ultrasonic signal feature is represented by N features, a total of MxN feature representations of ultrasonic signal features may be obtained during a single trip in the first scenario. Alternatively, a first distribution of each feature representation for each ultrasonic signal feature in the first scenario may be obtained based on the ultrasonic samples obtained by the first vehicle 400-1 during multiple trips in the first scenario. For example, if the first vehicle makes T trips in the first scenario, as described above, feature representations of MxN ultrasonic signal features may be obtained for each trip, and the first distribution of the feature representations of each of the MxN ultrasonic signal features in the T trips may be obtained. Furthermore, the first distribution of the N feature representations of all M ultrasonic signal features in the first scenario may constitute a first distribution set.

[0079] As shown in FIG. 4, ultrasonic data may also be collected from a second vehicle 400-2 equipped with an ultrasonic sensor 410-2 in a first scenario, and an ultrasonic sample set may be obtained based on the ultrasonic data. For example, the second vehicle 400-2 may make a single trip in the first scenario to obtain an ultrasonic sample set, such as the ultrasonic sample set 310 described above in conjunction with FIG. 3. For another example, the second vehicle 400-2 may make multiple trips in the first scenario to obtain an ultrasonic sample set corresponding to ultrasonic data collected multiple times. Since the second vehicle 400-2 is traveling in the same first scenario as the first vehicle 400-1 , as shown in FIG. 4, the second vehicle 400-2 will also travel along the path 430 at a speed of 10 km / h passing the lateral vehicle 440, the ground lock 450 and the vertical vehicle 460.

[0080] In one example, in a manner similar to that described above, each feature representation for each ultrasonic signal feature may be obtained based on the ultrasonic sample obtained by the second vehicle 400-2 during a single trip in the first scenario. Additionally, based on ultrasonic samples obtained when the second vehicle 400-2 makes multiple trips in the first scenario, a second distribution of each feature representation of each ultrasonic signal feature in the first scenario may be obtained, as well as a second distribution set of all feature representations of all ultrasonic signal features may be obtained.

[0081] It is to be understood that although a second vehicle of a different model than the first vehicle is shown in FIG. 4, in other examples, a second vehicle of the same model as the first vehicle may be used. However, the ultrasonic sensors 410-1 and 410-2 may have different installation arrangements on the first vehicle and the second vehicle. The installation arrangement may be different in one or more of the following: the height of the ultrasonic sensor from the ground, the relative position of the ultrasonic sensor to the center of the front axle of the vehicle, the relative position of the ultrasonic sensor to the center of the vehicle chassis, or the relative position of the ultrasonic sensor to the center of the rear axle of the vehicle. Additionally, while two vehicles are shown in FIG. 4, in other examples, more vehicles may be comprised, wherein the ultrasonic sensor may have a different installation arrangement on each vehicle.

[0082] In one example, the similarity of the first distribution and the second distribution of each feature representation of each ultrasonic signal feature may be compared, and the feature representation of the ultrasonic signal feature may be selected based on the similarity. Additionally, the first vehicle 400-1 and the second vehicle 400-2 may be used to collect ultrasonic data in different scenarios, and the similarity of each feature representation of each ultrasonic signal feature may be obtained for each scenario in a similar manner as described above. Subsequently, a similarity score of each feature representation of each ultrasonic signal feature may be obtained based on the similarities obtained in multiple different scenarios, and the feature representation of the ultrasonic signal feature to be selected may be determined by comparing the similarity score with a threshold.

[0083] A specific implementation of the method of the present disclosure will be described in detail below in conjunction with FIG. 5. FIG. 5 is a schematic diagram of a process for processing an ultrasonic signal according to examples of the present disclosure.

[0084] At block 510, the vehicle may make multiple trips in the first scenario, and obtain a first ultrasonic sample set through an ultrasonic sensor with a first installation arrangement.

[0085] The vehicle may be 400-1 as shown in FIG. 4. The first scenario may comprise arrangements regarding the following items: the vehicle driving route, the vehicle driving speed, the number of objects in the vehicle’s surroundings, the type of object in the vehicle’s surroundings, the position of object in the vehicle’s surroundings. The first installation arrangement may comprise arrangements regarding the following items: the height of the ultrasonic sensor from the ground, the relative position of the ultrasonic sensor to the center of the front axle of the vehicle, the relative position of the ultrasonic sensor to the center of the vehicle chassis, or the relative position of the ultrasonic sensor to the center of the rear axle of the vehicle. The first ultrasonic sample set may comprise multiple ultrasonic sample subsets, such as 511-1 to 511-L shown in FIG. 5, and each ultrasonic sample subset may correspond to the ultrasonic samples obtained by the vehicle during a single trip in the first scenario.

[0086] At block 520, each feature representation of each ultrasonic signal feature may be calculated based on each ultrasonic sample subset of the first ultrasonic sample set.

[0087] For example, the ultrasonic signal features of the echo sample may comprise [echo amplitude, echo significance, echo height], and the feature representations may comprise [mean value, variance, Fourier transform].

[0088] In one example, each feature representation of each ultrasonic signal feature may be calculated based on all echo samples in the ultrasonic sample subset 511-1 : [mean value of echo amplitude, variance of echo amplitude, Fourier transform of echo amplitude, mean value of echo significance, variance of echo significance, Fourier transform of echo significance, mean value of echo height, variance of echo height, Fourier transform of echo height]. Subsequently, each feature representation of each ultrasonic signal feature may be calculated based on all echo samples in the ultrasonic sample subsets 511-2 to 511-L as described above. It should be understood that the above description is only used as an example, and the ultrasonic sample may comprise M ultrasonic signal features, and each ultrasonic signal feature may be represented by N features, so a total of MxN feature representations of ultrasonic signal features may be obtained during a single trip in the first scenario.

[0089] In one example, each feature representation of each ultrasonic signal feature may be calculated for each object based on the echo samples corresponding to each object in the ultrasonic sample subset 511-1. For example, the ultrasonic sample subset 511-1 may be obtained in the scenario shown in FIG. 4, where there are three objects, and the ultrasonic samples in the ultrasonic sample subset 511-1 may be divided into three groups, wherein the sample group O1 corresponds to the lateral vehicle 440, the sample group 02 corresponds to the ground lock 450, and the sample group 03 corresponds to the vertical vehicle 460. In one example, the echo samples corresponding to each object may be identified by clustering. In another example, the echo samples corresponding to each object may be identified based on information collected by other types of sensors having the same or similar installation arrangements as the ultrasonic sensors, such as image sensors, lidars, etc. Therefore, [the mean value of echo amplitude, the variance of echo amplitude, the Fourier transform of echo amplitude, the mean value of echo significance, the variance of echo significance, the Fourier transform of echo significance, the mean value of echo height, the variance of echo height, the Fourier transform of echo height] for object 1 may be calculated based on the echo samples in sample group O1 ; [the mean value of echo amplitude, the variance of echo amplitude, the Fourier transform of echo amplitude, the mean value of echo significance, the variance of echo significance, the Fourier transform of echo significance, the mean value of echo height, the variance of echo height, the Fourier transform of echo height] for object 2 may be calculated based on the echo samples in sample group 02; and [the mean value of echo amplitude, the variance of echo amplitude, the Fourier transform of echo amplitude, the mean value of echo significance, the variance of echo significance, the Fourier transform of echo height, the mean value of echo height, the variance of echo height, the Fourier transform of echo height] for object 3 may be calculated based on the echo samples in sample group 03. Subsequently, each feature representation of each ultrasonic signal feature may be calculated for each object as described above based on the echo samples in the ultrasonic sample subsets 511-2 to 511-N. It should be understood that the above description is only used as an example. The ultrasonic sample may comprise M ultrasonic signal features, each ultrasonic signal feature may be represented by N features, and there may be K objects in the first scenario. Then, fora single trip in the first scenario, a total of K groups of MxN feature representations of ultrasonic signal features may be obtained for the K objects.

[0090] At block 530, the distribution of each feature representation of each ultrasonic signal feature may be calculated based on all ultrasonic sample subsets of the first ultrasonic sample set.

[0091] For example, as described above, each feature representation of each ultrasonic signal feature may be obtained for each of the ultrasonic sample subsets 511-1 to 511-L, totaling MxN feature representations, and then the distribution of each feature representation of each ultrasonic signal feature on all L ultrasonic sample subsets may be fitted. Those skilled in the art will appreciate that any applicable data distribution fitting method may be used to implement this operation.

[0092] In one example, the distribution of each feature representation of each ultrasonic signal feature may be obtained: [distribution of the mean value of echo amplitude, distribution of the variance of echo amplitude, distribution of the Fourier transform of echo amplitude, distribution of the mean value of echo significance, distribution of the variance of echo significance, distribution of the Fourier transform of echo significance, distribution of the mean value of echo height, distribution of the variance of echo height, distribution of the Fourier transform of echo height]. It should be understood that the above description is only used as an example, and MxN distributions may be obtained for the feature representations of MxN ultrasonic signal features, and the MxN distributions may be referred to as a first distribution set.

[0093] In one example, the distribution of each feature representation of each ultrasonic signal feature may be obtained for each object: object 1 [distribution of the mean value of echo amplitude, distribution of the variance of echo amplitude, distribution of the Fourier transform of echo amplitude, distribution of the mean value of echo significance, distribution of the variance of echo significance, distribution of the Fourier transform of echo significance, distribution of the mean value of echo height, distribution of the variance of echo height, distribution of the Fourier transform of echo height]; object 2 [distribution of the mean value of echo amplitude, distribution of the variance of echo amplitude, distribution of the Fourier transform of echo amplitude, distribution of the mean value of echo significance, distribution of the variance of echo significance, distribution of the Fourier transform of echo significance, distribution of the mean value of echo height, distribution of the variance of echo height, distribution of the Fourier transform of echo height]; and object 3 [distribution of the mean value of echo amplitude, distribution of the variance of echo amplitude, distribution of the Fourier transform of echo amplitude, distribution of the mean value of echo significance, distribution of the variance of echo significance, distribution of the Fourier transform of echo significance, distribution of the mean value of echo height, distribution of the variance of echo height, distribution of the Fourier transform of echo height]. It should be understood that the above description is only used as an example, and K groups of MxN distributions may be obtained for the feature representations of MxN ultrasonic signal features of K objects, and the K groups of MxN distributions may be referred to as a first distribution set.

[0094] Similar to blocks 510 to 530, at block 510-2, the vehicle may make multiple trips in the first scenario, and obtain a second ultrasonic sample set by using an ultrasonic sensor having a second installation arrangement. The installation arrangement may be different in at least one of the following: the height of the ultrasonic sensor from the ground, the relative position of the ultrasonic sensor to the center of the front axle of the vehicle, the relative position of the ultrasonic sensor to the center of the vehicle chassis, or the relative position of the ultrasonic sensor to the center of the rear axle of the vehicle. At block 520-2, each feature representation of each ultrasonic signal feature may be calculated based on each ultrasonic sample subset of the second ultrasonic sample set. At block 530-2, the distribution of each feature representation of each ultrasonic signal feature may be calculated based on all ultrasonic sample subsets of the second ultrasonic sample set.

[0095] Subsequently, at block 540, the similarity between the first ultrasonic sample set and the second ultrasonic sample set may be calculated based on the distribution of each feature representation of each ultrasonic signal feature obtained at blocks 530 and 530-2, such as the first distribution set and the second distribution set described above, to reflect the similarity of ultrasonic signal feature distribution between different installation arrangements in the first scenario.

[0096] In one example, the distribution similarity of each feature representation of each ultrasonic signal feature may be obtained: [distribution similarity of the mean value of echo amplitude, distribution similarity of the variance of echo amplitude, distribution similarity of the Fourier transform of echo amplitude, distribution similarity of the mean value of echo significance, distribution similarity of the variance of echo significance, distribution similarity of the Fourier transform of echo significance, distribution similarity of the mean value of echo height, distribution similarity of the variance of echo height, distribution similarity of the Fourier transform of echo height]. It should be understood that the above description is only used as an example, and MxN distribution similarities may be obtained for distributions of feature representations of MxN ultrasonic signal features under two installation arrangements, which may be referred to as a first similarity set. In one example, the distribution similarity of each feature representation of each ultrasonic signal feature for each object may be obtained: object 1 [distribution similarity of the mean value of echo amplitude, distribution similarity of the variance of echo amplitude, distribution similarity of the Fourier transform of echo amplitude, distribution similarity of the mean value of echo significance, distribution similarity of the variance of echo significance, distribution similarity of the Fourier transform of echo significance, distribution similarity of the mean value of echo height, distribution similarity of the variance of echo height, distribution similarity of the Fourier transform of echo height]; object 2 [distribution similarity of the mean value of echo amplitude, distribution similarity of the variance of echo amplitude, distribution similarity of the Fourier transform of echo amplitude, distribution similarity of the mean value of echo significance, distribution similarity of the variance of echo significance, distribution similarity of the Fourier transform of echo significance, distribution similarity of the mean value of echo height, distribution similarity of the variance of echo height, distribution similarity of the Fourier transform of echo height]; and object 3 [distribution similarity of the mean value of echo amplitude, distribution similarity of the variance of echo amplitude, distribution similarity of the Fourier transform of echo amplitude, distribution similarity of the mean value of echo significance, distribution similarity of the variance of echo significance, distribution similarity of the Fourier transform of echo significance, distribution similarity of the mean value of echo height, distribution similarity of the variance of echo height, distribution similarity of the Fourier transform of echo height]. It should be understood that the above description is only used as an example, and K groups of MxN distribution similarities may be obtained for K groups of MxN distributions under two installation arrangements, which may be referred to as a first similarity set.

[0097] In one example, the similarity may be calculated based on one or more of the following: cosine similarity, KL divergence, total variation. Those skilled in the art will appreciate that any applicable data distribution distance measurement method may also be used to implement this operation.

[0098] So far, the similarity of ultrasonic signal feature distribution between different installation arrangements in the first scenario is obtained at block 540. In addition, the process described above may be repeated for different scenarios to obtain the similarity of ultrasonic signal feature distribution between different installation arrangements in other scenarios.

[0099] Similar to block 540, at block 540-2, similarities between ultrasonic sample sets under different installation arrangements may be calculated based on the distribution of each feature representation of each ultrasonic signal feature obtained in the second scenario. For example, a second similarity set may be obtained in a similar manner as described above in conjunction with blocks 510 to 540 to reflect the similarity of ultrasonic signal feature distribution between different installation arrangements in the second scenario. The second scenario may be different from the first scenario in at least one of the following: the vehicle driving route, the vehicle driving speed, the number of objects in the vehicle’s surroundings, the type of object in the vehicle’s surroundings, the position of object in the vehicle’s surroundings.

[0100] At block 550, similarity scores may be determined based on the distribution similarities of each feature representation of each ultrasonic signal feature obtained at blocks 540 and 540-2 in different scenarios, such as the first similarity set and the second similarity set described above, and ultrasonic signal features and corresponding feature representations are selected based on the similarity scores to form an ultrasonic signal feature set for characterization.

[0101] In one example, distribution similarities obtained at block 540 or 540-2 may be distance metrics of data distribution, such as KL divergence values. For example, two groups of distribution similarities may be aggregated using any method applicable in the art, the distribution similarities in different scenarios may be averaged for each feature representation of each ultrasonic signal feature as the similarity score, the maximum or minimum value of the distribution similarities in different scenarios may be taken as the similarity score, or the maximum or minimum value of the distribution similarities in different scenarios may be removed and then averaged as the similarity score, etc.

[0102] In one example, the distribution similarities obtained at block 540 or 540-2 may be distribution similarities rankings, such as from high to low. For example, two groups of distribution similarities may be aggregated using any method applicable in the art, the distribution similarity rankings in different scenarios may be averaged for each feature representation of each ultrasonic signal feature as the similarity score, the maximum or minimum value of the distribution similarity rankings in different scenarios may be taken as the similarity score, or the maximum or minimum value of the distribution similarity rankings in different scenarios may be removed and then averaged as the similarity score, etc.

[0103] In one example, when the distribution similarities obtained at block 540 or 540-2 are obtained for an object, a similarity weight coefficient may be set according to the object type, and the distribution similarities of different object types may be weighted using the similarity weight coefficient before calculating the similarity scores. For example, in actual parking lot applications, vertical vehicles appear much more frequently than lateral vehicles, making it more important to accurately predict the position of vertical vehicles. In contrast, the likelihood of ground locks appearing in a public parking environment is relatively low. Therefore, the similarity weight coefficient for vertical vehicles may be set at 1 .1 , for lateral vehicles at 1 , and for ground locks at 0.9. The distribution similarities obtained at blocks 540 and 540-2 may be weighted according to the object type according to the determined weight coefficient before calculating the similarity scores.

[0104] In one example, when the distribution similarities obtained at block 540 or 540-2 are obtained for an object, the similarity scores may be determined separately according to the object type. Similarly, the distribution similarities obtained at block 540 or 540-2 may be distance metrics of the data distribution, such as KL divergence values; or, the distribution similarities may be distribution similarity rankings, such as from high to low. For example, two groups of distribution similarities may be aggregated using any method applicable in the art, the distribution similarities or rankings in different scenarios may be averaged for each feature representation of each ultrasonic signal feature as the similarity score, the maximum or minimum value of the distribution similarities or rankings in different scenarios may be taken as the similarity score, or the maximum or minimum value of the distribution similarities or rankings in different scenarios may be removed and then averaged as the similarity score, etc.

[0105] In one example, the ultrasonic signal features and the corresponding feature representations may be selected based on the comparison of the similarity score with the threshold to form the ultrasonic signal feature set for characterization. For example, the threshold may be an absolute value, such ultrasonic signal features and corresponding feature representations whose similarity scores are greater than a certain threshold value may be selected. For example, the threshold may be a relative value, such ultrasonic signal features and corresponding feature representations whose similarity scores are ranked from high to low by less than a certain threshold percentage may be selected.

[0106] In one example, the ultrasonic signal features and the corresponding feature representations may be selected according to the object type. For example, for object type 1 , [mean value of echo amplitude, Fourier transform of echo amplitude, mean value of echo significance, variance of echo height] may be selected, for object type 2, [variance of echo amplitude, Fourier transform of echo amplitude, mean value of echo significance] may be selected, and for object type 3, [variance of echo amplitude, mean value of echo significance, variance of echo significance, mean value of echo height, variance of echo height] may be selected. Those skilled in the art will appreciate that the ultrasonic signal features and corresponding feature representations described in conjunction with FIG. 4 and FIG. 5 are merely examples and are not to be considered as limitations.

[0107] Those skilled in the art should understand that, although only two scenarios are shown in FIG. 5, in other examples, any number of scenarios are possible. For example, a third similarity set may be obtained based on a third scenario, a fourth similarity set may be obtained based on a fourth scenario, and so on, and similarity scores may be obtained in the manner described above based on the similarity sets of all scenarios.

[0108] By obtaining ultrasonic sample sets in multiple scenarios with ultrasonic sensors having different installation arrangements, and comparing the distribution similarities of all feature representations of all ultrasonic signal features between different installation arrangements and different scenarios, partial feature representations of partial ultrasonic signal features that are insensitive to installation arrangement and scenario changes may be selected for characterization. Furthermore, feature representations of ultrasonic signal features may be selected according to the object type to achieve a balance between versatility and category characterization capabilities.

[0109] FIG. 6 is a flow chart of a method for processing an ultrasonic signal according to examples of the present disclosure.

[0110] At step 610, the vehicle may make multiple trips in the first scenario, and obtain a first ultrasonic sample set through an ultrasonic sensor with a first installation arrangement. For example, as described in conjunction with block 510 of FIG. 5.

[0111] At step 620, the vehicle may make multiple trips in the first scenario, and obtain a second ultrasonic sample set through an ultrasonic sensor with a second installation arrangement. For example, as described in conjunction with block 510-2 of FIG. 5.

[0112] In one example, the first scenario may be arranged by the following items: the vehicle driving route, the vehicle driving speed, the number of objects in the vehicle’s surroundings, the type of object in the vehicle’s surroundings, the position of object in the vehicle’s surroundings.

[0113] In one example, the first installation arrangement may be arranged by the following items: the height of the ultrasonic sensor from the ground; the relative position of the ultrasonic sensor to the center of the front axle of the vehicle; the relative position of the ultrasonic sensor to the center of the vehicle chassis; or the relative position of the ultrasonic sensor to the center of the rear axle of the vehicle.

[0114] In one example, each ultrasonic sample in the first ultrasonic sample set and the second ultrasonic sample set comprises multiple ultrasonic signal features, and each ultrasonic signal feature has multiple feature representations. For example, the ultrasonic sample may comprise an echo sample and an echo intersection sample, and the ultrasonic signal feature may comprise one or more of the following: echo point coordinates, echo amplitude, echo significance, echo distance, echo height, echo point deflection, echo intersection coordinates, echo intersection distance, echo intersection height, echo intersection deflection, echo intersection amplitude, echo intersection significance, adjacent echo intersection coordinates, adjacent echo intersection height, adjacent echo intersection distance, adjacent echo intersection deflection, adjacent echo intersection significance, adjacent echo intersection amplitude, adjacent echo intersection height, sensor coordinates, and echo time. For example, the feature representation may comprise one or more of the following: minimum, maximum, mean value, variance, median, quartiles, kurtosis, Fourier transform, piecewise approximation, autocorrelation, or cepstrum analysis.

[0115] At step 630, the distribution of each feature representation of each ultrasonic signal feature in the first scenario may be calculated based on the first ultrasonic sample set and the second ultrasonic sample set, respectively.

[0116] In one example, each feature representation of each ultrasonic signal feature may be obtained based on the ultrasonic samples corresponding to a vehicle’s single trip in the first ultrasonic sample set, and the first distribution set may be obtained based on each feature representation of each ultrasonic signal feature corresponding to the vehicle’s each trip. For example, as described in conjunction with blocks 520 and 530 of FIG. 5.

[0117] In one example, each feature representation of each ultrasonic signal feature may be obtained based on the ultrasonic samples corresponding to a vehicle’s single trip in the second ultrasonic sample set, and the second distribution set may be obtained based on each feature representation of each ultrasonic signal feature corresponding to the vehicle’s each trip. For example, as described in conjunction with blocks 520-2 and 530-2 of FIG. 5.

[0118] At step 640, an ultrasonic signal feature set for characterization may be determined based at least on the similarity in distribution of each feature representation of each ultrasonic signal feature in the first scenario between the first ultrasonic sample set and the second ultrasonic sample set.

[0119] In one example, similarity may be calculated based on the distribution corresponding to each feature representation of each ultrasonic signal feature in the first distribution set and the second distribution set to obtain a first similarity set. For example, as described in conjunction with block 540 of FIG. 5.

[0120] In one example, the similarity may be calculated based on one or more of the following: cosine similarity, KL divergence, total variation.

[0121] In one example, the first scenario may comprise one or more objects in the surroundings of the vehicle, and the first distribution set and the second distribution set and the first similarity set are further obtained for each of the one or more objects. For example, as described in conjunction with block 540 of FIG. 5.

[0122] In one example, the vehicle may also repeat the operations of steps 610 to 640 in a second scenario. For example, the vehicle makes multiple trips in the second scenario, and obtains a third ultrasonic sample set through the ultrasonic sensor with the first installation arrangement, wherein the second scenario comprises one or more objects in the surroundings of the vehicle; the vehicle makes multiple trips in the second scenario, and obtains a fourth ultrasonic sample set through the ultrasonic sensor with the second installation arrangement; the distribution of each feature representation of each ultrasonic signal feature in the second scenario is calculated for each object based on the third ultrasonic sample set and the fourth ultrasonic sample set; and the ultrasonic signal feature set for characterization is further determined based on a second similarity set of the distribution of each feature representation of each ultrasonic signal feature for each object in the second scenario between the third ultrasonic sample set and the fourth ultrasonic sample set. For example, as described in conjunction with block 540-2 of FIG. 5.

[0123] The first scenario and the second scenario may be different in at least one of the following: the vehicle driving route, the vehicle driving speed, the number of objects in the vehicle’s surroundings, the type of object in the vehicle’s surroundings, the position of object in the vehicle’s surroundings.

[0124] In one example, a similarity score may be determined for each feature representation of each ultrasonic signal feature at least in part based on the first similarity set and the second similarity set, and ultrasonic signal features and corresponding feature representations with similarities above a threshold may be selected based on the determined similarity scores to form the ultrasonic signal feature set for characterization. For example, as described in conjunction with block 550 of FIG. 5.

[0125] In one example, a similarity weight coefficient may be determined according to the object type, the similarity of each feature representation of each ultrasonic signal feature in the first similarity set and the second similarity set may be weighted according to the object type using the determined similarity weight coefficient, and the similarity score may be determined for each feature representation of each ultrasonic signal feature based at least in part on the weighted first similarity set and the second similarity set.

[0126] In one example, a similarity score may be determined for each feature representation of each ultrasonic signal feature according to the object type at least in part based on the first similarity set and the second similarity set, and ultrasonic signal features and corresponding feature representations with similarities above a threshold may be selected based on the similarity scores according to each object type to form the ultrasonic signal feature set for characterization, wherein the ultrasonic signal feature set for characterization comprises multiple ultrasonic signal feature subsets respectively used for different object types.

[0127] In one example, the threshold may be an absolute value of the similarity score, or a relative ratio value of the similarity score ranking. In one example, the ultrasonic signal features of an echo sample may comprise [echo point coordinates, echo amplitude, echo significance, echo distance], the ultrasonic signal features of an echo intersection sample may comprise [echo intersection coordinates, echo intersection distance, echo intersection height, echo intersection deflection], and the feature representation may comprise [minimum value, maximum value, mean value]. Therefore, there are 24 feature representations of all ultrasonic signal features. Based on the similarity scores obtained according to the method described herein, feature representations of ultrasonic signal features with scores ranking in the top 50%, i.e. , 12 types, may be selected for characterization. For example, the selected 12 feature representations of ultrasonic signal features may be used for all object types. Additionally or alternatively, the feature representations of the ultrasonic signal features with similarity scores ranking in the top 25%, i.e., 6 types, may be selected for characterization according to each object type. For example, 6 feature representations of ultrasonic signal features are selected for vertical vehicles, 6 feature representations of ultrasonic signal features are selected for lateral vehicles, and so on. The feature representations of ultrasonic signal features selected for different object types may be completely identical, partially identical, or completely different.

[0128] FIG. 7 is a flow chart of a method for training a machine learning model to determine an object in the surroundings of a vehicle based on an ultrasonic signal according to examples of the present disclosure.

[0129] At step 710, ultrasonic signal features and corresponding feature representations may be selected for characterization, and the ultrasonic signal features and corresponding feature representations may be selected from an ultrasonic signal feature set determined using the method described in the present disclosure.

[0130] At step 720, a training ultrasonic sample set may be processed based on the selected ultrasonic signal features and corresponding feature representations, and the training samples characterized by the selected ultrasonic signal features and corresponding feature representations may be input into a machine learning model set to be trained to output prediction information associated with objects in the surroundings of the vehicle.

[0131] In one example, the machine learning model in the machine learning model set may be implemented based on a classification algorithm and / or a regression algorithm.

[0132] In one example, the machine learning model set may comprise multiple machine learning models, wherein each machine learning model is used for a specific object type. For example, an applicable machine learning model may be selected from the machine learning model set based on the object type. Further, the ultrasonic signal features and corresponding feature representations selected at step 710 may be selected from a corresponding subset of ultrasonic signal feature set for characterization according to the object type, and used as input of the corresponding object type to a machine learning model at step 720.

[0133] At step 730, the machine learning model set may output the prediction information associated with objects in the surroundings of the vehicle, and the machine learning model set may be trained based on the output and the corresponding labels. FIG. 8 is a flow chart of a method for determining an object in the surroundings of a vehicle based on an ultrasonic signal according to examples of the present disclosure.

[0134] At step 810, an ultrasonic sample set may be obtained by an ultrasonic sensor during movement of the first vehicle, wherein each ultrasonic sample in the ultrasonic sample set comprises multiple ultrasonic signal features, and each ultrasonic signal feature has multiple feature representations. For example, as described in conjunction with the ultrasonic sample set 310 of FIG. 3.

[0135] At step 820, clustering may be performed based on the ultrasonic data set to obtain one or more sample clusters, wherein each sample cluster corresponds to an object in the surroundings of the first vehicle. For example, as described in conjunction with module 330 of FIG. 3.

[0136] At step 830, ultrasonic samples corresponding to each sample cluster may be processed based on ultrasonic signal features and corresponding feature representations to input a trained machine learning model set. For example, as described in conjunction with module 340 of FIG. 3. The ultrasonic signal features and the corresponding feature representations may be selected from an ultrasonic signal feature set determined using the method described in the present disclosure. The same ultrasonic signal features and the corresponding feature representations are used in the inference stage and the training stage.

[0137] In one example, the machine learning model set may comprise multiple machine learning models, wherein each machine learning model is used for a specific object type. For example, an applicable machine learning model may be selected from the machine learning model set based on the object type. For example, as described in conjunction with module 350 of FIG. 3. The selected ultrasonic signal features and corresponding feature representations are selected from different ultrasonic signal feature subsets of the ultrasonic signal feature set for characterization according to the object type, and the samples corresponding to each sample cluster may be processed separately based on the ultrasonic signal features and corresponding feature representations for a specific object type to serve as input to a machine learning model for the object type.

[0138] In one example, the trained machine learning model set may be trained using the method described in conjunction with FIG. 7.

[0139] In one example, the trained machine learning model set may be trained on the second vehicle based on an ultrasonic sample set obtained by an ultrasonic sensor during movement of the second vehicle, with the ultrasonic sensor having different installation arrangements on the first vehicle and second vehicle.

[0140] At step 840, prediction information associated with the object in the surroundings of the first vehicle is output by the machine learning model set. For example, as described in conjunction with module 360 of FIG. 3.

[0141] FIG. 9 is a block diagram of a system according to examples of the present disclosure.

[0142] The system 900 may comprise one or more processors 910 and a memory 920. The memory 920 may store executable instructions. The processor 910 may execute the executable instructions stored or encoded in the memory 920, thereby implementing the various operations and / or functions described above in conjunction with FIGs. 1 to 8. Although not shown in FIG. 9, those skilled in the art may appreciate that the system 900 may comprise various other components, such as various communication modules, bus modules, and possible user interface modules, and the like.

[0143] In some examples, the control system 900 may comprise the control unit 130 and / or the processing unit 115 shown in FIGs. 1 and 2.

[0144] The examples of the present disclosure further provide a computer-readable storage medium. The computer-readable storage medium may store executable instructions, and the executable instructions may implement the various operations and / or functions described above in conjunction with FIGs. 1 to 8 when executed by the processor. For example, the computer- readable storage medium may include, but is not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Electrically-Erasable Programmable Read-Only Memory (EEPROM), Static Random Access Memory (SRAM), hard disk, flash memory, and the like.

[0145] Examples of the present disclosure also provide a computer program product. The computer program product may comprise a computer program. The computer program, when executed by a processor, may implement the various operations and / or functions described above in conjunction with FIGs. 1 to 8.

[0146] Examples of the present disclosure also provides a vehicle. The vehicle may have an ultrasonic sensor such as that shown in FIG. 1 , which is used to transmit and receive ultrasonic signals. The vehicle may also have the system 900 of FIG. 9.

[0147] The above-described specific examples of the present disclosure have been described. Other examples are within the scope of the appended Claims. In some cases, actions or steps described in the Claims can be performed in a different order than that of the examples and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require specific or continuous sequences to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or advantageous.

[0148] Not all steps and units depicted in the above-mentioned flowcharts and system diagrams are required; certain steps or units may be omitted based on actual needs. The device structures described in the above-mentioned examples can be physical or logical structures. That is, some units may be realized by the same physical entity, while others may be realized by multiple physical entities or may be jointly realized by certain components in multiple separate devices.

[0149] Throughout the present Description, the term “exemplary” means “serving as an example, instance, or illustration” and does not imply “preferred” or “advantageous” over other examples. Specific examples comprise specific details to facilitate understanding of the described technology. However, these technologies may be implemented without these specific details. In some instances, to avoid causing difficulties in understanding the concepts of the described examples, known structures and devices are shown in block diagram form.

[0150] The aforementioned description of the present disclosure is provided to allow any person of ordinary skill in the art to implement or use the present disclosure. Various modifications to the present disclosure will be apparent to those of ordinary skill in the art, and the general principles defined herein may be applied to other variations without departing from the scope of protection of the present disclosure. Therefore, the present disclosure is not limited to the exemplary examples and designs described herein but is consistent with the broadest scope defined by the principles and novel features disclosed herein.

Claims

CLAIMS1 . A method for processing an ultrasonic signal, comprising: a vehicle making multiple trips in a first scenario, and obtaining a first ultrasonic sample set through an ultrasonic sensor with a first installation arrangement; a vehicle making multiple trips in the first scenario, and obtaining a second ultrasonic sample set through an ultrasonic sensor with a second installation arrangement, wherein each ultrasonic sample in the first ultrasonic sample set and the second ultrasonic sample set comprises multiple ultrasonic signal features, and each ultrasonic signal feature has multiple feature representations; calculating the distribution of each feature representation of each ultrasonic signal feature in the first scenario based on the first ultrasonic sample set and the second ultrasonic sample set respectively; and determining an ultrasonic signal feature set for characterization based at least on the similarity in distribution of each feature representation of each ultrasonic signal feature in the first scenario between the first ultrasonic sample set and the second ultrasonic sample set.

2. The method according to claim 1 , wherein the first installation arrangement and the second installation arrangement differ in at least one of the following: the height of the ultrasonic sensor from the ground; the relative position of the ultrasonic sensor to the center of the front axle of the vehicle; the relative position of the ultrasonic sensor to the center of the vehicle chassis; or the relative position of the ultrasonic sensor to the center of the rear axle of the vehicle.

3. The method according to claim 1 , wherein calculating the distribution of each feature representation of each ultrasonic signal feature in the first scenario based on the first ultrasonic sample set and the second ultrasonic sample set respectively comprises: obtaining each feature representation of each ultrasonic signal feature based on the ultrasonic samples corresponding to a vehicle’s single trip in the first ultrasonic sample set, and obtaining a first distribution set based on each feature representation of each ultrasonic signal feature corresponding to the vehicle’s each trip; and obtaining each feature representation of each ultrasonic signal feature based on the ultrasonic samples corresponding to a vehicle’s single trip in the second ultrasonic sample set, and obtaining a second distribution set based on each feature representation of each ultrasonicsignal feature corresponding to the vehicle’s each trip.

4. The method according to claim 3, further comprising: calculating similarity based on the distribution corresponding to each feature representation of each ultrasonic signal feature in the first distribution set and the second distribution set to obtain a first similarity set.

5. The method according to claim 4, wherein the similarity is calculated based on one or more of the following: cosine similarity, KL divergence, total variation.

6. The method according to claim 4, wherein the first scenario comprises one or more objects in the surroundings of the vehicle, and wherein the first distribution set and the second distribution set, and the first similarity set are further obtained for each of the one or more objects.

7. The method according to claim 6, further comprising: the vehicle making multiple trips in a second scenario, and obtaining a third ultrasonic sample set through the ultrasonic sensor with the first installation arrangement, wherein the second scenario comprises one or more objects in the surroundings of the vehicle; the vehicle making multiple trips in the second scenario, and obtaining a fourth ultrasonic sample set through the ultrasonic sensor with the second installation arrangement, wherein each ultrasonic sample in the third ultrasonic sample set and the fourth ultrasonic sample set comprises multiple ultrasonic signal features, and each ultrasonic signal feature has multiple feature representations; calculating the distribution of each feature representation of each ultrasonic signal feature for each object in the second scenario based on the third ultrasonic sample set and the fourth ultrasonic sample set respectively; and further determining the ultrasonic signal feature set for characterization based on the similarity in distribution of each feature representation of each ultrasonic signal feature for each object in the second scenario between the third ultrasonic sample set and the fourth ultrasonic sample set.

8. The method according to claim 7, wherein the first scenario and the second scenario are different in at least one of the following:the vehicle driving route, the vehicle driving speed, the number of objects in the vehicle’s surroundings, the type of object in the vehicle’s surroundings, the position of object in the vehicle’s surroundings.

9. The method according to claim 7, wherein further determining the ultrasonic signal feature set for characterization based on the similarity in distribution of each feature representation of each ultrasonic signal feature for each object in the second scenario between the third ultrasonic sample set and the fourth ultrasonic sample set comprises: calculating similarity based on the distribution corresponding to each feature representation of each ultrasonic signal feature in the third distribution set and the fourth distribution set to obtain a second similarity set; determining a similarity score for each feature representation of each ultrasonic signal feature at least partially based on the first similarity set and the second similarity set; and selecting ultrasonic signal features and corresponding feature representations with similarities above a threshold based on the determined similarity scores to form the ultrasonic signal feature set for characterization.

10. The method according to claim 9, further comprising: determining a similarity weight coefficient according to the object type; using the determined similarity weight coefficient to weight the similarity of each feature representation of each ultrasonic signal feature in the first similarity set and the second similarity set according to the object type; and determining the similarity score for each feature representation of each ultrasonic signal feature at least partially based on the weighted first similarity set and the second similarity set.11 . The method according to claim 7, wherein further determining the ultrasonic signal feature set for characterization based on the similarity in distribution of each feature representation of each ultrasonic signal feature for each object in the second scenario between the third ultrasonic sample set and the fourth ultrasonic sample set comprises: determining a similarity score for each feature representation of each ultrasonic signal feature according to the object type at least partially based on the first similarity set and the second similarity set; and selecting ultrasonic signal features and corresponding feature representations with similarities above a threshold based on the similarity scores according to each object type to formthe ultrasonic signal feature set for characterization, wherein the ultrasonic signal feature set for characterization comprises multiple ultrasonic signal feature subsets for different object types.

12. The method according to claim 1 , wherein the ultrasonic signal feature comprises one or more of the following: echo point coordinates, echo amplitude, echo significance, echo distance, echo height, echo point deflection, echo intersection coordinates, echo intersection distance, echo intersection height, echo intersection deflection, echo intersection amplitude, echo intersection significance, adjacent echo intersection coordinates, adjacent echo intersection height, adjacent echo intersection distance, adjacent echo intersection deflection, adjacent echo intersection significance, adjacent echo intersection amplitude, adjacent echo intersection height, sensor coordinates, and echo time.

13. The method according to claim 1 , wherein the multiple feature representations comprise one or more of the following: minimum, maximum, mean value, variance, median, quartiles, kurtosis, Fourier transform, piecewise approximation, autocorrelation, or cepstrum analysis.

14. A method for training a machine learning model to determine an object in the surroundings of a vehicle based on an ultrasonic signal, comprising: selecting ultrasonic signal features and corresponding feature representations from the ultrasonic signal feature set for characterization determined by the method according to claim 1 ; inputting a training ultrasonic sample set processed based on the selected ultrasonic signal features and corresponding feature representations into a machine learning model set to output prediction information associated with the object in the surroundings of the vehicle; training the machine learning model set based on the prediction information associated with the object in the surroundings of the vehicle and corresponding labels.

15. The method according to claim 14, wherein the machine learning model set comprises multiple machine learning models respectively used for different object types, and wherein the selected ultrasonic signal features and corresponding feature representations are selected from different ultrasonic signal feature subsets of the ultrasonic signal feature set for characterization according to the object type, and serve as input to a machine learning model for the object type.

16. A method for determining an object in the surroundings of a vehicle based on an ultrasonic signal, comprising: obtaining an ultrasonic sample set by an ultrasonic sensor during movement of the first vehicle, wherein each ultrasonic sample in the ultrasonic sample set comprises multiple ultrasonic signal features, and each ultrasonic signal feature has multiple feature representations; clustering based on the ultrasonic data set to obtain one or more sample clusters, wherein each sample cluster corresponds to an object in the surroundings of the first vehicle; processing the ultrasonic samples corresponding to each sample cluster based on the ultrasonic signal features and corresponding feature representations to input a trained machine learning model set, wherein the ultrasonic signal features and the corresponding feature representations are selected from the ultrasonic signal feature set for characterization determined by the method according to claim 1 ; and outputting prediction information associated with the object in the surroundings of the first vehicle by the machine learning model set.

17. The method according to claim 16, wherein the machine learning model set comprises multiple machine learning models for different object types, and wherein the selected ultrasonic signal features and corresponding feature representations are selected from different ultrasonic signal feature subsets of the ultrasonic signal feature set for characterization according to the object type, and wherein the ultrasonic samples in each sample cluster are processed using the selected ultrasonic signal features and corresponding feature representations based on the object type corresponding to each sample cluster as input to the machine learning model of the corresponding object type.

18. The method according to claim 16, wherein the trained machine learning model set is trained on the second vehicle based on an ultrasonic sample set obtained by an ultrasonic sensor during movement of the second vehicle, with the ultrasonic sensor having different installation arrangements on the first vehicle and second vehicle.

19. A system, comprising: at least one processor; a memory coupled to the at least one processor, the memory storing executableinstructions, wherein the executable instructions, when executed by the at least one processor, allow the at least one processor to implement the method according to any one of claims 1 to 18.

20. A computer-readable medium storing a computer program comprising instructions, wherein the instructions, when executed by the processor, cause one or more units to perform the method according to any one of claims 1 to 18.

21. A computer program product comprising a computer program, wherein the computer program, when executed by a processor, implement the method according to any one of claims 1 to 18.

22. A vehicle, comprising: an ultrasonic sensor for transmitting and receiving an ultrasonic signal; and one or more units for performing the method according to any one of claims 1-18.

Images