Method for outputting music

Abstract

Method for outputting music, wherein an internal processing unit (3) of a vehicle generates music description information (2), wherein the processing unit (3) uses at least one piece of context information (4) to generate the music description information (2), which piece of context information describes a context of a current driving situation, wherein at least one piece of context information (4) describes the following context: an environment traveled by the vehicle; wherein the processing unit (3) evaluates camera images recorded by a camera (6) of the vehicle to determine the context of the environment, and wherein the music is output via vehicle-internal loudspeakers (5), characterized in that the music is synthetically generated by a text-to-music model (1) by reading in the music description information (2); and at least two consecutive camera images used to generate different music description information (2) are stabilized.

Description

[0001] The invention relates to a method for outputting music according to the type defined in the preamble of claim 1.

[0002] Repetitive car journeys such as the daily commute to work or particularly long car journeys such as a trip on holiday can be monotonous and boring. To distract and entertain vehicle occupants, it has long been common practice to enable the playback of media content in vehicles, such as listening to the radio or playing music from a storage medium, such as playing a CD, reading an audio file from a computer-readable storage medium such as an SD card, a USB stick, or from a mobile device linked to the vehicle's infotainment system via Bluetooth. It is now also common practice to stream music, series, or films from the internet. Vehicle occupants can listen to their favorite genre of music or artist.

[0003] The disadvantage, however, is that the amount of music available is limited; in other words, vehicle occupants will eventually have "heard it all before." Furthermore, the music played in the vehicle interior may be inappropriate for the driving situation, which can impair user interaction with the vehicle. For example, classical music may be played at a low tempo while driving fast on a motorway, or drum and bass may be played while driving in a play street. The tempo of the music will then not match the driving speed, which can be oppressive for the driver. Furthermore, this could cause the driver to reduce their speed while overtaking or to drive at excessive speed on a section of road with a low speed limit. Such situations must be avoided.

[0004] The pace of development in the field of artificial intelligence is enormous. Artificial neural networks are capable of solving increasingly complex tasks in ever shorter time. With the help of large language models, particularly sophisticated user interaction is possible, which can already feel as if the user were communicating with a human. Generative AI models are also capable of generating content themselves, such as generating an image or video from a text description.

[0005] AI models are also known that can create audio content such as music from a text description. Examples include "MusicLM" and "Diffsound." For more information, see: https: / / google-research.github.io / seanet / musiclm / examples / and https: / / arxiv.org / abs / 2207.09983.

[0006] In addition, US 2023 / 0083346 A1 describes a system and method for the synthetic generation of audio content. This document describes a system for generating sounds based on a text description. Using natural language processing techniques, keywords are extracted from a corresponding text document, a relationship is established between the keywords, and sounds corresponding to the keywords are generated. Trained models for generating specific sounds are stored in a database. A disadvantage of this approach is that specially trained models must be provided to generate the various sounds, meaning that not all sounds can be generated artificially or the corresponding models may have to be retrained.

[0007] Furthermore, KR 10 2018 0 085 430 A describes a device and a method for providing a noise source in a vehicle using situation-specific context information of a vehicle and a driver. The device is capable of retrieving noises from an external database and outputting them in the vehicle depending on the driving context. However, the amount of noises that can be stored in a corresponding database is limited, so there is a risk that repetitive noises will be output in the vehicle or that noises that do not optimally match the respective situation cannot be found in the database.

[0008] Furthermore, DE 10 2020 127 433 A1 discloses a computer-implemented method for providing music to the interior of a motor vehicle. Information about the motor vehicle and / or a user of the motor vehicle is collected. Music data is generated based on this information. The music data includes information relating to the music to be played. Based on the music data, music is selected from a database and played in the vehicle.

[0009] Furthermore, DE 10 2021 112 020 A1 discloses an environmental perception system for experiencing an environment through music. Music is output via an audio system that changes depending on the environment. The music is remixed based on music composition templates. Artificial neural networks can be used for this purpose.

[0010] Furthermore, DE 10 2022 121 930 A1 discloses a method and a device for playing music in a motor vehicle. Based on sensor data, a usage situation of the motor vehicle is determined. A music signal is artificially synthesized based on the sensor data. The music signal is output in the motor vehicle for the appropriate usage situation.

[0011] Furthermore, DE 10 2021 118 311 A1 discloses the automatic perception and at least partially automatic driving of a vehicle. To improve perception in adverse weather conditions, normalization methods are applied to the point clouds generated by an active optical sensor system. The sensor system is a lidar.

[0012] Furthermore, US 2017 / 0126977 A1 discloses a robust image feature-based video stabilization and smoothing.

[0013] The present invention is based on the object of providing an improved method for outputting music, with the aid of which the interaction between a user and a vehicle can be improved.

[0014] According to the invention, this object is achieved by a method for outputting music having the features of claim 1. Advantageous embodiments and further developments emerge from the dependent claims.

[0015] A generic method for outputting music, wherein an internal computing unit of a vehicle generates music description information, wherein the computing unit uses at least one piece of context information to generate the music description information, which piece of context information describes a context of a current driving situation, wherein at least one piece of context information describes the following context: an environment driven through by the vehicle, wherein the computing unit evaluates camera images recorded by a camera of the vehicle to determine the context of the environment, and wherein the music is output via loudspeakers internal to the vehicle, is further developed in that the music is synthetically generated by a text-to-music model by reading in the music description information, and at least two consecutive camera images used to generate different pieces of music description information are stabilized.

[0016] The text-to-music model is a specially trained AI model capable of synthetically, i.e., artificially, generating music from music description information. The music description information is therefore a text or a text file. The music description information is generated by the vehicle's own computing unit. This requires at least one piece of context information. The context information can be provided to the computing unit, or the computing unit can also be capable of generating or collecting corresponding context information itself.

[0017] The music description information thus corresponds to the vehicle's current driving situation, so that appropriate music is played in the vehicle interior. This improves the interaction between the vehicle occupants and the driver with the vehicle. Thus, every journey becomes a personalized experience. The generated music differs not only from journey to journey, but also from driving situation to driving situation. This allows the use of the vehicle to be linked even more strongly with emotions. This also conveys a unique feeling of luxury. In particular, the music "fits" the current driving context, which improves the driver's cognitive ability to control the vehicle. If, for example, slow and relaxing music is played while driving slowly, this prevents unintentional exceeding of the speed limit or at least reduces the risk of this happening.In a driving situation where high acceleration is required, such as when overtaking or merging onto a motorway, music can be played at a faster tempo, allowing the overtaking maneuver or merging onto the motorway to be completed more quickly and safely. This also improves road safety.

[0018] The text-to-music model can be executed on the vehicle's own computing unit or on an external computing device such as a cloud server. Data connection between the computing unit and the cloud server is possible, for example, via mobile communications or Wi-Fi. In this case, the vehicle includes appropriate communication tools such as a telecommunications unit.

[0019] The environment through which the vehicle travels has a particular impact on the driving experience, as the driver and / or passengers perceive and process their surroundings visually and therefore with a particularly high cognitive level. For example, different music may be played in a city than when driving in the countryside, along a coastal road, through the mountains, or similar. Not only landscape features can be used to describe the surroundings, but also the design of the infrastructure in the vicinity of the vehicle. For example, the condition of the road surface, the type of road, the density of traffic signs or traffic lights, or even driving through a tunnel or underpass can be used to describe the environment. The level of abstraction can be as complex as desired.This also allows properties of objects detected in the vehicle's surroundings to be determined, and the environment to be described based on these properties. For example, the architectural style of a building, such as brutalism, can be identified. In this case, heavy metal music could be played in the vehicle interior. If, however, high-rise buildings with modern glass facades are detected, techno music could be played.

[0020] The vehicle is equipped with one or more cameras capable of capturing the vehicle's surroundings. For example, this can be a mono or stereo camera mounted behind the vehicle's windshield. This allows the vehicle to visually perceive its surroundings from a similar perspective to that of the driver. By evaluating the corresponding camera images, the processing unit is then able to assess the surroundings.

[0021] When determining contextual information from camera images, it can happen that consecutive camera images are blurred. This results in discrepancies between consecutive raw images from a camera, which could lead to the music intended to be played in the vehicle interior being adjusted even though the driving situation remains the same. To prevent this, the camera images are stabilized. This neutralizes blur and prevents inappropriate changes to the music.

[0022] An advantageous development of the method according to the invention further provides that at least one further piece of context information describes one of the following contexts: - a vehicle parameter; - a driving manoeuvre performed by the vehicle; - a traffic manoeuvre performed by another road user; - weather conditions; or - a volume of traffic.

[0023] By taking the listed contexts into account, the current driving situation can be described in particularly differentiated detail. This makes it possible to adapt the music played in the vehicle interior to a wide variety of boundary conditions. Vehicle parameters can be recorded by the computing unit itself. For this purpose, information transmitted from control units can be read via a vehicle field bus, such as a CAN bus or an Ethernet data line. For example, tire pressure, oil temperature, engine speed, wheel speed, the engaged gear, the vehicle's speed, the outside temperature, and the like can be recorded by the computing unit. The vehicle parameters themselves or information derived from the vehicle parameters, such as the risk of slippery conditions when outside temperatures are measured around freezing, can then be incorporated into the music description information accordingly.

[0024] The vehicle can also be capable of detecting and classifying completed driving maneuvers. For this purpose, vehicle parameters, among others, can be evaluated. For example, a brake pedal position, engine power demand, steering wheel position, and the like can be recorded and taken into account by the computing unit. For example, acceleration sensors included in the vehicle can also provide corresponding measured values that can be used to reconstruct driving maneuvers performed by the vehicle. This can, for example, enable the computing unit to identify fast cornering, an overtaking maneuver, an emergency braking maneuver, and the like. For example, relaxing music can be played while driving on a straight stretch at a constant speed, or exciting music can be played during a driving maneuver that requires a high level of reaction from the driver, such as an overtaking maneuver or emergency braking maneuver.

[0025] The behavior of other road users such as vehicles driving ahead or following, cyclists, or even pedestrians can also be taken into account to adapt the music played in the vehicle. For example, the music can be changed or special tones or sequences of tones, particularly played by a specific musical instrument, can be incorporated into the music played in the vehicle if a certain road user performs a certain traffic maneuver, for example if a pedestrian suddenly runs across the road, a cyclist runs a red light, the driver's own vehicle is cut off, or the like. For example, a certain electric guitar riff could be played in the music if the driver's own vehicle is stopped at a roundabout and another road user passes the driver's own vehicle in the roundabout.

[0026] This also makes the driver more aware of the behavior of other road users, which can also improve road safety.

[0027] Weather conditions and / or traffic volume can also be taken into account to adjust the music. For example, different music can be played when it's raining than when it's sunny. In particular, slow and soothing music can be played when there's a traffic jam or slow traffic, while music with a fast rhythm can be played when the road is clear.

[0028] It would also be conceivable to monitor the emotional state, level of distraction, and / or fatigue of vehicle occupants, for example, using camera-based gaze or blink frequency monitoring and / or taking into account vital parameters recorded using vital sensors. For example, it could also be detected that the driver is agitated, and appropriately soothing music could be played in the vehicle interior. This could prevent the driver from performing heated driving maneuvers, which could increase the risk of an accident.

[0029] The processing unit is capable of collecting the relevant context information and converting it into a character string. For example, if the speed of the drive motor is 6000 rpm, the processing unit can generate a text block such as: "The speed of the motor is 6000 rpm." To adapt the music accordingly, specific commands are incorporated into the music description information for the text-to-music model, such as: "Overlay the background music for the next 2 seconds with a chord XY played on an instrument AB." The processing unit can be programmed to translate between raw information and raw commands. For example, a motor speed of 6000 rpm can result in a command to the text-to-music model such as: "Play a generic drum and bass melody with a beat count of 150 BPM."

[0030] According to a further advantageous embodiment of the method according to the invention, the computing unit reads information from a digital road map of a navigation unit to determine the context of the environment. Thus, various methods are available for generating corresponding context information of the environment. In addition, the computing unit can also be in communication with a navigation unit. With the help of positioning means, for example, a transceiver for receiving signals transmitted by a global navigation satellite system, such as GPS, the vehicle can be located on the Earth's surface. The vehicle's position can thus be described using geocoordinates. The vehicle's position can then be compared with the digital road map.This allows the computing unit to determine, for example, whether the vehicle is traveling through a coastal region, a desert, a forest, a city center, or something similar. The respective road type, such as a play street, a highway, a country road, or something similar, can also be determined from the digital road map. Additional information, such as topographical information, can be stored in the digital road map. This allows the computing unit to also determine the geodetic altitude at which the vehicle is traveling.

[0031] A further advantageous embodiment of the method according to the invention further provides that the computing unit for generating a text describing the context of the environment: - applies machine vision methods to the camera images to detect and classify objects in the camera images, whereby the text describes a respective classified object; and / or - feeds respective camera images to an image-to-text model, which is trained to provide a textual description of a processed camera image.

[0032] Using machine vision methods, camera images can be processed in a differentiated manner, thus recognizing and deriving a wide variety of visual features. Objects can then be uniquely recognized and classified based on characteristic shapes, dimensions, and visual characteristics such as color, brightness, or contrast. For example, trucks can be distinguished from cars, a tree species can be identified, or the architectural style of a building can be determined. Corresponding algorithms can also be based on the use of artificial intelligence.

[0033] So-called image-to-text models are also known. These are appropriately trained AI models that can read a camera image and automatically generate text from it. See, for example, https: / / arxiv.org / pdf / 2205.14100.pdf.

[0034] According to a further advantageous embodiment of the method according to the invention, the computing unit for generating the music description information concatenates several pieces of context information or feeds the context information to a paraphrasing model trained to combine several individual text modules into a coherent text description. The music description information can thus correspond to a concatenation of context information, such as: "Sunny sky with little traffic." "Overtaking maneuver initiated by the driver. Driving on Federal Highway 3 across a field." "Vehicle settings set to sporty and engine speed set to 9000 rpm." "Target music style: alternative."

[0035] However, with the help of a properly trained paraphrasing model, this sequence of simple word strings can also be converted into continuous text. This can facilitate the processing of context information by the text-to-music model.

[0036] A further advantageous embodiment of the method according to the invention further provides that the computing unit stores at least one passage of the music to be played back in the vehicle as an audio file on a computer-readable storage medium, wherein the computing unit has write access to the computer-readable storage medium. This enables the vehicle occupants to re-experience sections of the music already played back in the vehicle. For example, the vehicle occupants may take a liking to the music in question and thus listen to it again later. For this purpose, a corresponding audio file is created on the computer-readable storage medium. The computer-readable storage medium can be integrated into the computing unit, for example embodied as a hard disk (HDD) or solid state drive (SSD). It can also be a portable storage medium such as an SD card or USB stick.It can also be a computing unit coupled to the computing unit, for example the memory of a mobile device coupled to the computing unit via Bluetooth, such as a smartphone, can be used.

[0037] To create the audio file, the music being played in the vehicle can be permanently recorded in a circular buffer. After the journey, the user can then select individual musical passages to be permanently saved. It would also be conceivable for the musical passages to be saved only when the vehicle occupant initiates a corresponding recording function in the vehicle, for example, by entering a corresponding control action via a touch-sensitive display, pressing a record button, or issuing a corresponding voice command.

[0038] According to a further advantageous embodiment of the method according to the invention, the computing unit generates the music description information again after an event occurs or after a specified period of time has elapsed and feeds it to the text-to-music model to adapt the music. This allows the music to adapt to the respective driving situation. The event can, for example, be a change in at least one context describing the respective driving situation. Additionally or alternatively, a specified period of time can be taken into account, such as every 15 seconds. This allows the music to be adapted even if the driving situation supposedly remains the same.

[0039] A further advantageous embodiment of the method according to the invention further provides that, in order to stabilize at least two consecutive camera images used to generate different music description information, the following steps are carried out: - Applying a corner detection algorithm to a first and a second camera image; - Extracting corner features for each corner; - comparing all corner features of the first camera image with all corner features of the second camera image; - as soon as two corner features are similar within a tolerance range: Mark these two corners as corresponding; and - Transform the second camera image into the coordinate system of the first camera image so that corresponding corners overlap.

[0040] To compare the corner features of the camera images, a distance measure such as a hemming distance can be taken into account.

[0041] According to a further advantageous embodiment of the method according to the invention, after determining corresponding corners, a RANSAC algorithm is applied to the first and second camera images to create an affine transformation between the first and second camera images. The computing unit transforms the second camera image into the coordinate system of the first camera image using the affine transformation. Thus, it may occur that corner features are considered to be corresponding, even though they actually are not. With the help of the RANSAC (Random Sample Consensus) algorithm, incorrectly assigned correspondences between corner features can be found and neutralized. For this purpose, point correspondences for respective corner features are collected, and valid inlier correspondences are searched for within them.

[0042] A further advantageous embodiment of the method according to the invention further provides that at least two consecutive camera images used to generate different music description information are filtered in time, wherein in particular the following step is carried out: - for each pixel of at least one channel of the camera images to be filtered, an average value is calculated for the brightness of corresponding pixels in the camera images.

[0043] Considering the generation of contextual information from camera images, an effect known as "temporal noise" can occur. This is a driving situation in which visual characteristics of camera images are constantly changing, such as alternating between shaded and sunny areas, for example, when driving through a forest or a tree-lined avenue. The generated text input can also vary significantly due to the significantly changed environmental influences, so that the music would change disproportionately even though the driving situation essentially remains unchanged. However, this can be reliably avoided by temporally filtering the camera images.

[0044] In this way, the pixels of the respective camera images can be smeared over time and only these smeared images can be used to analyze the environment. Each camera image can comprise several channels, for example a red, green and blue color channel. It is also conceivable for a corresponding camera to have an infrared channel. When a camera image is recorded, an individual brightness value is assigned to a respective pixel for each channel. For this purpose, the respective pixels can be divided into so-called subpixels. If the camera image is recorded at time t, for example, the brightness value of the respective pixel for the times t, t-1, t-2, ... tn is used for the respective channel and an average is calculated from this. Depending on the image acquisition frequency of the camera and the time period to be observed, any number of past camera images can be taken into account to calculate the corresponding average.For example, the time period to be considered may be five seconds, which, at a frame rate of 60 frames per second, results in 300 camera images.

[0045] Preferably, the individual brightness values of each pixel are multiplied by a weighting factor, with a smaller weighting factor being used for older camera images and a higher weighting factor for more recent camera images. This increases the influence of recently generated camera images on detecting a changing driving situation and reduces it for older camera images. If the driving situation does change, this can be detected more quickly, and the music can be adjusted accordingly.

[0046] A further advantageous embodiment of the method according to the invention further provides that the computing unit performs a polynomial regression over a series of measurements of at least one vehicle parameter measured by the vehicle, in particular by applying a Savitzky-Golay filter. By applying a polynomial regression, measurements can be smoothed. With an n-th order filter, at least n-1 support points are used for the regression. The advantage of a Savitzky-Golay filter is that, compared to other filters, high-frequency components are not simply removed but are also included in the calculation. Conventional methods such as calculating a moving average tend to flatten, shift, or distort the result, whereas a Savitzky-Golay filter preserves certain properties of the measured value distribution, such as relative maxima, minima, and / or measured value scatter.

[0047] For example, the suspension travel of a shock absorber can be considered as a vehicle parameter. When driving over a bumpy road, it wouldn't be desirable to adjust the music for every vibration. However, if a particularly large bump is encountered, it might be beneficial to react by adjusting the music. For example, a drum roll could be integrated into the music when driving over the high bump.

[0048] Further advantageous embodiments of the method according to the invention for outputting music also emerge from the exemplary embodiment which is described in more detail below with reference to the figure.

[0049] This shows Fig. 1 is a schematic representation of the system components used to provide a method according to the invention for outputting music in a vehicle.

[0050] A method according to the invention serves to synthetically generate music appropriate to a current driving situation and to output it via the vehicle's internal loudspeakers 5 (not shown in detail). The music is synthetically generated by a text-to-music model 1 based on music description information 2. By outputting music appropriate to the current driving situation, the user interaction of the vehicle occupants with the vehicle can be improved.

[0051] To generate the music description information 2, an in-vehicle computing unit 3 aggregates information describing the respective driving situation. For this purpose, the computing unit 3 can be connected, for example, to at least one camera 6 configured to capture the vehicle's surroundings, a navigation unit 7, a control unit 8 of a vehicle subsystem, and / or a telecommunications unit 9. Corresponding information can thus be supplied to the computing unit 3 via a field bus 10 of the vehicle.

[0052] Thus, the computing unit 3 processes camera images generated by the camera(s) 6 and optionally reads information from a digital road map of the navigation unit 7, queries vehicle parameters from corresponding control units 8, and / or obtains information from external sources via the telecommunications unit 9, such as a weather report, traffic information, and the like.

[0053] The processing unit 3 has a text generation module 11 for generating context information 4 from the respective information. These are short text modules that describe a respective context. This context information 4 is then combined by the processing unit 3 to form the music description information 2. For this purpose, the individual pieces of context information 4 can be simply arranged one after the other or combined into a coherent text using a paraphrasing model.

[0054] The respective music description information 2 is then fed to the text-to-music model 1 to generate the artificial music. Fig.In the embodiment shown in Figure 1, the text-to-music model 1 is executed on a vehicle-external computing device 12, for example in the form of a cloud server. In particular, this is a high-performance computing cluster. The text-to-music model 1 could generally also be executed on an in-vehicle computing unit, such as the computing unit 3. Communication between the computing unit 3 and the vehicle-external computing device 12 could also take place via the telecommunications unit 9. This allows the artificially generated music to be "streamed" back into the vehicle and output via the vehicle's internal loudspeakers 5.

[0055] Should the communication connection between the computing unit 3 and the computing device 12 fail, the computing unit 3 could also store "emergency music" that was generated once or can be updated by the text-to-music model 1 on the computing device 12 and transmitted to the computing unit 3 for permanent storage. The emergency music is then played via the loudspeakers 5, thereby informing the vehicle occupants that there is currently no communication connection to the computing device 12. The failure or restoration of the communication connection can also be considered a change in the driving situation.

Claims

[1] A method for outputting music, wherein an internal processing unit (3) of a vehicle generates music description information (2), wherein the processing unit (3) uses at least one piece of context information (4) to generate the music description information (2), which piece of context information describes a context of a current driving situation, wherein at least one piece of context information (4) describes the following context: an environment traveled by the vehicle; wherein the processing unit (3) evaluates camera images recorded by a camera (6) of the vehicle to determine the context of the environment, and wherein the music is output via vehicle-internal loudspeakers (5), characterized by , that the music is synthetically generated by a text-to-music model (1) by reading in the music description information (2); and at least two consecutive camera images used to generate different music description information (2) are stabilized. [2] Method according to claim 1, characterized by that at least one further piece of context information (4) describes one of the following contexts: - a vehicle parameter; - a driving manoeuvre performed by the vehicle; - a traffic manoeuvre performed by another road user; - weather conditions; or - a volume of traffic. [3] Method according to claim 2, characterized by that the computing unit (3) additionally reads information from a digital road map of a navigation unit (7) to determine a context of the environment. [4] Method according to one of claims 1 to 3, characterized by that the computing unit (3) for generating a text describing the context of the environment: - applies machine vision methods to the camera images to detect and classify objects in the camera images, whereby the text describes a respective classified object; and / or - feeds respective camera images to an image-to-text model, which is trained to provide a textual description of a processed camera image. [5] Method according to one of claims 1 to 4, characterized by that the computing unit (3) for generating the music description information (2) strings together several pieces of context information (4) or feeds the context information (4) to a paraphrasing model which is trained to combine several individual text modules into a coherent text description. [6] Method according to one of claims 1 to 5, characterized bythat the computing unit (3) stores at least one passage of the music to be played back in the vehicle as an audio file in a computer-readable storage medium, wherein the computing unit has write access to the computer-readable storage medium. [7] Method according to one of claims 1 to 6, characterized by that the computing unit (3) generates the music description information (2) again after the occurrence of an event or after the expiration of a specified period of time and feeds it to the text-to-music model (1). [8] Method according to one of claims 1 to 7, characterized by that in order to stabilize at least two consecutive camera images used to generate different music description information (2), the following steps are carried out: - Applying a corner detection algorithm to a first and a second camera image; - Extracting corner features for each corner; - comparing all corner features of the first camera image with all corner features of the second camera image; - As soon as two corner features are similar within a tolerance range: Mark these two corners as corresponding; and - Transform the second camera image into the coordinate system of the first camera image so that corresponding corners overlap. [9] Method according to claim 8, characterized by that after determining corresponding corners, a RANSAC algorithm is applied to the first and second camera images to produce an affine transformation between the first and second camera images, wherein the computing unit transforms the second camera image into the coordinate system of the first camera image using the affine transformation. [10] Method according to one of claims 1 to 9, characterized bythat at least two consecutive camera images used to generate different music description information (2) are filtered in time, wherein in particular the following step is carried out: - for each pixel of at least one channel of the camera images to be filtered, an average value is formed for the brightness of corresponding pixels in the camera images. [11] Method according to claim 10, characterized by that the individual brightness values of each pixel are multiplied by a weighting factor, whereby a smaller weighting factor is used for older camera images and a higher weighting factor is used for newer camera images. [12] Method according to one of claims 2 to 11, characterized by that the computing unit (3) carries out a polynomial regression over a series of measurements of at least one vehicle parameter collected by the vehicle, in particular by applying a Savitzky-Golay filter.

Images