DEVICE AND METHOD FOR DETERMINING THE DISTANCE OF A LIGHT SIGNAL GENERATING ELEMENT

Description

[0001] The invention relates to a device and a corresponding method for determining the distance of a light signal generator from a motor vehicle.

[0002] A vehicle can have one or more driving functions that assist the driver in controlling the vehicle, particularly in longitudinal and / or lateral control. An example of a driving function to assist with longitudinal control is Adaptive Cruise Control (ACC), which can be used to maintain a set speed and / or a predetermined distance from a vehicle ahead. This driving function can also be used in conjunction with a traffic signal (especially a traffic light) at a traffic junction (such as an intersection) to initiate automated longitudinal control, such as automated deceleration, at the traffic signal.

[0003] In the context of automated longitudinal guidance at a light signal transmitter, an estimated value of the distance of the light signal transmitter from the vehicle is typically determined, e.g. to determine the extent of the automated deceleration and / or to determine the initial time or position for the automated deceleration.

[0004] From US patent 2021 / 261152 A1, a method for operating a vehicle is known in which camera systems are used to detect traffic lights and other objects in the vicinity of the vehicle and to determine a driving trajectory of the vehicle based on this.

[0005] From DE 10 2015 216 979 A1 a method for operating a driver assistance system in an intersection situation is known, wherein a traffic scenario with relevant objects (e.g. traffic signal system) is detected using environmental sensors and the vehicle is located in the traffic scenario.

[0006] DE 10 2013 019550 B3 describes a method for driver assistance with regard to traffic light sequencing. DE 10 2013 001018 A1 describes a method for operating a vehicle in which the distance of the vehicle to a traffic signal is determined.

[0007] This document addresses the technical task of efficiently and precisely determining the estimated distance of a traffic signal transmitter in a traffic signal system, particularly to improve the quality of a driving function for automated longitudinal guidance at the traffic signal system.

[0008] The problem is solved by each of the independent claims. Advantageous embodiments are described, inter alia, in the dependent claims. It should be noted that additional features of a claim dependent on an independent claim, without the features of the independent claim itself or only in combination with a subset of the features of the independent claim, can constitute a separate invention independent of the combination of all features of the independent claim, which can be made the subject of an independent claim, a divisional application, or a subsequent application. This applies equally to technical teachings described in the description, which can constitute an invention independent of the features of the independent claims.

[0009] According to one aspect, a device for determining an estimated distance between a traffic signal and a motor vehicle is described. The traffic signal can be positioned in front of the vehicle in the direction of travel, and the vehicle can be approaching the traffic signal at a certain speed (e.g., greater than 0 km / h, in particular greater than 30 km / h). The traffic signal can be part of a traffic signal system that includes one or more traffic signal heads (e.g., for one or more different directions of travel). The traffic signal can be located at a traffic junction, in particular at an intersection. Furthermore, the traffic signal typically has one or more light signals, each of which can be individually activated or deactivated. The different light signals can have different colors (e.g., red, yellow, or green). The traffic signal can be a traffic light.

[0010] The device is designed to detect a number of objects in the vicinity of a traffic signal head positioned in front of the vehicle, based on image data from a camera mounted on the vehicle. Furthermore, the traffic signal head itself can also be detected based on the image data. The image data can comprise a sequence of temporally successive images. The individual images can be analyzed (e.g., using an image analysis method, possibly machine-learned) to identify one or more objects in the immediate vicinity of the traffic signal head. Object information can then be determined for each individual object. This object information can include, for example: the object's position (relative to the traffic signal head), its dimensions (e.g., width and / or height), and / or its type.

[0011] Examples of objects include: a road marking, in particular a stop line, in the vicinity of the signal head; a traffic sign in the vicinity of the signal head; a crossing lane at the intersection where the signal head is located; a vehicle stopped at the signal head; and / or a (horizontal or vertical) mast to which the signal head is attached.

[0012] For example, based on the image data, an image can be determined from the sequence of images that (superimposed on the image captured by the camera) contains a set of bounding boxes for the corresponding set of objects. Each bounding box can enclose one object. Furthermore, the object information for an object can be associated with its bounding box.

[0013] The device is further configured to assign at least one object from the set of objects to the signal transmitter. In particular, one or more objects from the set of objects that are (likely) closest to the signal transmitter (and thus at a similar distance from the vehicle) can be assigned. Alternatively or additionally, one or more objects that are larger and / or more visible than the signal transmitter can be assigned.

[0014] At least one object from the set of objects can be assigned to the signal generator, particularly using a machine-learned mapping unit. This mapping unit can comprise a trained, artificial neural network (especially a so-called deep neural network).

[0015] The matching unit can be trained using training data, where the training data (statistically) describes the target function of the matching unit. The target function can be designed to identify, from a set of detected objects located in the vicinity of a signal transmitter, the N objects that are each at a distance most similar to the distance of the signal transmitter. N can be, for example, 1, 2, 3, or 4.

[0016] The training data can consist of a large number of training datasets (e.g., 1000 or more, or 10000 or more). Each training dataset can contain input data (as input for the mapping unit) and target output data (which should be provided by the mapping unit at the output for the corresponding input data). The input data could, for example, include a set of objects in the vicinity of a signal transmitter that were detected based on image data. The target output data could, for example, indicate the one or more objects from the set of objects that should be used to determine the distance to the signal transmitter (e.g., because they each have a particularly similar distance to the distance to the signal transmitter).

[0017] The input data can, for example, consist of an image from a vehicle camera, in which the individual detected objects can be marked (e.g., as bounding boxes). Furthermore, the object information of each object can be passed to the assignment unit as input data. The assignment unit can then identify, as output data, the one or more objects from the set of objects in the vehicle camera image that are assigned to the signal transmitter. The assignment unit can thus classify the set of detected objects into the classes "assigned" and "unassigned".

[0018] The device is further configured to determine (particularly based on image data) an individual estimate of the distance of at least one assigned object from the vehicle. Thus, an individual estimate of the distance of each assigned object can be determined. This individual estimate of the distance of an object from the vehicle can be determined using a structure-from-motion method based on the image data, specifically on the sequence of images. Alternatively or additionally, an optical flow can be determined based on the sequence of images, and the individual estimate of the distance of each object from the vehicle can be determined based on this optical flow.

[0019] Therefore, an image analysis method can be used to determine, based on the image data, an individual estimate of the distance of the assigned object from the vehicle for each individual assigned object.

[0020] The device is further equipped to determine the estimated distance of the light signal generator based on the individual estimated distance of the at least one associated object from the vehicle.

[0021] As explained above, an associated object in the immediate vicinity of the signal transmitter is preferably larger and / or more visible than the signal transmitter itself. Consequently, the distance to the associated object can typically be determined with increased accuracy based on the image data from the vehicle camera. This, in turn, allows the distance to the signal transmitter to be determined with increased accuracy.

[0022] The device can be configured to determine an individual estimate of the distance between the light signal generator and the vehicle (particularly based on image data). This individual estimate can be determined, for example, using a structure-from-motion method based on the image data, especially on a sequence of images. Alternatively or additionally, the individual estimate of the distance between the light signal generator and the vehicle can be determined based on the optical flow.

[0023] The estimated distance of the light signal can then also be determined based on the individual estimated distance of the light signal from the vehicle (e.g., using an averaging and / or fusion method). This further increases the accuracy of the estimated distance to the light signal.

[0024] Individual distance estimates for one or more objects in the vicinity of the signal generator, as well as an individual distance estimate for the signal generator itself, can be determined. These individual estimates can then be fused using a (machine-trained) fusion unit to produce a single distance estimate for the signal generator. This fusion unit could, for example, be a trained neural network. The training can be performed using training data from a large number of datasets, where the training data (statistically) defines the objective function of the fusion unit. This objective function can be a fusion of the individual estimates that yields a particularly precise distance estimate for the signal generator. A dataset can contain individual estimates as input data for the fusion unit.Furthermore, the data set can display the estimated distance of the signal transmitter as output data for the fusion unit, which should be provided by the fusion unit based on the individual estimated values ​​of the corresponding input data.

[0025] The device can be configured to determine a single estimated distance of the assigned object from the light signal generator. This single estimated distance can be determined, for example, using a structure-from-motion method based on image data. Alternatively or additionally, the single estimated distance can be determined based on optical flux.

[0026] The estimated distance of the light signal can then also be determined based on the individual estimated distance of the associated object from the light signal. This further increases the accuracy of the estimated distance of the light signal.

[0027] The associated objects (or objects) may have a greater spatial distribution than the light signal generator. This can facilitate fusion with sensor data from one or more additional environmental sensors of the vehicle, thereby further increasing the accuracy of a single estimate of the associated object's distance and / or the accuracy of the estimated distance of the light signal generator.

[0028] The device can thus be configured to determine sensor data relating to an associated object using one or more additional environmental sensors of the vehicle, in particular a lidar sensor and / or a radar sensor. The individual estimated distance of the associated object from the vehicle can then be determined (also) based on the sensor data from the one or more environmental sensors. In particular, the sensor data from the one or more environmental sensors can be fused with the image data.

[0029] The device can be configured to automatically guide the vehicle longitudinally based on the estimated distance to the traffic signal. In particular, the timing and / or extent of automated deceleration of the vehicle can be determined and / or initiated based on this estimated distance. The increased accuracy of the estimated distance improves the quality of the automated longitudinal guidance.

[0030] Furthermore, the device can be configured to determine the signaling state of the traffic light (e.g., red or green), particularly based on image data. Depending on the signaling state, the vehicle can then be automatically decelerated (e.g., at red) to bring it to a standstill before the traffic light. Conversely, the vehicle can be automatically guided past the traffic light (e.g., at green). By taking the signaling state of the traffic light into account, the accuracy of the automated longitudinal guidance can be further improved.

[0031] According to another aspect, a (road) motor vehicle (in particular a passenger car or a truck or a bus or a motorcycle) is described that includes the device described in this document.

[0032] According to another aspect, a method for determining an estimated distance between a traffic signal and a motor vehicle is described. The method includes detecting, based on image data from a vehicle camera, a set of objects in the vicinity of the traffic signal located in front of the vehicle in the direction of travel, and assigning at least one object from this set to the traffic signal. Furthermore, the method includes determining, particularly based on the image data, an individual estimated distance of the assigned object from the vehicle. The method also includes determining the estimated distance of the traffic signal based on the individual estimated distance of the assigned object from the vehicle.

[0033] According to another aspect, a software (SW) program is described. The SW program can be configured to run on a processor (e.g., on a vehicle's control unit) and thereby execute the procedure described in this document.

[0034] Another aspect describes a storage medium. This storage medium can include a software program configured to run on a processor and thereby execute the procedure described in this document.

[0035] It should be noted that the methods, devices, and systems described in this document can be used both alone and in combination with other methods, devices, and systems described in this document. Furthermore, any aspect of the methods, devices, and systems described in this document can be combined with one another in a variety of ways. In particular, the features of the claims can be combined with one another in a variety of ways. Features listed in parentheses are to be understood as optional features.

[0036] The invention will now be described in more detail using exemplary embodiments. Figure 1 Examples of vehicle components; Figure 2 an exemplary traffic light system; Figure 3 an exemplary traffic situation; Figure 4 an exemplary intersection; and Figure 5A flowchart of an exemplary procedure for determining an estimated distance to a traffic signal system.

[0037] As stated at the outset, this document addresses the efficient and precise determination of the distance to a traffic signal positioned in front of a vehicle in the direction of travel. The distance is preferably determined solely based on image data from a camera mounted on the vehicle. The distances described in this document can refer to a distance along the vehicle's direction of travel. Alternatively or additionally, the distances between two entities can refer to the shortest distance between them.

[0038] Fig. 1Figure 1 shows exemplary components of a vehicle 100. The vehicle 100 comprises one or more environmental sensors 103 (e.g., one or more cameras, one or more radar sensors, one or more lidar sensors, one or more ultrasonic sensors, etc.) configured to acquire environmental data relating to the surroundings of the vehicle 100 (in particular, the surroundings in the direction of travel in front of the vehicle 100). Furthermore, the vehicle 100 comprises one or more actuators 102 configured to influence the longitudinal and / or lateral guidance of the vehicle 100. Examples of actuators 102 include: a braking system, a drive motor, a steering system, etc.

[0039] The (control) device 101 of the vehicle 100 can be configured to provide a driving function, in particular a driver assistance function, based on the sensor data from one or more environmental sensors 103 (i.e., based on the environmental data). For example, an obstacle on the vehicle 100's trajectory can be detected based on the sensor data. The control unit 101 can then activate one or more actuators 102 (e.g., the braking system) to automatically decelerate the vehicle 100 and thereby prevent a collision between the vehicle 100 and the obstacle.

[0040] Particularly within the context of automated longitudinal guidance of a vehicle 100, one or more traffic signal systems on the roadway or street being traveled by the vehicle 100 can be taken into account, in addition to a vehicle ahead. The status of a traffic signal system can be considered in particular, so that the vehicle 100 automatically decelerates at a red light relevant to its own (planned) direction of travel until it reaches the stop line of the traffic light and / or accelerates again when the traffic light turns green.

[0041] Fig. 2 shows an exemplary traffic signal system 200. The one in Fig. 2The illustrated traffic signal system 200 has four different signal heads 201, which are arranged at different positions at an approach to an intersection. The left signal head 201 has a left-pointing arrow 202, indicating that this signal head 201 applies to left-turning traffic. The two middle signal heads 201 have an upward-pointing arrow 202 (or no arrow 202), indicating that these two signal heads 201 apply to straight-ahead traffic. The individual traffic signals 203 of these two signal heads 201 form signal groups. Furthermore, the right signal head 201 has a right-pointing arrow 202, indicating that this signal head 201 applies to right-turning traffic.

[0042] Fig. 3Figure 1 shows an example of a vehicle 100 moving along a roadway towards a traffic signal 200. The one or more environmental sensors 103 of the vehicle 100 can be configured to acquire sensor data (especially image data) relating to the traffic signal 200. The sensor data can then be analyzed (e.g., using an image analysis algorithm) to determine the characteristics of one or more features of the traffic signal 200. In particular, the sensor data can be used to determine which signal element 201 of the traffic signal 200 is relevant for the (planned) direction of travel of the vehicle 100. Furthermore, the (signaling) state of the relevant signal element 201 (e.g., the color, such as red, yellow, or green) can be determined. Additionally, the distance 311 between the traffic signal 200 and the vehicle 100 can be determined.

[0043] The detection, classification, and / or positioning of light signal devices 201 can be performed based on image data from a camera 103. Methods based on "Structure from Motion" and / or the evaluation of optical flow in a sequence of images can be used. Furthermore, a detected light signal device 201 can be tracked over time using a Kalman filter. To determine an absolute distance value 311, heuristics based on the typical size of light signal devices 291 (e.g., a typical height and / or width) can be used. With a stereo camera system, the parallax between the two cameras of the stereo camera system can also be used for a (scaled) distance estimation.

[0044] The estimated distance 311 of a traffic signal 200, determined using the aforementioned methods, can exhibit a relatively high degree of inaccuracy, particularly at relatively large distances 311 (e.g., 80 meters or more) and / or relatively high driving speeds (e.g., 70 km / h). This inaccuracy may be so significant that a traffic signal 200 detected based on image data cannot be unambiguously assigned to a traffic signal mapped on a digital map, or may be incorrectly assigned to a mapped traffic signal. Furthermore, at relatively high driving speeds, a relatively long detection range of a signal 201 is typically required (e.g., up to 250 meters) to ensure reliable and / or convenient consideration of the signal 201 during automated longitudinal guidance. Consequently, the effectiveness of the vehicle's automated driving function 100 may be impaired.

[0045] Examples of reasons for the inaccuracy of a determined distance estimate could be: A relatively low resolution and / or a relatively small number of pixels of a light signal 203 of a light signal device 201 at relatively large distances 311. This can lead to a deviation in image evaluation of just one pixel having a relatively strong effect on distance estimation. A relatively low optical flux, especially if the light signal device 201 is located at the center of the image captured by a camera 103. An incorrect assumption regarding the size of a light signal device 201. A light signal device 201 is often positioned against the sky as a background, so that there are relatively few reference points in the image area of ​​the light signal device 201. An incorrect assumption regarding the size of a light signal device 201 due to effects such as lightsabers and / or blur (with a wet windshield, at dusk, or at night). In a situation with low ambient light, it may be necessary to...Only the active light signal 203 is recognized, and no longer the entire light signal generator 201. As a consequence, the available number of pixels of the light signal generator 201 is further reduced.

[0046] Fig. 4Figure 1 shows an example intersection 400 with a traffic signal system 200. The traffic signal system 200 can comprise several signal heads 201, each with one or more traffic lights 203, which may be assigned to different directions of travel. The vehicle 100 is positioned on an approach 410 to the intersection 400 and may be equipped to acquire environmental data relating to the area surrounding the vehicle 100. This environmental data (in particular, image data from a camera) can be displayed by the traffic signal system 200 (in particular, the one or more signal heads 201 and / or the one or more traffic lights 203 of the traffic signal system 200) at the approach 410 to the intersection 400, especially at the intersection itself. Furthermore, one or more additional objects 401, 402, 403, 404, 205 (in particular one or more landmarks) in the vicinity of the vehicle 100, especially at the intersection 400, can be displayed by the one or more environmental sensors 103 of the vehicle 100.Examples of objects are . a stop line 401 at the traffic signal system 200; a traffic sign 402 in the vicinity of the traffic signal system 200; a mast 205 on which the one or more signal heads 201 of the traffic signal system 200 are attached; the mast 205 being arranged vertically or horizontally; a lane 403 crossing the access road 410; and / or a vehicle 404 standing at the traffic signal system 200.

[0047] The (control) device 101 can be configured to assign one or more of the objects 401, 402, 403, 404, 205 detected on the basis of the environmental data (in particular on the basis of the image data) to the traffic signal system 200, in particular to the light signal head 201, for which an estimated distance value 311 is to be determined. In particular, the one or more objects that are located in the immediate vicinity of the traffic signal system 200 with a relatively high probability and / or that are located at a similar distance 411 from the vehicle 100 can be assigned.

[0048] The assignment of one or more of the detected objects 401, 402, 403, 404, 205 to the traffic signal system 200 can be achieved using a machine-trained assignment unit. The assignment unit can, for example, comprise a trained artificial neural network (in particular, a deep neural network).

[0049] To train the assignment unit, training data with a variety of training datasets can be used. A training dataset can contain as input data a list of objects 401, 402, 403, 404, 205 in the vicinity of a traffic signal 200. Furthermore, the training dataset can contain as target output data for the assignment unit a classification indicating which of the objects from the list of objects 401, 402, 403, 404, 205 should be assigned to the traffic signal 200.

[0050] The mapping unit can be trained using a learning algorithm (e.g., a backpropagation algorithm) to make the mapping unit exhibit the classification behavior described by the training data.

[0051] If necessary, an image from the surrounding camera 103 can be passed to the assignment unit as input data, whereby the one or more detected objects 401, 402, 403, 404, 205 in the vicinity of a traffic signal 200 are marked in the image (e.g., each as a bounding box). The assignment of the one or more objects can then be provided as output data.

[0052] The (control) device 101 can further be configured to determine object distance information for each of the one or more objects 401, 402, 403, 404, 205 that have been assigned to a detected traffic signal 200. The object distance information for an object 401 can include an (individual) estimated value of the distance 411 of the object 401 from the vehicle 100. The (individual) estimated value of the distance 411 can be determined based on environmental data, in particular on image data. The aforementioned methods, which are based on structure-from-motion analysis and / or optical flow analysis, can be used for this purpose. Alternatively or additionally, environmental data from one or more further environmental sensors 103 (such as a lidar sensor and / or a radar sensor) can be used.This allows for fusion with sensor data from one or more additional environmental sensors 103 (which is particularly possible with a relatively large object 401, and thus enables a particularly precise distance estimation).

[0053] In addition to an individually determined estimated distance 411 of signal transmitter 201, one or more individually determined estimated distances 411 of one or more objects 401, 402, 403, 404, 205 (assigned to signal transmitter 201) can be determined. Based on the majority of these individually determined estimated distances 411, the overall estimated distance 311 of signal transmitter 201 or signaling system 200 can then be determined with increased accuracy. For this purpose, a weighted average of the individual estimated distances 411 can be calculated, for example. Alternatively or additionally, a machine-trained fusion unit can be used to determine the overall estimated distance 311 based on the individual estimated distances 411.

[0054] This allows for context-based detection and assignment of objects (especially landmarks) to an intersection scene. A deep learning method can be used for this purpose. Objects, including one or more traffic signals, can be assigned to a location detected and classified as an intersection (400).

[0055] Based on the assigned objects, a fusion of distance measurements can be performed. The distance estimation can thus be supplemented by additional reference measurements (between vehicle 100 and object, and between object and traffic signal 200). By considering relatively large objects (with an increased number of pixels) and / or by considering one or more additional adjacent traffic signal 200, the accuracy of the estimated distance 311 of a traffic signal 200 can be increased. Considering a relatively large object, which may be located at the edge of an image, allows the use of a relatively large optical flow for determining the distance estimate, which in turn enables accelerated convergence of the Kalman filter. Examples of objects that can be considered are: traffic signs 402 (e.g.,Stop sign, priority sign) and / or directional sign; horizontal and vertical masts 205; another traffic signal system and / or another traffic signal head 201 at the traffic junction 400; a road marking (e.g. stop line 401); a stationary vehicle 404; a crossing lane 403; etc.

[0056] Fig. 5 Figure 500 shows a flowchart of an exemplary (possibly computer-implemented) procedure for determining an estimated distance 311 of a traffic signal 201 from a motor vehicle 100. The traffic signal 201 can be part of a traffic signal system 200 with one or more traffic signal units 201. The traffic signal 201 can have one or more light signals 203, each of which can be individually activated (so that light is emitted) or deactivated (so that no light is emitted).

[0057] Method 500 comprises the detection 501, based on image data from (at least) one camera 103 of the vehicle 100, of a set of objects 401, 402, 403, 404, 205 in the vicinity of the light signal device 201, which is arranged in front of the vehicle 100 in the direction of travel. For this purpose, an object recognition algorithm can be applied, which is configured to analyze the image data (which, for example, contains a temporal sequence of images) in order to detect objects. Object information can be determined for each object. Examples of object information include the object's position, size, and / or type.

[0058] Method 500 further comprises assigning 502 at least one object 401 from the set of objects 401, 402, 403, 404, 205 to the signal transmitter 201. The assignment 502 can be carried out based on the image data and / or on the object information relating to the individual objects 401, 402, 403, 404, 205. In particular, the assignment 502 can be carried out such that (if necessary, only) one or more objects 401 are assigned which have a relatively high probability (e.g., 50% or more) of a distance 411 from the vehicle 100 that deviates by less than a certain value (e.g., by 10% or less) from the distance 311 to be determined of the (light) signal transmitter 201.

[0059] Within the framework of the assignment 502, the one or more objects 401 from the set of objects 401, 402, 403, 404, 205 can thus be identified that have approximately the same distance 411 from the vehicle 100 as the signal transmitter 201. The assignment 502 can be carried out using a machine-learned assignment unit.

[0060] Method 500 further comprises determining 503, in particular based on image data, a single estimated distance 411 of the associated object 401 from the vehicle 100. The distance estimation can be performed using the optical flow in the temporal sequence of images and / or using a structure-from-motion method. Alternatively or additionally, the distance estimation can be performed based on sensor data from one or more additional environmental sensors 103 of the vehicle 100. In particular, fusion with sensor data from a lidar and / or radar sensor can be carried out.

[0061] Furthermore, the procedure 500 includes determining 504 the estimated value of the distance 311 of the light signal generator 201 based on the individual estimated value of the distance 411 of the associated object 401 from the vehicle 100.

[0062] The measures described in this document enable the distance of an upstream signal transmitter 201 to be determined efficiently and precisely. This makes it possible to provide a particularly reliable and robust driving function, in particular a driving function for automated longitudinal guidance, at a traffic junction 400.

Claims

1. A device (101) for determining an estimated value of the distance (311) of a light signal transmitter (201) from a motor vehicle (100); wherein the device (101) is configured to, • recognize a set of objects (401, 402, 403, 404, 205) in a surrounding area of the light signal transmitter (201) arranged in the direction of travel in front of the vehicle (100); • assign at least one object (401) from the set of objects (401, 402, 403, 404, 205) to the signal transmitter (201); • determine an individual estimated value of the distance (411) of the assigned object (401) from the vehicle (100); • determine, on the basis of the image data, an individual estimated value of the distance (411) of the light signal transmitter (201) from the vehicle (100); and • determine the estimated value of the distance (311) of the light signal transmitter (201) on the basis of the individual estimated value of the distance (411) of the assigned object (401) from the vehicle (100) and on the basis of the individual estimated value of the distance (411) of the light signal transmitter (201) from the vehicle (100); wherein the individual estimated value of the distance (411) of the assigned object (401) from the vehicle (100) and the individual estimated value of the distance (411) of the light signal transmitter (201) from the vehicle (100) are fused by a machine-learned fusion unit to form the estimated value of the distance (311) of the light signal transmitter (201).

2. The device (101) according to claim 1, wherein • the device (101) is configured to assign the at least one object (401) from the set of objects (401, 402, 403, 404, 205) to the signal transmitter (201) by means of a machine-learned assignment unit; and • the assignment unit in particular comprises a trained artificial neural network.

3. The device (101) according to any one of the preceding claims, wherein the device (101) is configured to determine the individual estimated value of the distance (411) of the assigned object (401) from the vehicle (100) by means of a structure-from-motion method on the basis of the image data.

4. The device (101) according to any one of the preceding claims, wherein the device (101) is configured to, • determine sensor data with respect to the assigned object (401) by means of one or more environment sensors (103) of the vehicle (100), in particular by means of a lidar sensor and / or by means of a radar sensor; and • determine the individual estimated value of the distance (411) of the assigned object (401) from the vehicle (100) on the basis of the sensor data of the one or more environment sensors (103).

5. The device (101) according to any one of the preceding claims, wherein • the image data comprise a sequence of temporally consecutive images; and • the device (101) is configured to, ∘ determine an optical flow on the basis of the sequence of images; and ∘ determine the individual estimated value of the distance (411) of the assigned object (401) from the vehicle (100) on the basis of the optical flow.

6. The device (101) according to one of the preceding claims, wherein the device (101) is configured to, • determine an individual estimated value of the distance of the assigned object (401) from the light signal transmitter (201); and • determine the estimated value of the distance (311) of the light signal transmitter (201) also on the basis of the individual estimated value of the distance of the assigned object (401) from the light signal transmitter (201).

7. The device (101) according to any one of the preceding claims, wherein the device (101) is configured to guide the vehicle (100) automatically in a longitudinal direction in dependence on the determined estimated value of the distance (311) of the light signal transmitter (201).

8. The device (101) according to claim 7, wherein the device (101) is configured to, • determine a signaling state of the light signal transmitter (201) on the basis of the image data; and • decelerate the vehicle (100) automatically in dependence on the signaling state in order to bring the vehicle (100) to a standstill in front of the light signal transmitter (201), or to guide it automatically past the light signal transmitter (201).

9. The device (101) according to any one of the preceding claims, wherein the set of objects (401, 402, 403, 404, 205) comprises one or more of, • a ground marking, in particular a stop line (401), in the surrounding area of the signal transmitter (201); • a traffic sign (402) in the surrounding area of the signal transmitter (201); • a crossing lane (403) at a junction (400) at which the signal transmitter (201) is arranged; • a vehicle (404) standing at the signal transmitter (201); and / or • a pole (205) to which the signal transmitter (201) is attached.

10. A method (500) for determining an estimated value of the distance (311) of a light signal transmitter (201) from a motor vehicle (100); wherein the method (500) comprises, • recognizing (501), on the basis of image data of a camera (103) of the vehicle (100), a set of objects (401, 402, 403, 404, 205) in a surrounding area of the light signal transmitter (201) arranged in the direction of travel in front of the vehicle (100); • assigning (502) at least one object (401) from the set of objects (401, 402, 403, 404, 205) to the signal transmitter (201); • determining (503), in particular on the basis of the image data, an individual estimated value of the distance (411) of the assigned object (401) from the vehicle (100); • determining, on the basis of the image data, an individual estimated value of the distance (411) of the light signal transmitter (201) from the vehicle (100); and • determining (504) the estimated value of the distance (311) of the light signal transmitter (201) on the basis of the individual estimated value of the distance (411) of the assigned object (401) from the vehicle (100) and on the basis of the individual estimated value of the distance (411) of the light signal transmitter (201) from the vehicle (100); wherein the individual estimated value of the distance (411) of the assigned object (401) from the vehicle (100) and the individual estimated value of the distance (411) of the light signal transmitter (201) from the vehicle (100) are fused by a machine-learned fusion unit to form the estimated value of the distance (311) of the light signal transmitter (201).