Apparatus and method for determining distance to traffic light

Abstract

To efficiently and accurately determine an estimate of the distance of a signal in a signal light facility. An apparatus for determining an estimate of a distance of a traffic light from a vehicle, the apparatus comprising: - based on image data of the camera of the vehicle 100, to recognize a certain number of objects 401, 402, 403, 404, 205 around a traffic light 201 located in front of the vehicle 100 in the direction of travel; - assigning at least one object 401 out of a quantity of objects 401, 402, 403, 404, 205 to a traffic light 201; - to determine an individual estimate of the distance 411 of the assigned object 401 from the vehicle 100; and determining an estimate of the distance of the traffic light 201 based on the individual estimates of the distance 411 of the assigned object 401 from the vehicle 100 It is composed.

Description

[Technical field]

[0001] The present invention relates to an apparatus and a corresponding method for determining the distance of a traffic light from a motor vehicle. [Background technology]

[0002] A vehicle may be equipped with one or more driving functions that assist the driver of the vehicle when driving (guiding) the vehicle, in particular when driving (guiding) longitudinally and / or laterally. An exemplary driving function that assists the longitudinal driving of the vehicle is an adaptive cruise control (ACC) function, which may be used to drive the vehicle longitudinally at a set driving speed and / or at a set target distance relative to a preceding vehicle traveling in front of the vehicle. Here, the driving function may also be used in connection with traffic lights (in particular traffic lights) at traffic junctions (e.g. intersections) to provide automated longitudinal driving, e.g. automated deceleration at the traffic lights.

[0003] In automated longitudinal maneuvering at traffic lights, an estimate of the distance of the traffic light from the vehicle is typically determined, for example, to set the magnitude of the automated deceleration and / or the starting time or starting position for the automated deceleration. Summary of the Invention [Problem to be solved by the invention]

[0004] The present specification relates to the technical problem of efficiently and accurately determining estimates for the distances of signals of a signal light installation, in particular in order to improve the quality of driving functions for automated longitudinal driving in the signal light installation. [Means for solving the problem]

[0005] The problem is solved by each independent claim. Advantageous embodiments are set out in particular in the dependent claims. It is to be pointed out that the additional features of the claims dependent on an independent claim may form an independent invention from the combination of all the features of the independent claim, either without the features of the independent claim or only in combination with some of the features of the independent claim, which may be made for the subject matter of an independent claim, a divisional application or a subsequent application. This also applies to technical suggestions given in the description, which may form an independent invention from one of the features of the independent claims.

[0006] According to one aspect, an apparatus is described for determining an estimate of a distance of a traffic light from a motor vehicle. The traffic light may be located ahead of the vehicle in the direction of travel, and the vehicle may move towards the traffic light at a predefined speed (e.g. greater than 0 km / h, in particular greater than 30 km / h). The traffic light may be part of a signal light installation with one or more traffic lights (e.g. for one or more different directions of travel). The traffic light may be located at a traffic junction, in particular at an intersection. Also, the traffic light typically has one or more light indications, each of which can be individually activated or deactivated. The different light indications may have different colors (e.g. red, yellow or green). The traffic light may be a traffic light.

[0007] The device is configured to recognize a quantity of objects around a traffic light arranged in front of the vehicle in the direction of travel based on image data of the vehicle's camera. Also based on the image data, the traffic light itself can be recognized. The image data can include a sequence of images consecutive in time. Each image can be analyzed (e.g. using (possibly machine-learned) image analysis techniques) to identify one or more objects in the direct surroundings of the traffic light. Object information can then be determined for each object. The object information for the object can include, for example, the position of the object (relative to the traffic light), the dimensions of the object (e.g. width and / or height) and / or the type of object.

[0008] Exemplary types of objects are road markings around the traffic lights, in particular stop lines, traffic signs around the traffic lights, intersecting lanes at the junction where the traffic lights are located, vehicles stopped at the traffic lights and / or the pole to which the traffic lights are fixed (horizontal or vertical).

[0009] For example, based on the image data, it is possible to identify images from a sequence of images having a number of bounding boxes (overlaid with the images detected by the camera) for a corresponding number of objects, where each bounding box may enclose one object, and to associate the bounding boxes for the objects with object information for the objects.

[0010] The device is also configured to assign at least one object of the number of objects to the traffic light, in particular the object or objects of the number of objects that are (presumably) closest to the traffic light (and thus at a similar distance from the vehicle), or alternatively or additionally, the object or objects that are larger and / or have better visibility than the traffic light can be assigned.

[0011] At least one object of the quantity of objects can be assigned to a traffic light, in particular by means of a machine-learned assignment unit, which can include a trained artificial neural network, in particular a so-called deep neural network.

[0012] The allocation unit may have been trained based on training data that (statistically) describes an objective function of the allocation unit, where the objective function may be adjusted to identify, among a quantity of known objects located around a traffic light, N objects that each have a distance closest to that of the traffic light, where N may be, for example, 1, 2, 3 or 4.

[0013] The training data may comprise a number of training data sets (e.g., 1000 or more or 10000 or more). Each training data set may then comprise input data (as input for the allocation unit) and target output data (to be provided by the allocation unit at the output for the corresponding input data). The input data may comprise, for example, a number of objects recognized based on the image data around the traffic light. The target output data may represent one or more objects of the number of objects used to determine the distance of the traffic light (e.g., because the objects have a particularly similar distance to the distance of the traffic light, respectively).

[0014] The input data may for example comprise a respective image of the vehicle camera, in which the recognized individual objects may be marked (for example respectively as bounding boxes). Furthermore, object information of the individual objects may be transmitted as input data to the allocation unit. As output data, one or more objects of a quantity of objects in the vehicle camera images that are assigned to a traffic light may be identified by the allocation unit. Thus, the allocation unit may classify the quantity of recognized objects into a classification of "assigned" and "not assigned".

[0015] The device is also configured to determine (in particular based on the image data) an individual estimate of the distance of at least one assigned object from the vehicle. Thus, for each assigned object, it is possible to determine a respective one individual estimate of the distance of each object. The individual estimate of the object's distance from the vehicle can then be determined based on the image data, in particular based on the sequence of images, based on a Structure-from-Motion method. Alternatively or additionally, an optical flow can be determined based on the sequence of images, and an individual estimate of the distance of each object from the vehicle can be determined based on the optical flow.

[0016] Thus, image analysis techniques can be used to determine, for each individually assigned object, a respective individual estimate of the assigned object's distance from the vehicle based on the image data.

[0017] The apparatus is also configured to determine an estimate of the distance of the traffic light based on the individual estimates of the distance of the at least one assigned object from the vehicle.

[0018] As mentioned above, the assigned objects in the direct surroundings of the traffic light are preferably, in some cases, larger and / or more visible than the traffic light. As a result, the distance of the assigned objects can be determined, typically with increased accuracy, based on the image data of the vehicle camera. On this basis, the distance of the traffic light can therefore be determined with increased accuracy.

[0019] The device may be configured to determine an individual estimate of the distance of the traffic light from the vehicle (in particular based on image data). The individual estimate of the distance of the traffic light from the vehicle may be determined based on image data, in particular based on a sequence of images, for example based on a Structure-from-Motion method. Alternatively or in addition, the individual estimate of the distance of the traffic light from the vehicle may be determined based on optical flow.

[0020] An estimate of the traffic light distance can then be determined based on the individual estimates of the traffic light distance from the vehicle (e.g., based on averaging and / or fusion techniques), thus further improving the accuracy of the traffic light distance estimate.

[0021] It is therefore possible to determine individual estimates of the distance of one or more objects in the surroundings of the traffic light and an individual estimate of the distance of the traffic light. The individual estimates can be fused to an estimate of the distance of the traffic light based on a (machine learned) fusion unit. The fusion unit can for example comprise a trained neural network. The learning can be based on training data comprising a number of training data sets, by which an objective function of the fusion unit is (statistically) described. The objective function can be a fusion of the individual estimates resulting in a particularly accurate estimate of the distance of the traffic light. A data set can have the individual estimates as input data of the fusion unit. The data set also represents an estimate of the distance of the traffic light as output data of the fusion unit, which estimate is to be provided by the fusion unit based on the individual estimates of the corresponding input data.

[0022] The device may be configured to determine an individual estimate of the distance of the assigned object from the traffic light. The individual estimate of the distance of the assigned object from the traffic light may be determined based on the image data, for example based on a Structure-from-Motion method. Alternatively or additionally, the individual estimate of the distance of the assigned object from the traffic light may be determined based on optical flow.

[0023] An estimate of the traffic light distance can then also be determined based on the individual estimates of the assigned object's distance from the traffic light, thus further improving the accuracy of the traffic light distance estimate.

[0024] The assigned object(s) may have a larger spatial extent than the traffic light, which may facilitate fusion with sensor data from one or more other surrounding sensors of the vehicle, which may further improve the accuracy of the individual estimates of the assigned object's distance and / or the estimate of the traffic light's distance.

[0025] The device can thus be configured to determine sensor data for the assigned object based on one or more further surrounding sensors of the vehicle, in particular based on a lidar sensor and / or based on a radar sensor. An individual estimate of the distance of the assigned object from the vehicle can then be (also) determined based on the sensor data of the one or more surrounding sensors. In particular, a fusion of the sensor data of the one or more surrounding sensors with image data can be performed.

[0026] The device can be configured to steer the vehicle longitudinally depending on the determined estimate of the distance of the traffic light, whereby in particular the time and / or magnitude of an automated deceleration of the vehicle can be determined and / or brought about depending on the determined estimate of the distance of the traffic light, with the quality of the automated longitudinal steering being improved by increased accuracy of the estimate.

[0027] Furthermore, the device can be configured to determine (in particular based on image data) the light status (e.g. red or green) of the traffic lights. The vehicle can then be automatically slowed down depending on the light status (e.g. if red) in order to bring the vehicle to a stop before the traffic light. On the other hand, the vehicle can be automatically passed through the traffic light (e.g. if green). By taking into account the light status of the traffic lights, the quality of the automated longitudinal driving can be further improved.

[0028] According to another aspect, there is described a (road vehicle) motor vehicle (in particular a car or a lorry or a bus or a motorcycle) comprising an apparatus as described herein.

[0029] According to another aspect, a method for determining an estimate of a distance of a traffic light from a motor vehicle is described. The method includes recognizing a quantity of objects around a traffic light located in front of the vehicle in a driving direction based on image data of a camera of the vehicle and assigning at least one object of the quantity of objects to the traffic light. The method further includes determining, inter alia, individual estimates of the distance of the assigned object from the vehicle based on the image data. The method also includes determining an estimate of the distance of the traffic light based on the individual estimates of the distance of the assigned object from the vehicle.

[0030] According to another aspect, a software (SW) program is described, the software program being configurable to execute in a processor (e.g., in a vehicle controller) and thereby to perform the methods described herein.

[0031] According to another aspect, a storage medium is described. The storage medium can include a software program configured to execute on a processor and thereby perform the methods described herein.

[0032] It should be noted that the methods, devices and systems described herein can be used alone or in combination with other methods, devices and systems described herein. Furthermore, aspects of the methods, devices and systems described herein can be combined with each other in various ways. In particular, the features of the claims can be combined in various ways. Also, features shown in parentheses should be understood to be optional features.

[0033] The present invention will be described in detail below with reference to examples. [Brief description of the drawings]

[0034] [Figure 1] FIG. 1 illustrates exemplary components of a vehicle. [Diagram 2] FIG. 1 illustrates an exemplary signal light installation. [Diagram 3] FIG. 2 illustrates an exemplary traffic situation. [Figure 4] FIG. 2 illustrates an exemplary junction. [Diagram 5] 4 is a flow chart of an example method for determining a distance estimate for a signal light fixture. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0035] As mentioned at the beginning, the present specification relates to an efficient and accurate determination of the distance of a signal light installation arranged in front of a vehicle in the direction of travel, where the distance should preferably be determined solely on the basis of image data of a camera of the vehicle. The distance described in the present specification may refer to the distance along the direction of travel of the vehicle. Alternatively or additionally, the distance between two entities may refer to the respective minimum distance between the two entities.

[0036] 1 shows exemplary components of a vehicle 100. The vehicle 100 includes one or more surrounding sensors 103 (e.g. one or more image cameras, one or more radar sensors, one or more lidar sensors, one or more ultrasonic sensors, etc.) configured to detect surrounding data related to the surroundings of the vehicle 100, in particular related to the surroundings in the driving direction ahead of the vehicle 100. Furthermore, the vehicle 100 includes one or more actuators 102 configured to influence the longitudinal and / or lateral driving of the vehicle 100. Exemplary actuators 102 are braking equipment, drive motors, steering, etc.

[0037] The (control) device 101 of the vehicle 100 can be configured to provide driving functions, in particular driver assistance functions, based on sensor data (i.e. based on ambient data) of one or more ambient sensors 103. For example, based on the sensor data it is possible to detect an obstacle in the driving trajectory of the vehicle 100. Based thereon, the control unit 101 can control one or more actuators 102 (e.g. braking equipment) in order to automatically slow down the vehicle 100 and thus avoid a collision between the vehicle 100 and the obstacle.

[0038] In particular, during automated longitudinal driving of the vehicle 100, in addition to the vehicle ahead, it is possible to take into account one or more signal light installations in the lane or road along which the vehicle 100 is travelling, in which case the current state of the signal light installation or traffic signal installation can in particular be taken into account, so that the vehicle 100 is automated to slow down to the traffic light stop line at a red light relevant to its (planned) direction of travel and / or to accelerate (possibly again) at a green light.

[0039] An exemplary signal light installation 200 is shown in Fig. 2. The signal light installation 200 shown in Fig. 2 comprises four different traffic lights 201, which are arranged at different positions on the approach to an intersection. The left traffic light 201 has an arrow 202 pointing to the left, thereby indicating that the traffic light 201 applies to left-turners (vehicles). The two middle traffic lights 201 have an arrow 202 pointing up (or no arrow), indicating that both traffic lights 201 apply to straight-through traffic. The individual light indications 203 of both traffic lights 201 form a signal group. Furthermore, the right traffic light 201 has an arrow 202 pointing to the right, thereby indicating that the traffic light 201 applies to right-turners (vehicles).

[0040] 3 exemplarily shows a vehicle 100 moving towards a signal light installation 200 on a travel path. One or more surrounding sensors 103 of the vehicle 100 can be configured to detect sensor data (particularly image data) with respect to the signal light installation 200. The sensor data can then be analysed to identify one or more characteristic features of the signal light installation 200. In particular, based on the sensor data it can be determined which traffic light 201 of the signal light installation 200 is associated with the (planned) travel direction of the vehicle 100. Furthermore, it is possible to detect the (signal indication) state (e.g. colour, such as red, yellow, green) of the associated traffic light 201. Besides, it is possible to determine the distance 311 of the signal light installation 200 from the vehicle 100.

[0041] The recognition, classification and / or localization of the traffic lights 201 can be based on image data of the camera 103. Methods based on "Structure from Motion" and / or evaluation of the optical flow in a sequence of images can be used. The tracking of the recognized traffic lights 201 over time can be based on a Kalman filter. To determine the absolute value of the distance 311, heuristics for the typical size of the traffic lights 291 (e.g. typical height and / or width) can be used. In a stereo camera system, the disparity between both cameras of the stereo camera system can be used for the (scaled) distance estimation.

[0042] The estimate of the distance 311 of the signal light installation 200 determined based on the above-mentioned technique may have a relatively large inaccuracy, especially if the distance 311 is relatively large (e.g. more than 80 meters) and / or if the driving speed is relatively high (e.g. 70 km / h). In some cases, the inaccuracy may be so large that the signal light installation 200 recognized based on the image data cannot be unambiguously assigned to the signal light installation depicted on the digital map, or is incorrectly assigned to the depicted signal light installation. Furthermore, a relatively large recognition range of the traffic light 201 (e.g. up to 250 meters) is typically required in the case of a relatively high driving speed in order to allow a comfortable and / or reliable consideration of the traffic light 201 in automated longitudinal driving. As a result, the quality of the automated driving function of the vehicle 100 may be affected.

[0043] Exemplary causes for the identified inaccuracies in the distance estimates may be: a relatively low resolution and / or a relatively small number of pixels of the light indication 203 of the traffic light 201 when the distance 311 is relatively large, which can lead to a deviation in the image evaluation for just one pixel leading to a relatively large effect on the distance estimation. - A relatively small optical flow, especially when the traffic light 201 is at the center point of the image detected by the camera 103. -Incorrect assumptions about the size of traffic light 201. Since the traffic light 201 is often positioned in the foreground sky as a background, only a relatively few reference points are located in the image area of ​​the traffic light 201. - Incorrect assumptions about the size of the traffic light 201 due to lightsaber and / or blurring effects (in the case of a wet windshield, at dusk or at night). In low ambient light situations, it may be possible that only the active light indications 203 can be recognized, but no longer the entire traffic light 201. As a result, the number of available pixels of the traffic light 201 is further reduced.

[0044] In Fig. 4, an exemplary junction 400 is shown with a signal light installation 200, which may include a number of traffic lights 201 each with one or more light indications 203, possibly assigned to different directions of travel. The vehicle 100 is on an approach 410 to the junction 400 and may be configured to detect surrounding data about the vehicle 100's surroundings. The surrounding data (particularly camera image data) may then represent the signal light installation 200 (particularly the traffic light or lights 201 and / or the light indication or indications 203 of the signal light installation 200) at the approach 410 to the junction 400, in particular the intersection. Furthermore, the surrounding data or data 103 of the vehicle 100 may represent one or more further objects 401, 402, 403, 404, 205 (particularly one or more landmarks) in the surroundings of the vehicle 100, in particular at the junction 400. Exemplary objects are: - a stop line 401 in the signal light installation 200; - traffic signs 402 around the signal light installation 200; - the signal light installation 200 has one or more signals 201 mounted on a fixed pole 205, which pole 205 can be arranged vertically or horizontally; - a lane 403 intersecting an access road 410; and / or A vehicle 404 stopped at the signal light installation 200.

[0045] The (control) device 101 can be configured to assign one or more of the recognized objects 401, 402, 403, 404, 205 based on the surrounding data (in particular based on image data) to the signal light installation 200, in particular to the traffic light 201, for which an estimate of the distance 311 should be determined. In particular, it is then possible to assign one or more objects which are in the immediate vicinity of the signal light installation 200 with a relatively high probability and / or have a similar distance 411 to the vehicle 100 with a relatively high probability.

[0046] The allocation of the recognized objects 401, 402, 403, 404, 205 to one or more signal light installations 200 can be based on a machine-learned allocation unit. The allocation unit can, for example, comprise a trained artificial neural network, in particular a deep neural network.

[0047] For the learning of the allocation unit, training data having a number of training data sets can be used. The training data set can have as input data a list of objects 401, 402, 403, 404, 205 in the surroundings of the signal light installation 200. The training data set can also have as target output data for the allocation unit a classification of which object from the list of objects 401, 402, 403, 404, 205 should be assigned to the signal light installation 200.

[0048] The allocation unit can be trained based on a learning algorithm (eg, based on a backpropagation algorithm) such that the allocation unit has the classification behavior described by the training data.

[0049] In some cases, images from the surrounding camera 103 can be communicated as input data to the assignment unit, in which one or more recognized objects 401, 402, 403, 404, 205 in the surroundings of the signal light installation 200 are characterized (e.g., each as a bounding box), and an assignment of the one or more objects can be provided as output data.

[0050] The (control) device 101 can also be configured to determine object distance information for each of one or more objects 401, 402, 403, 404, 205 assigned to the recognized signal light installation 200. The object distance information for the object 401 can then comprise an (individual) estimate of the distance 411 of the object 401 from the vehicle 100. The (individual) estimate of the distance 411 can then be determined based on ambient data, in particular on image data. The above-mentioned techniques based on structure from motion analysis and / or optical flow analysis can then be used. Alternatively or in addition, one or more further ambient sensors 103 (such as for example lidar and / or radar sensors) can be used. Thus, fusion with sensor data of one or more further ambient sensors 103 can be performed (this is possible in particular for relatively large objects 401, thus allowing a particularly accurate distance estimation).

[0051] It is therefore possible to determine (in addition to the individually determined (individual) estimate of the distance 311 of the traffic light 201) one or more individually determined (individual) estimates of the distance 411 of the objects 401, 402, 403, 404, 205 (assigned to the traffic light 201). Then, based on the individually determined individual estimates of the distance 411, it is possible to determine with higher accuracy an overall estimate of the distance 411 of the traffic light 201 or of the signal light installation 200. For this purpose, for example, the formation of a (weighted) average value of the individual estimates of the distance 411 can be performed. Alternatively or supplementarily, a (machine learned) fusion unit can be used to determine an overall estimate of the distance 311 based on the individual estimates of the distance.

[0052] Thus, context-based recognition and assignment of objects to intersection scenes can be performed. Deep learning methods can be used for this purpose. Objects, including one or more signal light installations 200, can then be assigned to locations detected and classified as intersections 400.

[0053] Based on the assigned object, a fusion of distance measurements can be performed. Thus, the distance estimation can be supplemented by additional reference measurements (between the vehicle 100 and the object, and between the object and the signal light installation 200). Consideration of a relatively large object (with an increased number of pixels) and / or consideration of one or more other adjacent traffic lights 200 can improve the accuracy of the estimate of the distance 311 of the traffic light 200. Consideration of a relatively large object, possibly located at the edge of the image, allows the use of a relatively large optical flow for the determination of the distance estimate, which again allows an accelerated convergence of the Kalman filter. Exemplary objects that can be considered are: traffic signs 402 (e.g. stop signs, priority signs) and / or directional signs; horizontal and vertical poles 205; other signal light installations and / or other traffic lights 201 at the traffic junction 400; road markings (e.g. stop lines 401); stopped vehicles 404; intersecting lanes 403; etc.

[0054] 5 illustrates a flow chart of an exemplary (possibly computer-implemented) method 500 for determining an estimate of a distance 311 of a traffic light 201 from a motor vehicle 100. The traffic light 201 may be part of a traffic light installation 200 having one or more traffic lights 201. The traffic light 201 may include one or more light indicators 203, each of which may be individually activated (to emit light) or deactivated (to not emit light).

[0055] The method 500 comprises recognizing 501, based on image data of (at least one) camera 103 of the vehicle 100, a certain number of objects 401, 402, 403, 404, 205 in the surroundings of a traffic light 201 arranged in front of the vehicle 100 in the driving direction. For this purpose, it is possible to apply an object recognition algorithm arranged to analyze the image data (e.g. comprising a temporal sequence of images) in order to recognize the objects. It is then possible to determine object information for each of the objects. Exemplary object information comprises the object position, the object size and / or the object type.

[0056] The method 500 further comprises assigning 502 at least one object 401 of the quantity of objects 401, 402, 403, 404, 205 to the traffic light 201. The assigning 502 can be performed based on image data and / or based on object information for each object 401, 402, 403, 404, 205. The assigning 502 can in particular be performed in such a way that one or more objects 401 (possibly only one) are assigned that have a distance 411 from the vehicle 100 with a relatively high probability (e.g. 50% or more) that differs from the to-be-determined distance 411 of the (light) traffic light 201 by only a predetermined value (e.g. 10% or less).

[0057] Thus, during the allocating 502, it is possible to identify one or more objects 401 of a quantity of objects 401, 402, 403, 404, 205 that have approximately the same distance 411 from the vehicle 100 as the traffic light 201. The allocating 502 can then be performed based on a machine-learned allocation unit.

[0058] The method 500 comprises determining 503 an individual estimate of the distance 411 of the assigned object 401 from the vehicle 100, in particular based on the image data. The distance estimation can be performed based on an optical flow in a temporal sequence of images and / or based on a Structure-from-Motion method. Alternatively or supplementarily, the distance estimation can be performed based on sensor data of one or more further surrounding sensors 103 of the vehicle 100. In particular, a fusion with sensor data of a LIDAR and / or a RADAR sensor can be performed.

[0059] Additionally, the method 500 includes determining 504 an estimate of the distance 311 of the traffic light 201 based on the individual estimates of the distance 411 of the assigned object 401 from the vehicle 100 .

[0060] The measures described herein make it possible to efficiently and precisely determine the distance of the traffic light 201 located ahead, which makes it possible to provide particularly reliable and stable driving functions at the traffic junction 400, in particular for automated longitudinal driving.

[0061] It should be noted that the invention is not limited to the embodiments shown, and in particular that the specification and drawings illustrate only the principles of the proposed methods, devices and systems.

Claims

1. An apparatus (101) for determining an estimate of a distance (311) of a traffic light (201) from a vehicle (100), the apparatus (101) comprising: - based on image data from the camera (103) of said vehicle (100), recognizing a certain number of objects (401, 402, 403, 404, 205) around said traffic light (201) located in front of said vehicle (100) in the direction of travel; - assigning at least one object (401) of a quantity of said objects (401, 402, 403, 404, 205) to said traffic light (201); - determining an individual estimate of the distance (411) of the assigned object (401) from the vehicle (100); and determining an estimate of the distance (311) of said traffic light (201) based on said individual estimates of the distance (411) of said assigned object (401) from said vehicle (100); 1. An apparatus comprising:

2. The device (101) - determining an individual estimate of the distance (411) of said traffic light (201) from said vehicle (100) based on said image data; and determining an estimate of the distance (311) of said traffic light (201) also based on said individual estimate of the distance (411) of said traffic light (201) from said vehicle (100); 2. The apparatus of claim 1, wherein the apparatus is configured to:

3. the device (101) is configured to assign at least one object (401) of a quantity of the objects (401, 402, 403, 404, 205) to the traffic light (201) based on a machine-learned assignment unit; and - the allocation unit includes a specially trained artificial neural network 3. Apparatus (101) according to claim 1 or 2.

4. The device (101) according to claim 1 or 2, characterized in that the device (101) is configured to determine the individual estimate of the distance (411) of the assigned object (401) from the vehicle (100) based on a structure-from-motion method based on image data.

5. The device (101) - determining sensor data for the assigned object (401) based on one or more ambient sensors (103) of the vehicle (100), in particular based on lidar and / or radar sensors; and - determining the individual estimate of the distance (411) of the assigned object (401) from the vehicle (100) based on the sensor data of the one or more surrounding sensors (103); 3. The device (101) according to claim 1 or 2, characterized in that it is configured

6. - said image data comprises a sequence of temporally consecutive images; said device (101) - determining an optical flow based on said sequence of images; and - determining, based on the optical flow, the individual estimate of the distance (411) of the assigned object (401) from the vehicle (100); 3. The device (101) according to claim 1 or 2, characterized in that it is configured

7. The device (101) - to determine an individual estimate of the distance of the assigned object (401) from the traffic light (201); and determining an estimate of the distance (311) of said traffic light (201) also based on said individual estimate of the distance of said assigned object (401) from said traffic light (201); 3. The device according to claim 1, wherein the device is configured as follows:

8. The device (101) according to claim 1 or 2, characterized in that the device (101) is configured for automated longitudinal driving of the vehicle (100) depending on the determined estimated value of the distance (311) of the traffic light (201).

9. The device (101) - determining the light display status of said traffic light (201) based on image data; and - to automatically slow down said vehicle (100) depending on said light display status in order to bring said vehicle (100) to a standstill before said traffic light (201) or to automatically pass through said traffic light (201); 9. The device (101) according to claim 8, characterized in that it is configured

10. A certain number of the objects (401, 402, 403, 404, 205) - road markings around said traffic lights (201), in particular stop lines (401); - traffic signs (402) around said traffic lights (201); - the lanes (403) intersecting at the junction (400) where said traffic light (201) is located; - a vehicle (404) stopped at said traffic light (201); and / or - a pole (205) to which said traffic light (201) is fixed; 3. The device (101) according to claim 1 or 2, characterized in that it comprises one or more of:

11. A method (500) for determining an estimate of a distance (311) of a traffic light (201) from a vehicle (100), the method (500) comprising: - recognizing (501) a number of objects (401, 402, 403, 404, 205) around the traffic light (201) located in front of the vehicle (100) in the direction of travel, based on image data from the camera (103) of the vehicle (100); - assigning (502) at least one object (401) of a quantity of said objects (401, 402, 403, 404, 205) to said traffic light (201); - determining (503) an individual estimate of the distance (411) of said assigned object (401) from said vehicle (100), in particular based on image data; and - determining (504) an estimate of the distance (311) of the traffic light (201) based on the individual estimates of the distance (411) of the assigned object (401) from the vehicle (100); A method comprising:

Images