Method and device for determining the detection range of a sensor of a motor vehicle

Abstract

The invention provides a method for determining a detection range of a sensor mounted at a motor vehicle, the sensor scanning the surroundings of the motor vehicle. The method is characterized in that it comprises identifying, by means of the sensor, a predetermined object in the surroundings of the motor vehicle, determining a distance of the identified object to the motor vehicle, and determining the detection range on the basis of the determined distance.

Description

Technical Field

[0001] The present invention relates to a method for determining the detection range of a sensor installed in a motor vehicle (which scans the environment of the motor vehicle) and a motor vehicle designed to perform the method. Background Technology

[0002] Driver assistance systems, such as Highway Navigator, take over the longitudinal and lateral guidance of motor vehicles on highways at speeds up to 130 km / h in varying weather conditions, relying on reliable identification of objects in the vehicle's environment.

[0003] To this end, sensors are installed on vehicles that can scan their surroundings and identify predetermined objects within that environment. These sensors output a list of identified objects, where each object corresponds to a list of real-world objects identified by the respective sensor. In lidar and radar sensors, these objects typically include other road users, but due to their high reflectivity, they can also include guardrails or, especially in lidar sensors, traffic signs. Because cameras have far greater capabilities, they can also identify objects beyond road markings, traffic lights, tunnels, etc. Sensors typically provide a classification of the object type (e.g., pedestrians, cars, trucks, cyclists, signs, etc.) and sometimes even a probability distribution over predetermined object types to model ambiguities in various recognition algorithms, such as when a sensor is unsure whether an identified object is a cyclist or a motorcyclist but is very certain it is not a truck.

[0004] The problem here is, in particular, determining the sensor's detection range or field of view for different types or categories of objects, because, for example, under changing weather conditions, the rated sensor power, i.e., the maximum detection range that the sensor can provide, cannot always be expected.

[0005] Therefore, if sensor data based on a sensor at a certain place or location cannot identify an object, it may be because there is actually no object at that location, or the sensor may be unable to scan or observe the location due to hardware errors, software errors, and / or external influences such as weather and / or location obstruction, with the latter being the more critical case.

[0006] To overcome this situation, it is generally recommended to check whether the sensors can identify static objects in the motor vehicle environment that exist in the location according to map data and should be identified by the sensors at rated power.

[0007] This map-based approach reaches its limits, particularly in heavy traffic conditions, especially when real-world objects begin to obscure intended objects to be identified through map data, such as traffic signs. Such systems also depend on the environment having a sufficiently accurate variety of intended objects required for distance estimation. If such infrastructure elements or intended objects are sparse, the identification range cannot be determined with sufficient reliability. Furthermore, this system relies on the precise location of the intended objects within the map and on the map being up-to-date.

[0008] Typically, the field of view of a sensor is also determined based on statistical data derived offline from the labeled data.

[0009] However, this data-driven approach requires a large amount of labeled training data, which must include a good, distortion-free representation of the expected environment in which the sensor will be exposed. Such training data is difficult to obtain and label. Furthermore, there is the residual risk that the sensor may encounter situations outside the training data where the field of view can no longer be estimated.

[0010] Therefore, conventional sensors cannot reliably identify or determine their detection range based solely on the data they detect during operation. Summary of the Invention

[0011] In the context of the prior art, the object of the present invention is to provide an apparatus and a method that are respectively adapted to overcome at least the aforementioned disadvantages of the prior art.

[0012] In particular, it is designed to enable the actual field of view or detection range of a sensor to be determined during its operation based on the sensor's detection data.

[0013] This objective is achieved through the features of the independent claim. The dependent claims contain preferred modifications of the invention.

[0014] This objective is then achieved by a method for determining the detection range of a sensor installed at a motor vehicle, wherein the sensor scans the environment surrounding the motor vehicle.

[0015] One consideration is that the sensor is a camera, particularly a video camera. Detection range can also be used to determine the detection range of, for example, a sensor designed to be semi-circular, particularly with a 150° field of view and a specific detection range as its radius.

[0016] The method includes the step of identifying predetermined objects in the surrounding environment of a motor vehicle, particularly another vehicle (e.g., another motor vehicle) and / or pedestrians, using sensors.

[0017] In addition, the method also includes determining the distance from the identified object to the motor vehicle, and determining the detection range based on the determined distance.

[0018] It is also possible to determine whether the predetermined object is continuously and repeatedly identified by the sensor in the environment surrounding the vehicle, particularly without interruption within a predetermined time period. It is also possible to determine the detection range based on the distance to the predetermined object only when the predetermined object is continuously and repeatedly identified by the sensor in the environment surrounding the vehicle.

[0019] To determine the detection range, the distance of the predetermined object can be determined by the time point at which the predetermined object is first identified by the sensor in multiple detections.

[0020] In addition, the method may also include determining the road segments traveled by the motor vehicle since the first time the predetermined object was identified by the sensor in multiple instances.

[0021] Then, the detection range can be determined based on the distance between the predetermined object and the time point at which the predetermined object is first identified by the sensor in multiple tests, and the difference between the determined distance and the road segment traveled by the vehicle during that period.

[0022] In addition, the method may also include assigning the identified predetermined objects to one of a plurality of predetermined object categories, such as passenger cars, trucks, motorcycles and / or pedestrians.

[0023] Then, the detection range can be determined based on the determined distance of the predetermined object to the object category to which the identified predetermined object is assigned.

[0024] Then, the detection range can be determined based on the determined distance of the predetermined object and the corresponding conversion value, for the object category to which the identified predetermined object has not been assigned.

[0025] In addition, the (current) detection range can be determined based on a previously determined detection range, which was specifically determined using the methods described above prior to the detection range.

[0026] One approach is to determine the (current) detection range by using a weighted filter that assigns a higher weight to the determined detection range and a lower weight to the previously determined detection range.

[0027] The detection range of a sensor can be divided into multiple regions, particularly multiple horizontal regions, where the detection range of each region is determined based on the predetermined object detected within the corresponding region.

[0028] In other words, the above can be summarized as follows:

[0029] In the first step of the above method, data is collected from one or more sensors.

[0030] In the second step, all irrelevant identifications can be filtered out, that is, all identifications that cannot be sufficiently guaranteed by the sensor to be true or correct.

[0031] The criterion for this could be to filter out the following corresponding identification objects, namely those that are identified in a continuous dataset acquired by sensors in a quantity less than a predetermined number.

[0032] For each initial object identification, i.e., the first appearance of the object, a timestamp t0 and the object's relative position to the sensor or vehicle at time t0 can be determined. Then, after the same object has been identified N times consecutively, for example 9 times, the object can be used as a candidate at time t0+N to determine the detection range.

[0033] Once an object has been identified N times, especially if there is a list of multiple objects identified N times, the sensor's field of view or detection range can be calculated in the third step of the method.

[0034] Therefore, the first step is to determine when the corresponding object is first identified. It is advisable to consider not declaring the identification of the object at that point in time as valid, as the uncertainty regarding whether the object actually exists is too great at that time.

[0035] To determine the detection range, movement or road segments traversed between the relative positions at time point t0 and the current time point t0+N can be considered to obtain the new relative position of the object. The distance of the object identified at time point t0 minus the road segments traversed by the vehicle between time points t0 and t0+N is determined as the sensor's detection range for that specific type of object.

[0036] Therefore, so-called vehicle self-movement compensation can be performed. That is, at time point t0, a previously unidentified object is identified at its x,y position relative to the vehicle. It can now be determined whether this object is a real object or a so-called ghost object, i.e., a false positive. This determination requires several computation cycles (in this case, N cycles) and object tracking or tracing of the identified object during these cycles. However, the vehicle may continue to move during these computation cycles. This relative self-movement of the vehicle can now be subtracted from the relative x,y position of the object identified at time point t0. Therefore, the position of the first identified object is converted to its relative position relative to the vehicle at the current time point t0+N. For example, consider an object first identified at position (100,0) at time point t0. The object is confirmed after N cycles, during which the vehicle has moved forward 30m. Thus, the converted or recalculated relative position is at (70,0). Therefore, the detection range will be 70m.

[0037] Since multiple objects can be observed in parallel, the sensor's field of view, or detection range, can be horizontally divided into individual segments, and the detection range for each segment can be derived separately. If multiple object candidates fall into the same segment, the dataset can be reduced by applying a max operation to the list of candidates in the same segment.

[0038] Furthermore, the detection range can be determined for objects that do not belong to other categories, based on the detection range defined by the identified object. For this purpose, a model can be used to model the relationship between the sensor's ability to identify different object types. For example, if a sensor can see or identify a car at a distance of 200m, it can identify smaller objects within 150m, such as motorcyclists, and even smaller objects within only 70m, such as pedestrians.

[0039] This is advantageous because determining the pedestrian detection range cannot always assume there are enough pedestrians near motor vehicles, such as when traveling on a highway. Based on these models, if other object types, such as cars, are actually identified, the pedestrian detection range can be inferred.

[0040] In the fourth step of the algorithm, the new detection range or the currently determined detection range can be combined with the previously estimated detection range of the corresponding sensor. This can be achieved, for example, by using a weighted filter that assigns higher weight to new measurements and lower weight to existing measurements. These filters allow for soft variations in the detection range, and appropriate adjustments to the weighting parameters allow for faster or slower updates to the detection range, depending on the specific application of the sensor.

[0041] Along with determining the detection range, this method also allows for the determination of accuracy and information regarding the last update of the detection range. For example, the uncertainty of the determined detection range can be assessed using the weighted variance of the last scan of the sensor's detection range. Similarly, information about the time of the last update of the detection range provides a set of information for higher-level software, which can then determine the extent to which the determined detection range can be trusted. This can be crucial, for example, for adjusting parameters related to safe distances. For instance, the detection range is periodically used to calculate the maximum safe speed of a motor vehicle. It is conceivable that the maximum speed could be further reduced based on the estimated uncertainty of the detection range.

[0042] In addition, a motor vehicle, particularly a passenger car, is provided. The motor vehicle includes sensors and control devices connected to the sensors, the sensors being arranged and configured to scan the surrounding environment of the motor vehicle.

[0043] The control device is designed to determine the sensor's detection range using the method described above.

[0044] It is conceivable that the motor vehicle is at least partially automated, i.e., designed to at least temporarily and automatically take over the lateral and / or longitudinal guidance of the motor vehicle.

[0045] The same applies to the apparatus as described in the reference method, and vice versa. Attached Figure Description

[0046] The following will refer to Figure 1 Describe one implementation method.

[0047] Figure 1 A flowchart is shown for a method of determining the detection range of a sensor installed at a motor vehicle, the sensor scanning the environment surrounding the motor vehicle. Detailed Implementation

[0048] like Figure 1 As shown, the method for determining the detection range of a sensor installed on a motor vehicle basically includes four steps S1-S4, in which the sensor scans the surrounding environment of the motor vehicle.

[0049] The motor vehicle has a control device that is connected to a sensor in such a way that it can receive data from the sensor corresponding to the surrounding environment of the motor vehicle scanned by the sensor.

[0050] Based on this data, the control device is designed to perform the following method. It is conceivable that, after performing the following method, the control device outputs a control signal to the vehicle based on the sensor detection range determined by the method, for example, to control its speed.

[0051] In the first step S1 of the method, a predetermined object in the surrounding environment of the motor vehicle is identified by a sensor, which here has a camera or is implemented as a camera.

[0052] Then, in the first step S1 of the method, it will be determined whether the predetermined object is identified continuously and repeatedly, particularly without interruption, by the sensor in the environment surrounding the motor vehicle for a predetermined period of time.

[0053] Furthermore, at least once the intended object is identified, the distance between that object and the vehicle, and especially to the sensor, must be determined.

[0054] If this is the case, i.e., the sensor identifies the predetermined object multiple times, for example, nine times, then in the second step S2, the distance from the identified object to the motor vehicle is determined. This distance was determined in the first step S1 at the time when the predetermined object was first identified by the sensor.

[0055] In the third step S3 of the method, the road segment traveled by the motor vehicle since the time point when the predetermined object was first identified by the sensor in multiple attempts is determined.

[0056] In addition, the identified predefined objects are assigned to one of the multiple predefined object categories.

[0057] Then, in the third step S3, the detection range of the object category to which the object is assigned is determined based on the distance determined in the second step S2 and the road segment through which the vehicle has traveled.

[0058] In other words, the detection range is determined only when the sensor identifies the predetermined object multiple times consecutively near the vehicle, i.e., when there is a predetermined probability of actual identification, based on the distance between the predetermined object and the sensor at the time when the predetermined object is first identified.

[0059] More precisely, the detection range to which the identified predetermined object is assigned is determined based on the distance determined at the time point when the predetermined object is first identified by the sensor in multiple tests, and the difference between the determined road segments.

[0060] By using the corresponding conversion value or the corresponding conversion formula, the detection range can now be determined for other object categories to which the identified predetermined object has not been assigned, based on the determined distance of the predetermined object, and in particular, based on the detection range determined for the object category to which the object has been assigned.

[0061] It is also possible to determine or estimate the sensor's detection range, specifically horizontally, i.e., the 2D range, based on the sensor's determined detection range, i.e., the 1D value. Furthermore, it is possible to divide the detection range into multiple regions, specifically horizontally, where the detection range of each region is determined based on the predetermined object detected within the respective region.

[0062] In the fourth step S4, the detection range determined in the third step S3 (or multiple detection ranges determined in the third step S3 if there are multiple object categories) is combined with the previously determined detection range (or multiple previously determined detection ranges).

[0063] This combination can be achieved by using one or more weighted filters that assign higher weights to the determined detection range and lower weights to the previously determined detection range.

[0064] List of reference numerals

[0065] Steps of the S1–S4 method

Claims

1. A method for determining the detection range of a sensor installed at a motor vehicle, the sensor scanning the environment surrounding the motor vehicle, Its features are, The method includes: The sensor identifies predetermined objects in the surrounding environment of the motor vehicle. Determine the distance from the identified object to the motor vehicle. The sensor identifies the predetermined object multiple times consecutively in the surrounding environment of the motor vehicle. Determine the road segments traveled by the motor vehicle since the first time the predetermined object was identified by the sensor in a plurality of instances. The detection range of the sensor is determined based on the distance between the predetermined object and the determined road segments at the time point when the predetermined object is first identified by the sensor in multiple tests.

2. The method according to claim 1, Its features are, The method further includes: The identified predefined objects are assigned to one of several predefined object categories. The detection range is determined based on the distance determined by the predetermined object to which the identified predetermined object is assigned to the object category.

3. The method according to claim 2, Its features are: The detection range is determined based on the determined distance to the predetermined object and the corresponding conversion value, and the object category to which the identified predetermined object has not been assigned.

4. The method according to any one of claims 1 to 3, Its features are, The method further includes: The current detection range is also determined based on a previously determined detection range, which was determined prior to the detection range.

5. The method according to claim 4, Its features are: The current detection range is determined by using a weighted filter based on a combination of the determined detection range and a previously determined detection range, wherein the weighted filter assigns a higher weight to the determined detection range and a lower weight to the previously determined detection range.

6. The method according to any one of claims 1 to 3, Its features are, The detection range of the sensor is divided into multiple regions, wherein the detection range of each region is determined based on the predetermined object detected in the corresponding region.

7. The method according to claim 6, Its features are, The multiple regions refer to multiple horizontal regions.

8. A motor vehicle, The vehicle includes sensors and control devices connected to the sensors, the sensors being arranged and configured to scan the vehicle's surrounding environment. Its features are, The control device is designed to determine the detection range of the sensor by means of any one of claims 1 to 7.

Images