Method for classifying objects in the environment of a vehicle as objects drivable under or on the carriageway, computing device and driver assistance system

Abstract

The invention relates to a method for classifying an object in the surroundings of a vehicle, comprising the following steps: receiving sensor data describing the environment from an environmental sensor of the vehicle; identifying an object in a region of a lane in which the vehicle is located from the sensor data; determining an object region on the lane associated with the object; associating a base point with the identified object from the sensor data; determining a height of the base point relative to a vehicle vertical direction of the vehicle; determining a lane height in the object region relative to a vehicle heading direction; wherein the lane height is determined under the assumption that the lane has a predetermined slope between a front region of the lane located in front of the vehicle and the object region; and classifying the object as an object drivable underneath if a difference between the lane height and the base point height exceeds a predetermined threshold.

Description

Technical Field

[0001] This invention relates to a method for classifying objects in a vehicle environment. It also relates to a computing device for a driver assistance system and a driver assistance system for a vehicle. Finally, it relates to a computer program. Background Technology

[0002] Modern vehicles include driver assistance systems that enable autonomous driving. These systems typically incorporate multiple environmental sensors to detect objects in the vehicle's environment. A key challenge in perception for autonomous driving is classifying the measurements or data from these sensors. Distinguishing between static obstacles (such as cargo left in a lane or the end of a traffic jam) and structures that can be driven underneath (such as road signs, speedometers, and traffic guidance systems) is particularly difficult.

[0003] Different methods are known from existing technologies to address this problem. For example, sensor data provided by a lidar sensor or radar sensor can be projected into a camera image. Here, object classification can be performed based on a corresponding object recognition algorithm in the camera image. Furthermore, the plausibility of the identified object traveling below with a moving object can be checked. For example, if the object moves to some extent through a structure in the sensor data.

[0004] DE 10 2017 112 939 A1 describes a radar device for a vehicle that determines an indication from a probabilistic module based on a probabilistic relationship between a stationary vehicle and an overlying object. For each detection range, known first and second associations are modeled in the probabilistic model, where the probabilistic relationship corresponds to derived parameters and detection ranges. The known first association correlates the parameters with the probability of a stationary vehicle, and the known second association correlates the parameters with the probability of an overlying object. The radar device determines a threshold for the calculated indication to determine whether the target is a stationary vehicle or an overlying object.

[0005] Furthermore, a sensor system for identifying bridge or tunnel entrances for vehicles is known from DE 10 2015 213 701 A1. The sensor system includes a lateral lidar sensor disposed on a first side of the vehicle, the lateral lidar sensor having a detection area covering the lateral environment of the vehicle. Here, the lateral lidar sensor is rotatably disposed about a vertical axis, such that the forward portion of the detection area of ​​the lateral lidar sensor in the driving direction detects a spatial region positioned above the front of the vehicle at a predetermined distance. Furthermore, the lateral lidar sensor is tilted relative to the horizontal line about a transverse axis, such that the forward portion of the detection area of ​​the lateral lidar sensor in the driving direction detects a spatial region located at a predetermined height above the vehicle. Summary of the Invention

[0006] The purpose of this invention is to provide a simple yet reliable solution for object classification, particularly for distinguishing between objects that can be driven below and important objects on the lane.

[0007] This objective is achieved by the method, computing device, driver assistance system, and computer program according to the invention. Advantageous improvements of the invention are described below.

[0008] The method according to the invention is used to classify objects in a vehicle environment. The method includes receiving sensor data describing the environment from environmental sensors of the vehicle. Furthermore, the method includes identifying objects in a region of the lane where the vehicle is located based on the sensor data, and determining an object region associated with the object in the lane. Furthermore, the method includes associating a base point with the identified object based on the sensor data, and determining the height of the base point relative to the vehicle's vertical direction. Furthermore, the method includes determining the lane height in the object region relative to the vehicle's vertical direction, wherein the lane height is determined assuming a predetermined slope between the lane's forward region located in front of the vehicle and the object region. Furthermore, the method includes classifying the object as an object that can be driven below if the difference between the lane height and the base point height exceeds a predetermined threshold.

[0009] This method allows for the classification of objects in a vehicle environment. Specifically, it distinguishes between objects that can be driven underneath by vehicles and important objects on the lane. Objects that can be driven underneath can be, for example, structural or infrastructure installations that extend beyond the lane. Examples of such objects include bridge-shaped road signs, speedometer displays, traffic guidance systems, bridge or tunnel entrances. Important objects on the lane can be, for example, goods or other traffic participants left on the lane. In particular, stationary or low-speed traffic participants can be considered important objects on the lane. For example, such important objects on the lane can be associated with the end of a congestion. This method is particularly suitable for objects whose distance from a vehicle exceeds a specific minimum distance, such as 50 meters.

[0010] This method can be executed using the vehicle's corresponding computing device. The computing device can receive sensor data from the vehicle's environmental sensors. Environmental sensors can be, for example, distance sensors, lidar sensors, or radar sensors. Environmental sensors can also be designed as cameras. The sensor data can include, for example, multiple measurements or measurement points describing the environment and objects within it. Based on the sensor data or measurements, the computing device identifies objects located in or on the lane area. Furthermore, the object area is associated with objects on the lane. The object area describes the area of ​​the lane associated with the object or the area where the object is located.

[0011] Furthermore, a base point is associated with the object based on sensor data. Here, it is not necessary to rigidly interpret the base point as a point on the object that also contacts the lane surface. Instead, the base point can be understood as the point or measurement of the object with the minimum distance from the lane surface. Therefore, the base point of the object is specifically the measurement of the object's lowest point relative to the vehicle's vertical direction or the vertical line. A height is determined for this base point, specifically relative to the vehicle's vertical direction.

[0012] Furthermore, according to the present invention, the lane height or ground height in the target area is determined. The lane height is also determined relative to the vertical direction of the vehicle or relative to the mounting position of the environmental sensors or a reference point of the vehicle. Here, the lane height is determined under the assumption that the lane is at a predetermined slope between the forward area and the target area in front of the vehicle. Here, the forward area describes the area in front of the vehicle in the forward travel direction. The height relative to the vertical direction of the vehicle can be determined from the forward area or is known. Within the scope of the present invention, the term "slope" can be understood as lane uphill and lane downhill. The lane slope is positive when uphill and negative when downhill. Slope can also be referred to as gradient. The lane height in the target area associated with the object is determined as follows: that is, assuming that the lane rises or falls between the forward area and the target area.

[0013] Furthermore, if the difference between the lane height and the base point height exceeds a predetermined threshold, the object is classified as a vehicle that can travel underneath. Therefore, if the base point height of an object is significantly higher than the assumed lane height in the object area, the object is determined to be a vehicle that can travel underneath. This is done under the previously described assumption that the lane height in the object area is greater than the lane height in the preceding area or the lane area, where the vehicle is currently in the preceding area or the lane area. Therefore, real congestion ends or real static obstacles can be reliably classified, and braking can be applied due to correct classification. In this way, the false alarm rate for classifying important objects in the lane can be reduced, thereby reducing the false braking rate. Therefore, overall, object classification can be performed in a simple and reliable manner, in terms of false alarm rate or unidentified real objects in the lane.

[0014] Preferably, if the difference between the lane height and the base point height is less than a predetermined threshold, the object is preferably classified as an important object in the lane. If the height of the object's base point is only slightly different from the assumed lane height, or if the height of the base point is lower than the determined lane height, then the object is determined to be an important object in the lane. Specifically, the object is determined to be a static object or obstacle in the lane. This also reliably prevents important objects in the lane from being incorrectly classified as vehicles that can be driven underneath by assuming that the lane height in the object area is higher than the lane height in the area in front of the vehicle. This could lead to a collision between the vehicle and the important object in the worst case.

[0015] In one implementation, the slope is predetermined based on an assumption of a predetermined maximum gradient and / or a predetermined maximum curvature change of the lane. For example, it can be assumed that the lane height increases or decreases linearly between the area ahead and the target area. For example, a maximum gradient of the lane can be assumed. For example, the maximum gradient can be between +1% and +10%. The maximum gradient can be determined based on the lane's geographical location or geographical conditions in the environment. For example, the gradient in mountainous areas can be chosen to be greater than on flat ground.

[0016] However, the slope between the forward region and the object region can also be predetermined based on a pre-defined maximum curvature change. This, for example, means that the lane's orientation in the vehicle's vertical direction is not linearly extrapolated from the forward region, but rather extrapolated using a worst-case curvature assumption. Therefore, it is possible to reliably prevent objects that can be driven underneath from being incorrectly classified as important objects in the lane.

[0017] According to another embodiment, a predetermined slope is determined based on digital map data, wherein the digital map data describes the slope of the lane between the area ahead and the object area. The digital map data can be appended to or used in place of a predetermined maximum slope and / or a predetermined maximum curvature change. For example, information describing the lane slope or curvature can be extracted only from high-precision 3D map data or atlases. Therefore, in particular, no evaluation of the geometry of the high-precision 3D map data or individual lanes is required. This reduces the amount of data required to determine the lane slope or lane height. Classification can be further improved by considering map data describing the lane slope.

[0018] In principle, it can be assumed that the lane ascends between the area ahead and the target area. If map data is used to extrapolate the slope, and the map data describes the lane descending between the area ahead and the target area, then a downhill slope of the lane can be assumed at least locally. For a downhill lane, a linear slope, such as a slope between -1% and -10%, can be assumed, or a predetermined curvature change can be assumed.

[0019] Furthermore, it is advantageous to determine the threshold based on a predetermined maximum height of the lower edge of the object, which is detected based on sensor data. As mentioned earlier, the base point associated with the object may not describe the area where the object also contacts or touches the lane. For example, if the object is a passenger car or a truck, it may be based on measurements using lidar sensors and / or radar sensors, rather than detecting the vehicle's wheels or tires. For example, in the case of a truck, the underlying load-bearing beam can be identified based on sensor data or measurements and thus considered as the object's base point. The threshold is therefore determined such that it is greater than this lower edge height. Here, the maximum height of the detectable lower edge of a typical traffic participant or object can be used as a basis. Here, the threshold can correspond to the maximum height or can be selected to be greater than the predetermined maximum height. Therefore, misclassification can be prevented reliably.

[0020] Furthermore, it is specifically mentioned that uncertainties in the sensor data and / or tolerances in determining lane height are considered when determining the threshold. Uncertainties or tolerances exist in the sensor data or measurements when detecting the base point of an object. Additionally, tolerances exist when determining lane height based on a predetermined slope. These uncertainties or tolerances can be considered when determining the threshold. For example, the threshold can be determined based on the previously described maximum height of the object's lower edge and a height value to compensate for uncertainties and tolerances.

[0021] In another embodiment, to determine the lane height, the height of the lane surface relative to the vehicle's vertical direction in the forward region is determined based on sensor data. In other words, the height of the lane or lane surface in the forward region can be determined based on sensor data. In a specific forward region in front of the vehicle, the geometry of the ground or lane surface, especially its elevation changes, can be reliably detected using lidar or radar sensors. For example, the lane height in the forward region at a distance of up to 50 meters from the front of the vehicle can be determined using lidar or radar sensors.

[0022] The method according to the invention is particularly applicable to objects having a predetermined minimum distance from a vehicle or environmental sensor. For example, the minimum distance can be greater than 50 m. Specifically, the method can be used for objects outside the forward area. The minimum distance is related to the design of the environmental sensor, the sensor principle, the installation height of the environmental sensor, and / or environmental conditions.

[0023] The computing device for a driver assistance system according to the invention is designed to perform the method according to the invention and its advantageous design. For example, the computing device may include one or more controllers.

[0024] The driver assistance system for vehicles according to the invention is designed to operate the vehicle in a manner that is at least partially automated, based on the classification of objects in the environment. The driver assistance system includes a computing device according to the invention. Furthermore, the driver assistance system may have at least one environmental sensor. The environmental sensor may preferably be designed as a lidar sensor or a radar sensor. The environmental sensor may also be designed as a camera. By classifying objects as objects that can be driven below or important objects in the lane, the computing device can output corresponding control signals for the partially automated motor vehicle. For example, if an object is classified as an important object in the lane, braking can be performed.

[0025] The vehicle according to the invention includes a driver assistance system according to the invention. The vehicle is particularly designed as a passenger car.

[0026] Another aspect of the invention relates to a computer program comprising instructions that, when executed by a computing device, cause the computing device to perform the method according to the invention and its advantageous design. Furthermore, the invention relates to a computer-readable (storage) medium comprising instructions that, when executed by a computing device, cause the computing device to perform the method according to the invention and its advantageous design.

[0027] The preferred embodiments and advantages described with reference to the method according to the invention are accordingly applicable to computing devices according to the invention, driver assistance systems according to the invention, vehicles according to the invention, computer programs according to the invention, and computer-readable (storage) media according to the invention.

[0028] Other features of the invention are derived from the accompanying drawings and description. The features and combinations of features mentioned in the above description, as well as the features and combinations of features mentioned in the following description of the drawings and / or shown separately in the drawings, can be used not only in the combinations described separately, but also in different combinations or individually, without departing from the scope of the invention. Attached Figure Description

[0029] The invention will now be explained in more detail with reference to preferred embodiments and the accompanying drawings. Herein lies:

[0030] Figure 1 A schematic diagram of a vehicle is shown, including a driver assistance system for classifying objects as objects that can be driven below or important objects in a lane; and

[0031] Figure 2 It shows a diagram of vehicles in the lane, measurements of the objects being described, and the assumed lane direction.

[0032] Components that are identical or have the same function are given the same reference numerals in the accompanying drawings. Detailed Implementation

[0033] Figure 1 A schematic diagram of vehicle 1 (here, a passenger car) is shown in top view. Vehicle 1 includes a driver assistance system 2, by means of which vehicle 1 can be operated at least partially automatically. Driver assistance system 2 includes a computing device 3, which may be formed, for example, by at least one controller of vehicle 1. Furthermore, driver assistance system 2 includes an environmental sensor 4, which may be designed as, for example, a radar sensor or a lidar sensor. The environmental sensor 4 can provide sensor data describing the vehicle's environment 5. Sensor data can be transmitted from the environmental sensor 4 to the computing device 3.

[0034] Here, vehicle 1 is located in lane 6. Based on sensor data provided by environmental sensor 4, an object 7 currently in front of vehicle 1 in the direction of travel on lane 6 can be identified. For example, the distance between vehicle 1 and object 7, as well as their relative positions, can be determined based on the sensor data. Furthermore, an object region 8 is defined based on the sensor data, and this object region is associated with object 7 on lane 6. Additionally, lane 6 or its lane surface 9 can be detected in the forward region 10 in front of vehicle 1 in the direction of travel based on the sensor data. Specifically, uphill or downhill sections of lane 6 can be detected in the forward region 10 based on the sensor data from environmental sensor 4.

[0035] Figure 2 Another view showing the measurement 11 of vehicle 1 and object 7. Here, the sensor data provided by environmental sensor 4 includes the measurement 11. For clarity, only three measurement 11s are shown here. The figure does not show the distance between vehicle 1 and object 7 to scale. For example, the distance between vehicle 1 and object 7 could be 150m. In the case of such a distance, it is difficult to distinguish a static object (e.g., cargo left on lane 6 or a stationary end of a traffic jam) from a horizontal structure that can be driven from below (e.g., a bridge-shaped road sign or speedometer) when measured by environmental sensor 4 such as a lidar sensor or radar sensor.

[0036] The base point 12 of object 7 is determined based on measurement value 11. Here, base point 12 describes the point or measurement value 11 of object 7 with the minimum distance from the lane surface 9 relative to the vertical direction z of the vehicle. In other words, base point 12 describes the lowest point of object 7. As already explained, the slope of lane 6 or lane surface 9 in the forward region 10 can be detected based on sensor data. Here, the slope of lane 6 is extrapolated in region 13. Region 13 of lane 6 extends between the forward region 10 and the object region 8 associated with object 7. Here, the slope is calculated based on the worst-case assumption of slope or curvature change for lane 6. For example, the slope or curvature change can be assumed to be 2%. For example, a height difference of 2m can be derived over a distance of 100m. Thus, in the example, a lane height h1 is obtained in object region 8, which is 2m higher than the measured ground height or the height h0 of lane 6 in the forward region 10.

[0037] Now check whether the difference between the height h2 of base point 12 and the lane height h1 is greater than a predetermined threshold T. If so, object 7 is classified as an object that can be driven underneath. Otherwise, object 7 is classified as a significant object on lane 6. The threshold T can be determined based on a predetermined maximum height of the lower edge of the object that can be detected by means of sensor data. The height of the lower edge can, for example, correspond to the typical height of the load-bearing beam of a truck. Furthermore, uncertainty in sensor values ​​and / or tolerances in determining lane height h1 can be included when determining the threshold. Overall, objects 7 in the environment 5 of vehicle 1 or on lane 6 can be classified in a simple and reliable manner.

Claims

1. A method for classifying objects (7) in the environment (5) of a vehicle (1), comprising the following steps: Receive sensor data describing the environment (5) from the environmental sensor (4) of the vehicle (1); Based on the sensor data, identify objects (7) in the area of ​​the lane (6) where the vehicle (1) is located; Determine the object region (8) on the lane (6) that is associated with the object (7); The base point (12) is associated with the identified object (7) based on the sensor data; and Determine the height (h2) of the base point (12) relative to the vehicle vertical direction (z) of the vehicle (1). Its features are, Determine the lane height (h1) in the object area (8) relative to the vertical direction (z) of the vehicle; The lane height (h1) is determined by assuming that the lane (6) has a predetermined slope between the front area (10) of the lane (6) located in front of the vehicle (1) and the object area (8); and If the difference between the lane height (h1) and the height (h2) of the base point (12) exceeds a predetermined threshold (T), the object (7) is classified as an object that can travel below.

2. The method according to claim 1, Its features are, If the difference between the lane height (h1) and the height (h2) of the base point (12) is less than the predetermined threshold (T), then the object (7) is classified as an important object on the lane (6).

3. The method according to claim 1 or 2, Its features are, The slope is predetermined, assuming a predetermined maximum slope and / or a predetermined maximum curvature change of the lane (6).

4. The method according to claim 1 or 2, Its features are, The slope is predetermined based on digital map data, wherein the digital map data describes the slope of the lane (6) between the area ahead (10) and the object area (8).

5. The method according to claim 1 or 2, Its features are, The threshold (T) is determined based on a predetermined maximum height of the lower edge of the object, and the predetermined maximum height can be detected based on the sensor data.

6. The method according to claim 1 or 2, Its features are, The uncertainty in the sensor data and / or the tolerance in determining the lane height (h1) are also taken into account when determining the threshold (T).

7. The method according to claim 1 or 2, Its features are, In order to determine the lane height (h1), the height (h0) of the lane surface (9) of the lane (6) in the front area (10) relative to the vertical direction (z) of the vehicle is determined based on the sensor data.

8. A computing device (3) for a driver assistance system (2) for a vehicle (1), wherein the computing device (3) is designed to perform the method according to any one of claims 1 to 7.

9. A driver assistance system (2) for a vehicle (1) having a computing device (3) according to claim 8, wherein the driver assistance system (2) is designed to operate the vehicle (1) at least partially automatically based on the classification of objects (7) in the environment (5) of the vehicle (1).

10. A computer program product comprising instructions that, when executed by a computing device (3), cause the computing device to perform the method according to any one of claims 1 to 7.

Images