Methods for object tracking

Abstract

The invention relates to a method for object tracking, wherein a radar sensor (2) transmits radar signals in successive measuring cycles, which radar signals are reflected by an object and detected by the radar sensor (2) as a radar target (5), wherein object motion information for object tracking is determined based on the radar target (5), and a search window for the radar target (5) of the object is defined based on the motion information, wherein the search window is expanded if a change in the motion information is determined in successive measuring cycles which exceeds a predefinable limit value and / or no more radar targets (5) are detected for the tracked object.

Description

Technical Field

[0001] The present invention relates to a method, in particular implemented by a computer, for object tracking and accident detection by means of a radar sensor of an assistance system or driver assistance system, and to a driver assistance system, in which object tracking and accident detection are performed, in particular, according to the method of the invention, and also to a computer program product for executing the method and a removable computer-readable storage medium on which the computer program product for executing the method is stored. Background Art

[0002] Modern vehicles, such as motor vehicles and motorcycles, are increasingly equipped with driver assistance systems. These systems use suitable sensors or sensor systems to detect the surroundings, identify traffic situations, and support the driver through, for example, braking and / or steering interventions or by issuing visual, haptic, or audio warnings. Radar sensors, lidar sensors, camera sensors, ultrasonic sensors, or similar sensors are often used as sensor systems to detect the surroundings. Conclusions about the surroundings can then be drawn from the sensor data detected by the sensors. Detection of the surroundings using radar sensors is based, for example, on a combination of electromagnetic waves emitted and their reflections from objects such as other road users, obstacles in the roadway, or buildings at the edge of the roadway. Individual reflections or detections associated with an object are detected by the radar sensor as radar targets and assigned to the corresponding object using, for example, suitable algorithms. Such objects can be tracked or observed, with object tracking (object tracking) being performed continuously, meaning that the tracked object should not be lost, for example, due to so-called "tracking interruptions," as this could lead to erroneous conclusions or misinterpretations of the traffic scene. This in turn can lead to the vehicle's driver assistance systems not intervening in this situation, or intervening too late, such as by braking, to prevent a rear-end collision.

[0003] One of the most complex situations that can arise for driver assistance systems is an accident involving a vehicle ahead. Here, for example, the vehicle ahead strikes an obstacle, while the host vehicle itself is not yet involved. However, this presents an increased risk for the host vehicle, as the movement of the preceding accident vehicle changes rapidly, to which the host vehicle must react. While the tracking algorithms stored in the driver assistance system can adapt within their dynamic range, the dynamics occurring in this situation represent an extreme case, leading to a break in tracking and, consequently, to an interruption in the expected continuity of the detected object.

[0004] For example, tracking algorithms for accidents involving moving objects ahead present a challenge: these situations often occur suddenly, without warning, making it difficult for the tracking algorithm to adapt and react accordingly. Furthermore, the physical variables present during an accident can be many times greater than those encountered in normal traffic flow.

[0005] DE 10 2011 001 248 A1 discloses a method for supporting the driver of a motor vehicle equipped with a driver assistance system, in which object motion information is detected using a radar measuring device for object tracking. A problem with this is that some motion parameters, such as relative acceleration, cannot be detected by the radar measuring device or the object model is inaccurate, which can lead to object loss. When an object is re-detected, a new object is initialized, even though the measured values ​​originate from an already tracked object. This type of measurement situation particularly arises when the measured values ​​deviate significantly from the position predicted by the object model. Object loss due to sudden changes in acceleration is a significant problem for general assistance functions. Such highly dynamic situations are particularly important in accident avoidance applications. Object loss and reinitialization mean a loss of precious fractions of a second. To address this problem, it is proposed to detect object information in the vehicle's surroundings using a camera, and to use the detected object information to correct object tracking, thereby improving object tracking. However, this requires additional hardware (camera and its control unit) and additional computing effort (data fusion and correction calculations) and does not eliminate the "weaknesses" of the radar measurement device. Summary of the Invention

[0006] The object of the present invention is therefore to provide an improved method for object tracking and accident detection and a corresponding assistance system which overcome the disadvantages of the prior art and improve object tracking in a simpler and more cost-effective manner.

[0007] The above object is solved by the overall theory of claim 1 and the independent claims. In the dependent claims, the design solutions of the present invention are required according to the purpose of use.

[0008] In the object tracking method according to the present invention, a radar sensor transmits radar signals in a plurality of consecutive measurement cycles. These radar signals are then reflected by objects and detected by the radar sensor as radar targets. Motion information of the object is then determined based on these radar targets for object tracking, with the motion information defining a radar target search window for the object. If a change in the motion information exceeding a predeterminable threshold is detected in consecutive measurement cycles and / or no more radar targets are detected for the tracked object, the search window is expanded. This has the advantage that the tracked object remains continuously detected, i.e., there are no tracking interruptions that could lead to loss of tracking, which could result in undesirable control interventions in the vehicle, for example.

[0009] The object's velocity and / or acceleration are preferably used as motion information. This information is typically already determined in such radar sensors or driver assistance systems that include radar sensors, so that no additional hardware and / or computational effort is required or only a negligible amount. For example, the search window can be expanded with respect to velocity, for example, from 1 to 2 meters per second (initial state) to at least 5 meters per second, preferably 7 meters per second, and in particular 10 meters per second. Accordingly, the threshold value for the motion information can also be defined as a speed difference between two measurement cycles exceeding 1 meter per second, preferably exceeding 3 meters per second, and in particular exceeding 5 meters per second.

[0010] Depending on the purpose of use, after the search window has been extended, either the current measurement cycle can be repeated or the next measurement cycle can be started.

[0011] According to an advantageous embodiment of the present invention, motion information patterns for object tracking are stored, for example, in a vehicle memory or in a driver assistance system control unit. This allows, for example, for classification of objects and / or traffic situations by matching and comparing the motion information of detected objects with the stored motion information patterns. A motion information pattern refers, in particular, to specific data, parameters, and / or variables that indicate an object class (e.g., an accident vehicle) or a particular traffic scenario. For example, if the speed of a vehicle ahead suddenly changes dramatically—for example, from 50 km / h to 0 km / h within a few seconds—and the trajectory correspondingly ends abruptly and shortens dramatically, an accident scenario can be inferred. Detecting such traffic situations thus confirms an accident hypothesis, allowing the corresponding vehicle to be classified as an accident vehicle.

[0012] Furthermore, if a radar target detected in the expanded search window corresponds to a motion information pattern, the radar target can be assigned to a specific object.

[0013] Depending on the intended use, the extension of the search window can be canceled if no radar target within the extended search window can be assigned to the specified object within a predefinable number of measurement cycles.

[0014] The extension of the search window is preferably limited to a predefinable number of measurement cycles, for example to the next three measurement cycles, in particular to the next five measurement cycles, in particular to the next ten measurement cycles or a similar number of measurement cycles.

[0015] According to an advantageous embodiment of the invention, the acceleration of an object is determined based on the velocity difference quotient and assigned to the object. For example, highly dynamic behavior of an accident object or vehicle can be reported to the host vehicle in this way, since the extremely high acceleration is directly transmitted to the accident object as a characteristic without filtering via the difference quotient.

[0016] Furthermore, a device can be provided for forwarding or transmitting information about the movement of objects, the classification of objects, and / or the classification of traffic situations. This has the advantage that a vehicle classified as an accident object or an accident situation is notified to other road users via an interface, for example, by radio transmission or the like (particularly by vehicle-to-vehicle communication or vehicle-to-external information exchange), so that other road users can also react accordingly to the situation (e.g., braking, accelerating, evasive maneuvers, replanning trajectories, making an emergency call, issuing visual, audio, or haptic warnings, etc.).

[0017] Depending on the purpose of use, the classification can be limited to a predeterminable range of motion information.

[0018] A plausibility test of the determined movement information and / or object classification and / or traffic situation classification is preferably predefined based on a plurality of measurement cycles (for example three, five, ten, etc. measurement cycles).

[0019] The present invention also includes a driver assistance system that implements object tracking, in particular, according to the method of the present invention. To this end, the driver assistance system includes a radar sensor for object tracking that transmits radar signals during successive measurement cycles. These radar signals are reflected by the object to be tracked and detected by the radar sensor as radar targets. Object motion information, such as velocity and / or acceleration, can be determined using the radar targets for object tracking. This motion information is used to define a search window for the object's radar targets. For example, the search window can be defined based on the object's velocity so that, when predicting motion or trajectory, the object remains within the search window in subsequent measurement cycles. If the motion information determined during successive measurement cycles changes beyond a predeterminable limit (e.g., if a predeterminable velocity is exceeded or exceeded, or if velocity changes), and / or if no radar targets or detectors are detected or can no longer be detected for the tracked object, the search window is extended.

[0020] A radar sensor is preferably a sensor that detects objects using transmitted electromagnetic waves that are reflected by and re-received from an object. The electromagnetic waves can have various wavelengths and frequency ranges. For example, the wavelength range of the electromagnetic waves can be 1 mm to 10 km or a frequency range of 300 GHz to 30 kHz. A preferred wavelength range is 1 cm to 1000 m or a frequency range of 30 GHz to 300 kHz. A further preferred wavelength range is 10 cm to 100 m or a frequency range of 3 GHz to 3 MHz. A particularly preferred wavelength range is 1 m to 10 m or a frequency range of 300 MHz to 30 MHz. Furthermore, the electromagnetic waves may be in a wavelength range of 10 nanometers to 3 millimeters or in a frequency range of 30 petahertz to 0.1 terahertz, preferably in a wavelength range of 380 nanometers to 1 millimeter or in a frequency range of 789 terahertz to 300 gigahertz, preferably in a wavelength range of 780 nanometers to 1 millimeter or in a frequency range of 385 terahertz to 300 gigahertz, particularly preferably in a wavelength range of 780 nanometers to 3 micrometers or in a frequency range of 385 terahertz to 100 terahertz.

[0021] The present invention also includes a computer program product with program code, which, when implemented on a computer or other programmable computing device known in the art, performs the method according to the present invention. Therefore, the method can also be configured as a purely computer-implemented method, wherein the term "computer-implemented method" within the meaning of the present invention describes a process plan or method steps that are implemented or executed by a computer. In this case, a computing device such as a computer, a computer network, or another programmable device known in the art (e.g., a computer device comprising a processor, a microcontroller, or the like) can perform data processing using programmable computing rules. The basic features of the method can be influenced, for example, by a new program, a plurality of new programs, an algorithm, or the like.

[0022] Furthermore, the present invention also comprises a computer-readable storage medium containing instructions, which cause a computer executing the instructions to implement the method according to at least one of the preceding claims.

[0023] The invention also explicitly includes combinations of features or claims that are not explicitly mentioned, so-called dependent combinations. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The present invention will be described in more detail below based on advantageous embodiments, wherein:

[0025] Figure 1 A simplified schematic diagram showing a traffic situation in which the host vehicle follows a vehicle in front and tracks the vehicle in front with the aid of suitable sensors;

[0026] Figure 2Show the basis Figure 1 A simplified schematic diagram of a traffic situation with predicted objects;

[0027] Figure 3 Shown in Figure 1 Simplified schematic diagram of a traffic situation following the illustrated traffic situation, wherein an accident occurs with the vehicle traveling ahead;

[0028] Figure 4 Show Figure 2 A simplified diagram of a traffic situation, including the predicted object positions of the vehicle ahead;

[0029] Figure 5 a simplified schematic diagram showing the long-range radar sensor scan modes "near scan" and "long scan"; and

[0030] Figure 6 A simplified schematic diagram shows a radar scan of a preceding accident vehicle that has been detected continuously and that suddenly experiences negative acceleration. DETAILED DESCRIPTION

[0031] Figure 1Reference numeral 1 in the figure depicts a vehicle equipped with a driver assistance system. This system can implement or control functions such as ACC (Adaptive Cruise Control or Abstandsregeltempomat) and / or EBA (Emergency Breaking Assist or Emergency Brake Assist) and / or LKA (Lane Keep Assist or Spurhalte / Spurwechsel Assist) and can detect the surroundings or vehicle surroundings using suitable sensors and preferably classify them using a classification device. To implement these functions, the driver assistance system includes a central control unit (ECU: Electronic Control Unit, ADCU: Assisted and Autonomous Driving Control Unit), not shown in the figure. The classification device can be stored as a standalone module or as a software application or algorithm on the central control unit of the driver assistance system. A radar sensor 2, in particular a long-range radar sensor, is provided as a sensor for the vehicle 1 having a forward detection range 3. In addition, there is another vehicle 4 traveling in front of the host vehicle 1, which is detected by the host vehicle's driver assistance system during object tracking with the help of radar sensor 2. Vehicle 4 can then be tracked based on the sensor data of radar sensor 2 through reflection detection from vehicle 4 or motion information (such as the speed or acceleration of vehicle 4) determined by radar target 5. Using radar target 5 and related motion information, the host vehicle 1 can predict the subsequent movement or trajectory of vehicle 4 and align or adjust the expected detection or search window related to vehicle 4 accordingly. Figure 2 Predicted object 6 is shown. Furthermore, vehicle 4 can be classified by a classification device (e.g., as a car, truck, or, in the event of an accident, as an accident vehicle, etc.). Furthermore, this classification can also be included in the motion prediction. This allows the traffic situation to be determined, allowing for timely responses to changes or dangers by means of braking and / or steering interventions, speed adjustments, warnings, or the like.

[0032] In the following traffic situation, if Figure 3As shown, the vehicle 2 traveling ahead is involved in an accident due to an obstacle 7. For example, during the accident, at a speed of approximately 50 km / h, the average acceleration here is approximately 200 m / s². In comparison, the maximum absolute acceleration during start-up is, for example, only approximately 3 to 7 m / s², and the maximum absolute acceleration during emergency braking is approximately -10 m / s². Furthermore, the duration from impact to parking is approximately 72 milliseconds, and the speed change per calculation cycle (assuming cycle = 70 milliseconds) is approximately 7 m / s. Such variations can already lead to object loss or tracking interruption, because the search window for detected objects or reported radar targets is no longer within the expected range of "normal driving", because in an accident situation, the changes in speed and position are very fast and very drastic. For this reason, as a comparison, in Figure 4 Shown in Figure 2 The expected object position or predicted object 6 in , and Figure 3 The radar targets of the accident vehicle 4 are actually detected in the image. These radar targets 5 are now outside the search area (here, within the area of ​​the predicted object 6). As a result, the original object or vehicle 4 ahead is lost. If vehicle 4 is subsequently detected again, a new object or a stationary object with a significantly lower speed is detected. Information about the motion transition of vehicle 4 from "moving" to "stationary" is lost.

[0033] According to the method described herein, a vehicle 4 is flagged as an accident object if its motion information falls below a certain threshold or limit value. For example, if its absolute acceleration falls below -12 meters per second squared, meaning it can no longer be interpreted as an emergency braking event. This is achieved by detecting in a first step that the velocity of an object or vehicle that has been stably tracked has changed excessively significantly between two consecutive measurement cycles. In this case, the radar target 5 must still be within the normal velocity search window. Alternatively, it can be detected that a previously stably tracked object is no longer being measured for no apparent reason—that is, no radar target 5 is assigned to the object in the current calculation cycle. This could be because, for example, the velocity of the radar target 5 has changed so dramatically that it is no longer within the search window. If such a situation is detected, the velocity search window for the object is expanded to include radar targets 5 that are currently outside the velocity search window. In this case, the radar target 5 must be located in front of the object. These measures further reduce the probability of false alarms.

[0034] After adjusting the adaptive search window, the search is repeated in the current measurement cycle or starting from the next measurement cycle for radar detection results that correspond to the object accident hypothesis or the motion information pattern stored in the memory. If corresponding detection results are found, they are assigned to the accident candidate. Here, the acceleration a can be determined from the velocity v and the time t by the difference quotient. The applicable method is

[0035] a=(v(n-1)–v(n)) / Δt

[0036] This acceleration is then transferred or assigned to the object so that the correct kinematic prediction can be made for the next calculation cycle, meaning the new velocity is reduced accordingly and the position shifted accordingly. Since a typical accident scenario only lasts approximately 70 milliseconds, the entire accident scenario is complete after only a few measurement cycles, or possibly even a single measurement cycle (for a 70 millisecond cycle time), and the accident object reaches a standstill. Therefore, whenever possible, the search window extension should be limited to just a few measurement cycles. If no accident is confirmed within this time, the search window can be normalized again.

[0037] The method according to the invention can also be applied to all driver assistance systems, in particular radar-based ones, which implement functions such as Emergency Brake Assist (EBA), Active Lane Keeping Assist with Steering Support (LKA), Adaptive Cruise Control (ACC) or similar functions. However, the main focus is on forward-facing sensor systems (front radar or long-range radar). For example, a universal radar sensor can have different scanning modes, which can also include different angles of detection range, so that Figure 5 It is shown that for the respective application the short-range area (short-range scanning, SR) and / or the long-range area (long-range scanning, FR) are illuminated separately.

[0038] To this end, the traffic situation is detected in two radar scans (a short-range scan SR and a long-range scan FR). For these two independent scans, the speed difference between two consecutive measurements must be greater than a predeterminable value, preferably greater than 1 meter per second. This triggers an expansion of the speed search window, for example, to 7 meters per second over the next five measurement cycles (the standard is approximately 1 to 2 meters per second). By the end of this cycle, the potential accident is completely resolved. If the radar detection is assigned to an accident candidate in the next measurement cycle, the acceleration can also be determined using the difference quotient. If the acceleration value exceeds an absolute value of, for example, -12 meters per second squared, the object is classified as an accident object and the acceleration is assigned to the tracked object. Furthermore, a second scan is not necessarily required for plausibility testing, but this can reduce the number of false triggering events or even prevent them.

[0039] Such information can be provided, for example, as data information, as radar scans or the like, in a practical manner via interfaces (vehicle-to-vehicle communication or vehicle-to-external information exchange) to other vehicles or recipient clients. Figure 6 The radar scan is shown as a measurement result (depending on the application of acceleration a (in meters per second) and time t (in seconds)). An object or a vehicle involved in an accident ahead is continuously detected, and an acceleration of -50 meters per second squared occurs within a short period of time. The given parameters can also be varied to suit the situation by limiting the accident to a specific speed range, thereby further increasing the reliability of the determination.

[0040] List of reference numerals:

[0041] 1 vehicle

[0042] 2 radar sensors

[0043] 3 Detection Area

[0044] 4 vehicles

[0045] 5 Radar Targets

[0046] 6 Predicting Objects

[0047] 7 Obstacles

[0048] SR Close Range Scan or Short Range Scan

[0049] FR Long Range Scan or Remote Scan

[0050] a Acceleration

[0051] v Speed

[0052] t time

Claims

1. A method for object tracking, wherein: The object is another vehicle (4) traveling in front of the vehicle (1), and the radar sensor (2) of the vehicle (1) sends radar signals in successive measurement cycles, which are reflected by the object and detected by the radar sensor as a radar target (5), wherein object motion information for object tracking is determined based on the radar target (5), and the motion information is the object speed and / or acceleration, and A search window for a radar target (5) of the object is defined based on the motion information, wherein The search window is extended if the following occurs during consecutive measurement cycles: - a change in the movement information of an object exceeding a predeterminable limit value is detected due to an accident, and / or - No more radar targets (5) are detected for the tracked object due to an incident with the object.

2. The method according to claim 1, characterized in that After extending the search window, either the current measurement cycle is repeated or a subsequent measurement cycle is started.

3. The method according to claim 1 or 2, characterized in that Movement information patterns are stored for object tracking, by means of which objects and / or traffic situations can be classified.

4. The method according to claim 3, characterized in that By matching the motion information of the object with the motion information pattern, the object and / or the traffic situation can be classified based on the motion information pattern.

5. The method according to claim 3, characterized in that If a radar target (5) detected in the expanded search window corresponds to a motion information pattern, the radar target is assigned to the object.

6. The method according to claim 5, characterized in that If no radar target (5) of the expanded search window can be assigned to the object within a predeterminable number of measuring cycles, the expansion of the search window is canceled.

7. The method according to claim 1 or 2, characterized in that Limits the extension of the search window to a defined number of measurement cycles.

8. The method according to claim 1 or 2, characterized in that The acceleration of the object is determined from the difference quotient of the velocities and assigned to the object.

9. The method according to claim 3, characterized in that Devices are provided in advance that can forward object movement information, object classification and / or traffic situation classification.

10. The method according to claim 3, characterized in that Limit the classification to the scope of motion information that can be specified.

11. The method according to claim 3, characterized in that A plausibility test of the determined movement information and / or object classification and / or traffic situation classification is provided based on a plurality of measurement cycles.

12. A driver assistance system having an object tracking function by means of the method according to any one of claims 1 to 11, wherein the object is another vehicle (4) traveling in front of the own vehicle (1), the driver assistance system having A radar sensor (2) for object tracking, wherein The radar sensor (2) transmits radar signals in successive measurement cycles, which are reflected by objects and detected by the radar sensor (2) as radar targets (5), wherein object motion information for object tracking is determined from the radar targets (5), the motion information being the object's velocity and / or acceleration, and A radar target (5) search window for the object is defined based on the motion information, wherein The search window is extended if the following occurs during consecutive measurement cycles: - Detection of a change in motion information exceeding a predeterminable limit value due to an accident involving an object, and / or - No more radar targets (5) are detected for the tracked object due to an incident with the object. 13 . A computer program product with a program code, which, when executed on a computer, carries out the method according to claim 1 .

14. A computer-readable storage medium comprising instructions, wherein the instructions enable a computer executing the instructions to implement the method according to any one of claims 1 to 11.

Images