OBJECT TRACKING METHOD

Abstract

An object tracking method, in which a radar sensor emits radar signals in successive measurement cycles, said radar signals being reflected by the object and captured by the radar sensor as radar targets, wherein movement information about the object for object tracking is determined on the basis of the radar targets and a search window for the radar targets of the object is defined on the basis of the movement information, wherein the search window is widened if a change in the movement information which exceeds a definable limit value is determined in successive measurement cycles and/or if no radar targets of the tracked object are captured anymore.

Claims

1. An object tracking method comprising: emitting radar signals in successive measurement cycles; receiving said radar signals being reflected by an object; determining movement information about the object for object tracking on the basis of the radar signals and a search window for the object defined on the basis of the movement information; and widening the search window if a change in the movement information exceeds a definable limit value in successive measurement cycles.

2. The method according to claim 1, wherein the movement information comprises speed and/or acceleration.

3. The method according to claim 1, further comprising: following the widening of the search window, repeating a current measurement cycle or starting a following measurement cycle.

4. The method according to claim 1, further comprising saving movement information patterns for object tracking by means of which the object can be classified by aligning the movement information about the object with the movement information patterns.

5. The method according to claim 4, further comprising assigning to the object radar targets captured in the widened search window based on the movement information patterns.

6. The method according to claim 5, further comprising canceling the widening of the search window is if no radar targets of the widened search window can be assigned to the object in a definable number of measurement cycles.

7. The method according to claim 1, wherein the widening of the search window is restricted to a definable number of measurement cycles.

8. The method according to claim 2, further comprising determining acceleration of the on the basis of a difference quotient of the speed.

9. The method according to claim 1, further comprising transmitting the movement information about the or classification of the object.

10. The method according to claim 9, wherein the classification is restricted to a definable region of the movement information.

11. The method according to claim 10, further comprising providing a plausibility check of the movement information and/or of the classification of the object on the basis of multiple measurement cycles.

12. A driver assistance system comprising: a radar sensor configured to emit radar signals in successive measurement cycles, receive said radar signals being reflected by an object, determine movement information about the object for object tracking on the basis of the radar signals and a search window for the object defined on the basis of the movement information, and widen the search window if a change in the movement information exceeds a definable limit value in successive measurement cycles.

13-14. (canceled)

15. A non-transitory computer-readable recording medium having embodied thereon computer-readable codes which when executed by a radar sensor cause the radar sensor to execute an object tracking method, the method comprising: emitting radar signals in successive measurement cycles; receiving said radar signals being reflected by an object; determining movement information about the object for object tracking on the basis of the radar signals and a search window for the object defined on the basis of the movement information; and widening the search window if a change in the movement information exceeds a definable limit value in successive measurement cycles.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The embodiments are explained in greater detail below with reference to the accompanying drawings, in which:

[0025] FIG. 1 shows a simplified schematic representation of a traffic situation in which an ego vehicle follows a vehicle driving in front and tracks it by means of suitable sensor technology;

[0026] FIG. 2 shows a simplified schematic illustration of a traffic situation following the one in FIG. 1, in which the vehicle driving in front has an accident;

[0027] FIG. 3 shows a simplified schematic representation of the traffic situation from FIG. 2, with the predicted object position of the vehicle driving in front;

[0028] FIG. 4 shows a simplified schematic representation of the scan modes “near scan” and “far scan” of a far-range radar sensor; and

[0029] FIG. 5 shows a simplified schematic representation of a radar scan of a vehicle driving in front which is involved in an accident, which has been continuously captured and suddenly exhibits negative acceleration.

[0030] Reference numeral 1 in FIG. 1 describes an ego vehicle which is equipped with a driver assistance system which can execute or control functions such as, e.g., ACC (Adaptive Cruise Control) and/or EBA (Emergency Braking Assist) and/or LKA (Lane Keep Assist) and can capture the environment or the vehicle environment by means of suitable sensor technology and can preferably classify said environment by means of a classifier. For executing the functions, the driver assistance system comprises a central control unit (ECU—Electronic Control Unit, ADCU Assisted & Automated Driving Control Unit) which is not depicted in the figures. The classifier can be saved as an independent 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 far-range radar sensor, which has a forward-facing detection range 3, is provided as a sensor in the ego vehicle 1. Furthermore, in front of the ego vehicle 1, there is a further vehicle 4 driving in front of the ego vehicle 1, which is captured by the driver assistance system in the course of the object tracking by means of the radar sensor 2. On the basis of the sensor data of the radar sensor 2, the vehicle 4 can then be tracked by determining movement information (e.g., the speed or the acceleration of the vehicle 4) from the reflected detections or radar targets 5 of the vehicle 4. By means of the radar targets 5 and the associated movement information, the ego vehicle 1 can predict the following movement or the trajectory of the vehicle 4 and align or adapt the search area to expected detections associated with the vehicle 4, or, accordingly, the search window to the predicted object 6 depicted in FIG. 2. Furthermore, the vehicle 4 can then be classified by the classifier (e.g., as a car, truck, or in the event of an accident as a vehicle involved in an accident and the like). The classification can in addition also be included in the movement prediction. Consequently, the traffic situation can be established so that alterations or hazards can be reacted to in good time with braking and/or steering interventions or speed adjustments, by emitting warnings or the like.

[0031] In the following course of the traffic situation, the vehicle 2 driving in front has an accident due to an obstacle 7, according to FIG. 3. For example, the average acceleration here during the accident at a speed of approx. 50 km/h is approx. 200 m/s.sup.2. For example, by contrast, the maximum absolute acceleration when starting is only about 3-7 m/s.sup.2 and the maximum absolute acceleration during full braking is about −10 m/s.sup.2. In addition, the duration from the impact up to standstill is around 72 ms and the change in speed per computing cycle (assumed cycle=70 ms) is around 7 m/s. Such variables can already lead to object losses or track breaks, as the search windows for detected objects or reported radar targets no longer lie within the expected range of a “normal journey”, since both the speed and position change very quickly and very significantly in an accident situation. As a comparison, FIG. 4 depicts the expected object position or the predicted object 6 from FIG. 2 as well as the actually captured radar targets of the vehicle 4 which has had an accident from FIG. 3. These radar targets 5 now lie outside the search area (which is located here in the region of the predicted object 6). As a consequence, the original object or the vehicle driving in front 4 is lost. If the vehicle 4 is then recognized again, a new object with a very much lower speed or a stationary object is created. The information about the transition of the movement of the vehicle 4 from “moving” to “stationary” is lost.

[0032] In accordance with the method according to the present application, the vehicle 4 is marked as an object involved in an accident if the movement information thereof lies below a certain threshold or limit value. For example, if the absolute acceleration thereof is below −12 m/s.sup.2, i.e., it can no longer be explained by full braking. This is effected in that, in a first step, it is recognized that a disproportionately large change in the speed of the assigned radar targets 5 exists between two successive measurement cycles in the case of an object or vehicle that has hitherto been stably tracked. However, the radar targets 5 must still lie within the normal search window for speeds. Alternatively, it is recognized that an object that has hitherto been stably tracked is not measured anymore without any obvious reason, i.e., no radar targets 5 are assigned to the object anymore in the current computing cycle. This can, e.g., be due to the fact that the speed of the radar targets 5 has already changed so significantly that they do not lie within the search window anymore. Should such a situation be recognized, the speed search window for this object is enlarged in such a way that radar targets 5 are also sought far outside the previous speed search window. The position of the radar targets 5 must lie in front of the object. Such measures can additionally reduce the probability of false positive events.

[0033] After the search window has been adapted, a search is accordingly made again for radar detections either in the current measurement cycle or beginning in the next measurement cycle, which radar detections correspond to the accident hypothesis of the object or a movement information pattern saved in a memory. If corresponding detections are found, they are assigned to the accident candidate. The acceleration a can be determined via the difference quotient on the basis of the speed v and time t, in that

a=(v(n−1)−v(n))/Δt

[0034] This acceleration is then transferred or assigned to the object in order to be able to carry out a correct kinematic prediction for the next computing cycle, i.e., the new speed decreases accordingly and the position is displaced accordingly. Since a typical accident scenario usually only lasts around 70 ms, the entire accident scenario is over after just a few measurement cycles, possibly already after a single measurement cycle (with a cycle time of 70 ms), and the object involved in the accident comes to a standstill. Therefore, if possible, the widening of the search window should only be limited to a few measurement cycles. If no accident is confirmed within this time, the search window can be normalized again.

[0035] Furthermore, the method according to the present application can be applied to all, in particular, radar-based driver assistance systems, e.g., EBA (Emergency Brake Assist), LKA (Lane Keeping Assist), ACC (Adaptive Cruise Control) or the like, wherein the focus is however primarily on forward-facing sensor systems (front radar or long-range radar). Generic radar sensors can, for example, have different scan modes which can also comprise different aperture angles of the detection area in order to illuminate the near range (near scan SR) and/or the far range (far scan FR) for the respective application, as depicted in FIG. 5.

[0036] One embodiment of this is that the described traffic situation is detected in both radar scans (near-scan SR and far-scan FR), wherein the difference in speed between two successive measurements must be greater than a definable value for both scans independently of one another, preferably greater than 1 m/s. This activates the widening of the speed search window, e.g., to 7 m/s (approx. 1-2 m/s being standard) for the following five measurement cycles. A potential accident should in any case be completely terminated by the end of this cycle time. If radar detections are assigned to the accident candidate in the following measurement cycles, the acceleration can likewise be formed by means of the difference quotient. If the acceleration value exceeds e.g., an absolute value of −12 m/s.sup.2, the object is classified as an object involved in an accident and the acceleration is assigned to the tracked object. Furthermore, the use of a second scan to check the plausibility is not absolutely necessary, but this can reduce the quantity of incorrectly triggered events or even prevent them.

[0037] In a practical way, this information can then be provided via an interface, e.g., to other vehicles or subscribers (Car-2-Car communication or Car-to-X communication), for example as data information, as a radar scan or the like. A radar scan (plotted as a function of acceleration a in m/s and time t in s) is depicted as a measurement result in FIG. 6, in which the object or the vehicle driving in front involved in an accident has been continuously captured and the acceleration of −50 m/s.sup.2 occurs for a short period of time. In a practical way, the indicated parameters can also be varied in such a way that accidents are limited to a certain speed range, as a result of which the determination reliability can be improved even further.