DE102013011239A1 - Method for determining a movement of an object - Google Patents

Abstract

The invention relates to a method for determining a movement of an object (O1 to On), wherein the object (O1 to On) is detected by means of at least one radar sensor (1). According to the invention, at least one velocity profile (P) of all measurement points (M1 to Mm) associated with the object (O1 to On) is determined based on data acquired by means of the radar sensor (1), the velocity profile (P) being determined from radial velocities (vr) of the measurement points ( M1 to Mm), an absolute velocity (vabs) and / or a direction of movement of the object (O) being determined on the basis of the velocity profile (P).

Description

The invention relates to a method for determining a movement of an object, wherein the object is detected by means of at least one radar sensor.

Methods for determining a movement of an object by means of one or more radar sensors are generally known from the prior art.

Such so-called tracking methods use Kalman filters to assign objects in several successive radar recordings of several individual measurements.

Methods known for orientation estimation of objects using the least squares method, also known as least squares fit, represent a simple, one-dimensional calculation method based on radial measurement points. All measuring points are used so that there is a risk of errors due to so-called clutter or micro-Doppler, caused by arm and leg movements of pedestrians or wheels of vehicles.

Methods for determining a movement of an object are, for example, " Rohling, H. et al.: Lateral velocity estimation for automotive radar applications; Hamburg University of Technology, Department of Telecommunications, Eißendorfer Strasse 40, 21073 Hamburg, Germany "," Fölster, F., Rohling, H .: Lateral velocity estimation based on automotive radar sensors; Hamburg University of Technology TUHH, Germany " and " Fölster, F., Rohling, H .: Observation of lateral moving traffic with an automotive radar; Hamburg University of Technology (TUHH), Eißendorfer Strasse 40, 21073 Hamburg, Germany " known.

The invention is based on the object to provide a comparison with the prior art improved method for determining a movement of an object.

The object is achieved by a method having the features specified in claim 1.

Advantageous embodiments of the invention are the subject of the dependent claims.

In the method for determining a movement of an object, the object is detected by means of at least one radar sensor.

According to the invention, at least one velocity profile of all measurement points assigned to the object is determined on the basis of data acquired by means of the radar sensor, wherein the velocity profile is formed from radial velocities of the measurement points, wherein an absolute velocity and / or a direction of movement of the object is or are determined based on the velocity profile.

The determination of the characteristic velocity profile enables a particularly reliable and robust determination of the absolute velocity and the direction of movement of the object in a single radar measurement without model assumptions and without a reference point.

Embodiments of the invention are explained in more detail below with reference to drawings.

Showing:

1 schematically different movements of an object and associated velocity profiles of the object as well as an enlarged detail,

2 schematically a flow of a first embodiment of a method according to the invention for determining a velocity profile according to 1 .

3 schematically a means of the first embodiment of the inventive method according to 2 generated diagram with a profile of radial velocities of the object,

4 schematically a flow of a second embodiment of a method according to the invention for determining a velocity profile according to 1 .

5 schematically recorded by means of a radar sensor measured data and a determined from these measurement data course of radial velocities of the object,

6 schematically a first embodiment of a procedure for the spatial allocation of measuring points of the radar sensor in clusters,

7 schematically the object according to 1 with different clusters of measuring points,

8th schematically a velocity profile of the cluster according to 7 .

9 schematically a second embodiment of a procedure for the spatial allocation of measuring points of the radar sensor in clusters,

10 schematically two objects with different speeds,

11 schematically a means of the second embodiment according to 9 generated velocity profile of the objects according to 10 .

12 schematically a first embodiment of a sequence for the temporal and spatial assignment of measuring points of the radar sensor in clusters,

13 schematically measured by a radar sensor measured data of an object,

14 schematically a from the measured data according to 13 determined course of radial velocities of the object,

15 schematically a second embodiment of a process for the temporal and spatial assignment of measuring points of the radar sensor in clusters,

16 schematically three objects with different speeds,

17 schematically a means of the second embodiment according to 15 generated velocity profile of the objects according to 16 .

18 schematically a section of the velocity profile according to 1 in an analytical calculation of the same,

19 schematically an object located in the sensor field of a radar sensor and determined by different methods absolute speeds of the object and object orientations, and

20 schematically a crossing situation with several crossing objects in the sensor field of a radar sensor.

Corresponding parts are provided in all figures with the same reference numerals.

In 1 is a means of an in 19 illustrated radar sensor 1 captured and trained as a vehicle object O1 represented in different motion situations. The radar sensor 1 is arranged for example on a vehicle, not shown, and has an example in 10 shown coordinate system with the direction axes x, y on. The radar sensor 1 is arranged in the coordinate origin of the coordinate system.

The object O1 moves at a speed v. To determine a movement of the object O1, this is done by means of the radar sensor 1 detected. In this case, the object O1 is starting from a position of the radar sensor 1 considered.

The object O1 is regarded as an extended object, so that an azimuth angle θ of a position vector, not shown, of a measurement changes over the course of the object O1. At the azimuth angle θ becomes a measuring angle of the radar sensor 1 Understood. In the illustrated middle embodiment, of which an enlarged detail in the range of the azimuth angle θ _{1 is shown} on the left, the object O1 moves at the moment of detection by means of the radar sensor 1 with the speed v in one Direction which deviates at an angle α from the directional axis y. The angle α corresponds to an object orientation G.

It is by means of the radar sensor 1 due to the Doppler effect, a radial velocity v _{r of} the object O1 is measured directly. As a result of the azimuth angle θ of the position vector of a measurement changing over the course of the object O1, an angle between the constant velocity vector of the velocity v in a rectilinear motion of the object O1 and the position vector consequently changes over the azimuth angle θ.

This results in a cosinusoidal velocity profile P, in which the respective radial velocity v _{r is} plotted as a function of the associated azimuth angle θ ₁ to θ ₅ .

That is, the measured radial velocity v _{r of} a traversing object O1 with rectilinear motion, depending on the direction of movement and position in an in 19 Sensor field SF of the radar sensor shown in more detail 1 is positive or negative and subject to a sinusoidal or cosinusoidal profile over the position angle or azimuth angle θ ₁ to θ ₅ . The zero crossing shifts with the direction of movement at an angle α. At the zero crossing, the direction of movement is perpendicular to the current position vector with azimuth angles θ ₁ to θ ₅ . The amplitude of the sinusoidal or cosinusoidal course corresponds to one in 19 closer absolute velocity v _{abs of} the object O1. The sinusoidal or cosinusoidal course is also referred to as kinematic expansion or as the velocity profile P below.

2 shows a sequence of a first embodiment of a method according to the invention for determining the velocity profile P according to FIG 1 ,

Prior to the determination of the velocity profile P, so-called clutter and micro-Doppler, for example caused by wheels of the object O1, are removed in a first step S1 from the azimuth angles θ and radial velocities v _r supplied as measurement data. For this purpose, a robust method is used which uses a so-called RANSAC algorithm.

By means of the RANSAC algorithm, as in 3 represented, a regression line through the measuring points M1 to Mm laid. For this purpose, a corridor K is first determined in a second step S2, in which all measurements of the velocity profile P would have to be safe. This is done on the basis of the considered angular range of the azimuth angle θ and a maximum speed v.

Optionally, in a third step S3, the measuring points M1 to Mm are measured on the basis of their amplitude, i. H. their reflected performance, or weighted by their density in their environment. A weighting factor, a so-called smooth kernel density, is formed.

In a fourth step S4, measuring points M1 to Mm which lie outside the corridor K are excluded. These measuring points M1 to Mm are in 3 shown as circles.

In a fifth step S5, the exact determination of the velocity profile P is carried out. The exact velocity profile P is determined analytically or by means of a least squares method, a so-called least-square-fit method, using the remaining measurement points M1 to Mm. Subsequently, a quality value of the velocity profile P is preferably calculated. This is done based on a number of measured different discrete velocity values and the slope. In addition, the ratio of measuring points M1 to Mm within the corridor K, so-called in-liners, to measuring points M1 to Mm outside the corridor K, so-called outliers, is also used.

In 4 a sequence of a second embodiment of a method according to the invention for determining the velocity profile P is shown.

This embodiment is based on that the sinusoidal or cosinusoidal velocity profile P directly from all measuring points M1 to Mm or from a in 5 section A shown this can be determined, if an assembly of spatially contiguous measuring points M1 to Mm to so-called clusters according to the 6 to 11 can be waived.

Since a large number of objects O1 to On are recorded simultaneously with different speeds v, a particularly robust method is required for this purpose.

For this purpose, in a step S6 the measuring points M1 to Mm or the section A of measuring points M1 to Mm are detected and fed in a further step S7 to a RANSAC algorithm, which is based on a model designed as a sinusoidal model according to the following equation: v _r = v _abs · sin (Θ + α) (1)

The advantage of the RANSAC algorithm is that it remains functional even if the measurement points M1 to Mm include outliers of 70% to 80%. This means that in about 20% to 30% of all examined measuring points M1 to Mm have to come from the same object O1 to On. For this reason, the examination of a section A is more useful. After an object O1 to On has been found, its measurement points M1 to Mm are removed and the algorithm is executed again.

Furthermore, in step S7, the most suitable sine curve SV is determined with the absolute velocity v _abs as the amplitude and the angle α as the phase shift.

In a further step S8, the measuring points M1 to Mm are determined, which are within a predetermined distance to the sinusoidal curve SV and within a certain range.

In the subsequent step S9, the velocity profile P and from the latter the absolute velocity v _abs and from the object orientation G its direction of movement are determined on the basis of the sine curve SV associated measuring points M1 to Mm.

In 5 are the means of the radar sensor 1 detected measuring points M1 to Mm and from these measuring points M1 to Mm directly determined speed profile P shown. The determination is made as in 4 described.

6 schematically shows a first embodiment of a procedure for the spatial allocation of measuring points M1 to Mm of the radar sensor 1 in clusters C1 to C4. In the 7 and 8th an object O1 is shown with different clusters C1 to C4 of measuring points M1 to Mm and a velocity profile P of the clusters C1 to C4.

Spatially coherent measurement points M1 to Mm can generally be summarized in clusters C1 to C4, also referred to as reflection centers. These arise due to the different reflection properties of the object O1. If the clusters C1 to C4 belong to a rigid object O1 to On, they have an identical velocity profile P.

One goal is to recognize the clusters C1 to C4 of an object O1 to On so as to have an extent and position expressed by a reference point such as a reference point. B. to be able to determine the focus of an extended object O1 to On. In the prior art model assumptions exist which predict a position and size of the clusters C1 to C4 of an object O1 to On, for example of a vehicle, depending on the orientation and position in the sensor field SF. On such model assumptions, which are very object-specific and must be determined consuming, can be dispensed with in the application of the method and its embodiments in an advantageous manner.

For each cluster C1 to C4, a spatial velocity profile P is determined in a step S10. Thereafter, it is checked whether two or more clusters C1 to C4 can be combined to form a common velocity profile P. It can either be checked whether a common sine or cosine can be formed or whether parameters of the sine or cosine match. The latter is done by deriving a mean speed.

Is an extension of the clusters C1 to C4 in the direction of the azimuth angle θ too low to determine a velocity profile P, such. As in the cluster C3, it is checked whether the associated speed value can be assigned to a speed profile P.

In a step S11, the clusters C1 to C4 are merged into an object O1 having a velocity profile P. For multiple objects O1 to On, clusters C1 to C4 become as in the following 9 to 11 shown separated into several objects O1 to On with different velocity profiles P.

9 schematically shows a second embodiment of a procedure for the spatial allocation of measuring points M1 to Mm of the radar sensor 1 in clusters C1 to C4. In the 10 and 11 Two objects O1, O2 with a common cluster C1 of measuring points M1 to Mm and several velocity profiles P of the cluster C1 are shown.

For the cluster C1, a spatial velocity profile P is determined in a step S12. Thereafter, it is checked whether the cluster C1 has a plurality of different velocity profiles P.

In a step S13, the cluster C1 is separated into two objects O1, O2, each having a velocity profile P.

This means that if two objects O1, O2 are in spatial proximity at different speeds, they are combined to form a cluster C1. If there are a large number of outliers in the calculation of the velocity profile P, an attempt is made to fit two velocity profiles P into the cluster C1 and thus spatially separate the cluster C1.

In the illustrated embodiment, the object O2 is a pedestrian who moves next to an object O1 formed as a stationary vehicle. Both objects O1, O2 are detected in a cluster C1 because of their spatial proximity. However, two speed profiles P can be determined and thus the pedestrian can be spatially separated from the vehicle.

In 12 is a first embodiment of a process for the temporal and spatial assignment of measuring points M1 to Mm of the radar sensor 1 represented in clusters C1 to C5. The 13 and 14 show several temporally successively detected measuring points M1 to Mm of an object O1 and a common velocity profile P of the clusters C1 to C5.

For the clusters C1 to C5 of measuring points M1 to Mm of an object O1 determined in several time steps, which move in a straight line with approximately the same speed v _abs , a velocity profile P is determined in a step S14. Subsequently, it is checked whether the multiple clusters C1 to C5 can be described by a common velocity profile P. If this is the case, as shown, a spatial assignment takes place in step S15. In this case, an object O1 can be assigned to several frames by means of a data association.

Alternatively or additionally, the clusters C1 to C5 are already spatially assigned in each time step before the feed in step S14.

That is, when an object O1 moves rectilinearly at an approximately constant speed, the measurement points M1 to Mm also form a common velocity profile P from successive measurements.

In the 13 and 14 Measuring points M1 to Mm of a vehicle are shown in different measurements. They all have the same velocity profile P, so they can all be assigned to the same object O1, ie to the vehicle.

The illustrated reduction of the number of clusters C1 by combining objects O1 spatially split over several detections enables a particularly fast tracking of the object O1 without model assumptions, such as the reflection model, while at the same time being robust against occlusion. On the other hand, occlusion leads to a displacement of the reference point or center of gravity of the object O1 to On in the case of methods known from the prior art and thus to an erroneous velocity estimation, for example by a Kalman filter.

In 15 is a second embodiment of a process for the temporal and spatial assignment of measuring points M1 to Mm of the radar sensor 1 represented in a cluster C1. 16 shows three objects O1 to O3, which are in close proximity to each other. In 17 the objects O1 to O3 are shown with a common cluster C1 of measuring points M1 to Mm and several velocity profiles P of the cluster C1. The measuring points M1 to Mm of each object O1 to O3 are recorded in chronological succession.

For the cluster C1 of measuring points M1 to Mm of the objects O1 to O3 determined in several time steps, a velocity profile P is determined in a step S16. Subsequently, it is checked whether the cluster C1 has a plurality of different velocity profiles P. If this is the case, as shown in step S17, a separation of the cluster C1 into the three objects O1 to O3 and the assignment of the objects O1 to O3 with a part of the cluster C1.

In other words, in the case of objects O1 to O3 standing close to one another, only one cluster C1 results from the contiguous measuring points M1 to Mm of the objects O1 to O3. In contrast to the spatial distribution, information about the velocity profile P can already be used from the previous measurement. The velocity profile P is then searched in the cluster C1.

The example with the pedestrian located next to the vehicle according to the 10 and 11 is extended here. In addition, the previous measurement is considered, in which the pedestrian was recorded as a separate object O3, ie as a separate cluster.

The velocity profile P is determined in the subsequent measurement in the assembled cluster C1.

Thus, there is still the advantage that on the basis of the velocity profile P also incorrectly assembled clusters C1 to C5 can be separated again.

18 shows a section of the velocity profile P according to 1 , For the analytical calculation of the velocity profile it is necessary that a zero crossing at the azimuth angle θ _{0 is} known.

If the zero crossing is not included in the measurement, this is estimated from the measured course of the measuring points M1 to Mm. For this purpose, a slope v _r '(θ) of the velocity profile P is determined and an associated mean value v (θ) _mean determined. From this, the zero crossing is analytically determined and from this the object orientation G and thus its direction of motion. Based on the zero crossing and the mean value v (θ) _middle , the absolute velocity v _{abs is determined} as the amplitude of the sine or cosine.

The method described above can be described by the following equations:

Alternatively or additionally, the determination of the velocity profile P takes place by means of the least squares method, the so-called least-square-fit method, wherein, in contrast to the methods known from the prior art, for example to that of FIG. Rohling, H. et al.: Lateral velocity estimation for automotive radar applications; Hamburg University of Technology, Department of Telecommunications, Eißendorfer Strasse 40, 21073 Hamburg, Germany "Known methods, but no outlier are taken into account and thus the robustness and accuracy of the method increases. In particular, as described, errors due to clutter and micro Doppler are avoided since not all measuring points M1 to Mm are used to determine the velocity profile P.

In 19 a comparison of an absolute speed v _abs determined by means of the method according to the invention with an absolute speed v _{abs of} an object O1 determined by means of a method according to the prior art is shown. Furthermore, a comparison of an object orientation G determined by the method according to the invention with an object orientation G of the object O1 determined by means of a method according to the prior art is shown. The values determined by means of the method according to the invention are shown with a solid line and the values determined by means of the method according to the prior art with a broken line.

Both methods have in common is that the sensor field SF of the radar sensor 1 crossing object O1 is detected. It can be seen that the values of the absolute velocity v _abs and the object orientation G determined by means of the robust method according to the invention are more accurate and almost straightforward without Outliers or Outlier are formed. This is already evident from speed profiles P, not shown here, of the various methods.

20 shows an application of the method according to the invention for detecting crossing objects O1 to On formed as vehicles on a road intersection. The objects O1 to On are located in the sensor field SF of the radar sensor 1 ,

The absolute velocity v _abs and direction of movement, ie object orientation G, of a crossing object O1 to On at an intersection can already be detected with the first measurement. Known from the prior art algorithms, for example, use a Kalman filter, measure a lateral displacement of the objects O1 to On over several measurements and therefore require significantly longer for a meaningful estimate of the direction of movement than the inventive method.

Especially in crossing vehicles, by a limited opening angle of the sensor field SF of the radar sensor 1 be detected relatively late, resulting in this time savings significant advantages in the calculation of a collision probability.

In addition, the quality factor can be included in the situation analysis so that, for example, the probability of collision can be determined more quickly and reliably. Furthermore, the velocity estimate can also be used to initialize a Kalman filter.

LIST OF REFERENCE NUMBERS

1

radar sensor

A

neckline

C1 to C5

cluster

G

object orientation

K

corridor

M1 to Mm

measuring point

O1 to On

object

P

velocity profile

SF

sensor field

SV

sine wave

S1 to S17

step

v

speed

v (θ) _medium

Average

v _abs

absolute speed

v _r

radial speed

v _r '(θ)

pitch

x

direction axis

y

direction axis

α

angle

θ

azimuth

θ ₀ to θ ₅

azimuth

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited non-patent literature

• Rohling, H. et al.: Lateral velocity estimation for automotive radar applications; Hamburg University of Technology, Department of Telecommunications, Eißendorfer Strasse 40, 21073 Hamburg, Germany [0005]
• Fölster, F., Rohling, H .: Lateral velocity estimation based on automotive radar sensors; Hamburg University of Technology TUHH, Germany [0005]
• Fölster, F., Rohling, H .: Observation of lateral moving traffic with an automotive radar; Hamburg University of Technology (TUHH), Eißendorfer Strasse 40, 21073 Hamburg, Germany [0005]
• Rohling, H. et al.: Lateral velocity estimation for automotive radar applications; Hamburg University of Technology, Department of Telecommunications, Eißendorfer Strasse 40, 21073 Hamburg, Germany [0082]

Claims (6)

Method for determining a movement of an object (O1 to On), wherein the object (O1 to On) is detected by means of at least one radar sensor ( 1 ), characterized in that by means of the radar sensor ( 1 ) at least one speed profile (P) of all the object (O1 to On) associated measuring points (M1 to Mm) is determined, the speed profile (P) of radial velocities (v _r ) of the measuring points (M1 to Mm) is formed wherein on the basis of the velocity profile (P) an absolute velocity (v _abs ) and / or a direction of movement of the object (O) is / are determined. A method according to claim 1, characterized in that a quality value is determined for each radial velocity (v _r ). A method according to claim 1 or 2, characterized in that the speed profile (P) over a measuring angle of the radar sensor ( 1 ) is spatially evaluated. A method according to claim 3, characterized in that in the evaluation all measuring points (M1 to Mm) of the velocity profile (P) to an object (O1 to On) are summarized. Method according to one of the preceding claims, characterized in that speed profile (P) is spatially and temporally evaluated over at least an approximately rectilinear movement of the object (O1 to On) over a plurality of time steps. Method according to one of the preceding claims, characterized in that the determination of the velocity profile (P) is carried out analytically or by means of a least squares method.

Images