DE102011121560A1 - Method for detection and classification of objects based on radar data, involves forming spacing cells for equal space or angle cells for equal angles, where temporal velocity curves are determined for multiple point targets of object - Google Patents

Abstract

The method involves determining a space (R), velocity (v-1) or an angle of an object (O1) from the radar data. Spacing cells are formed for equal space or angle cells are formed for equal angles. Temporal velocity curves are determined for multiple point targets of an object based on sequential velocity profiles determined successively and temporally. The object is classified based on the temporal velocity curves of multiple point targets associated to the object.

Description

The invention relates to a method for detecting and classifying objects based on radar data, wherein a distance, a velocity and / or an angle of an object with respect to a reference object are determined from the radar data.

From the DE 10 2010 045 980 A1 For example, a radar method and a radar system are known. By means of the radar method, a distance, a velocity and an angle of an object with respect to a reference object are determined. The object is a pedestrian and the reference object is a motor vehicle. In this case, a transmission signal is transmitted by means of a radar transmitter, by means of a radar antenna element, a reflected from the object transmission signal received from a current transmission signal and the received transmission signal generates a baseband signal. The baseband signal and a signal derived from the baseband signal are continued by an estimation method. The estimation method is a spectral estimation method.

The invention is based on the object of specifying a novel method for the detection and classification of objects.

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 a method for detecting and classifying objects based on radar data, a distance, a velocity and / or an angle of an object with respect to a reference object are or are determined from the radar data.

According to the invention, distance cells and / or angular cells for the same distances are formed, and for speeds in the distance cells and / or angle cells of at least one object, the velocities and speed profiles are determined at defined times, wherein for a plurality of point targets of the at least one object based on a concatenation Temporal velocity profiles are determined temporally successively detected speed profiles, wherein based on the temporal velocity curves of a plurality of point targets belonging to an object, the at least one object is classified.

The distance of the object to the reference object, for example a vehicle, is determined in particular from a signal propagation time of a transmitted signal to the object and from this back to the reference object. Due to the so-called Doppler effect, a radial, d. H. a relative velocity oriented in the direction of a connecting line between the reference object and the reflecting object, in a frequency of the reflected and detected signal. This frequency is obtained in particular by means of a Fourier transformation of a baseband signal and can be directly related to a distance value, i. H. a point target of an angle and distance cell.

On the basis of the evaluation of a plurality of speed profiles belonging to a single object, the method according to the invention makes it possible in a particularly advantageous manner to reliably classify moving pedestrians and to distinguish vehicles and pedestrians with the aid of radar sensors. As a result, accidents can be avoided in a particularly advantageous manner, or at least their severity can be reduced.

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

Showing:

1 schematically a traffic situation with a reference object and an object,

2A to 2C schematically an influence of movements of an object to be detected in a detection of the object based on radar data in a fast ramping process,

3A to 3C schematic representations for determining and processing a speed of the object,

4 schematically a juxtaposition of temporally successively detected velocities and a temporal velocity profile determined therefrom and

5A to 5B schematically an adaptive increase of a velocity resolution.

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

In 1 there is shown a traffic situation with a reference object RO and an object O1, wherein each of the reference object RO and the object O1 is a vehicle. The object O1 moves at a speed v ₁ to the reference object RO.

In embodiments not shown, the object O1 is a pedestrian, cyclist, two-wheeler, inline skaters or other road users.

The reference object RO includes a radar device 1 , by means of which a signal S is emitted, which is reflected by the object O1. The radar device 1 includes not shown radar sensors with a preferred frequency of 77 GHz and / or 79 GHz. Alternatively or additionally, radar sensors with a frequency of 24 GHz or other values are also possible. The radar sensors with the higher frequencies, however, are more suitable for the detection of pedestrians because of their higher resolution, since they have only relatively small movement speeds. The radar sensors may continue to be used in the automotive industry radar sensors.

A distance R of the object O1 to the reference object RO is determined from a signal propagation time Δt _{1 of} the emitted signal S to the object O1 and from this back to the reference object RO.

The 2A to 2C show an influence of movements of the object to be detected O1 in a detection of the object O1 based on radar data in a fast ramping process with a so-called FMCW modulation.

In 2A It is shown that for detecting the object O1 by means of the radar device 1 Radar data as signal S in the form of transmitter lamps Tx with a transmission frequency f _Tx and a bandwidth B are emitted.

By reflection at the object O1, the transmitter lamps Tx are reflected and as reception ramps Rx with a reception frequency f _Rx after the signal propagation time Δt ₁ from the radar device 1 detected.

Due to the signal propagation time Δt _{1 of} the signal S, the distance R is formed in the reception _frequency f _Rx .

For this purpose, an intermediate frequency f _{IF, O1 of} the object O1 is determined as the difference between the transmission frequency f _Tx and the reception frequency f _Rx at a defined time, the intermediate frequency f _{IF, O1} according to _FIG f _{IF, O1} ~ R [1] proportional to the distance R is.

The intermediate frequency f _{IF, O1} as a function of the time t is in 2 B shown.

2C shows an amplitude A of the signal with the intermediate frequency f _{IF, O1} as a function of the time t, wherein it can be seen that each transmitter lamp Tx and receiving ramp Rx causes a phase rotation in the baseband signal BS.

In the 3A to 3C Representations for determining and processing a speed v _{1 of} the object O1 are shown.

Due to the in 3A Doppler effect shown forms in addition to the distance R a radial, ie in the direction of a connecting line between the radar device 1 and the reflective object O1 oriented relative velocity in the receiving frequency f _Rx and thus in the intermediate frequency f _{IF, O1} . The intermediate frequency f _{IF, O1} can be obtained by means of a Fourier transformation of the baseband signal and be assigned directly to a location or the distance R.

In 3B is shown how the relative velocity of the object O1 is determined with respect to the reference object RO. In this case, the relative velocity is determined from the baseband signal BS by detecting a phase change Δφ between successive reception ramps Rx. The baseband signal BS is evaluated for this purpose at defined times, which is characterized by a ramp index k. Between the respective time points or ramp indices k, a time duration is formed, a so-called ramp repetition duration T _RRI .

3C shows the baseband signals BS for the ramp indices k = 1, 2, 3, 4 as a function of the time t.

The phase change Δφ is according to Δφ ~ ν _R [2] proportional to the relative velocity, which is indicated by the symbol v _R.

The absolute velocity v _{1 of} the object O1 is determined from the relative velocity and the intrinsic velocity of the reference object RO.

Conditions for the illustrated determination of the velocity v ₁ are that the object O1 or at least a part of the object O1, from which the velocity v _{1 is to} be determined, for each ramp index k in a same in 5B Distance cell RZ is shown and that a movement of the object O1 during the measurement period is less than a range resolution, ie less than a change in the distance R. Another condition is that the phase change Δφ between two consecutive receiving ramps Rx is less than π.

If the object O1 is a pedestrian, as opposed to a vehicle, several speeds v ₁ , ie a so-called Doppler spread, occur due to movements of the pedestrian. The occurrence of several speeds v ₁ is significantly caused by different torso and leg movements of the pedestrian. Due to the movement of the pedestrian, therefore, a marked course of speed occurs in the form of a prominent Doppler signature, which is tracked over a predetermined measurement period.

In order to determine this speed profile and to distinguish or classify the object O1, in particular for distinguishing whether the object O1 is a pedestrian or a vehicle, a point target identified with the determined distance, ie a part of the object O1, at defined times, as described above, determines the speed v ₁ and formed a two-dimensional distance / speed map RvK.

The two-dimensional distance / speed map RvK shows 4 , The prerequisite is that the speed v ₁ and the distance R can be processed independently of each other.

In the distance / speed map RvK, for objects O1, O2 removed at different distances R from the reference object RO, speeds v ₁ , v ₂ for different point targets PZ _1.1 , PZ _1.2 , PZ _1.3 , PZ _2.1 , PZ _2.2 , PZ _{2.3 are} in range cells RZ1 entered to RZn.

In this case, the distance cells RZ1 to RZn, which are each arranged in the same row of the distance / speed card RvK, have the same distances E. For the point targets PZ _1.1 to PZ _1.3 , PZ _2.1 to PZ _{2.3 of} the objects O1, O2 located for the distance cells RZ1 to RZn, the speeds v ₁ , v _{2 are} determined as described at defined times.

For each object O1, O2 and the associated point targets PZ _1.1 to PZ _1.3 , PZ _2.1 to PZ _2.3 , velocity profiles GP1, GP2 are determined. For several point targets PZ _1.1 to PZ _1.3 , PZ _2.3 to PZ _{2.3 of} the objects O1, O2 temporal velocity profiles are determined in a so-called tracking of the objects O1, O2 over several sequences based on a sequence of successively detected speed profiles GP1, GP2.

On the basis of an extraction of the temporal speed profiles of several point targets PZ _1.1 to PZ _1.3 , PZ _2.1 to PZ _2.3 belonging to an object O1, O2, which represent a so-called Doppler signature, the objects O1, O2 are classified.

The illustrated cyclic speed curves are characteristic of pedestrian-trained objects O1, O2. The point targets PZ _1.3 , PZ _2.3 when running the pedestrian forward facing legs and the point targets PZ _1.1 , PZ _2.1 when running backward facing legs. The point targets PZ _1.2 , PZ _2.2 are each part of the upper body of the pedestrian.

By contrast, a vehicle has a constant speed profile or a constant Doppler profile.

In addition to pedestrians and vehicles, such characteristic speed profiles also include cyclists, cyclists, inline skaters or other road users, so that a simple and reliable classification and differentiation of these is possible.

The pedestrian moves longitudinally to the reference object RO, ie to the radar device 1 , this is from the radar device 1 detectable and classifiable by the corresponding speed profiles or the Doppler signature.

However, if the pedestrian moves obliquely or laterally to the radar device, there is only a small Doppler shift, so that they are difficult or impossible to detect by conventional speed processing. Therefore, there are high demands on the hardware of the radar device 1 posed. In order to be able to classify the pedestrian safely nevertheless, a speed resolution is additionally increased adaptively.

The adaptive increase of the speed resolution has the great advantage that the speed profiles or Doppler shifts can be extracted more easily and in better quality if only a few speed samples are available. This means that if only a very short measuring time is available for the acquisition of measured data, speed profiles can nevertheless be extracted or improved.

In the in 4 As shown, the adaptive increase in velocity resolution has already occurred.

A possible procedure for the adaptive increase of the velocity resolution in the fast ramping process with the so-called FMCW modulation is in the 5A and 5B shown.

Two possibilities are used for increasing the speed resolution during the fast ramping process.

In a first way, the radar sends 1 a sufficient number of fast transmitter lamps Tx to achieve a high resolution. Depending on the number of ramps, the number of different speeds v ₁ , v ₂ , v ₃ per range cell RZ1 to RZn increases, the number of ramps being proportional to the number of different speeds v ₁ , v ₂ , v ₃ per range cell RZ1 to RZn. In contrast, the total duration of the ramp transmission determines the Doppler frequency, so that as the time duration increases, a distance and angular resolution is increased.

In a second possibility, the reception ramps Rx are generated virtually. This is done, for example, after the from DE 10 2010 045 980 A1 known methods. In order to save computing capacity, this method is significantly improved in combination with the adaptive adaptation of the speed resolution. The adaptive adaptation is carried out, for example, with a so-called CFAR algorithm in combination with an autoregressive linear prediction. As an alternative or in addition to the linear prediction, other high-resolution algorithms can also be used.

A combination of both possibilities shows 5A , A course of an intermediate power P _IF as a function of the intermediate frequency f _IF and a profile of the intermediate frequency f _IF as a function of the time t are formed by means of a Fourier transformation, in particular a so-called fast Fourier transformation FFT.

This course shows 5B in the left-hand illustration for three objects O1 to O3, where the objects O1, O2 are the pedestrians and the third object O3 is a vehicle.

However, only for the objects O1, O2 the increase in the velocity resolution is performed.

The two-dimensional distance / speed map RvK is formed by means of a further fast Fourier transformation FFT. Due to the increase in the velocity resolution, velocity cells vZ1 to vZm are divided or reduced. This makes the speed course more accurate and easier to extract and more accurate. As a result, the classification of the objects O1, O2 is possible in a very reliable manner.

Since the third object O3 designed as a vehicle has a uniform course of speed, a subdivision of its velocity cell vZ2 is not required, yet a classification of the object O3 is still possible in a very reliable manner.

Furthermore, since only a few measuring cycles are available for tracking the objects O1, O2 designed as crossing pedestrians, the speed resolution of the measurement is reduced. The adaptive speed resolution enhancement also allows for an increase in resolving power to improve the quality and predictive value of the velocity trajectories or Doppler signatures.

If it is not possible to adaptively increase the speed resolution, alternatively a radar sensor of the radar device, which is normally oriented in the direction of movement of the reference object RO, can be used 1 be directed by up to 90 ° to the side of the reference object RO. This also makes crossing pedestrians easier to recognize, as the Doppler shift increases during a movement phase. Alternatively or additionally, a plurality of differently oriented radar sensors are provided.

Alternatively or in addition to that in the 1 to 5B Determining the speed profiles using the distance cells RZ1 to RZn these are determined using not shown angle cells. The determination is carried out in almost the same way, in contrast to the distance cells RZ1 to RZn angle cells are formed for the same angle and alternatively or in addition to the distance cells RZ1 to RZn for located in the angle cells point targets at least the objects O1, O2 at defined times the speeds v ₁ , v ₂ and from this the velocity profiles GP1, GP2 are determined. The ascertainment of the speed profiles also takes place here on the basis of the succession of speed profiles GP1, GP2, which are acquired sequentially in time.

Furthermore, it is additionally or alternatively possible in a manner not shown that from the distance R and an associated angle, a two-dimensional angle / distance map is formed. In this case, angular / distance cells are formed for the same distances R and the respectively associated angle, and for speeds in the angle / range cells of the objects O1, O2, the speeds v ₁ , v _{2 are determined} at defined times. From the speeds v ₁ , v ₂ , in turn, the velocity profiles GP1, GP2 are formed, which are strung together to determine the speed profiles in the manner described. By tracking an object O1, O2, O3 z. The classification of the object O1, O2, O3 is carried out, for example, from one distance / angle cell to the next, while the speed profiles are juxtaposed and the resulting speed profiles. For the speed analysis is thereby a high speed resolution capability, which is ensured even with crossing objects O1, O2, O3 by the adaptive increase in the speed resolution.

Furthermore, the illustrated method for determining the speed characteristics and increasing the speed resolution in alternative radar applications with different modulation methods, such as the so-called LFMSK method (linear frequency modulated shift keying) or the so-called FMFSK method (frequency modulated frequency shift keying ), feasible.

• – ein Objekt O1, O2, O3 wird in einer Entfernungszelle RZ1 bis RZn detektiert;
• – ein Geschwindigkeitsprofil GP1, GP2, GP3 bzw. Dopplerspektrum wird für die entsprechende Entfernungszelle RZ1 bis RZn prozessiert und abgespeichert
• – falls eine adaptive Steuerung der Geschwindigkeitsauflösung erforderlich ist, Kombination vom so genannten CFAR-Algorithmus und der autoregressiven linearen Prädiktion
• – Tracking des Objektes O1, O2, O3 mittels Aneinanderreihung der Geschwindigkeitsprofile GP1, GP2, GP3 über mehrere Sequenzen
• – Ermittlung der Klasse des Objektes O1, O2, O3 aus dem zeitlichen Geschwindigkeitsverläufen

In summary, a chronological sequence of a possible embodiment of the method according to the invention is shown in a highly simplified manner:

• An object O1, O2, O3 is detected in a range cell RZ1 to RZn;
• A speed profile GP1, GP2, GP3 or Doppler spectrum is processed and stored for the corresponding range cell RZ1 to RZn
• - if an adaptive control of the velocity resolution is required, combination of the so-called CFAR algorithm and the autoregressive linear prediction
• Tracking the object O1, O2, O3 by sequencing the velocity profiles GP1, GP2, GP3 over several sequences
• - Determination of the class of the object O1, O2, O3 from the temporal velocity curves

LIST OF REFERENCE NUMBERS

1

radar device

B

bandwidth

BS

Baseband signal

f

frequency

FFT

Fast Fourier Transformation

f _IF

intermediate frequency

f _{IF, O1}

intermediate frequency

f _Rx

receiving frequency

f _Tx

transmission frequency

GP1, GP2, GP3

velocity profile

k

ramp Index

O1 to O3

object

P _IF

between performance

PZ _1.1 to PZ _1.3

point target

PZ _2.1 to PZ _2.3

point target

PZ ₃

point target

R

distance

RO

reference object

RvK

Range / speed map

Rx

reception ramp

RZ

Distance cell

RZ1 to RZn

Distance cell

S

signal

t

Time

T _RRI

Ramp repetition period

Tx

send ramp

v ₁ , v ₂ , v ₃

speed

vZ1 to vZm

speed cell

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 patent literature

• DE 102010045980 A1 [0002, 0053]

Claims (5)

Method for detecting and classifying objects (O1, O2, O3) on the basis of radar data, wherein a distance (R), a velocity (v ₁ , v ₂ , v ₃ ) and / or an angle of at least one object (O1 , O2, O3) with respect to a reference object (RO) are or is characterized , characterized in that for the same distances (R) distance cells (RZ1 to RZn) and / or angular angles for the same angle are formed and for in the distance cells (RZ1 to RZn) and / or angular cells located point targets (PZ _1.1 to PZ _1.3 , PZ _2.1 to PZ _2.3 , PZ ₃ ) at least one object (O1, O2, O3) at defined times the velocities (v ₁ , v ₂ , v ₃ ) and From this, velocity profiles (GP1, GP2) can be determined, wherein for a plurality of point targets (PZ _1.1 to PZ _1.3 , PZ _2.1 to PZ _2.3 , PZ ₃ ) of the at least one object (O1, O2, O3) based on a sequence of speed profiles (GP1 , GP2) temporal Geschwin be determined on the basis of the temporal velocity curves of several to an object (O1, O2, O3) associated point targets (PZ _1.1 to PZ _1.3 , PZ _2.1 to PZ _2.3 , PZ ₃ ) the at least one object (O1, O2, O3) becomes. A method according to claim 1, characterized in that from the distance (R) and the angle, a two-dimensional angle / distance map is formed, for equal distances (R) and a respective associated angle / distance cells are formed and for in the angle - / distance cells located point targets (PZ _1.1 to PZ _1.3 , PZ _2.1 to PZ _2.3 , PZ ₃ ) at least one object (O1, O2, O3) at defined times the velocities (v ₁ , v ₂ , v ₃ ) and from there velocity profiles ( GP1, GP2) are determined, wherein for a plurality of point targets (PZ _1.1 to PZ _1.3 , PZ _2.1 to PZ _2.3 , PZ ₃ ) of the at least one object (O1, O2, O3) based on a sequence sequentially timed speed profiles (GP1, GP2) temporal velocity curves are determined, based on the temporal velocity curves of several to an object (O1, O2, O3) associated point targets (PZ _1.1 to PZ _1.3 , PZ _2.1 to PZ _2.3 , PZ ₃ ) the lower an object (O1, O2, O3) is classified. A method according to claim 1 or 2, characterized in that the object (O1, O2, O3) is classified at a cyclic speed course as a pedestrian. Method according to one of the preceding claims, characterized in that the object (O1, O2, O3) is classified as a vehicle at a substantially constant speed course. Method according to one of the preceding claims, characterized in that a size of a speed resolution is adjusted adaptively.

Images