阅读说明：本技术 一种基于快速fmcw雷达的道路车辆低运算量监测方法 (Road vehicle low-computation monitoring method based on rapid FMCW radar ) 是由 房晓辉 韩海峰 于 2019-10-10 设计创作，主要内容包括：本发明公开了一种基于快速FMCW雷达的道路车辆低运算量监测方法,利用快速FMCW雷达获取道路上车辆的反射回波,与雷达的发射波进行混频后得到此时刻的拍频,拍频信号经过ADC采样发送给计算模块；在计算模块中只在估计目标车辆存在的情况下进行角度检测,而不是在每帧中检测角度,从而进一步降低了计算复杂度。由于运动目标车辆并不总是存在,因此本发明在运动目标车辆不存在的情况下减少了不必要的处理。(The invention discloses a road vehicle low computation amount monitoring method based on a rapid FMCW radar, wherein the rapid FMCW radar is used for acquiring a reflection echo of a vehicle on a road, the reflection echo is mixed with a transmission wave of the radar to obtain a beat frequency at the moment, and a beat frequency signal is sent to a calculation module through ADC sampling; the angle detection is performed only in the case where the presence of the target vehicle is estimated in the calculation module, instead of detecting the angle in each frame, thereby further reducing the calculation complexity. Since the moving object vehicle does not always exist, the present invention reduces unnecessary processing in the case where the moving object vehicle does not exist.)

1. A road vehicle low computation monitoring method based on a rapid FMCW radar is characterized by comprising the following contents:

the method comprises the steps that a fast FMCW radar is used for obtaining a reflection echo of a vehicle on a road, a beat frequency at the moment is obtained after the reflection echo is mixed with a transmission wave of the radar, and a beat frequency signal is sent to a calculation module after ADC sampling; in the calculation module:

the transmitted wave signal may be represented by x (t) as:

T_Frepresenting a period of one frame, having N_FA frame, wherein the ith frame contains L swept frequencies chirp, then x⁽ⁱ⁾(t) represents the emission at the ith frame, expressed as:

then x₀(t) represents a chirp, i.e., chirp, which can be expressed as:

x₀(t)＝exp(j2π(f₀t+μt²/2))

0≤t≤T，f₀carrier frequency, μ is sweep frequency, i.e., μ ═ B/T;

the radar receiving antenna array is a linear array formed by K antennas, a single vehicle target is arranged in front of a radar, the ith chirp signal of a certain frame is reflected by the single vehicle target in front of the radar, and a vehicle echo received by the kth antenna is represented as:

wherein a is₀In order to be the amplitude value,white Gaussian noise signal, f_DIs Doppler frequency shift, theta is the included angle between the antenna and the vehicle, and d is the distance between the antennas; then the velocity term is v^l(f_D)＝exp(j2πf_DlT) and the term of the direction angle is

then beat signal y_l,k(t) can be expressed as:

where f is_bμ τ is the beat frequency, τ is the time interval between the transmitted wave and the echo; in addition, w_l,k(t) and a can be expressed as:

carrying out ADC on the beat signal to obtain a sampled beat signal:

y_l,k(n)＝ax_b(n)v^lψ^k+w_l,k(n)

therefore, when multi-target detection is performed, if M vehicles are arranged in front of the radar, the mth vehicle is monitored, and the beat frequency signal of the ith frame can be expressed as:

here, the

Selecting two beat signals, the L-th beat signal₁1 and l₂A beat signal, y_l1,k[n]And y_l2,k[n]Subtracting them from each other and using d_l1l2,k[n]Represents:

d_l1l2,k[n]＝y_l1,k[n]-y_l2,k[n]＝ax_b[nT_s]ψ^k(v^l1-v^l2)+w_l1,k[n]-w_l2,k[n]

omit noise term, let:

d_l1l2,k[n]＝βx_b[nT_s]ψ^k

where β is a (v)^l1-v^l2) (ii) a And for the above formula, performing distance FFT and angle FFT on the distance term and the angle term respectively to obtain distance and angle information.

2. The fast FMCW radar-based road vehicle low computation monitoring method of claim 1 wherein two chirp, i.e./, are chosen correctly₁,l₂So that β ≠ 0.

3. A fast FMCW radar-based low computation monitoring method for road vehicles as claimed in claim 2, wherein the speed of the target vehicle is known by the speed term v^l＝exp(j2πf_DlT); the fields composed for the velocity term are referred to below as chirp fields; one frame consists of L chirp, so that L sampling points exist in a chirp domain, the interval time between the sampling points is the period T of the chirp, and the T can be regarded as a sampling period of a phase; in addition, the period can be obtained from the doppler shift formula as follows:

wherein [.]Is a rounding symbol;

but for multi-target vehicles or target vehicles with unknown speed,

Therefore:

4. the fast FMCW radar-based road vehicle low computation monitoring method of claim 1, wherein the computing module is an FPGA chip or a DSP chip.

Technical Field

The invention relates to a road vehicle low-computation monitoring method based on a rapid FMCW radar, and belongs to the technical field of radar monitoring.

Background

Radar sensor has been by automobile autopilot, aspect fields wide application such as road monitoring, for ultrasonic sensor, optical sensor, the radar is hardly influenced by environmental factor such as haze, weather, light. In a radar road monitoring system, a continuous modulation wave (FMCW) radar is more popular because of its lower hardware requirements and simple design than a pulse wave radar.

In the traditional FMCW radar, a modulation waveform mainly comprises a sawtooth wave and a triangular wave, and the period T of a single sweep chirp is usually 1 to 10ms, so that when the chirp is the sawtooth wave, the distance-speed coupling phenomenon exists due to the fact that the time width is large. The FMCW of the triangular wave modulation can eliminate the distance-speed coupling phenomenon through ascending and descending sweep frequencies, and determine the unique value of the single target distance and speed. However, for multi-target monitoring, because the radar parameters are the same, a plurality of groups of parallel straight lines are generated in frequency bands of the transmitted signals of a plurality of targets above and below the frequency sweep, for example, four intersections exist between two targets, so that two false alarm targets exist, and as the number of targets increases, false targets multiply. In order to overcome the multivalue, another modulation straight line with scanning frequency is often added, but the design difficulty and the requirement of radar hardware in road monitoring are increased.

The above problem can be solved by modulating a fast FMCW wave by a sawtooth wave with an increased scanning frequency. Thus, the chirp period is in a microsecond level, and the bandwidth is in a GHz range, so that the vehicle speed frequency shift generated by the Doppler effect is ignored relative to the distance frequency shift, and the influence of the Doppler effect is ignored when the range is calculated by using beat frequency, thereby eliminating the distance-speed coupling phenomenon; meanwhile, the phase in each distance unit represents the speed change, and the information of the speed and the distance is obtained by utilizing two-dimensional FFT.

The full-dimensional FFT firstly carries out the FFT operation of beat frequency aiming at each chirp of each frame, namely a distance domain, and obtains the distance; then, performing phase FFT on all chirp, namely chirp domains in a frame to obtain speed information; when the AOA calculation is needed, FFT calculation is needed to be carried out on the data of the whole receiving antenna array, and the FFT calculation is 3D-FFT. Obviously, the data volume is large, and the number of the FFT operation points is limited, so that the calculation is complex, the operation amount is large, the operation time delay is increased, more resources are consumed, and the FMCW radar is not suitable for a simple FMCW radar with low cost and high time delay requirement.

Disclosure of Invention

The purpose of the invention is as follows: aiming at the problems and the defects in the prior art, the invention provides a road vehicle low-computation monitoring method based on a rapid FMCW radar.

The technical scheme is as follows: a road vehicle low computation monitoring method based on a rapid FMCW radar comprises the following contents:

the method comprises the steps that a fast FMCW radar is used for obtaining a reflection echo of a vehicle on a road, a beat frequency at the moment is obtained after the reflection echo is mixed with a transmission wave of the radar, and a beat frequency signal is sent to a calculation module after ADC sampling; in the calculation module:

the transmitted wave signal may be represented by x (t) as:

T_Frepresenting the period of a frame, N_FRepresenting the number of frames, containing L swept frequencies chirp in the ith frame, then x⁽ⁱ⁾(t) represents the emission at the ith frame, expressed as:

then x₀(t) represents a chirp which can be expressed as:

x₀(t)＝exp(j2π(f₀t+μt²/2))

where 0. ltoreq. t.ltoreq.T, f₀The carrier frequency, T is the sweep period, and μ is the sweep frequency, i.e., μ ═ B/T, B denotes the bandwidth. The fast FMCW radar wave is embodied in the frequency band B of GHz magnitude; the sweep period, i.e., one chirp period T, is us magnitude.

The radar receiving antenna array is a linear array formed by K antennas, a single vehicle is arranged in front of a radar, the ith chirp signal of a certain frame is reflected by the single vehicle in front of the radar, and the echo of the vehicle received by the kth antenna is represented as:

wherein a is₀For amplitude, λ represents the wavelength,white Gaussian noise signal, f_DAnd theta is the Doppler frequency shift, and theta is the included angle between the antenna and the vehicle. Then the velocity term is v^l(f_D)＝exp(j2πf_DlT) and the term of the direction angle is

Then the echo function x_l,k(t) can be simplified as:

then the beat signal y obtained after mixing_l,k(t) can be expressed as:

where f is_bμ τ is the beat frequency (beat frequency) obtained after mixing,

representing the transmitted wave x₀(t) and τ represents the time interval between the emission and the reception of the echo, x_b(t) is the sinusoid of the beat signal. In addition, w_l,k(t) and a can be expressed as:

by subjecting the beat signal to ADC at a sampling rate of f_sObtaining a beat frequency signal after sampling:

y_l,k(n)＝ax_b(nT_s)v^lψ^k+w_l,k(n)

therefore, when multi-target detection is performed, if M vehicles are arranged in front of the radar, the mth vehicle is monitored, and the beat frequency signal of the ith frame can be expressed as:

here, the

For the distance term, FFT is performed for n, and the peak position is the beat frequency, which is the frequency shift due to the distance, to obtain the distance, hereinafter referred to as distance FFT,the FFT is performed for l as a velocity term, reflecting the phase change caused by the velocity, thereby obtaining the velocity, hereinafter referred to as velocity FFT,for the angle-of-arrival term, an FFT is performed for k, and the angle can be obtained, hereinafter referred to as the FFT,

in order to be a noise term, the noise term,

representing the amplitude value of the beat signal of the mth vehicle in the ith frame. L beat signals in one frame pass through ADC to obtain L groups of N_sPoint discrete digital signal data

Selecting two beat signals, the L-th beat signal ₁1 and l₂A beat signal, y_l1,k[n]And y_l2,k[n]Subtracting them from each other and using d_l1l2,k[n]Represents:

d_l1l2,k[n]＝y_l1,k[n]-y_l2,k[n]＝ax_b[nT_s]ψ^k(v^l1-v^l2)+w_l1,k[n]-w_l2,k[n]

if there is no movementTarget vehicle, Doppler frequency due to velocity being zero, i.e. f_D0; at this time, in the above formula, only clutter or noise terms exist; this means that only the noise term is distance FFT; on the other hand, in the case of a moving target vehicle, v^l1And v^l2Not equal, the Doppler frequency is not zero, unless l₁＝l₂Or l₂＝l₁+ lT, l is an integer and T is the chirp period. Therefore, omitting the noise term (here, the signal-to-noise ratio can be increased, or the filter can reduce the noise, so as to omit the noise term), let:

d_l1l2,k[n]＝βx_b[nT_s]ψ^k

where β is a (v)^l1-v^l2)。

From the above formula, it can be seen that the beat signal x_b[nT_s]Containing distance information and a direction angle term psi^kStill present. And performing distance FFT and angle FFT on the distance term and the angle term respectively to obtain distance and angle information.

Further, two chirp, i.e./, need to be selected correctly₁,l₂So that β ≠ 0. This is done because situations where the algorithm does not work properly are to be avoided. If two l₁And l₂The phase change is symmetrical and the subtracted output may be zero as if there were no moving objects, which gives rise to a so-called velocity blind spot. In order to overcome the above problems, there are two modes: in one case, the target vehicle is single, its speed is known; in another case, the target vehicle is not unitary or the speed of the target vehicle is unknown.

The fields composed for the velocity term are referred to below as chirp fields; one frame consists of L chirp, so that L sampling points exist in a chirp domain, the interval time between the sampling points is the period T of the chirp, and the T can be regarded as a sampling period of a phase; knowing the speed of the target vehicle, the velocity term v for the chirp domain^l＝exp(j2πf_DlT), the period can be obtained from the doppler shift formula as follows:

for the m-th target vehicle speed

The doppler shift of (1). Within one cycle, the number of samples can be recorded as:

wherein [.]Is a rounded symbol. In the first case of the above-mentioned case,

is that

Uniform distribution in the interval. But for multi-target vehicles or target vehicles with unknown speed,

it cannot be determined that in this case, a chirp is randomly selected over the entire frame, i.e.

Therefore:

randomly selecting two beat signals in L sweep frequencies within the frame, and even if the difference value is zero in one frame, the calculation of the next stage is not performed, but two beat signals are randomly selected in the next frame, and the calculation of the next stage is performed at the momentWill change such that

At this time, other calculation of FFT is carried out to obtain a monitoring conclusion, so that the problem that the target cannot be identified at certain specific speed is avoided.

Has the advantages that: compared with the prior art, the road vehicle low-computation monitoring method based on the rapid FMCW radar only carries out angle detection under the condition that the target vehicle is estimated to exist, and does not detect the angle in each frame, so that the calculation complexity is further reduced. Since the target vehicle does not always exist, the present invention reduces unnecessary processing in the case where the target vehicle does not exist.

The invention solves the problem that the previous algorithm can not detect the target vehicle with specific speed. The probability of a slow or fast target vehicle being lost is reduced by randomly selecting two chirp per frame. Secondly, the distance and angle are detected without performing two-dimensional FFT for each frame, and when it is determined that a moving target vehicle exists, the distance and angle of the target are detected by performing two-dimensional FFT.

The algorithm of the invention is realized on an FPGA chip, only one ADC serial input can be used, thus reducing the application of pins and generating loose margin on the time sequence, because two beat frequency signals are random, the interval time of the two beat frequency signals has no strict requirement, when the data acquisition of the beat frequency signal ADC in one sweep frequency period is finished, the data acquisition can be finished after the buffer memory is finished, and then the beat frequency data of the other sweep frequency period is acquired, thus the resources are multiplexed in this respect, the area is saved, and the resource space can be reserved for the IP of the algorithm.

Drawings

FIG. 1 is a schematic diagram of a conventional detection method;

fig. 2 is a schematic diagram of a low complexity algorithm based on two frequency sweep chirp;

fig. 3 is a flow chart of a modified scheme for performing a low-complexity algorithm based on two swept-frequency chirp.

Detailed Description

The present invention is further illustrated by the following examples, which are intended to be purely exemplary and are not intended to limit the scope of the invention, as various equivalent modifications of the invention will occur to those skilled in the art upon reading the present disclosure and fall within the scope of the appended claims.

The road vehicle low computation monitoring method based on the rapid FMCW radar comprises the following contents:

the method comprises the steps that a fast FMCW radar is used for obtaining a reflection echo of a vehicle on a road, a beat frequency at the moment is obtained after the reflection echo is mixed with a transmission wave of the radar, and a beat frequency signal is sent to a calculation module after ADC sampling; in the calculation module:

the transmitted wave signal may be represented by x (t) as:

T_Frepresenting a period of one frame, having N_FA frame, wherein the ith frame contains L swept frequencies chirp, then x⁽ⁱ⁾(t) represents the emission at the ith frame, expressed as:

then x₀(t) represents a chirp which can be expressed as:

x₀(t)＝exp(j2π(f₀t+μt²/2))

where 0. ltoreq. t.ltoreq.T, f₀The carrier frequency, T, is the sweep period, and μ is the sweep frequency, i.e., μ ═ B/T.

The radar receiving antenna array is a linear array formed by K antennas, a single vehicle is arranged in front of a radar, the ith chirp signal of a certain frame is reflected by the single vehicle in front of the radar, and the echo of the vehicle received by the kth antenna is represented as:

wherein a is₀In order to be a complex amplitude value,

white Gaussian noise signal, f_DAnd theta is the Doppler frequency shift, and theta is the included angle between the antenna and the vehicle. Then the velocity term is v^l(f_D)＝exp(j2πf_DlT) and the term of the direction angle is

Then the echo function x_l,k(t) can be simplified as:

then the beat signal y obtained after mixing_l,k(t) can be expressed as:

where f is_bμ τ is the beat frequency, x_b(t) is the sinusoid of the beat signal. In addition, w_l,k(t) and a can be expressed as:

by subjecting the beat signal to ADC at a sampling rate of f_sObtaining a beat frequency signal after sampling:

y_l,k(n)＝ax_b(nT_s)v^lψ^k+w_l,k(n)

therefore, when multi-target detection is performed, if M vehicles are arranged in front of the radar, the mth vehicle is monitored, and the beat frequency signal of the ith frame can be expressed as:

here, the

In the form of a distance term, the distance term,

in order to be the term of the speed,

in order to arrive at the angle term,is a noise term.

As shown in fig. 1, the conventional 3-dimensional FFT algorithm performs FFT on a distance domain, a chirp domain, and an antenna array domain sequentially. L beat signals 4 in one frame pass through ADC to obtain L groups of N_sPoint discrete digital signal data

First, for each groupTo N_RFFT of the points.

Wherein N is_sThe number of sampling points of the beat signal after sampling is obtained. It can be seen that N is done here_RPoint FFT for distance term

FFT derived N_RThe frequency resolution of the point result is called the range unit, since the frequency resolution is to distinguish the two peaks by the minimum frequency difference, and the corresponding range resolution is to distinguish the two vehicles by the minimum distance. Since the row of the matrix 5 stores the output data after FFT, the horizontal direction is a distance domain and the vertical direction marks the number of chirp, and since there are L chirp in the matrix 5, the vertical direction is a chirp domain.

Then, FFT is performed on the same range bin, i.e. on matrix 5To do N_cAnd (6) point FFT.

It can be seen that N is done here_cPoint FFT is for velocity term in radar systems

Thereby obtaining corresponding information of the speed. The 2-dimensional FFT performed by the above two equations yields N_R×N_cSince there are K receiving antennas, the matrix of (1) is calculated K times to obtain N_R×N_cAnd outputting the xK two-dimensional FFT data. For matrix 6 antenna array directions N_AAnd inputting the point FFT to obtain angle information.

Obviously, the 3-dimensional FFT has higher operation complexity and more required hardware resources, and if the 3-dimensional FFT is implemented in an FPGA, on one hand, the requirement on logic resources is high, and the requirement on timing sequence is high, so that the FFT is not suitable for a project with lower cost, high stability requirement and fast operation response speed.

For FMCW radar monitoring systems, a first important issue is the rapid determination of the presence of a target vehicle of interest, here a moving target, for monitoring purposes. Without the target vehicle, we do not need to use resources unnecessarily. That is, the present invention should only operate when there is a target vehicle of interest, that is, only the moving target vehicle will generate doppler shift of radar wave, the velocity terms of two chirp periods will be different, because of the fast FMCW wave used herein, the chirp period is very short, the distance terms and angle terms of different chirp periods are not affected by doppler shift, so that when the difference signals are subtracted, the difference value of velocity terms will be obtained, and when the difference value is zero, it indicates that the object has no doppler shift, and is a stationary object and is not the target of interest. To this end, the present invention first determines whether a moving target vehicle is present, and if there is no moving target vehicle, the doppler frequency is zero, i.e., f_D0. The parameter detection operation is performed after determining that the target exists. This is achieved byThe problem of high complexity of full-dimensional FFT operation is effectively solved, and resources are saved. Two beat signals are selected from the L beat signals. For example, the first₁1 and l₂A beat signal, y_l1,k[n]And y_l2,k[n]Subtracting them from each other and using d_l1l2,k[n]Shows the following in FIG. 2:

d_l1l2,k[n]＝y_l1,k[n]-y_l2,k[n]＝ax_b[nT_s]ψ^k(v^l1-v^l2)+w_l1,k[n]-w_l2,k[n]

if there is no moving target vehicle, the Doppler frequency is zero, i.e. f_D0; at this time, in the above formula, only clutter or noise terms exist; this means that only the noise term is FFT; on the other hand, in the case of a moving target vehicle, v^l1And v^l2Not equal, the Doppler frequency is not zero, unless l₁＝l₂Or l₂＝l₁+ lT, l is an integer and T is the chirp period. Thus, the noise term is omitted, let:

d_l1l2,k[n]＝βx_b[nT_s]ψ^k

where β is a (v)^l1-v^l2)。

From the above formula, it can be seen that the beat signal x_b[nT_s]Containing distance information and a direction angle term psi^kStill present. And performing FFT on the distance term and the angle term respectively to obtain distance and angle information.

Further, two chirp, i.e./, need to be selected correctly₁,l₂So that β ≠ 0. This is done because situations where the algorithm does not work properly are to be avoided. If two l₁And l₂The phase change is symmetrical and the subtracted output may be zero as if there were no moving object, which gives rise to a so-called velocity blind zone, which is a decision to go further for β being 0 within a frame. Since the system operation time is in the frame period, each frame is selected to be l₁,l₂Therefore, the state of erroneous determination is always present. To overcome the above problems, the above problems are avoidedIn case of misjudgment, the invention judges in a plurality of frames, when two chirp in one frame obtain beta as 0, then compares the second frame, and randomly selects other chirp periods l₃,l₄If a limited number of frames have β ═ 0, it is determined that there is no moving object, otherwise, it is determined that there is a moving object, and then correlation operations such as FFT are performed, as shown in fig. 3.

The selection interval of Chirp can be divided into two modes: in one case, the target vehicle is single, its speed is known; in another case, the target vehicle is not unitary or the speed of the target vehicle is unknown.

Knowing the speed of the target vehicle, in the chirp domain, by the speed term v^l＝exp(j2πf_DlT), the period is calculated as:

for the m-th target vehicle speed

The doppler shift of (1). In addition, T is the period of one sweep chirp, and then in the chirp domain, T is the sample interval. The number of points in a cycle can be written as:

wherein [.]Is a rounded symbol. In the first case of the above-mentioned case,

is that

Uniform distribution in the interval. But for multi-target vehicles or target vehicles with unknown speed,it cannot be determined that in this case, a chirp is randomly selected over the entire frame, i.e.

Therefore:

two beat signals are randomly selected among the L swept frequencies chirp within the frame, even if the difference is made zero in one frame, but in the next frame,

will change such thatThis avoids the problem of the target not being identifiable at certain speeds.