Sie sind auf Seite 1von 4

Moving Target Range Detection Algorithm for FMCW Radar

Eugin Hyun *, Jonghun Lee **


*
ART (Advanced Radar Technology) Lab. Robotics Division,
DGIST (Daegu Gyeongbuk Institute of Science & Technology), Korea
Primary author, email: braham@dgist.ac.kr
**
ART (Advanced Radar Technology) Lab. Robotics Division,
DGIST (Daegu Gyeongbuk Institute of Science & Technology), Korea
Corresponding author, email: jhlee@dgist.ac.kr

Abstract: In the typical FMCW radar, the two-step FFT method is mainly used to extract
the target velocity and range. While this approach is very useful to detect the
simultaneous range and velocity, the computational effort is very high because all range
bins are computed by FFT (Fast Fourier Transform) to extract the target velocity.
Moreover, in some radar applications, the only moving target range is the range of
interest, and the velocity information is not required. In this paper, we propose a type of
low-complexity hardware architecture for FMCW radar that detects the moving target’s
range. Because this method uses two simple detectors an all-target detector and a
stationary-target detector instead of the two-step FFT, the computational complexity is
reduced. By means of a simulation, we verify the proposed method in a comparison with
a typical method.

1. Introduction
FMCW radar has been used in the various applications, such as vehicles, security, defense,
ships, and measurements. In FMCW radar, a frequency-modulated continuous signal is
transmitted and the returned signal reflected from a target is mixed with the transmitted signal.
A beat signal generated by a mixer is used to measure the target range and velocity. FMCW
radar is very effective, but radar ambiguities occur due to the Doppler shift. That is, there are
some limitations when detecting simultaneously the range and velocity in multi-target
situations [1][2][3].
Typically, in order to resolve range and velocity ambiguities, the fast ramp based method
has been very popular [4][5]. Figure 1 presents the proposed method. In the conventional
method, the FFT is used instead of the MTD (moving target detection) algorithm). That is, in
the typical approach, because a range-Doppler matrix map is generated through a two-step
FFT process, extra algorithms such as moving target detection and a clutter cancellation
algorithm are not required. Because all stationary targets including clutter are gathered at the
range positions with zero Doppler, we can easily determine the moving targets.
This approach is very useful when used to detect the range and velocity of a moving target
and suppress clutter, including the stationary targets. However, the computational complexity
is very high and the required hardware resources are considerable. Moreover, in some
applications such as security systems, only the range information of moving targets is
required. In such a case, information to identify whether or not the detected target is a moving
object is required. That is, the target’s velocity value is not required. Therefore, for these
applications, the two-step FFT-based fast ramp approach has overhead that increases the
processing effort.
In this paper, we propose a signal processing architecture for moving target range detection
with low complexity.
2. The proposed method
Figure 1 presents the proposed signal processing concept to identify the moving target and
to detect the target range.
First, sample data received through the ADC (analog digital converter) is generated into the
frequency spectrum with range information by the first FFT processors. Here, R (m,n )
presents the data of the m th range bin the on nth ramp.
In the typical architecture, the second FFT processors extract the Doppler spectrum of the
target using data from the same range bins. Here, PRI is the pulse repletion time. The
maximum detectable Doppler frequency can be expressed as 1/ 2  PRI .
However, in the proposed method, instead of a second FFT process, the proposed MTD
processor is used. The proposed MTD is divided two parts: the ATD (all-target-detector) part
and the STD (stationary-target-detector) part. First, in the ATD, all targets including moving
targets, stationary targets, and clutter are detected in every range bin. Second, the STD detects
stationary targets and clutter.

Fig. 1. The proposed signal processing concept to detect moving target range. In the typical method, the FFT is employed
instead of MTD

In the ATD, non-coherent integration is used for every N ramp in order to enhance the
SNR (signal-to-noise ratio). The STD consists of coherent integration with ego-speed
compensation. The MTD and STD are expressed by equation (1), where a(m ) and s (m )
constitute the range profile of the mth range bin. If the installed radar system is fixed, the ego-
velocity is zero and k (ego-Doppler frequency index) is zero. Otherwise, k has the
corresponding value of the ego-speed. The value of k can be determined using a LUT (look-
up table), as shown in Table I, where the velocity step V is calculated using the Doppler
frequency step fd  1/ PRI  M .
Last, the range of the moving target can be detected by comparing the ATD and STD
results.

N 1 N 1 j 2n
 k
a (m )  
n 0
R (m,n ) , s (m )  
n 0
R (m ,n )  e N (1)
Table I. Look-up table used to calculate the ego-Doppler frequency index
M M 
Ego-Velocity  V ...  V 0 V ...   1  V
2 2 
Ego-Doppler M M
frequency index ... M 1 0 1 ... 1
(k) 2 2

3. The simulation results


The typical method and the proposed algorithm are simulated using MATLAB. The
transmitted bandwidth is 200 MHz, the center frequency is 76.5GHz, the PRI is 0.1ms, and
the number of ramps is 64. Moreover, the ADC sampling rate is 5MHz, and the first FFT
point is 512. Thus, the number of range bins is 256 (=first FFT point * 0.5). To simulate the
typical approach, the second FFT point is set to 64. These system parameters are intended to
meet the following performance targets: a maximum range of 50m, a maximum velocity of
30km/h, a range resolution of 1m, and a velocity resolution of 2km/h.
In this simulation, the ranges and velocities of objects are assumed as shown in Figure 2. In
this scenario, there are three targets inside the radar detection range: two moving targets and
one stationary target. One moving target is moving away and another target is coming into the
radar system.

Fig. 2. The simulation example

Figure 3 shows the detection results for the typical method. First, Figure 3(a) shows the
range profile, where the x-axis is the range and the y-axis is the PSD (power spectrum
density). Two moving targets located at 10m and 50m are detected with multiple scatter
points. The velocity information and the target range are marked in Figure 3(b). Using the
target’s distance and speed, we can easily determine whether or not the detected target is a
moving target.

Fig. 3. The detection result for the typical method, (a) the Fig. 4. The detection result for the proposed method, (a)
range profile, (b) the corresponding range-velocity map ATD and (b) STD
result
Figure 4 presents the proposed method results. Figures 4(a) and (b) show the results of the
ATD and the STD. These figures show that the result of the ATD is similar to that of the
typical method, as shown Figure 3(a). That is, in the ATD, all targets are detected using non-
coherent integration. It was also noted that only the stationary target is detected in the STD
result by means of coherent summation. Finally, the moving targets are determined as 10m
and 50m using the ATD and STD results.

4 Conclusion
In this paper, we proposed a low-complexity type of hardware architecture for moving
target range detection. While the size of the memory depends on the range bins and the
velocity bins in the typical method, only three memory spaces for the size of the range bins
are required in the proposed method. Moreover, the computational effort is also reduced
compared with the typical method, which extracts all velocity values. Therefore, we expect
that the proposed method will be useful for applications such as security systems, in which
only moving target range detection is required. The next research step is a field test under
various environments. In addition, comparisons with other methods can be another subject.

Acknowledgments

This work was supported by the DGIST R&D Program of the Ministry of Education, Science
and Technology of Korea (13-RS-02).

References:
[1] Shunji Miyahara, “New Algorithm for Multiple Object Detection in FM-CW Radar”, SAE
Technical paper series, 2004, 2004-01-0177
[2] Eugin Hyun, Jong-Hun Lee, “A Method for Multi-target Range and Velocity Detection in
Automotive FMCW Radar”, Proceedings of the 12th International IEEE Conference on
Intelligent Transportation System, 2009, pp. 7-11
[3] Eugin Hyun, Woojin Oh, Jong-Hun Lee, “Development of FMCW Radar Signal Processing
Module for Civilian and Defense Vehicle Applications”, Proceedings of 2011 International
Symposium on Embedded Technology, 2011, pp. 1-4
[4] Winkler, V.: Range Doppler detection for automotive FMCW Radar. Proceedings of the 4th
IEEE European Radar Conference, pp. 166-169. (2011)
[5] Eugin Hyun, Sang-Dong Kim, Yeong-Hwan Ju, Jong-Hun Lee, Eung-Noh You, Jeong-Ho
Park, Dong-Jin Yeom, Sang-Hyun Park, Seung-Gak Kim, “FPGA Based Signal Processing
Module Design and Implementation for FMCW Vehicle Radar Systems”, Proceedings of
2011 IEEE CIE International Conference on Radar, 2011, pp. 237-275

Das könnte Ihnen auch gefallen