Sie sind auf Seite 1von 3

Cooperative target localization and tracking for random sparse array

through-wall-radar

Xiang Yuan, Lingjiang Kong*, Yong Jia

*School of Electronic Engineering,


University of Electronic Science and Technology of China
Chengdu, Sichuan, P.R.China, 611731
Email: lingjiang.kong@gmail.com

In applications of through wall moving target tracking,


Abstract handheld sensors are applied immediately in the place of
operation. The requirement results in that the handheld
To improve the precision of target localization and system usually uses small antenna array necessary for
tracking, a novel through-wall-radar system based on random detection and tracking of the target. The configuration with
sparse array is developed and an approach of cooperative one stationary transmitter and multiple receivers is adopted in
target localization and tracking is proposed. Simulation the TWR system. The stationary transmitter has good
results validate the approach for the novel system. transmitting power and illuminates the target area. The
receivers move as randomly as possible around the target area
Keywords: target localization and tracking; through-wall- and keep widely separated to collect echo signals over a wide
radar; random sparse array. angular area. Therefore we can expect that the target position
estimates can be improved if these additional measurements
from multiple receivers are taken into account. The scenario
1. Introduction adopted here is illustrated in Fig.1, where T represents the
Through-wall-radar (TWR) is an emerging technology, transmitter, R1 , R2 , R3 , R4 represent the receivers. The
permitting seeing through various obstacles and most adopted system mode poses a considerable strain on the
common building materials including reinforced concrete, method employed to pinpoint the receiver positions in order
sheet rock, brick, wood, plastic [1-2]. In most practical to assure a sufficient coherence to the system. All the
applications, TWR systems are developed to locate and track transceivers are connected through a wireless network which
the moving target during security operations. provides timing and control information and transmits data to
Various works are carried out to solve the problem of the processing unit. Besides, using wireless network
through-wall localization and tracking for moving target. In technology allows receivers to be added or removed from the
[3], the target is localized by the TWR system with the fixed network giving it flexibility for use in different types of field
and uniform linear array including one transmitter and two operations.
receivers. This radar system, estimating target position based
on the non-coherent ellipse-cross-correlation (ECL) method, R4
collects target echoes from a single view. Therefore, the ECL
method is subjected to fluctuation of the target radar-cross-
section (RCS), which causes the low accuracy of target
localization and tracking. In [4], the TWR radar system with a R1
single transmitter and multiple dispersed receivers is R3
presented and a coherent imaging method is used for target
localization and tracking. In this method, however, the weak
coherence of the target echoes leads to the decline of image
quality, which makes it unfavorable to locate and track target
through imaging method. R2
In this paper, random sparse array is adopted for TWR to
implement target localization and tracking, which collect
target echoes in multiple views to reduce the effect of target T
RCS fluctuation. Based on the linear data fusion for the Fig.1. Geometry of the detecting scene.
estimations of target position acquired by non-coherent ECL
method, the cooperative target localization and tracking is 3. Raw data pre-processing
applied to the TWR system, which is demonstrated to achieve
the higher estimation precision for target trajectory by Raw radar data can be interpreted as a set of impulse
simulation results. responses of surrounding, through which the emitted signals
were propagated. The intention of raw data pre-processing is
2. System configuration to obtain better data for detection and localization of target.
The background subtraction is utilized for improvement of
signal to noise ratio as a commonly used class of techniques
___________________________________
978-1-4244-8443-0/11/$26.00 ©2011 IEEE


that strive to enhance the performance of detection and The intention of data fusion is to find a unbiased estimation
localization. value x12 f ( x1 , x2 ) whose standard deviation is smaller
The background refers to a signal that is contributed in
than V x1 and V x 2 . Motivated by this, the method of linear
all measured impulse response. It contains especially the
antenna cross-talk, which is very huge in comparison with the weighted sum of x1 and x2 is considered, which can be
power of the electromagnetic waves reflected back from the described as
target, and waves reflected from static objects. The moving
target indicator (MTI) filter is utilized, which differentiates x12 ax x1  bx x2 (4)
the moving target from other static objects. where a x and bx are the coefficients which minimize the
variance of x12 , denoted as Var[ x12 ] E[ x12  x]2 .
4. Linear data fusion algorithm Meanwhile, ax  bx 1 is added as the constraint condition
For brevity, the system configuration with one to achieve the unbiased estimation. Then the conditional
transmitter and four receivers is adopted. Theoretically, there extremum can be obtained based on Lagrange multiplier
are 6 pairs of localization results can be obtained through approach. The obtained coefficients a x and bx are given by
ellipse-cross-localization (ECL) method using the V x22 V x21
measurements from every two receivers. In order to improve ax ,b (5)
the localization accuracy, the data fusion algorithm is V x21  V x22 x V x21  V x22
considered. The variance of x12 is expressed as
Var[ x12 ] (1 V x21  1 V x22 )1 , which is smaller than each of
4.1. Error analysis of ECL the original one. By applying the same processes, y1 and y2
The principle of the ECL method is shown in Fig.2. can also be fused as y12 a y y1  by y2 , where
p ( x, y )
V y22 V y21
l1 ay , by (6)
l0 l2
V y21  V y22 V y21  V y22

R1 ( x1 , y1 ) T (0, 0) R2 ( x2 , y2 ) y Corresponding to four receivers, every two of the four


x
propagation ranges r1 , r2 , r3 , r4 related to one target can be
used to determine the target position. To assure the
independence of the input coordinates for fusion, the way of
the combination of the propagation ranges can be considered
Fig.2. ECL schematic diagram. as
In this figure, the target coordinates are given by p( x, y ) ,
and the transmitter and two receivers are respectively located (r1r2 , r3 r4 ), (r1r3 , r2 r4 ), (r2 r3 , r1r4 ) (7)
at T (0, 0), R1 ( x1 , y1 ), R2 ( x2 , y2 ) . Let r1 l0  l1 and Then the fused data can be acquired as
r2 l0  l2 denote the propagation ranges in air, then the ( x1' , y1' ),( x2' , y2' ), ( x3' , y3' ) (8)
intersection can be obtained by solving the following equation
After fusion, the coherence of the three pairs of coordinates
­ x 2  y 2  ( x  x )2  ( y  y )2 r becomes relatively weak, or they can be considered as
° 1 1 1
® 2 (1) independent with each other. The three pairs can be fused
°̄ x  y  ( x  x2 )  ( y  y2 )
2 2 2
r2 according to the aforementioned linear data fusion method.
Then the final estimates ( x ', y ') are found.
The propagation range errors 'r1 , 'r2 can be acquired based
on total differential method, given by 5. Tracking
wr1 wr
'r1 'x  1 'y The target trajectory estimated by the localization
wx wy methods can be further processed by tracking algorithms.
(2)
wr2 wr They provide a new estimation of target localization based on
'r2 'x  2 'y foregoing positions of target. Usually, the target tracking will
wx wy result in the target trajectory error estimation decreasing
According to [5], the variances of location errors, including trajectory smoothing.
denoted as V x2 , V y2 , can be calculated from equation (2). The most of tracking systems utilize Kalman filtering for
target tracking. Another solution for target tracking and target
4.2. Data fusion trajectory smoothing can be provided by the application of
sinc function method [6]. For the data processing within our
x1 , x2 both are the independent estimations of the x scenario, the D  E filtering for target tracking is used. It
coordinates of target. The expectations and variances are assumes that the target is in linear motion with a constant
assumed as following velocity along each coordinate.
E[ x1 ] x Var[ x1 ] V x21
(3) 6. Simulation results
E[ x2 ] x Var[ x2 ] V x22


In order to illustrate the performance of the proposed 7. Conclusions
TWR system based on the adopted method in this paper, the
simulation scenario of through wall localization and tracking This paper deals with the target localization and tracking
of two moving persons has been analyzed. through TWR system. Firstly, the novel system configuration
In this simulation, two targets move along two different with random sparse array is introduced, which enables the
diagonal lines respectively in the area which is receivers to collect target echoes from multiple views. Then
x u y (10m ~ 10m) u (15m ~ 40m) . By using the ECL the linear data fusion is proposed to combine multiple
estimations of target position. Compared with the traditional
method for the conventional system configuration with one system configuration with one transmitter and two receivers,
transmitter and two receivers, the localization results are the simulation results show that the estimations of target
shown in Fig.3. The moving targets trajectory, estimated by trajectories are more precise with the novel configuration.
exploiting the proposed sparse array TWR system based on Further, the trajectories are smoothed after tracking.
the linear data fusion algorithm, is depicted in Fig.4. It can be
seen from the two figures, that the results in Fig.4, in
Acknowledgments
comparison with results in Fig.3, correspond much better to
true trajectory. It demonstrates that the proposed approach in This work was sponsored by National Natural Science
this paper can achieve higher localization accuracy than the Foundation of China (60802063) and by Sichuan Youth
traditional one. The results after target tracking are shown in Science and Technology Foundation (2011JQ0024 ).
Fig.5. It can be seen that the target trajectories are smoothed
after tracking, which follow the true trajectories better. References
40 [1] P.Withington,H.Fluhler, S.Nag, “Enhancing homeland
Target1
Target2
security with advanced UWB sensors,” IEEE Microwave
35 Magezine, vol. 4, pp. 51-58, 2003.
[2] S.S.Ram, H.Ling, “Through-Wall Tracking of Human
30
Movers Using Joint Doppler and Array Processing,”
IEEE Geoscience and Remote sensing letters, vol.5,
y/m

pp.537-541, 2008.
25
[3] R.Zetik, S.Crabbe, J.Krajnak, P.Peyerl, J.Sachs,
R.Thoma, “Detection and localization of persons behind
20 obstacles using M-sequence through-the-wall radar,” in
Proceedings of SPIE-Sensors, and Command, Control,
15 Communications, and Intelligence Technologies for
-10 -5 0 5 10
x/m Homeland Security and Homeland Defense, vol.6201,
Fig.3. Localization with traditional method. May 2006.
40 [4] D.Kocur, J.Gamec, M.Svecova, M.Gamcova,
Target1 J.Rovnakova, “Imaging Method: A Strong Tool for
Target2
Moving Target Tracking by A Multistatic UWB Radar
35
System,” in IEEE international Symposium on Applied
Machine Intelligence and Informatics, Herlany, Slovakia,
30 2010, pp.11-19.
y/m

[5] S.G.Jia, L.J.Kong, B.Liu, “Ellipse-Cross-Localization


25 Accuracy Analysis of Through-the-Wall Radar,” in
IEEE Radar Conference, 2009, pp.1-4.
20 [6] K.Yu, J.P.Montillet, A.Rabbachin, P.Cheong, I.Op-
permann, “UWB location and tracking for wireless
15
embedded networks,” in Signal Processing, vol. 86, pp.
-10 -5 0 5 10 2153 – 2171, 2006.
x/m
Fig.4. Localization with data fusion.
40
Target1
Target2
35

30
y/m

25

20

15
-10 -5 0 5 10
x/m
Fig.5. Target tracking.



Das könnte Ihnen auch gefallen