Sei sulla pagina 1di 4

362 IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 7, NO.

2, APRIL 2010

Underwater Object Detection Based


on Gravity Gradient
Lin Wu, Xin Tian, Jie Ma, and Jinwen Tian

AbstractA novel method of underwater object detection based the central axes of mass bodies can be identified effectually
on gravity gradient is presented, which can be used on autonomous [14][18].
underwater vehicles (AUVs) to detect abnormal objects underwa- Underwater object detection technology based on gravity
ter. Gravity gradient anomalies of partial area, which are caused
by the object, can be measured by a gravity gradiometer on an gradient, as a new technology, can perceive an abnormal object
AUV. Then, anomalies can be inversed with a gravity gradient in- underwater effectively. The gravity gradiometer is not easy to
version algorithm, so the mass and barycenter of an object can be be detected and interfered with. It does not need an AUV to
estimated. Simulation results show that approximate information float near the surface and is nonemanating and so completely
of an object can be provided by the proposed method. covert [19], [20].
Index TermsGravity gradient anomaly, gravity gradient
inversion, gravity gradient measurement, underwater object
detection. II. G RAVITY G RADIENT I NVERSION
FOR O BJECT D ETECTION
I. I NTRODUCTION
The z-component of gravity anomaly at the field point
P (x, y, z) caused by the object can be described as
A UTONOMOUS underwater navigation is an area of re-
search with broad commercial and military applications.
The precise detection of underwater abnormal objects is one g(x, y, z)

of the primary challenges in underwater navigation research (, , )(, , )( z)
nowadays [1][3]. At present radar, laser and sonar are de- =G ddd.
[( x)2 + ( y)2 + ( z)2 ]3/2
ployed on autonomous underwater vehicles (AUVs) for object
detection [4][7]. However, similar to the radar detection close (1)
to the surface, laser or active sonar transmission also causes the
damage of covertness of AUVs. In the aforementioned formula, (, , ) is the coordinates of
For several years, considerable efforts have been put into pro- the position of the object. (, , ) is the geometrical function;
ducing underwater object detection systems with geophysical it is determined by the boundary of the object; it is one on the
technologies such as gravity, gravity gradient, or magnetic [8] boundary and zero otherwise. (, , ) is the density contrast
[10]. Gradiometers allow measurement from a moving vehicle; distribution function. The geometrical and physical parameters
they circumvent a fundamental limitation which the equiv- of the object can be acquired by (, , ) and (, , ). G is
alence principle imposes on gravimeters. In addition, better the universal gravitational constant.
spatial resolution and increased information can be obtained The underwater object detection technology based on gravity
through gravity gradiometry, and a gradiometer is sensitive gradient regards the object as 3-D body with density contrast
to topographic effects or mass changes. Therefore, it is more and the object detection as gravity gradient inversion. There-
appropriate for object detection [11][13]. fore, the problem is converted into gravity gradient inversion
Partial- or full-tensor gravity gradiometers have been de- for geometrical and physical parameters of the object.
ployed for a long time in geological exploration. Currently, The Cartesian reference frame used to specify source and
gradiometers in the world achieve a sensitivity of 104 E (1 E = field points is shown in Fig. 1. The xy plane is taken as
109 s2 ) with a frequency from 1 to 103 Hz. Airborne gravity horizontal and the z-axis as vertical, with increasing values of
gradiometry allows rapid surveying of large areas. Results z representing increasing height. If a particle with its mass m
from the surveys indicate that geological targets with density- is at point O(, , ) and i is assigned a value for its density,
contrast-covered 150200-m subsurface can be detected, and the gravity potential at the field point P (x, y, z) caused by
the object (inertial centrifugal force is negligible here) can be
described as
Manuscript received January 12, 2009; revised June 16, 2009, August 2, Gm Gm
2009, and September 18, 2009. Date of publication December 4, 2009; date = = . (2)
of current version April 14, 2010. r ( x) + ( y)2 + ( z)2
2
The authors are with the National Key Laboratory of Science and Technology
on Multi-spectral Information Processing Technology, Huazhong University of By differentiating with respect to x, y, and z, it is possible
Science and Technology, Wuhan 430074, China (e-mail: wulin1108@gmail.
com; majie.hust@sohu.com). to derive the six components of the gravity gradient tensor gij
Digital Object Identifier 10.1109/LGRS.2009.2035455 at the field point P (x, y, z) caused by the object (where gij =

1545-598X/$26.00 2009 IEEE


WU et al.: UNDERWATER OBJECT DETECTION BASED ON GRAVITY GRADIENT 363

Fig. 2. Model of the object.

The formula for gravity gradient inversion is


 
(x, y)gij (x, y)dxdy. (9)
Fig. 1. Variables required to calculate the gravity gradient response of the
object.
When (x, y) = x or (x, y) = y, substituting (3)(8) into
gi /j is the derivative of the ith component of g with respect the aforementioned expression, it is possible to get M , which is
to spatial coordinate j; i, j = x, y, z) as follows: the mass of the object
   
2 xgxz (x, y) dxdy = ygyz (x, y) dxdy = 2GM. (10)
gxx =
x2

2( x)2 ( y)2 ( z)2 Thus, M is equal to 1/2G of the integral of xgxz (x, y)
=G i ddd or ygyz (x, y) on the xy plane; the value range of x and y is
[( x)2 + ( y)2 + ( z)2 ]5/2
(, +) theoretically.
(3) If the distance between the object and the field point is big
2
enough to ignore the size of the object as though it was a
gxy =
xy particle, (3)(8) can be written as

3( x)( y) gxx = GM 3(x)
2
R2
=G i ddd
R5
[( x)2 + ( y)2 + ( z)2 ]5/2
3(y)2 R2

gyy = GM

R5
2
R2
(4) gzz = GM 3(z) R5 (11)
2

gxy = GM 3xy
gxz =
R 5
xz
gxz = GM 3xz

R5
3( x)( z)
=G i ddd gyz = GM 3yzR5 .
[( x)2 + ( y)2 + ( z)2 ]5/2
In the formulas, M is the mass of the object, and R is the
(5) distance given by R = (x)2 + (y)2 + (z)2 . Replacing
2 x, y, and z with ( x), ( y), and ( z) enables the
gyy =
y 2 calculation of these formulas at any point.

2( y)2 ( x)2 ( z)2 By combining subequations of (11), the orientation of the
=G i ddd object can be calculated as follows:
[( x)2 + ( y)2 + ( z)2 ]5/2 

(6)
(x, y, z) = arccos z(x,y,z)


R(x,y,z)
2

gyz =
= arccos gxy (x,y,z) 2 1 gxy (x,y,z) 2
yz
+ g (x,y,z) +1
gyz (x,y,z)
  xz

3( y)( z) (x, y) = arctan y(x,y,z)
= arctan
gyz (x,y,z)
=G i ddd
z(x,y,z) g (x,y,z)
xz

[( x)2 + ( y)2 + ( z)2 ]5/2


R(x, y, z)3 = GM
gxy (x,y,z) 2 3gxy (x,y,z) 2 1 .
(7) gzz (x,y,z)
gyz (x,y,z)
+ gxz (x,y,z)
+1
2
(12)
gzz = 2
z
 M is obtained from (10) with survey on the xy plane.
2( z)2 ( x)2 ( y)2
=G i ddd. Under some given conditions, for example, symmetrical density
[( x)2 + ( y)2 + ( z)2 ]5/2 contrast distribution, regular shape, etc., the center of gravity of
(8) the object can be gained.
364 IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 7, NO. 2, APRIL 2010

TABLE I
G EOMETRICAL AND P HYSICAL PARAMETERS OF THE M ODEL

TABLE II
I NFLUENCES OF THE S URVEY A REA ON THE E STIMATION
OF THE M ASS OF THE O BJECT

Fig. 3. Components of the gravity gradient tensor responses of the model.

III. S IMULATIONS
A simulated model is built up from many cubic blocks
(voxels) stacked together to create the effect of a large 3-D
volume of an object. Each voxel is 1 m3 in size and is assigned a
value for its density. This creates a 3-D model of the object with
realistic features. Fig. 2 shows the model, and Table I exhibits
the size and density contrast values assigned to the object units.
The mass distribution of the object is assumed symmetrical, so
the density contrast is a constant, not a spatial distribution here.
Fig. 3 shows the gravity gradient tensor components of the
model (gxx , gxy , gxz , gyy , gyz , and gzz ) as measured on the xy
plane, the height of which is constant 50 m lower than that of
the object.
Fig. 4. Curve of the estimation of the mass of the object with varying survey
area.
IV. R ESULTS AND D ISCUSSION
If the object was in a state of rest, the gravity gradient tensor can be gotten with just much small survey area, according to
responses on the measurement plane could be measured by a which the detection may be predigested.
gravity gradiometer which was carried on an AUV. In the survey If the object was in a state of motion, the sequence of varying
area, the AUV moved, and data were acquired at every survey gravity gradient tensor responses at the position of the gravity
point which was spaced 1 m with both x and y orientations, gradiometer could be measured synchronously. Substituting
which means that the bigger the survey area is, the more data these responses into (12), the orientation of the object can
are acquired. gxz or gyz in (10) can be substituted by the gxz or be calculated with the estimation of the mass of the object.
gyz responses at every point, so the mass of the object could be The actual track of the object and the position of the gravity
estimated. Table II shows the influence of the survey area on the gradiometer are marked with solid line and cross in Fig. 5.
estimation of the mass of the object; the corresponding curve is According to the estimation discussed earlier, the mass of the
shown in Fig. 4, where the actual mass of the object is denoted object cannot be very precise here, so 36t was assigned to the
by the dashed line. It can be seen that the bigger the survey area estimated mass here, and a 0.01-E rms measurement error of
is, the more exact the estimation is. The magnitude of the mass the gradiometer was added to the simulation. The estimated
WU et al.: UNDERWATER OBJECT DETECTION BASED ON GRAVITY GRADIENT 365

R EFERENCES
[1] J. Lean and D. A. Pratt, Multi-sensor marine geophysical profiling and
digital acquisition using SAS1, Explor. Geophys., vol. 22, no. 1, pp. 235
242, 1991.
[2] G. J. Heald and H. D. Griffiths, A review of underwater detection tech-
niques and their applicability to the landmine problem, in Proc. 2nd Int.
Conf. Detection Abandoned Land Mines, 1998, pp. 173176.
[3] M. Montanari, J. R. Edwards, and H. Schmidt, Autonomous underwater
vehicle-based concurrent detection and classification of buried targets
using higher order spectral analysis, IEEE J. Ocean. Eng., vol. 31, no. 1,
pp. 188199, Jan. 2006.
[4] S. Panagopoulos and J. J. Soraghan, Small-target detection in sea clut-
ter, IEEE Trans. Geosci. Remote Sens., vol. 42, no. 7, pp. 13551361,
Jul. 2004.
[5] D. P. Eickstedt and H. Schmidt, A low-frequency sonar for sensor-
adaptive, multistatic, detection and classification of underwater targets
with AUVs, in Proc. MTS/IEEE OCEANS, San Diego, CA, Sep. 2226,
2003, vol. 3, pp. 14401447.
[6] H. Mizushima, T. Maki, T. Ura, T. Sakamaki, H. Kondo, and
M. Yanagisawa, Autonomous recognition of bubble plumes for nav-
igation of underwater robots in active shallow vent areas, in Proc.
MTS/IEEE OCEANS, Vancouver, BC, Canada, Sep. 29Oct. 4, 2007,
pp. 16.
Fig. 5. Estimation of the track of the object. [7] M. J. Buckingham, E. M. Giddens, F. Simonet, and T. R. Hahn, Propeller
noise from a light aircraft for low-frequency measurements of the speed of
track of the object is signed in Fig. 5 with pentagrams. sound in a marine sediment, J. Comput. Acoust., vol. 10, no. 4, pp. 445
464, 2002.
Statistically, the average distance between the actual position [8] G. R. Mumaw and C. Murphy, Airborne 3D FTG data proves utility for
of the object and the gravity gradiometer is 192.17 m, with both minerals and oil & gas exploration/exploitation, ASEG Ext. Abstr.,
48.57% estimation error of the mass of the object; the average vol. 2003, no. 2, pp. 112113, 2003.
[9] J. Vrbancich, Instrumentation for detection and generation of ELF emis-
estimation error of the track in the x- and y-direction is 18.89% sions in seawater, Explor. Geophys., vol. 24, no. 2, pp. 167174, 1993.
and 18.01%, respectively, here. The performance is much more [10] A. Salem, T. Hamada, J. K. Asahina, and K. Ushijima, Detection of
encouraging than the results in [8][10]. unexploded ordnance (UXO) using marine magnetic gradiometer data,
Explor. Geophys., vol. 36, no. 1, pp. 97103, 2005.
Additionally, the proposed inversion algorithm code was [11] B. Pawlowski, Gravity gradiometry in resource exploration, Lead. Edge,
written in MATLAB and run on a personal computer; it costs vol. 17, no. 1, pp. 5152, Jan. 1998.
437.5 ms with 271 times estimation of the orientation. There- [12] D. Chapin, Gravity instruments: Past, present, future, Lead. Edge,
vol. 17, no. 1, pp. 100112, Jan. 1998.
fore, it is possible to estimate in real time with this method [13] E. H. van Leeuwen, BHP develops worlds first airborne gravity gra-
according to the frequency of the gradiometer. diometer for mineral exploration, Preview (ASEG), vol. 86, pp. 2830,
Jun. 2000.
[14] M. H. Dransfield, M. J. Buckingham, C. Edwards, F. J. van Kann,
V. C ONCLUSION A. G. Mann, R. Matthews, and P. J. Turner, Gravity gradiometry for
geophysical prospecting, Explor. Geophys., vol. 22, no. 1, pp. 107110,
The underwater object detection method based on gravity 1991.
[15] D. DiFrancesco, D. Kaputa, and T. Meyer, Gravity gradiometer systems,
gradient has been proposed. The gravity gradient anomalies Preview (ASEG), vol. 133, pp. 3039, 2008.
caused by the object can be measured by a gravity gradiometer [16] J. B. Lee, FALCON gravity gradiometer technology, Explor. Geophys.,
on an AUV, then can be inversed to get the estimation of the vol. 32, no. 3/4, pp. 247250, 2001.
[17] M. Dransfield, A. Christensen, M. Rose, P. Stone, and P. Diorio,
mass and barycenter of the object. This research attempted to FALCON test results from the Bathurst mining camp, Explor. Geo-
investigate the feasibility of this technology. Simulation results phys., vol. 32, no. 4, pp. 243246, 2001.
show that, based on the assumption that the density contrast [18] S. Hammond and C. Murphy, Air-FTG: Bell Geospaces airborne gravity
gradiometer, Preview (ASEG), vol. 105, pp. 2224, 2003.
is a constant, when the mass of an object had been estimated, [19] J. Moryl, H. Rice, and S. Shinners, The universal gravity module for
the orientation could be calculated via the gravity gradient enhanced submarine navigation, in Proc. IEEE Position Location Navig.
inversion algorithm. Furthermore, this technology may be used Symp., 1998, pp. 324331.
[20] H. Rice, S. Kelmenson, and L. Mendelsohn, Geophysical navigation
for terrain estimation and avoidance in underwater navigation technologies and applications, in Proc. IEEE Position Location Navig.
in the future. Symp., 2004, pp. 618624.

Potrebbero piacerti anche