09-04-2013, 04:32 PM
Airborne Radar STAP using Sparse Recovery of Clutter Spectrum
[attachment=52916]
Abstract:
Space-time adaptive processing (STAP) is an effective tool for detecting a moving
target in spaceborne or airborne radar systems. Statistical-based STAP methods generally need
sufficient statistically independent and identically distributed (IID) training data to estimate the
clutter characteristics. However, most actual clutter scenarios appear only locally stationary and
lack sufficient IID training data. In this paper, by exploiting the intrinsic sparsity of the clutter
distribution in the angle-Doppler domain, a new STAP algorithm called SR-STAP is proposed,
which uses the technique of sparse recovery to estimate the clutter space-time spectrum. Joint
sparse recovery with several training samples is also used to improve the estimation performance.
Finally, an effective clutter covariance matrix (CCM) estimate and the corresponding STAP filter
are designed based on the estimated clutter spectrum. Both the Mountaintop data and simulated
experiments have illustrated the fast convergence rate of this approach. Moreover, SR-STAP is
less dependent on prior knowledge, so it is more robust to the mismatch in the prior knowledge
than knowledge-based STAP methods. Due to these advantages, SR-STAP has great potential for
application in actual clutter scenarios.
Introduction
An airborne/spaceborne (A/S) space time adaptive processor (STAP) attempts to detect a
moving target in the presence of a Doppler/angle spread clutter environment [1-2]. Due to the
motion of the radar platform, one dimensional processing in neither angle nor Doppler domain can
effectively distinguish the moving target from the surrounding clutter environment. Therefore, it is
necessary to carry out joint angle-Doppler processing. The fundamental component of STAP is the
effective estimation of the clutter covariance matrix (CCM), which is used to construct the optimal
linear weighting of the adaptive matched filter such that the output signal-to-clutter ratio is
maximized [1].
Spectrum Estimation and Clutter Suppression
As stated above, the CCM estimation could be obtained with high performance as long as the
clutter spectral characteristics are accurately acquired. Based on this idea, a new STAP algorithm
is developed via sparse recovery to obtain the clutter spectrum and estimate the CCM with much
less IID snapshots [23].
The sparsity of the space-time clutter
As stated above, the angle-Doppler domain is discretized into , s d N N cells along the angle
and Doppler axes, respectively. Each cell in this discretized plane corresponds to a certain
space-time steering vector and all these vectors make up the overcomplete basis Ψ . Because the
noise does not correspond to a certain space-time steering vector, its distribution is reflected as a
small noise floor in the angle-Doppler plane. As shown in Fig. 1, the angle-Doppler dependence in
(1) focuses the clutter distribution only along the clutter ridge, whose slope is determined by the
radar parameters. The STAP clutter scenario usually has a high CNR [1-2], therefore the
distribution of the training data in the angle-Doppler plane is mainly determined by the clutter
component. Moreover, the clutter scatters from the sidelobe are much smaller (more than 10dB
lower) than that from the mainlobe due to the effect of antenna azimuth weighting. In this way, the
significant clutter scatters only exist along the clutter ridge within the mainlobe [ ] min max θ ,θ .
Conventionally, the significant elements of the actual solution are used to express the degree of
sparsity in the sparse recovery [13-14]. In our problem of spectrum estimation, the sparsity of the
space-time clutter is equal to the number of cells occupied by the clutter ridge in the discretized
angle-Doppler plane, which is marked by the slash cells in Fig. 1.
Simulation experiments
In the actual clutter scenario, where the clutter is only locally stationary, the CCM estimation
with fast convergence rate has a great advantage because it can avoid the problem of the
heterogeneity in the training samples [2, 5]. Therefore, the simulations in this subsection are
presented to evaluate the convergence rate for CCM estimation. In addition, the influence of the
mismatch in the prior knowledge is also considered because both the SR-STAP and KB-STAP
approaches utilize this information in some form. Normally, the prior knowledge includes both the
radar parameters and the scattering properties of the actual clutter scenario [9-10].
Velocity mismatch
In the first scenario, the velocity mismatch is considered and other radar parameters are kept
in accordance with the actual scenario. Figure 6 gives the Loss IF performance versus the number
of IID snapshots. The convergence rate of LSMI is slow at about 12 (twice the clutter rank)
because it does not use any prior knowledge. For the CL algorithm, when the assumed velocity
ass v coincides with the actual scenario 300 / assv = m s , the assumed CCM in (11) is identical to
the actual one, so the convergence rate is exactly one. However, when there is some velocity
mismatch, such as 285 / assv = m s , which is common airborne radar systems, the assumed
clutter ridge will deviate from the actual one such that the assumed CCM could hardly contain all
the actual clutter scatters along the actual clutter ridge. As a consequence, the convergence rate in
CL will decrease to about 6. This indicates that the convergence rate of KB-STAP is sensitive to
the velocity mismatch. On the other hand, SR-STAP could effectively accelerate the convergence
rate in the ideal case where 300 / assv = m s because it has the capacity of obtaining the spectral
characteristics even with a few snapshots.
[attachment=52916]
Abstract:
Space-time adaptive processing (STAP) is an effective tool for detecting a moving
target in spaceborne or airborne radar systems. Statistical-based STAP methods generally need
sufficient statistically independent and identically distributed (IID) training data to estimate the
clutter characteristics. However, most actual clutter scenarios appear only locally stationary and
lack sufficient IID training data. In this paper, by exploiting the intrinsic sparsity of the clutter
distribution in the angle-Doppler domain, a new STAP algorithm called SR-STAP is proposed,
which uses the technique of sparse recovery to estimate the clutter space-time spectrum. Joint
sparse recovery with several training samples is also used to improve the estimation performance.
Finally, an effective clutter covariance matrix (CCM) estimate and the corresponding STAP filter
are designed based on the estimated clutter spectrum. Both the Mountaintop data and simulated
experiments have illustrated the fast convergence rate of this approach. Moreover, SR-STAP is
less dependent on prior knowledge, so it is more robust to the mismatch in the prior knowledge
than knowledge-based STAP methods. Due to these advantages, SR-STAP has great potential for
application in actual clutter scenarios.
Introduction
An airborne/spaceborne (A/S) space time adaptive processor (STAP) attempts to detect a
moving target in the presence of a Doppler/angle spread clutter environment [1-2]. Due to the
motion of the radar platform, one dimensional processing in neither angle nor Doppler domain can
effectively distinguish the moving target from the surrounding clutter environment. Therefore, it is
necessary to carry out joint angle-Doppler processing. The fundamental component of STAP is the
effective estimation of the clutter covariance matrix (CCM), which is used to construct the optimal
linear weighting of the adaptive matched filter such that the output signal-to-clutter ratio is
maximized [1].
Spectrum Estimation and Clutter Suppression
As stated above, the CCM estimation could be obtained with high performance as long as the
clutter spectral characteristics are accurately acquired. Based on this idea, a new STAP algorithm
is developed via sparse recovery to obtain the clutter spectrum and estimate the CCM with much
less IID snapshots [23].
The sparsity of the space-time clutter
As stated above, the angle-Doppler domain is discretized into , s d N N cells along the angle
and Doppler axes, respectively. Each cell in this discretized plane corresponds to a certain
space-time steering vector and all these vectors make up the overcomplete basis Ψ . Because the
noise does not correspond to a certain space-time steering vector, its distribution is reflected as a
small noise floor in the angle-Doppler plane. As shown in Fig. 1, the angle-Doppler dependence in
(1) focuses the clutter distribution only along the clutter ridge, whose slope is determined by the
radar parameters. The STAP clutter scenario usually has a high CNR [1-2], therefore the
distribution of the training data in the angle-Doppler plane is mainly determined by the clutter
component. Moreover, the clutter scatters from the sidelobe are much smaller (more than 10dB
lower) than that from the mainlobe due to the effect of antenna azimuth weighting. In this way, the
significant clutter scatters only exist along the clutter ridge within the mainlobe [ ] min max θ ,θ .
Conventionally, the significant elements of the actual solution are used to express the degree of
sparsity in the sparse recovery [13-14]. In our problem of spectrum estimation, the sparsity of the
space-time clutter is equal to the number of cells occupied by the clutter ridge in the discretized
angle-Doppler plane, which is marked by the slash cells in Fig. 1.
Simulation experiments
In the actual clutter scenario, where the clutter is only locally stationary, the CCM estimation
with fast convergence rate has a great advantage because it can avoid the problem of the
heterogeneity in the training samples [2, 5]. Therefore, the simulations in this subsection are
presented to evaluate the convergence rate for CCM estimation. In addition, the influence of the
mismatch in the prior knowledge is also considered because both the SR-STAP and KB-STAP
approaches utilize this information in some form. Normally, the prior knowledge includes both the
radar parameters and the scattering properties of the actual clutter scenario [9-10].
Velocity mismatch
In the first scenario, the velocity mismatch is considered and other radar parameters are kept
in accordance with the actual scenario. Figure 6 gives the Loss IF performance versus the number
of IID snapshots. The convergence rate of LSMI is slow at about 12 (twice the clutter rank)
because it does not use any prior knowledge. For the CL algorithm, when the assumed velocity
ass v coincides with the actual scenario 300 / assv = m s , the assumed CCM in (11) is identical to
the actual one, so the convergence rate is exactly one. However, when there is some velocity
mismatch, such as 285 / assv = m s , which is common airborne radar systems, the assumed
clutter ridge will deviate from the actual one such that the assumed CCM could hardly contain all
the actual clutter scatters along the actual clutter ridge. As a consequence, the convergence rate in
CL will decrease to about 6. This indicates that the convergence rate of KB-STAP is sensitive to
the velocity mismatch. On the other hand, SR-STAP could effectively accelerate the convergence
rate in the ideal case where 300 / assv = m s because it has the capacity of obtaining the spectral
characteristics even with a few snapshots.