Informasi Jawa Tengah

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering
Copyright c 1999 John Wiley & Sons, Inc.
RADAR TARGET RECOGNITION
Recent advances in radar provide sufficient resolution and enough information to recognize tactical targets
from radar returns (1). For target recognition applications, different types of radar, such as synthetic aperture
radar (SAR), millimeter wave (MMW) real aperture radar (RAR), and interferometric synthetic aperture radar

(IFSAR), have been explored. Modern radar systems also provide extensive data, including fully polarimetric
andDoppler channels, but there still are many challenges for target recognition using radar returns. The special
characteristics of radar and recent advances in radar technology make the recognition of targets difficult. For
example, the radar profile changes drastically for a small change in look angle, and the recent development of
stealth technology significantly alters the radar signature. This article covers many of the different approaches
for different radar types of radar target recognition.
Synthetic aperture radar provides high resolutions for both range and azimuth directions and has been
used widely for target recognition applications. Since SAR is an imaging radar, it can provide detailed images
of target areas with cloud-penetrating characteristics.When its fully polarimetric data are utilized, targets can
be recognized with high accuracy. The SAR target recognition is typically done in multiple stages (2). The first
stage of a typical SAR target recognition algorithm is a prescreener, where regions of interest are located. A
constant false-alarm rate (CFAR) detection algorithm is often used in the first stage. The second stage may be
a discriminator, where natural clutter false alarms are rejected. In this stage, textural features are extracted
from the target-size windows applied to the neighborhood of pixels located by CFAR detection. The third stage
is typically a classifier, where the targets are classified and human-made discretes are rejected. Extensive
studies in statistical pattern recognition have been done, and many different pattern classifiers are used in
target recognition applications. For SAR target recognition, spatial matched filters are investigated (2) for the
recognition of ground vehicles. Novak’s (2) spatial matched filter approach in target recognition is explained in
later sections.
Although SAR technology provides many advantages over RAR there are many applications for which
RAR is important. The azimuth resolution of SAR is much better than that of RAR, but it is typically based
on the assumption that the movement of the radar is at constant speed and direction. When the radar is
moving at a rapidly changing speed and trajectory, as in the case of the radar mounted on fighter airplanes
or self-guided weapons, the assumptions for SAR processing are not correct. Millimeter wave RAR represents
the next generation of military smart sensors for detection, tracking, and surveillance due to its high-range
resolution, which is critical in target recognition. The MMW RAR technology is sufficiently advanced, and the
range resolution of MMW radar is sufficiently high to discriminate tactical targets at a distance of several
kilometers (3). The target recognition by hierarchical modeling of high-range resolution (HRR) MMW radar
signatures is discussed in this article.
Artificial neural networks (ANN) have been widely used in target recognition. The use of ANN in radar
target recognition is also discussed in this article. ANNs are used as pattern classifiers and feature extractors
and in model adaptation and fuzzy classification in radar target classification. Feed-forward neural networks
have been used as pattern classifiers after being “trained” with known target samples. The leaning algorithm of
neural networks provides a powerful tool for adaptation of the classifier to input vectors (4). A self-organizing
1
2 RADAR TARGET RECOGNITION
feature map (SOM) has been used as a feature extractor (5) for radar target recognition. In combination
with Kohonen’s learning vector quantizer (LVQ) for supervised classification, SOM has been applied to the
recognition of ground vehicles from MMW HRR radar signatures (5). Perlovsky, et al. (6) suggested the modelbased
neural network to include a priori information to an ANN. This approach can reduce the search space
of the neural network by incorporating a priori information to the adaptability of an ANN. The fuzzy neural
network approach (7) is also suggested to classify targets that may belong to more than one class. The advances
in neural network approaches can improve the performance of target recognition algorithms further. There are
many approaches to incorporating information from many different sources for radar target recognition. By
fusing the information from more than one sensor, the accuracy of radar target recognition may be improved.
Two approaches in utilizing information from multiple sensors are discussed in this article. IFSAR provides
elevation information, in addition to two-dimensional radar images, by processing interference between radar
returns received by two different antennas. By processing IFSAR images and fusing to SAR of visual images,
the accuracy of the target recognition can be improved substantially. The approach of combining IFSAR and
visual images using image registration approach is discussed in this article. There are statistical approaches
in data fusion, and Bayesian data fusion approaches are used in radar target recognition (8). In this approach,
features from polarimetric SAR images are fused to improve the recognition accuracy.
Radar target recognition is a complex problem, and no single algorithm performs better than other
algorithms with different types and modes of radar. In this article, different approaches for radar target
recognition are discussed in terms of radar types and approaches.
Hierarchical Modeling Approach for Noncooperative Target Recognition
Target recognition using hierarchical modeling of MMW RAR radar signatures is considered in (9), and the
algorithm is tested with MMW radar data from a noncooperative target identification (NCTI) database. In
radar target recognition problems, the targets may be shifted and scaled from the trained data. The change in
look angle results in a scale change in radar signature, and the change in azimuth angle results in a shifting
of radar signature. Therefore, a classifier using features extracted from the radar signatures should classify
targets reliably even when radar signatures are shifted or scaled. In general, parameters of time series models
such as ARMA models are not scale invariant. Thus, the classification of scaled targets with model parameter
features can result in poor classification accuracy. There are a few approaches to make a classifier more reliable
to shifting and scaling. For example, the use of scale invariant features or a training with scaled samples can
make a classifier more robust to scaling.
The hierarchical ARMAmodeling approach is briefly discussed in the following. Suppose that a continuous
signal x(t) is a training sample. The classifier needs to classify the scaled signal of x(t) correctly as the same
class as x(t). The scaled signal of x(t) is given by
One approach to achieve this is to train a classifier with features extracted from scaled signals of x(t). For
example, different features at m different scales are extracted from scaled signals yα1(t), yα2(t), . . ., yαm(t); then
the classifier is trained with these multiscale features. If the number of scales included in the training is large
enough, the classifier will classify signals having large-scale changes.
However, there are at least two potential problems with this approach if the signal is a discrete signal {x(i),
i=1, . . ., N}. First, the original signal is defined only at discrete points, and the signal at the finer scale is not
defined at certain points. Second, feature extraction is performed multiple times with a single training sample,
RADAR TARGET RECOGNITION 3
Fig. 1. A decimation filter.
and the computational complexity increases linearly as the number of scales increases. These difficulties can
be solved by the hierarchical modeling approach. The hierarchical modeling approach presented in this section
extracts multiscale features without adding much computational complexity.
A discrete signal can be scaled to a coarser scale or a finer scale by decimation filtering or interpolation,
respectively. We will first consider the decimation filtering of a signal and its effect on the statistical model,
and then we will consider the scaling to a finer scale as a modeling process. A decimation filter is defined as a
local averaging (finite impulse response [FIR] filtering) followed by a down-sampling process, as shown in Fig.
1. If the down-sampling rate is m, the decimation-filtered signal represents the signal at the scale reduced by
the factor of m. Let H be a FIR filter of length r and ↓ be the down-sampling operator of factor m.
Suppose that a signal at a coarser scale ym(i) is obtained by decimation-filtering of the original signal x(i).
Suppose that the signal {x(i), i=1, . . ., N} follows an ARMA(p,q) model.
where {w(i)} is a zero mean white noise sequence with variance σ2
w, and aj’s and bj’s are real coefficients.
Equation (5) can be rewritten as
where
and δ is the unit delay operator, and we assume that the roots of Ap(δ) and Bq(δ) lie inside of the unit circle for
stability and invertability of the model.
4 RADAR TARGET RECOGNITION
To find features at coarser scale, the model at a coarser scale should be considered. The following theorem
summarizes the results on the modeling of a decimation-filtered ARMA process.
The decimation-filtered process {ym(i)} defined in Eq. (4) follows an ARMA(p,q∗)model, where the order of
AR polynomial is p, the order of the MA polynomial is q∗=[(p(m−1)+r+q−1)/m], and the model parameters
can be obtained from the model parameters of x(i).
where
Also, the AR polynomial Cp(δ) of aggregated data satisfies the following relation:
where r1, . . ., rp are roots of Ap(δ).
The proof can be found in (9).
The ARMA model parameters are shift invariant because this model’s parameters depend only on mean
and correlations. Power spectral density can be estimated by using ARMA model parameters and is also shift
invariant and provides features that are intuitively appealing. For example, spectral peaks or notches represent
the presence or absence of a frequency component in the signal. For radar signal classification, ARMA power
spectrum features at multiple scales are used. Power spectral density of an ARMA process can be estimated
by an extended least squares (ELS) method.
Suppose that x(i) is an ARMA(p,q) process and w(i) is the input white sequence with variance σ2 as defined
in Eq. (5). Let Rxx(k) be the autocorrelation functions of x(i), and Rxw(k) be the cross correlation between x(i)
and w(i). The Yule–Walker equation for the ARMA process x(i) is given by the following:
The AR parameters are estimated by solving the above Yule–Walker equations. By using the estimated
AR parameters, the MA component of x(i) can be obtained by filtering AR component from x(i).
The power spectral density of the ARMA process x(t) is estimated from the correlations of xma(t) and the AR parameters
estimated by Yule–Walker equations. The ELS power spectrum estimation algorithm is summarized
as follows.
RADAR TARGET RECOGNITION 5
ELS Spectrum Estimation Algorithm
Step 1: Compute sample correlations Rxx(k) for k=0, . . ., p+q.
Step 2: Estimate AR parameters a1, . . ., ap by
Step 3: Compute the sample correlation of MA component xma(i) that is obtained by removing the AR
component.
Step 4: Compute the ARMA power spectrum.
For each training sample x(i), the models at the other scales (both coarser and finer scales) are obtained
by the hierarchical modeling approach presented in the previous section. The model at a coarser scale is
obtained using Theorem 1. The AR polynomial is obtained by Eq. (10), and the correlation of the signal at
the coarse scale is obtained with a proper choice of smoothing filter H, such as a Gaussian filter. Thus, the
spectral density of the signal at a coarser scale is obtained by the ELS algorithm. The model at a finer scale is
obtained by the approach explained in step 2. The AR polynomial of the signal at a finer scale is obtained under
no-hidden-periodicity assumption. The correlation function at a finer scale is obtained by disaggregation (9),
6 RADAR TARGET RECOGNITION
and the ARMA spectrum at a finer scale is obtained by the ELS estimation algorithm. The multiscale feature
extraction algorithm is summarized as follows.
Multiscale Spectral Feature Extraction Algorithm:
Step 1: Each radar return is normalized to zero mean and unit variance by
where ˆ mand ˆσ2 are sample mean and sample variances of x(i). M K-dimensional features from M scales
(including coarser and finer scales) are obtained from the normalized radar returns by the following
procedure.
Step 2: For each training sample, the AR parameters and correlations are estimated by the ELS algorithm.
For k=0, 1, . . ., K −1, the power spectrum is estimated at ω=πk/K. The logarithm of the power spectral
density forms a K-dimensional feature vector.
Step 3: At each coarser scale, a feature vector is obtained by estimating the power spectrum using the
ELS method, with model parameters obtained by the hierarchical modeling approach. The logarithm of
the power spectral density forms a K-dimensional feature vector at a coarser scale. Feature vectors at
multiple scales are obtained by repeating this step at coarser scales.
Step 4: At each finer scale, a feature vector is obtained by estimating the power spectrum using the ELS
method, with model parameters obtained by the hierarchical modeling approach. This is repeated for
other finer scales, and multiple K-dimensional feature vectors are obtained from the logarithm of the
power spectral density.
Classification is done by a minimum distance classifier with multiple prototypes. In this approach, each
training sample generates M prototypes corresponding to M scales. Therefore, if there are N training signals
for each class, then NM prototypes will be available for each class. Let us assume that there are Nk prototypes
z1k
, . . ., zk
Nk in the class k ∈ {1, . . ., K}. For a test pattern x, the distance to the class k is defined by
where the intersample distance d(x, z) is the Eucledian distance between x and z. The distance Di is the
smallest of the distances between x and each of the prototypes of the class k. The test pattern x is classified by
the minimum distance decision rule: x is classified into class k if Dk In Ref. 9, the hierarchical model-based features are tested with NCTI data. Figure 2 shows a typical
NCTI radar signature and estimated power spectral density. In Ref. 9, about 95% of classification accuracy is
reported with 5000 MMW RAR radar signatures.
Neural Network Approaches
Neural networks have been widely used in radar target recognition applications (4,5,6,7). A good survey on
the use of neural networks in target recognition can be found in Ref. 4. Neural network approaches have been
popular because they have many promising qualities in target recognition applications. Neural networks have
the capability for adaptation to additional targets and environments. The existence of powerful learning algorithms
is one of the main strengths of the neural network approach (4). Neural networks also offer techniques
RADAR TARGET RECOGNITION 7
Fig. 2. A HRR radar signature from NCTI database and its power spectrum estimated by hierarchical modeling.
for selecting, developing, clustering, and compressing features into a useful set, and they provide automatic
knowledge acquisition and integration techniques for target recognition systems.
Feed-forward neural networks have been used as a pattern classifier for the target recognition. (10). Let
xi be the multidimensional feature vector extracted from a radar image, and let S be the index set of the target
patterns.
The optimum receiver (10) defined by
8 RADAR TARGET RECOGNITION
can be implemented as a neural network:
where T is the threshold and (x) is the heavyside step function. Roth (10) showed that detection of target
patterns out of a set of P of patterns can be handled by the preceding feed-forward neural network.
Neural networks have been also used as feature extractors for target recognition. Kohonen’s selforganizing
map (SOM) and LVQ have been used in the two-stage target recognition approach (5). SOM is
based on unsupervised competitive learning where only one output node, or one per local group of nodes at
a time, gives the active response to the current input signal, and it clusters input vectors into preselected C
classes by adapting connection weights to nodes in the network, and is used as a feature extractor in Ref. 5. At
each iteration of the SOM algorithm, the best matching node c is selected by
where x is the current input vector, and {m1, . . ., mC} is the set of nodes (cluster centers). Then each node mi
located in the neighborhood of the node c is adapted by the learning rule:
where the gain ηi(t) can be a simple monotonically decreasing function of time or a Gaussian gain function
defined by
The learning rate α(t) and the kernel width σ(t) are monotonically decreasing functions, and their exact forms
are not critical. LVQ is used as a supervised classifier of the features extracted by SOM in Ref. 5. In Ref. 5,
more than 94% of accuracy is reported in the target recognition experiment, with MMW data having five types
of ground vehicles.
Recently, the model-based neural network (MBNN) was introduced (6) to combine a priori knowledge
of models of data with adaptivity to changing data properties. The learning and adaptation of the MBNN
is done by iterative estimation of association weights and model parameters. Different statistical models for
different physical processes, background clutter, outlier, target pixels, and so on, are also introduced in Ref. 6.
This approach has the potential to improve target recognition performances by allowing inclusion of a priori
information in addition to the adaptability of a neural network.
Fuzzy neural networks are also used in radar target recognition. Fuzzy ARTMAP and EMAP neural
networks are suggested (7) for radar target recognition. Fuzzy neural networks allow us to make soft decisions
in classifying a target, and each input vector can belong to more than one class. The fuzzy association between
the input vector and the classified target can improve the performance and the complexity of the adaptation.
Exploitation of Elevation Data from IFSAR
IFSAR is a technique to generate high-resolution digital elevation model (DEM) based on the phase difference
in SAR signals received by two spatially separated antennas (11). There are drawbacks in height maps derived
RADAR TARGET RECOGNITION 9
Fig. 3. An IFSAR image.
from IFSAR data: The data are noisy and the spatial resolution is much inferior to that of visual data. The
spatial resolution is further degraded by the noise removal step. Figure 3 shows a height map produced by a
real IFSAR. A typical IFSAR elevation image is noisy and needs to be filtered before it can be reliably used.
Also, there are regions with “no data” that result either from the fact that the original scene was not on a
rectangular grid or from radar geometry effects, which cause some points not to be mapped. Interpolation and
nonlinear filtering techniques are used to filter the elevation data.
Positioning of IFSAR and visual data allows for the fusion of clues from both sensors for target recognition.
It is needed to overcome various difficulties resulting from the limitations of the sensor. For example, building
detection requires the extraction and grouping of features such as lines, corners, and building tops to form
buildings (12). The features extracted from visual data usually contain many unwanted spurious edges, lines,
and so on that do not correspond to buildings. The grouping stage requires complex and computationally
intensive operations. Further, the height of a building is typically estimated by extracting shadows and sun
angle when available and is not reliable when the shadows are cast on adjacent buildings. Another drawback
of methods based exclusively on visual data lies in their sensitivity to imaging conditions.
IFSAR elevation data can be used in conjunction with visual data to overcome the aforementioned dif-
ficulties. Current IFSAR technology provides sufficient elevation resolution to discriminate building regions
from surrounding clutter. These building regions are not well defined from a visual image when the buildings
have the same intensity level as their surrounding background. Similarly, a building having different colors
may be wrongly segmented into several buildings. IFSAR data are not affected by color variations in buildings
and therefore are better for building detection.
Figure 4 shows a visual image and edges detected by the Canny operator for the area shown in Fig. 3. The
top part of Fig. 4 shows a building with two different roof colors and roof structures on many buildings. Many
spurious edges not corresponding to the building appear in the edge map shown on the bottom right of Fig. 4.
Using the IFSAR elevation map shown in Fig. 3, buildings and ground regions are labeled using a two class
10 RADAR TARGET RECOGNITION
Fig. 4. Visual image and edges detected by the Canny operator.
classifier. The IFSAR and visual images are registered. Figure 5 shows the result of registration of a visual
image and the segmented elevation image. Features corresponding to roads, parked cars, trees, and so on are
suppressed from the visual images using the segmented buildings derived from the IFSAR image.
The locations and the directions of edges in the segmented image are estimated and are used to locate
edges of buildings in the visual image. In the visual image, an edge pixel corresponding to each edge pixel in
the registered height image is searched in the direction perpendicular to the estimated direction in the height
RADAR TARGET RECOGNITION 11
Fig. 5. Buildings segmented from the IFSAR image overlaid to visual image.
image. If an edge is found within a small neighborhood, the edge pixel is accepted as a valid edge of a building.
If such a pixel is not found in the neighborhood, the edge is not accepted. Figure 6 shows the refined edges
obtained by searching in the neighborhoods of height edges. Most of building edges in the height image are
found while the unwanted edges are removed.
Spatial Matched Filter Classifiers for SAR (2)
A typical target recognition using SAR is done in multiple stages and is illustrated by the block diagram in
Fig. 7 (13). In the first stage, a CFAR detector prescreens by locating potential targets on the basis of radar
amplitude. Since a single target may produce multiple detections, the CFAR detections are clustered (grouped
together). Then a region of interest (ROI) around the centroid of each cluster is passed to the next stage of the
algorithm for further processing.
The second stage takes each ROI as its input and analyzes it. The goal of this discrimination stage
is to reject natural-clutter false alarms while accepting real targets. This stage consists of three steps: (1)
determining the position and orientation of the detected object, (2) computing simple texture features, and (3)
combining the features into a discrimination statistic that measures how “targetlike” the detection object is.
The third stage is classification, where a 2-D pattern-matching algorithm is used to (1) reject clutter false
alarms caused by man-made clutter discretes (buildings, bridges, etc.) and (2) classify the remaining detected
objects. Those detected objects that pass the second stage are matched against stored reference templates of
targets. If none of the matches exceeds a minimum required score, the detected object is classified as clutter;
otherwise, the detected object is assigned to the class with the highest match score.
Matched filters are investigated in 2 as pattern-matching classifiers in the target recognition system
shown in Fig. 7. They are synthetic discriminant function (SDF), the minimum average correlation
energy (MACE) filter, the quadratic distance correlation classifier (QDCC), and the shift-invariant 2-D
12 RADAR TARGET RECOGNITION
Fig. 6. Edges refined using IFSAR result.
pattern-matching classifier. The basic structure of the SDF and MACE filter is characterized in the frequency
domain by
where H denotes the DFT of the spatial matched filter. The matrix X is composed of a set of target training
vectors obtained by taking the DFT of the target training images. The vector U represents a set of constraints
imposed on the values of the correlation peaks obtained when the training vectors are run through the spatial
matched filter. The matrix A represents a positive definite weighting matrix. A is an identity matrix for SDF
RADAR TARGET RECOGNITION 13
Fig. 7. Block diagram of a typical baseline target recognition system. [Adapted from Novak et al. (13)].
and is the inverse of the following matrix D.
where N is the number of training images and p is the dimension of the training vectors.
In the QDCC, the DFT of the spatial matched filter is expressed by
where m1 and m2 are means of the DFTs of the training images for classes 1 and 2, respectively. S is a diagonal
matrix defined by
where M1 and M2 are matrices with elements of m1 and m2 placed on the main diagonal, and Xi and Yi are ith
training vectors from classes 1 and 2, respectively.
In the shift-invariant 2-D pattern-matching classifier, the correlation scores are calculated by
where T is the DFT of the dB-normalized test image and Ri is the ith reference template.
Novak et al. (2) did extensive experiment with the high-resolution (1 ft×1 ft) fully polarimetric SAR data.
In the four-class classification experiment using four types of spatial matched filter classifiers, it is reported
that all targets are correctly classified (2).
Multisensor Fusion (8,14,15)
Research has been conducted on multisensor fusion for target recognition. Some of the motivating factors of
such research are increased target illumination, increased coverage, and increased information for recognition.
Significant improvement in target recognition performance has been reported (8) when multiple radar sources
are utilized using sensor fusion approaches. Tenney and Sandell (14) developed a theory for obtaining the
distributed Bayesian decision rules. Chair and Varshney (15) presented an optimal fusion structure given that
14 RADAR TARGET RECOGNITION
Fig. 8. A typical data fusion approach for target recognition. [Adapted from Heuter et al. (8)].
detectors are independently designed. The target recognition using multiple sensors is formulated as a twostage
decision problem in Ref. 8. A typical radar target recognition approach using data fusion is illustrated in
Fig. 8. After the prescreening, single-source classifications are performed first; then the fusion of decision are
performed.
The data fusion problem is treated as an m-hypothesis problem with individual source decisions being
the observations. The decision rule for m-hypothesis is written as
For Bayes’ rule, gi(u) is a posterior probability. That is,
Since the prior probability and the distribution of features cannot be estimated accurately, a heuristic function
is used (8). It is a direct extension of Bayesian approach introduced by Varshney (16), and the function gi(·) is
generalized to include the full threshold range:
where P0 and P1 are prior probabilities; 1 and 0 are the sets of all i such that {gi(u)≥Ti} and {gi(u)
respectively, with Ti being the individual source threshold for partitioning decision regions; and the probabilities
Pf
i and Pd
i are false alarm rates and probabilities of detections of each local sensor. The probabilities
Pf
i and Pd
i are defined by the cumulative distribution functions (CDF) for each decision statistic. In practice,
the CDFs are quantized and estimated from training on the individual sensor’s classifier error probabilities.
In a distributed scenario, the weighting can be computed at each sensor and transmitted to the fusion center,
where they will be summed and compared to the decision threshold. In 8, the data fusion approach is applied
to multiple polarimetric channels of a SAR image, and substantially improved classification performance is
reported.
Summary
In radar target recognition, different types of radar are employed for different applications. In this article,
radar target recognition approaches for different radar systems are discussed.
RADAR TARGET RECOGNITION 15
BIBLIOGRAPHY
1. A. W. Rihaczek, S. J. Hershkowitz, Radar Resolution and Complex-Image Analysis, Norwood, MA: Artech House, 1996.
2. L. M. Novak, Radar target identification using spatial matched filters, Pattern Recognition, 27 (4): 607–617, 1994.
3. J. D. Wald, D. B. Krig, T. DePersia, ATR: Problems and possibilities for the IU community, Proc. ARPA Image Understanding
Workshop, January 1992, San Diego, CA, pp. 255–264.
4. M.W. Roth, Survey of neural network technology for automatic target recognition, IEEE Trans. Neural Netw., 1: 28–43,
1990.
5. A neural clustering approach for high resolution radar target classification, Pattern Recognition, 27 (4): 503–513, 1994.
6. L. I. Perlovsky et al., Model-based neural network for target detection in SAR images, IEEE Trans. Image Process., 6:
203–216, 1997.
7. M. A. Rubin, Application of fuzzy ARTMAP and ART-EMAP to automatic target recognition using radar range profiles,
Neural Netw., 8: 1109–1116, 1995.
8. A. Hauter, K. C. Chang, S. Karp, Polarimetric fusion for synthetic aperture radar target classification, Pattern Recognition,
30 (5): 769–775, 1997.
9. K. B. Eom, R. Chellappa, Non-cooperative target classification using hierarchical modeling of high range resolution
radar signatures, IEEE Trans. Signal Process., 45: 2318–2327, 1997.
10. M. W. Roth, Neural networks for extraction of weak targets in high clutter environments, IEEE Trans. Syst. Man
Cybern., 19: 1210–1217, 1989.
11. H. A. Zebker, R. M. Goldstein, Topographic mapping from interferometric synthetic aperture radar observations, J.
Geophys. Research, 91: 4993–4999, 1986.
12. R. Chellappa et al., On the positioning of multisensor imagery for exploitation and target recognition, Proc. IEEE, 85:
120–138, 1997.
13. L. M. Novak, A comparison of 1-D and 2-D algorithms for radar target classification, Proc. IEEE Int. Conf. Syst. Eng.,
August 1991, pp. 6–12.
14. R. R. Tenney, N. R. Sandell, Detection with distributed sensors, IEEE Trans. Aerosp. Electron. Syst., 17: 501–510, 1981.
15. Z. Chair, P. K. Varshney, Optimal data fusion in multiple sensor detection systems, IEEE Trans. Aerosp. Electron. Syst.,
22: 98–101, 1986.
READING LIST
J. S. Baras, S. I. Wolk, Model-based automatic target recognition from high-range-resolution radar returns, SPIE Proc.,
2234: 57–66, 1994.
M. Basseville, A. Benveniste, A. S. Willsky, Multiscale autoregressive processes, Part I & II, IEEE Trans. Signal Process.,
40: 1915–1954, 1992.
B. Bhanu, Automatic target recognition: State of the art survey, IEEE Trans. Aerosp. Electron. Syst., 22: 364–379, 1986.
V. Cantoni et al., Recognizing 2D objects by a multi-resolution approach, Proc. Int. Conf. Pattern Recognition, Vol. 3,
Jerusalem, Israel, October 1994, pp. 310–316.
N. C. Currie, R. D. Hayes, R. N. Trebits, Millimeter-Wave Radar Clutter, Norwood, MA: Artech House, 1992.
I. Daubechies, The wavelet transform, time-frequency localization and signal analysis, IEEE Trans. Inf. Theory, 36: 961–
1005, 1990.
D. M. Dunn, W. H. Williams, T. L. DeChaine, Aggregate versus subaggregate models in local area forecasting, J. American
Statistical Assoc., 71: 68–71, 1976.
J. Geweke, Temporal aggregation in the multiple regression model, Econometrica, 46: 643–662, 1978.
C. W. J. Granger, M. J. Morris, Time series modeling and interpretation, J. Royal Statistical Soc., A-139: 246–257, 1976.
S. Kingsley, S. Quegan, Understanding Radar Systems, New York: McGraw-Hill, 1992.
D. C. McKee et al., Model-based automatic target recognition using hierarchical foveal machine vision, SPIE Proc., 2755:
70–79, 1996.
R. A. Mitchell, R. Dewall, Overview of high range resolution radar target identification, Proc. Automatic Target Recognition
Working Group Conf., Monterey, CA, November 1994.
16 RADAR TARGET RECOGNITION
F. A. Pino, P. A. Morettin, R. P. Mentz, Modelling and forecasting linear combinations of time series, Int. Statistical Rev.,
55: 295–313, 1987.
O. Rioul, A discrete-time multiresolution theory, IEEE Trans. Acoust. Speech Signal Process., 41: 2591–2606, 1993.
M. L. Skolink, Introduction to Radar Systems, New York: McGraw-Hill, 1980.
N. S. Subotic et al., Multiresolution detection of coherent radar targets, IEEE Trans. Image Process., 6: 21–35, 1997.
L. G. Telser, Discrete samples and moving sums in stationary stochastic processes, J. American Statistical Assoc., 62:
484–499, 1967.
D. R. Wehner, High Resolution Radar, Norwood, MA: Artech House, 1987.
W. Wei, The effect of temporal aggregation of parameter estimation in distributed lag model, J. Econometrics, 8: 237–246,
1978.
W. W. S. Wei, D. O. Stram, Disaggregation of time series models, J. Royal Statistical Soc., B-52: 453–467, 1990.
M. A. Wincek, G. C. Reinsel, An exact likelihood estimation procedure for regression ARMA time series models with
possibly non-consecutive data, J. Royal Statistical Soc., B-48: 303–313, 1986.
KIE B. EOM
George Washington University