Sei sulla pagina 1di 11

Feature Article: DOI. No. 10.1109/MAES.2019.

2933972

Deep Learning for Classification of Mini-UAVs Using


Micro-Doppler Spectrograms in Cognitive Radar
Albert Huizing, Matijs Heiligers, Bastiaan Dekker, Jacco de Wit, Radar
Technology Department, TNO, The Hague, The Netherlands
Lorenzo Cifola, Ronny Harmanny, Advanced Development Thales Nederland
B.V., Delft, The Netherlands

operate effectively and robustly in a dynamic environ-


INTRODUCTION
ment. In addition to the feedback from the receiver, a cog-
Military radar can not only provide kinematic information nitive radar can also use prior knowledge about the
(position, speed, and course) of land, sea, and air targets environment and targets of interest to optimize its perfor-
during day and night in all weather conditions, but also mance [2]. This prior knowledge can be provided by
provides information about the type of target using micro- human experts in the form of heuristics, terrain databases,
Doppler properties. The micro-Doppler properties of a tar- or computer models of the interaction between the radar
get are determined by the moving parts on the target body. and the environment, including target signatures, such as
The number, location, and type of motion of these parts micro-Doppler spectrograms. Alternatively, prior knowl-
are specific for a given target type and consequently the edge can also be derived from previous radar measure-
related micro-Doppler signature can be exploited for clas- ments in the same or similar environments [3]. Examples
sification. Analysis of the micro-Doppler signature may of this are a land clutter map and micro-Doppler spectro-
provide detailed properties of rotating parts, such as the grams of targets of interest.
rotation rate, number of blades, and blade length. How- Figure 1 shows a high-level block diagram that incor-
ever, for operators and/or image analysts, the interpreta- porates two of the key elements of a cognitive radar: a per-
tion and understanding of radar micro-Doppler ception–action cycle and priori knowledge stored in long-
spectrograms is much more difficult and time-consuming term memory, i.e., a knowledge base (KB). This diagram
than the analysis of optical images because of the different has been adapted from the cognitive radar architecture
nature of the radar imaging principle and target scattering proposed by Guerci [2]. A knowledge-aided processor
mechanisms. Consequently, there is a need for automatic (KAP) uses prior knowledge in the KB to control the radar
target recognition (ATR) in radar using micro-Doppler scheduler, the waveforms transmitted by the adaptive
spectrograms. transmitter, the radio frequency filter parameters and the
An ability to automatically recognize complex pat- sampling rate in the adaptive receiver, and the parameters
terns in radar signals is one of the key elements of the per- and configuration of the radar signal processing. The sig-
ception–action cycle in cognitive radar as identified by nal processing typically includes functions, such as pulse
Haykin et al. [1]. A cognitive radar uses the feedback compression, Doppler filtering, detection, tracking, and
from the receiver to adjust the transmitted waveforms to classification. The KAP may decide to update the KB
based on information that is extracted from the radar sig-
nal. For example, if the number of false alarms in an area
Authors’ current addresses: Albert Huizing, Matijs is too large due to a change in the clutter distribution, the
Heiligers, Bastiaan Dekker, Jacco de Wit, Radar KAP may decide to update the clutter map.
Technology Department, TNO, 2594 AK, The Hague, In conventional radars, the recognition of targets, such
The Netherlands, (E-mail: albert.huizing@tno.nl). Lor- as mini-UAVs typically relies on the extraction of prede-
enzo Cifola, Ronny Harmanny, Advanced Develop- fined target features from the received radar signal. These
ment, Thales Nederland B.V. 2628 XH, Delft, The
target features are defined and implemented by a human
Netherlands.
expert in the radar signal processing, often after a long
Manuscript received March 27, 2019, revised July 12,
2019, and ready for publication August 6, 2019. and tedious analysis of large amounts of radar data.
Review handled by P. Willett. Instead of this hand-crafted feature engineering approach,
0885-8985/19/$26.00 ß 2019 IEEE a cognitive radar can use deep learning techniques to

46 IEEE A&E SYSTEMS MAGAZINE NOVEMBER 2019


automatically extract the relevant features from a set of CLASSES” describes how the spectrograms of target classes
radar measurements or radar simulations that are stored in that are not represented in the training set, i.e., targets
the KB. The extracted features can then be used by the unknown by the cognitive radar, can be detected with
KAP to classify the target. deep neural networks. The detection of an unknown target
Deep learning has become the preferred method for can provide a trigger for the radar scheduler to collect
many pattern recognition applications, such as segmenta- micro-Doppler spectrograms of this target for inclusion in
tion, speech recognition, and face recognition since the the training set, i.e., the knowledge base. “DENOISING OF
quantum leap in image recognition performance realized SPECTROGRAMS” shows how a trained neural network
with the AlexNet convolutional neural network (CNN) in can be used to denoise spectrograms. This denoising process
the ImageNet 2012 challenge [4]. Despite the major may enable a cognitive radar to recognize targets at a longer
achievements of deep learning techniques, such as CNNs range. “ADVERSARIAL TRAINING FOR SPECTRO-
in the past few years, it is not obvious that the success of GRAM GENERATION” describes the generation of new
deep learning in the commercial domain can be replicated spectrograms for training a classifier using generative adver-
in the military domain. Large labeled datasets, which are sarial networks. Finally, in “CONCLUSION,” conclusions
the key to the success of most commercial applications of are drawn with respect to the application of deep learning for
deep learning, are often not available in military applica- classification of mini-UAVs with cognitive radar.
tions. In addition, the cost of decision errors in the military
domain is typically much higher than in the commercial
applications. This makes it mandatory to achieve a robust CLASSIFICATION OF MINI-UAVS
performance with a low error rate. Furthermore, military
The commercial availability of compact electronics and
ATR algorithms should not only be accurate, but their
advanced open source software has led to a proliferation
behavior should also be predictable to gain the trust of
of mini-UAVs that can be used for many different pur-
military commanders.
poses including criminal, terrorist, and military activities.
One of the first applications of deep learning for ATR
Owing to their relatively slow speed and small size, they
using synthetic aperture radar imagery was presented by
are hard to distinguish from natural targets, such as birds.
Morgan [5]. Since then many papers on the application in
Recently, radar signal processing techniques using fea-
radar of CNNs and other types of deep neural networks
tures extracted from spectrograms and cepstrograms have
have been published, including the application of deep
been developed to discriminate birds and mini-UAVs
learning for target classification, and human gait and ges-
[10], [11]. The recognition of different classes of mini-
ture recognition using micro-Doppler spectrograms [6]–[9].
UAVs is, however, a more difficult problem due to the
This paper investigates the potential of deep learning
overlapping features and rapidly changing characteristics
techniques for the classification of mini-UAVs using micro-
of mini-UAVs. Therefore, an enhanced capability for
Doppler spectrograms in the context of cognitive radar. The
radars is needed to recognize the type of mini-UAV for
paper also presents preliminary results on the use of deep
threat evaluation and assignment of countermeasures.
learning in the preprocessing and training process of a classi-
fier. This paper is organized as follows. “CLASSIFICATION
OF MINI-UAVS” shows how deep neural networks can be
applied for the classification of mini-UAVs using measured
MINI-UAV MEASUREMENTS
or simulated sets of micro-Doppler spectrograms that are To investigate the potential of deep learning techniques
stored in the long-term memory (or knowledge base) of a for the recognition of mini-UAVs, radar measurements of
cognitive radar. “DETECTION OF UNKNOWN TARGET five different types of mini-UAVs have been acquired

NOVEMBER 2019 IEEE A&E SYSTEMS MAGAZINE 47


Deep Learning for Classification of Mini-UAVs Using Micro-Doppler Spectrograms in Cognitive Radar

Figure 1.
High-level block diagram of a cognitive radar (adapted from Guerci [2]).

during a measurement campaign at the NLR RPAS Test UAV body is removed to center the Doppler spectrum
Centre in Marknesse, The Netherlands. The mini-UAVs around 0 Hz. Examples of the measured micro-Doppler
included two types of fixed wing aircraft (Robbe Air spectrograms for all 5 mini-UAVs can be seen in Figure 2.
Trainer 140 and Sky Walker X8), an Align T-REX 550 The coherent integration time of the spectrograms is
helicopter, a Mikrokopter QuadroXL quadcopter, and a 10.7 ms. The modulations of the radar signal due to the
Mikrokopter OktoXL octocopter, see Figure 2. moving parts can clearly be observed.
Measurements of the mini-UAVs were conducted with
an experimental continuous wave radar that operates at a
frequency of 9.5 GHz and samples the radar signal at a
rate of 96 kHz. The complex radar signals are converted
MICRO-DOPPLER SIGNAL SIMULATIONS
to micro-Doppler spectrograms using a short time fourier In addition to the measurements, micro-Doppler signals
transform (STFT). The Doppler frequency of the mini- have been simulated using computer models of mini-

Figure 2.
Photos (1st row), measured spectrograms (2nd row), computer models (3rd row), and simulated spectrograms (4th row) of five mini-UAVs.

48 IEEE A&E SYSTEMS MAGAZINE NOVEMBER 2019


Huizing et al.

Figure 3.
Workflow for the classification of mini-UAVs with deep neural networks using micro-Doppler spectrograms.

UAVs to test the idea that a deep neural network can be CLASSIFICATION OF MINI-UAVS WITH DEEP NEURAL
trained with simulated micro-Doppler spectra. The trained
neural network can then be used to classify micro-Doppler NETWORKS
spectra measured with a real radar system during opera-
Figure 3 shows the workflow for the classification of mini-
tions. This concept would reduce the need for expensive UAVs with deep neural networks. The training phase con-
and time-consuming radar measurements to populate the sists of generating a training set with (a) simulated radar
knowledge base of a cognitive radar. It also could enable signals using target models or (b) real target signals mea-
the classification of mini-UAVs of which physical charac- sured with a radar. Micro-Doppler spectrograms are cre-
teristics are available, but no measurements can be made ated by applying an STFT successively to overlapping
due to security reasons. sequences of the time-domain signals. The preprocessing
The simulation of micro-Doppler spectra of targets, includes removal of the speed of the mini-UAV body, and
such as mini-UAVs is difficult because of the complex
normalization of the power spectrogram. In the test phase,
mechanical structure of these targets and the time varying the trained neural network is used to classify the measured
characteristic of the radar cross section (RCS) due to the radar signals.
moving parts. To limit the computation time for micro- Table 1 shows the number of micro-Doppler spectro-
Doppler signal simulations, an approach was chosen in grams that have been extracted from the measured
which the mini-UAVs are represented with so-called and simulated radar signals for the training and test sets
principal scatterers for which the analytic expressions of in the mini-UAV classification experiments described
their RCS exist [12]. Examples of these principal scatter- below.
ers are point scatterers, cylinders, ellipsoids, flat triangu-
lar plates, and thin wires. For each sample of the radar
signal, the position and attitude of each principal scat- Table 1.
terer is determined and the associated RCS for this scat-
terer is computed. Finally, the reflected radar signal for Micro-Doppler Spectrograms for the Five Mini-UAVs
the entire target is calculated by adding the contributions
of all principal scatterers. This simulation approach is rel- Training Test
atively fast, but neglects the occlusion of some scatterers Target
by other scatterers on the body, multiple reflections, dif- Simulated Measured Simulated Measured
fraction, and materials other than metal. Figure 2 shows
the computer models of the five mini-UAVs considered Air trainer 506 570 525 499

in this paper and examples of the associated simulated Skywalker 475 886 545 775
micro-Doppler spectrograms. Figure 2 clearly shows a
T-REX 550 965 1284 1080 1092
resemblance between the measured and the simulated
micro-Doppler spectrograms, but also some measured QuadroXL 477 841 546 869
characteristics that are not present in the simulated OktoXL 480 1370 555 1222
spectrograms.

NOVEMBER 2019 IEEE A&E SYSTEMS MAGAZINE 49


Deep Learning for Classification of Mini-UAVs Using Micro-Doppler Spectrograms in Cognitive Radar

Table 2.

Deep Neural Network Test Cases for Classification of Mini-UAVs Using Micro-Doppler Spectrograms

Case Training Test data Neural Sampling Integration Accuracy


data network rate time

A Simulated Simulated CNN 96 kHz 10.7 ms 91.7%


B Simulated Measured CNN 96 kHz 10.7 ms 36.7%
C Measured Measured CNN 96 kHz 10.7 ms 82.8%
D Measured Measured CNN 24 kHz 10.7 ms 87.1%
E Measured Measured CNN 24 kHz 2.7 ms 97.7%
F Measured Measured LSTM-RNN 24 kHz 2.7 ms 89.0%

shadowing, multiple reflections, diffraction, and material


CLASSIFICATION EXPERIMENTS AND RESULTS
properties. More advanced RCS modeling techniques
To investigate the accuracy of different types of deep neu- using a hybrid finite element method and method of
ral networks for the classification of mini-UAVs, several moments approach may improve the fidelity of the micro-
test cases were defined, see Table 2. Test case A involves Doppler spectrograms at the expense of computational
the use of a CNN with a high sampling rate (96 kHz) of cost [13].
the radar signal and a medium coherent integration time A second conclusion is that the decimation of the sam-
(10.7 ms) in the Doppler filtering and simulated training pling rate by a factor of 4 improves the mean classification
and test data. Test case B gives an idea if the simulated accuracy for the measured spectrograms considerably.
spectrograms can be used to classify measured spectro- This is due to the fact that the part of the spectrogram that
grams from real mini-UAVs. does not contain any relevant information is removed by
Test case C uses measured spectrograms for training
and testing. Test case D also uses measured spectrograms
for training and testing, but employs a reduced sampling Table 3.
rate by decimating the radar signal with a factor 4. Test
case E uses a low sampling rate and a short integration CNN Configuration
time (2.7 ms). Finally, test case F uses a long short-term
memory (LSTM) recurrent neural network (RNN) instead Layer Type Size Kernel Filters
of a CNN as a classifier [14].
Table 3 shows the configuration of the CNN used in the 1 Input 1024  177 – –
first three test cases (A, B, and C). All convolutional and 2 Convolution 1024  177 55 32
fully connected layers (except for layer 13 which is linear)
3 Convolution 1024  177 55 32
use a rectified linear unit as a nonlinear activation function.
The convolutional and max-pool layers of the CNNs in test 4 Max-pool 512  88 22 32
cases D and E have smaller sizes than the configuration in 5 Convolution 512  88 55 64
Table 3 due to the reduced sampling rate and integration
6 Convolution 512  88 55 64
time. The CNN is trained with a categorical cross-entropy
loss function and the ADAM optimizer [5]. 7 Max-pool 256  44 22 64
The LSTM RNN in test case F does not use a spectro- 8 Convolution 256  44 55 64
gram as an input, but operates on an input sequence of
9 Convolution 256  44 55 64
Doppler spectra. The LSTM-RNN configuration has four
layers and a hidden state size of 256 for the LSTM cells. 10 Max-pool 128  22 22 64
A comparison of the mean classification accuracy in 11 Flatten 1  180224 – –
Table 2 shows that good results are achieved except for
the case where simulated spectrograms are used for train- 12 Fully con. 1  512 – –
ing a CNN and measured spectrograms are used during 13 Fully con. 15 – –
testing. This is perhaps not surprising due to the lack of
14 Soft-max 15 – –
fidelity in the target models, which do not include

50 IEEE A&E SYSTEMS MAGAZINE NOVEMBER 2019


Huizing et al.

Figure 4.
Adversarial training of an auto-encoder with the GANomaly method to detect spectrograms of unknown targets.

the decimation. This simple and effective preprocessing application of a threshold to the output of the Soft-max
step is apparently difficult to learn by the CNN from the layer of a CNN [15]. If the maximum output of the Soft-
limited number of training samples in test case C. max layer does not exceed the threshold, an unknown tar-
Another conclusion is that the reduction in coherent get class is declared. This Soft-max approach is also
integration time in the Doppler filtering from 10.7 ms to referred to as a reject option for a classifier [16]. By vary-
2.7 ms also gives a significant improvement in classifica- ing the value of the threshold, a receiver operation charac-
tion accuracy. This improvement can be explained by the teristic (ROC) curve can be obtained, which shows the
fact that the modulations by the rotating propeller or rotor probability of detecting an unknown target class (true pos-
blades are less smeared in the Doppler spectrum due to itive rate) as a function of the probability of falsely declar-
the shorter integration time, and the changes in the Dopp- ing an unknown target class (false positive rate).
ler spectrum become clearer. An alternative to the Soft-max approach called GANo-
Although the use of an LSTM-RNN proved to be less maly is based on an auto-encoder that is trained in an
accurate than the CNN in the overall classification, an adversarial manner [17]. This technique has been used to
important advantage of an LSTM-RNN is that it can pro- investigate the potential of deep learning to screen avia-
vide good classification results already after a few coher- tion luggage for anomalous items using X-ray screening.
ent processing intervals, whereas the CNN only provides a Figure 4 shows the configuration of the GANomaly net-
classification result after an entire spectrogram has been work that consists of three subnetworks. The auto-encoder
processed [14]. Another advantage of the LSTM-RNN is acts as a generator network that learns a compact latent
that it is capable to deal with transitions in target behavior, space representation z of the input spectrograms x by try-
e.g., a fixed wing mini-UAV that takes off vertically, and ing to reconstruct the input spectrogram. The encoder
then, proceeds to horizontal flight. compresses the reconstructed spectrogram to an estimate
of the latent representation using the same architecture as
the encoder in the auto-encoder. The discriminator deter-
mines, with an encoder architecture, if the input is a real
DETECTION OF UNKNOWN TARGET CLASSES spectrogram or a fake spectrogram.
During military operations, cognitive radars will often be The detection of a spectrogram originating from an
employed in conditions where novel target classes are unknown target assumes that the auto-encoder is not able
observed of which there are no examples of micro-Dopp- to reconstruct such a spectrogram accurately because the
ler spectrograms in the training set of the deep neural net- network is trained only with spectrograms from known
work. In this case, the cognitive radar should detect the target classes. A reconstructed spectrogram for an
presence of this unknown target class and may decide to unknown spectrogram will also lead to discrepancies
schedule specific radar measurements to add spectrograms between the latent vector and its estimate.
of the unknown target class to the training set. This discrepancy is used by the GANomaly method
The detection of unknown target classes can be during the test phase to detect a spectrogram of an
achieved in several ways. The simplest approach is the unknown target class.

NOVEMBER 2019 IEEE A&E SYSTEMS MAGAZINE 51


Deep Learning for Classification of Mini-UAVs Using Micro-Doppler Spectrograms in Cognitive Radar

Figure 5.
ROC curve showing the true positive rate versus the false positive
rate for detecting an unknown Align T-REX 550 mini-UAV with Figure 6.
the Soft-Max and GANomaly methods. AUC results for the detection of unknown target classes with Soft-
max and GANomaly. The error bars represent variations due to the
use of random seeds for the initialization of the neural networks.
The objective function for training the GANomaly net-
works consists of three loss functions. The adversarial loss
function Ladv computes the L2 distance between features network and hyperparameters are as described in
computed by the function f from an intermediate layer of “CLASSIFICATION OF MINI-UAVS.” The models for
the discriminator network for the real and generated spec- GANomaly and Soft-max are trained for, respectively, 15
trograms. This feature matching approach reduces the and 50 epochs.
instability of GANs during training [18]. The context loss Figure 6 shows that both the Soft-max and the GANo-
function Lcon penalizes the auto-encoder for reconstruction maly achieve AUC ranging from 0.5 to 0.8. Note that
errors by measuring the L distance between the real and comparing the performance of deep learning methods is
generated spectrograms. Finally, the encoder loss Lenc often tricky because the result can highly depend on the
measures the L1 distance between the latent representations effort that has been put in optimizing the networks, train-
of the input and reconstructed spectrograms. ing parameters, and data preprocessing steps. Especially
Experiments with the Soft-max and GANomaly methods adversarial training methods, such as GANomaly, are
for the detection of unknown target classes have been con- notorious for instability issues. In the experiments, it was
ducted using the dataset with mini-UAV spectrograms, as observed that the final performance of the GANomaly
described in “CLASSIFICATION OF MINI-UAVS.” Each method was not yet fully satisfactory, and perhaps better
of the mini-UAVs has been used alternatively as an unknown performances can be achieved with other GAN configura-
target class when training the neural networks on the micro- tions for the detection of unknown target classes.
Doppler spectrograms from the remaining ‘“known” mini-
UAVs. Figure 5 shows an example of an ROC curve for a
test where the T-REX 550 helicopter was used as an
unknown target class. The GANomaly method performs bet-
DENOISING OF SPECTROGRAMS
ter at low and high false positive rates, whereas the Soft-max The micro-Doppler spectrograms collected during the
method works better at intermediate false positive rates. measurement campaign described in “CLASSIFICATION
Figure 6 shows the area-under-curve (AUC) for the OF MINI-UAVS” have a relatively high signal-to-noise
detection of unknown target classes with the Soft-max and ratio (SNR) due to the short range between the radar and
GANomaly methods. Each experiment is performed three the mini-UAVs. However, the SNR will be lower at longer
times to capture the performance variations due to the random ranges, and therefore, the performance of a target classifier
initialization of the neural networks and due to the random will degrade. To investigate if deep learning can be used
selection of the data samples in the test set and training set. to improve the SNR of the spectrograms in the preprocess-
The above results are obtained using the measured ing before classification, and hence, increase the range at
spectrograms of test case E. All spectrograms are truncated which a cognitive radar can recognize a mini-UAV, an
after the first 64 time samples resulting in square input experiment was conducted with a denoising adversarial
data of 6464 pixels. This input size enabled the use of auto-encoder. Adversarial trained auto-encoders can be
network configuration and training parameters described used to restore input data, such as images that are cor-
in the paper by Akcay et al. [17] that are proven to be rupted by noise [19]. The architecture of the denoising
successful for the GANomaly method. For Soft-max, the adversarial auto-encoder is like the architecture of the

52 IEEE A&E SYSTEMS MAGAZINE NOVEMBER 2019


Huizing et al.

Figure 7.
Denoising adversarial auto-encoder applied to spectrograms corrupted with synthetic noise.

GANomaly network, as shown in Figure 4, except that the imbalance in the training set used in this paper
separate encoder is omitted and the objective function for (see Table 1) is approximately a factor of two. This imbal-
training of the auto-encoder only consists of the mean ance can have a negative impact on the convergence of
square error between noise corrupted versions of the spec- the classifier during training and the overall accuracy [20].
trograms and the uncorrupted spectrograms. There are several ways to mitigate the effect of an imbal-
Figure 7 shows two examples of measured micro- anced training set. The first category concerns methods
Doppler spectrograms of a T-REX 550 helicopter with a that leave the training set intact and modify the training
high SNR in the first column. The second column shows procedure or the classifier to deal with the imbalance. The
the same spectrograms to which synthetic Gaussian noise second category involves changes to the training set itself
has been added to lower the SNR. The third column shows by generating data for classes that are underrepresented.
the reconstructed spectrograms using a denoising adver- In this section, a generative network called InfoGAN has
sarial auto-encoder. Although some details of the micro- been investigated for the generation of realistic training
Doppler modulations are distorted, it is still possible to data for underrepresented target classes [21]. When com-
recognize the micro-Doppler contribution due to the tail pared with a standard GAN, an InfoGAN has the advan-
rotor and, up to a certain extent, the blade tip contribution. tage that it learns in an unsupervised way, interpretable
In a second experiment, the denoising adversarial and disentangled representations of challenging datasets.
auto-encoder has been tested with measured spectrograms Figure 9 shows the architecture of an InfoGAN network
that are characterized by a lower SNR. As apparent from
the reconstructed spectrograms shown in Figure 8, a signal
to background ratio enhancement in the order of 20 dB is
achieved by the denoising auto-encoder. Furthermore,
information associated to tail rotor signature and blade tip
rotation is partially restored. Although the SNR of the
spectrograms seems to be significantly enhanced by the
denoising adversarial auto-encoder, the impact of this pre-
processing on the accuracy of a target classifier, such as a
CNN still has to be investigated.

ADVERSARIAL TRAINING FOR SPECTROGRAM


GENERATION
Training sets for target classifiers based on machine learn- Figure 8.
ing are often characterized by an imbalance in the number Denoising adversarial auto-encoder applied to measured low SNR
of examples for different target classes. For example, the spectrograms.

NOVEMBER 2019 IEEE A&E SYSTEMS MAGAZINE 53


Deep Learning for Classification of Mini-UAVs Using Micro-Doppler Spectrograms in Cognitive Radar

Based on an information-based regularization term


within the overall training cost function, a structure is
introduced in the latent variable that enables the aggrega-
tion of spectrograms characterized by similar characteris-
tics. This is apparent in the generated spectrograms shown
in Figure 10 for nine different categories associated with
nine different states for the latent code (c). In addition,
since clustering is based on spectrogram appearance,
instances of the same class characterized by differing
spectrograms will be associated to different categories.
For example, both examples from category 2 and 3 belong
to the T-REX 550 helicopter class, although characteri-
zed by a different Doppler bandwidth. When the specified
categories are higher than the level of diversity among the
dataset examples, then overlap between categories can be
expected, for instance as observed for categories 5 and 9.
Figure 9.
InfoGAN architecture.

that consists of a generator, a discriminator, and an auxil-


CONCLUSION
iary unit. The results of several experiments in this paper have dem-
The generator input vector consists of the concatena- onstrated the potential of deep learning techniques for the
tion of two parts: a noise vector (z) and a latent code (c). classification of mini-UAVs with a cognitive radar. Deep
The generator tries to fool the discriminator by generating neural networks, such as CNNs and RNNs can provide an
a realistic spectrogram, whereas the discriminator tries to accurate classification performance after being trained on
distinguish the synthetic and real spectrograms. In addi- a set of measured micro-Doppler spectrograms that are
tion to the probability that an input is a real or a fake spec- stored in the long-term memory of a cognitive radar.
trogram, the discriminator also computes a distribution The analysis of the classification performance of dif-
Q(c j G(c,z)) that measures the mutual information bet- ferent types of deep neural networks and different prepro-
ween the latent code c and the generated spectrogram. cessing parameters showed that the choice of the coherent

Figure 10.
InfoGAN applied to micro-Doppler spectrograms with examples of synthesized spectrograms per category of the latent code.

54 IEEE A&E SYSTEMS MAGAZINE NOVEMBER 2019


Huizing et al.

integration time in the Doppler filtering had a higher Dutch Radar Center of Expertise (D-RACE), a strategic
impact on the classification performance than the type of alliance between Thales Nederland B.V. and TNO.
the deep neural network (CNN versus LSTM-RNN) in the
classifier. The use of simulated spectrograms to classify
measured spectrograms requires a higher fidelity modeling REFERENCES
of the micro-Doppler signals than the modeling approach
[1] S. Haykin, Y. Xue, and P. Setoodeh, “Cognitive radar: Step
based on principal scatterers used in this paper.
toward bridging the gap between neuroscience and engi-
When a cognitive radar measures a micro-Doppler
neering,” Proc. IEEE, vol. 100, no. 11, pp. 3102–3130,
spectrogram from a target class that is not represented in
Nov. 2012.
the training set, an anomaly detection algorithm could pro-
[2] J. R. Guerci, Cognitive Radar: The Knowledge-Aided
vide a trigger to the cognitive radar scheduler to collect
Fully Adaptive Approach, Norwood, MA, USA: Artech
micro-Doppler spectrograms from this unknown target.
House, 2010.
The performance of two deep learning techniques, Soft-
[3] M. S. Greco, F. Gini, P. Stinco, and K. Bell, “Cognitive
max and GANomaly, for the detection of spectrograms
radars: On the road to reality: Progress thus far and possibil-
from unknown target classes has been compared. Both
ities for the future,” IEEE Signal Process. Mag., vol. 35,
techniques had a similar performance with respect to the
no. 4, pp. 112–125, Jul. 2018.
probability of detecting unknown classes with the GANo-
[4] Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,”
maly technique showing a larger performance variation
Nature, vol. 521, no. 7553, pp. 436–444, 2015.
due to the difficulty in training adversarial networks. Dif-
[5] D. A. E. Morgan, “Deep convolutional neural networks for
ferent GAN architectures for anomaly detection that are
ATR from SAR imagery,” Proc. SPIE, vol. 9475, 2015,
easier to train, such as a Wasserstein GAN may provide a
Art. no. 94750F.
better performance [22].
[6] Y. Kim and T. Moon, “Human detection and activity clas-
A denoising adversarial auto-encoder has been trained
sification based on micro-Doppler signatures using deep
to remove noise from spectrograms. An improvement of
convolutional neural networks,” IEEE Geosci. Remote
20 dB in the SNR of a measured spectrogram has been
Sens. Lett., vol. 13, no. 1, pp. 8–12, Jan. 2016.
observed. If this denoising technique does not alter the €
[7] M. S. Seyfioglu, S. Z. G€urb€uz, A. M. Ozbayo glu, and
micro-Doppler characteristics significantly, a considerable
M. Y€uksel, “Deep learning of micro-Doppler features for
increase in the range at which a cognitive radar can recog-
aided and unaided gait recognition,” in Proc. IEEE Radar
nize mini-UAVs could be achieved.
Conf., 2017, pp. 1125–1130.
To mitigate the effects of class imbalances in the num-
[8] R. P. Trommel, R. I. A. Harmanny, L. Cifola, and
ber of training examples on the performance of a classi-
J. N. Driessen, “Multi-target human gait classification
fier, an InfoGAN can be used to generate additional
using deep convolutional neural networks on micro-
spectrograms of underrepresented classes. A quantitative
Doppler spectrograms,” in Proc. Eur. Radar Conf., 2016,
study concerning the benefits of this approach will be con-
pp. 81–84.
ducted in the future research.
[9] B. Dekker, S. Jacobs, A. S. Kossen, M. C. Kruithof,
The overall conclusion of this paper is that deep learn-
A. G. Huizing, and M. Geurts, “Gesture recognition with a
ing techniques have a high potential for improving the
low power FMCW radar and a deep convolutional neural
classification of mini-UAVs using micro-Doppler spectro-
network,” in Proc. Eur. Radar Conf., 2017, pp. 163–166.
grams in cognitive radar. However, the selection and tun-
[10] P. Molchanov, K. Egiazarian, J. Astola, R. I. A. Harmanny,
ing of the appropriate neural network architectures for
and J. J. M. de Wit, “Classification of small UAVs and birds
different processing tasks, such as classification and
by micro-Doppler signatures,” in Proc. Eur. Radar Conf.,
denoising is tedious and time consuming. A better under-
2013, pp. 172–175.
standing of the theory behind deep neural networks, poten-
[11] R. Harmanny, J. J. M. de Wit, and G. Premel-Cabic,
tially enabled by the link with compressive sensing and
“Radar micro-Doppler mini-UAV classification using
sparse signal representations, would not only accelerate
spectrograms and cepstrograms,” Int. J. Microw. Wireless
the development process, but also improve the acceptance
Technol., vol. 7, no. Special Issue 3/4, pp. 469–477, 2015.
of these advanced techniques in military radar
[12] V. C. Chen, F. Li, S.-S. Ho, and H. Wechsler, “Micro-
applications.
Doppler effect in radar: Phenomenon, model, and simula-
tion study,” IEEE Trans. Aerosp. Electron. Syst., vol. 42,
no. 1, pp. 2–21, Jan. 2006.
ACKNOWLEDGMENT
[13] P. J. Speirs, A. Schr€oder, M. Renker, P. Wellig, and
TNO has received funds for this study from the Netherlands A. Murk, “Comparisons between simulated and measured
Ministry of Defense, under Grant V1512 and Grant V1908. X-band signatures of quad-, hexa- and octocopters,” in
The study has been conducted within the scope of the Proc. 15th Eur. Radar Conf., 2018, pp. 325–328.

NOVEMBER 2019 IEEE A&E SYSTEMS MAGAZINE 55


Deep Learning for Classification of Mini-UAVs Using Micro-Doppler Spectrograms in Cognitive Radar

[14] G. Klarenbeek, R. I. A. Harmanny, and L. Cifola, “Multi- [19] A. Creswell and A. A. Bharath, “Denoising adversarial
target human gait classification using LSTM recurrent neu- autoencoders,” IEEE Trans. Neural Netw. Learn. Syst.,
ral networks applied to micro-Doppler,” in Proc. 14th Eur. vol. 30, no. 4, pp. 968–984, Apr. 2019.
Radar Conf., 2017, pp. 167–170. [20] M. Buda, A. Makib, and M. A. Mazurowski, “A system-
[15] A. Bendale and T.E. Boult, “Towards open set deep atic study of the class imbalance problem in convolutional
networks,” in Proc. IEEE Conf. Comput. Vis. Pattern Rec- neural networks,” Neural Netw., vol. 106, pp. 249–259,
ognit., Jun. 2016, pp. 1563–1572. 2018.
[16] D. M. J. Tax and R. P. W. Duin, “Growing a multi-class [21] X. Chen, Y. Duan, R. Houthooft, J. Schulman, I. Sutskever,
classifier with a reject option,” Pattern Recognit. Lett., and P. Abbeel, “InfoGAN: Interpretable representation
vol. 29, no. 10, pp. 1565–1570, 2008. learning by information maximizing generative adversarial
[17] S. Akcay, A. Atapourabarghouei, and T. P. Breckon, nets,” in Proc. 30th Int. Conf. Neural Inf. Process. Syst.,
“GANomaly: Semi-supervised anomaly detection via 2016, pp. 2172–2180.
adversarial training,” in Proc. ACCV, Perth, Australia, [22] T. Schlegl, P. Seeböck, S.M Waldstein, G. Langs, and
Dec. 2018. U. Schmidt-Erfurth, “f-anogan: Fast unsupervised anomaly
[18] T. Salimans, I. Goodfellow, W. Zaremba, W. Cheung, detection with generative adversarial networks,” Medical
A. Radford, and X. Chen, “Improved techniques for train- image analysis, vol. 54, pp. 30–44, May 2019.
ing GANs,” in Proc. 30th Int. Conf. Neural Inf. Process.
Syst., 2016, pp. 2234:2242.

56 IEEE A&E SYSTEMS MAGAZINE NOVEMBER 2019

Potrebbero piacerti anche