IEEE Aerospace and Electronic Systems Magazine - December 2020 - 41

Zhang et al.

Figure 11.
Detection results by the spatiotemporal SEG-MSAKBOT
extractor [27].

surface roughness of asphalt. It shows rather specular
behavior. In a higher frequency band (e.g., W-band),
however, due to the short wavelength, the specular
behavior is significantly reduced, the road's echo is stronger, and thus a potential shadow of a vehicle on the road
shows significantly more contrast. Of course, the shadow
of moving vehicles off-road on different backgrounds,
such as grass, should even present more contrast level
[12]. Abovementioned influences arise the strict demands
on radar system design and signal acquisition. It will be
a difficult task and more realistic for the detection and
tracking of indistinct shadow signatures, particularly for
the smaller vehicles. Accordingly, the vehicle shadows
may not be easy to capture by our developed NUAAMiniSAR system.
Nevertheless, vehicle motion across the video screen
remains some bright lines. Significant defocusing effect
represents the backscattering reflectivity of moving vehicle that can cause space-variant phase information, and
hence its geographical location is considerably different
from the actual trajectory. Compared with the cast shadow
region lacking of direct backscattering energy, defocused
vehicle information indicates a powerful echo-location
characteristic in the cases of Doppler shift and range
smear. In terms of visual effect in the SAR image domain,
there are the format of a higher gray value in favor of recognizable detection. Moving bright lines independent of
SCR, in a sense, are also propitious to improve the robustness and universality for a practical VideoSAR system.
Herein, by exploiting computer vision technology,
we provide two perspectives to illustrate the potential
advantages in terms of automatic vehicle detection with
defocusing effect. Measured CSL-VideoSAR observation
data containing a cooperative moving target is utilized to
test the detection performance of our developed NUAAMiniSAR system.
Vehicle keyframe represents the effective contents in
processing massive VideoSAR data used for indexing,
interactive browsing, semantic storage, and information
retrieval. As proposed in our previous work [27], a spatiotemporal extractor, called subaperture energy gradient
together with modified statistical and knowledge-based
object tracker (SEG-MSAKBOT), is designed to robustly
discriminate the keyframe information. It can be employed
to the detection of vehicle keyframe, as shown in Figure 11.
DECEMBER 2020

Figure 12.
Detection results by the TTD-PEADMM extractor for the singlechannel single-pass VideoSAR-LRSD framework [28].

Another computer vision-based detection perspective
is built on the single-channel single-pass VideoSAR lowrank plus sparse decomposition (LRSD). Scattering
objects often present some nondeterministic anisotropy
behaviors. Spatially and temporally varying disturbances
render the vehicle detection extremely difficult for the
VideoSAR-LRSD framework, and inhomogeneous video
disturbance represents an ill-posed problem. Consequently, the three-term decomposition with the proximal
exchange-based alternating directions method of multipliers (TTD-PEADMM) is designed for a more robust
vehicle detection. As shown in Figure 12, we provide
some detection results to explain this issue under the VideoSAR-LRSD framework. Comprehensive demonstration
and comparisons with the state-of-the-art algorithms have
been presented in our previous work [28] in detail. The
TTD-PEADMM-based VideoSAR-LRSD perspective can
offer the improved accuracy and competitive results.
Accordingly, our developed NUAA-MiniSAR system can
be a valuable instrument for detecting the moving target
with multitemporal defocusing effect in the automatic
monitoring task of dynamic events.

CONCLUSION
In this article, a multirotors NUAA-MiniSAR system and its
associated unified VideoSAR signal processing architecture
are presented in detail. The main limitations and technical
challenges of system design and video processing with such
a highly maneuverable platform have been taken into
account. Experimental results show that the cost-effective
UHR system is capable of obtaining the fine-focused video
sequences by the FPGA processor that provides a rapid
response and perception of observation scene.
Owing to its operation flexibility and public data release
service, not only does the useful teaching instrument in college curriculum assist students with more detailed insight
into the rationale of a prototype radar system and relevant

IEEE A&E SYSTEMS MAGAZINE

IEEE Aerospace and Electronic Systems Magazine - December 2020

Table of Contents for the Digital Edition of IEEE Aerospace and Electronic Systems Magazine - December 2020