Published in IET Radar, Sonar and Navigation
Received on 28th May 2011
Revised on 7th August 2011
doi: 10.1049/iet-rsn.2011.0190
ISSN 1751-8784
Adaptive range-spread target detection based
on modified generalised likelihood ratio test in
non-Gaussian clutter
T. Jian Y. He F. Su C. Qu
Research Institute of Information Fusion, Naval Aeronautical and Astronautical University, Yantai 264001,
People’s Republic of China
E-mail: iamjiantao@yahoo.com.cn
Abstract: Adaptive detection of a range-spread target is addressed for a possibly singular estimated covariance matrix, in non-
Gaussian clutter modelled as a spherically invariant random vector. Firstly, a modified generalised likelihood ratio test with
recursive estimator (MGLRT-RE) is derived. To improve the adaptability and to reduce the computational complexity of
MGLRT-RE, a simplified MGLRT (SMGLRT) is proposed and is proved to be constant false alarm rate (CFAR) to the
statistics of the texture theoretically. Based on secondary data, the heuristic SMGLRT-CA (cell-averaging) and MGLRT-RE-
CA are also designed. The SMGLRT outperforms the MGLRT and MGLRT-RE; similarly, the SMGLRT-CA with fully
CFAR properties outperforms the MGLRT-CA and MGLRT-RE-CA. The performance assessment conducted by Monte
Carlo simulation confirms the effectiveness of the proposed detectors.
1 Introduction
A high-resolution radar (HRR) can resolve a target into
a number of scatterers, by using pulse compression
techniques. Moreover, the multiple dominant scatterers
(MDS) in isolated range cells are usually referred to as a
so-called range-spread target [1]. As the detection strategies
of point-like targets may fail for range-spread targets [2],
the adaptive detection of range-spread targets has gained
more and more attention among the radar signal-processing
community. Moreover, the algorithms to detect range-
spread targets can also be used to detect a formation of
point-like targets with the same velocity that are spatially
distributed in range [3].
At present, range-spread target detection in Gaussian
background has been investigated largely, and many
valuable results have been achieved. For instance, with
some a priori statistical knowledge about the range-spread
target [4], the detection performance has been enhanced in
Gaussian white noise. In [5], when the estimated covariance
matrix of Gaussian clutter is possibly singular, a modified
generalised likelihood ratio test (MGLRT) detector is
derived with the upper-bounded constant false alarm rate
(CFAR) property. Furthermore, the MGLRT is applied to
the polarisation diversity [6], and the performance gain is
achieved for Gaussian scenarios. In particular, the detection
of moving range-spread target for airborne radar is
investigated in Gaussian ground clutter [7]. By constraining
the covariance matrix structure of Gaussian disturbance,
adaptive range-spread target detection is addressed without
secondary data in [8, 9].In[10], a double threshold
decision rule is devised, and the rejection performance is
emphasised particularly on. In [11], the range-spread target
detection is studied, by exploiting the image features of
cross time-frequency distribution of a pair of adjacent
received signals.
In particular, the partially homogeneous Gaussian
environment is also considered for range-spread target
detection, where the secondary data s hare the same
covariance matrix of the primary data, but possess possibly
different power levels. For example, the detectors relying
on the generalised likelihood ratio test (GLRT) or on a two-
step GLRT-based design procedure are proposed in [12]
and are assessed in [13].In[14], to cope with the a-priori
uncertainty, all the clutter data are assumed to be
clustered into groups of cells sharing the same disturbance
power value, and a certain degree of a-priori knowledge
about the rate of change of the power level in range is
retained. In addition, based on the GLRT, adaptive range-
spread target detection is addressed in homogeneous and
partially homogeneou s noise plus subspace interference
in [15].
In fact, the background clutter may no longer be modelled
accurately as a Gaussian random variable (RV) in the HRR
situations [16]. More specifically, at the higher-range
resolution, the radar system receives target-like spikes
representing non-Gaussian observations, which can be
better described as a spherically invariant random vector
(SIRV) [17]. For instance, in [18], with the known clutter
covariance matrix, two GLRT-based detectors for a range-
spread, Doppler-shifted target in SIRV clutter are
developed. Moreover, the GLRT-based adaptive detection
schemes of range-spread targets in SIRV clutter are
addressed partly in [19, 20].
970 IET Radar Sonar Navig., 2011, Vol. 5, Iss. 9, pp. 970–977
&
The Institution of Engineering and Technology 2011 doi: 10.1049/iet-rsn.2011.0190
www.ietdl.org